Home

The Michigan Model for Diabetes User Manual

1. a CappedGaussian3 10 60 The following table explains how the template distribution is set up to help the users understand how to set up and modify these distributions Corr_SBP_DBP Diabetes_Type_2 Do not change Alive to Deo not change Demographics Characteristics Age 60 2 6 8 CappedGaussian3 Duration_Of_Diabetes Male Race Bernoulli 0 10 1 1 White 2 Black BMI Height lif Male 1 7602 0 0742 CappedGaussian3 1 6281 0 0699 CappedGaussian3 Ce SBP ss 149 8421 4 CappedGaussian3 83 44 11 3 21 4 Corr_SBP_DBP SBP The function is 149 8 CappedGaussian3 1 mean_DBP SD _ Corr_SBP_DBP 2 11 3 DBP SBP_SD Co rr SBP_DBP SB P mean_SBP Cap pedGaussian3 1 Corr_SBP_DBP Max 0 Min 20 Michigan Model for Diabetes User Manual Exp Ln 1 7 0 45 CappedGaussian3 TotalCholesterol HDLCholesterol LDLCholesterol Triglycerides 0 456 HbA1c Max 5 7 Min 30 Exo CappedGaussian3 0 07 1 98 AF Bernoulli 0 05 be set to one No Cerebrovascular Cerebrovascu Disease lar disease Survive_ Stroke 1 No Cerebrovascular Disease sub model Coronary Angina heart disease CHFwoMI O S submodel CADwProc Oe Survive_Ml lif No_CVD Angina O 1 CHF O O No_ Nephropathy Bernoulli 0 9 Nephropathy sub model Micro _Albuminuria lif No Nephropathy O Bernoulli 0 30 Proteinuria 1 Micro Albuminuria No _Nephropath ESRD_Dialysis WO O o o Nl ESRD Transplant No_Neuropathy Bernoulli 0 9 Neuropathy Cli
2. e Alc over time SBP over time Alc Alc e 10 o 200 2 180 8 8 160 7 7 140 120 6 100 porting i eee eee eee ee a f Time Time Total Cholesteroal over time T C _ ee Se T 9 8 6 5 4 3 IIIT rrr rrr rrr rrr 0 1 2 3 4 5 NuU RUA AX OD THM Hite Ata ae ad Time 55 Michigan Model for Diabetes User Manual The following table shows the simulated incidence rate for the simulate population in this Complication Incidence rate 1000 PY example Example 4 To obtain confidence intervals for life expectancy and quality adjusted life expectancy stroke Revascularization Amputation Blind In Both Eyes Cardiovascular Death Cumulative Incidence 3 2 estimates This feature is currently only available to internal users The MMD group is working on providing it to external users 66 Michigan Model for Diabetes User Manual 9 Appendices Appendix A Michigan Model for Diabetes Disease Progression Model A1 Model Structure and Transition Probabilities Coronary Heart Disease 1 5 No CHD CHD Death i Competing Death Figure A1 Overall Structure of Michigan Model for Diabetes Keys Regular State Event State C Module C Terminal State Transition e gt Hidden transitions shown in Figures A2 A3 and A5 to A8 e gt Splitting transition to multiple sub processes No transi
3. Occurrence Probability Or CVD_Procedure_Ente Bernoulli 0 67 Cost_Comment YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlvFventC ost 1 Or Proliferative_left_Ent ESRD_Transplant_Entered Amputation_Entered Angina_Entered Or CVD_Procedure Ente Or And MI_Entered 1 C Or And MI_Entered CVD Or CHFwoML Entered E Stroke Fntered YearlyEventCost 1 0 1 YearlyEventCost 1101 YearlyEventCost 138071 YearlyEventCost 42929 YearlyEventCost 8282 YearlyEventCost Iif PC YearlyEventCost 41744 YearlyEventCost 60865 YearlyEventCost 34635 VearlvFventCost 55778 Set Yearly Cost to 0 ADD UP EVENT COST Add cost of amputation ESRD_transplant cost accordin Add cost of amputation Event Cost for Angina Procedure Cost 3 1 are CABG Mycardial Infarction without pr Mycardial Infarction with proc Add cost of CHF hospitalization Fvent cost for Stroke Cost QoL parameter 2 Change the event cost for amputation in the Function cell You can also modify the text Occurrence Probability a y Cr in the Notes cell to keep notes of this change Function Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Cost Quality of Life update rules
4. Treatment Parameters In the examples included in the MMD zip file we have set the value for treatment related parameters according to standard of practice in the US To change them click on Stage 0 Initiation to bring the following tab to the front lt PROJECT DEFINITION J a File Help Project Definition Simulation Save Gij I TEPE Name My First Diabetes Simulatic Created On 2015 07 28 12 32 34 591000 Notes How to simulate an i i Observational Stud Derived From Qbervational Study Templ Last Modified 2015 07 28 12 32 34 591000 Primary Model Michigan Model For Diabetes 201 v am No of Simulation Steps 10 ae Population Set My population m No of Repetitions 4 ae Racal al EE Stage 0 A lization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the imulation Affected Parameter Function Notes Threshold_SBP 140 Threshold for increasing the le Threshold_Alc T3 Threshold for increasing the le Threshold_LDL 150 38 67 Threshold for increasing the le Max_Level_ACE 8 Highest level of treatment with Max_Level_Diabetes Trt 5 Max_Level_Statin 2 Highest level of treatment with Max_Level_Aspirin 1 YearlyRateOfQuittingSm 0 03 YearlyRateOfStartAspirin 0 05 Equ_CompetingDeath 10 75 10 Use modified death rates f Discount_FirstlL0Years 0 03 L Discount After OVears
5. 18 Michigan Model for Diabetes User Manual The following table shows the definition for the six compliance parameters in the model program Definition and suggested range Function _ _ Compliance _ CVD Compliance_diabet es Compliance_ACE Compliance_statin Compliance_beta Compliance_ Aspirin The proportion of patients who are willing to comply with treatment for hyperglycemia dyslipidemia and hypertension and using aspirin when there is a CVD event This number should be relatively high and higher than all the rest of the compliance parameters The proportion of patients who comply with treatment for hyperglycemia regardless of history of CVD event The proportion of patients who comply with treatment for hypertension regardless of history of CVD event The proportion of patients who comply with treatment for dyslipidemia using statin regardless of history of CVD event The proportion of patients who comply with treatment using beta blocker regardless of history of CVD event The proportion of patients who comply with aspirin therapy regardless of history of CVD event Each parameter should be set to either equal 0 or one of the following pre set covariates Compliance_ 100 Compliance_95 Compliance_90 Compliance_ 10 Compliance_5 Number at the end of the name of each of the above covariates indicates the rate of compliance For example if you wish to set the prop
6. Stage 3 Update Treatment Stage 4 Update Costs Cost Quality of Life update rules in the simulation Notes Procedure 3 1 are CABG 3 2 ar Set Yearly Cost to 0 ADD UP EVENT COST Add cost of amputation ESRD transplant cost accordin Add cost of amputation Event Cost for Angina Procedure Cost 3 1 are CABG Mycardial Infarction without pr Mycardial Infarction with proc Add cost of CHF hospitalization Fwent cost for Stroke The MMD provides a utility module that can calculate yearly and cumulative values Table C1 in Appendix C shows the utility penalties related to patient characteristics and conditions Users can modify utility scores following the steps below using blind in both eyes as an example 24 Michigan Model for Diabetes User Manual 1 Highlight the utility score you would like to modify and click on the Down Arrow at the bottom of the window to bring down the parameter line to the editing cells Cost Quality of Life update rules in the simulation Affected Parameter Cost_Comment HealthUtilityScoreT his ear HealthUtilityScoreT his ear HealthUtilityScoreThis ear HealthutilityScoreT his ear HealthUtilityScoreT his ear HealthUtilityScoreT his ear HealthUtilityScoreT his ear HealthutilityScoreT his ear HealthutilityScoreThisVear ESRD_Dialysis HealthUtilityscoreThisVear ESRD Transplant Occurrence Probability 1 1 Female Ge BMI 30 Orf Metformin OtherOralMedication Or
7. The Michigan Model for Diabetes User Manual COPYRIGHT 2015THE REGENTS OF THE UNIVERSITY OF MICHIGAN Version 2 0 September 17 2015 Produced by the University of Michigan Michigan Center of Diabetes Translational Research MCDTR Disease Modeling Group http Awww med umich edu mdrtc cores MCDTR_MMCore DiseaseModel index html Michigan Model for Diabetes User Manual Condition of Use and Copyright Both the IEST software and THE MICHIGAN MODEL FOR DIABETES MMD COPYRIGHT 2015 THE REGENTS OF THE UNIVERSITY OF MICHIGAN are being released for use by researchers under a general public license Permission is granted to use create derivative works of copy and distribution of IEST and MMD only within the original licensee s organization for noncommercial education and research purpose subject to the following copyright and conditions No charge is made to academic organizations This tool is provided as is No condition is made or implied nor is any warranty given or to be implied as to the accuracy of this tool or that it will be suitable for any particular purpose or for use under any specific conditions The Regents of the University of Michigan disclaim all responsibility for the use which is made of this tool The University of Michigan shall not be liable for any damages including special indirect incidental or consequential damages with respect to any claim arising out of or in connection with the use of the tool ev
8. Update Costs Covariate update rules in the simulation Affected Parameter Compliance_15 Compliance_10 Compliance _5 COMMENT Compliance _ACE Compliance _beta If In State Eq Time 1 Eq Time 1 Eq Time 1 Occurrence Probability Function Bernoulli 15 20 Compliance_20 Bernoulli 10 15 Compliance_15 Bernoulli 5 10 Compliance_10 1 Compliance levels for diseases are defined Compliance _30 Compliance levels for treatment for hypertension Compliance _30 Compliance levels for beta_blocker Compliance _30 Compliance levels for treatment for dysalycemia Compliance _30 Compliance levels for treatment for dyslipidemia Compliance levels for treatment for dysalycemia dysli Age increase Generate random number used for correlations If in State 20 Michigan Model for Diabetes User Manual 5 You should see that the parameter is back in the list of parameters above with new value Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Covariate update rules in the simulation Affected Parameter If In State Occurrence Probability Function Compliance _15 Eq Time 1 Bernoulli 15 20 Compliance _20 Compliance _10 Eq Time 1 Bernoulli 10 15 Compliance_15 Compliance_5 Eq Time 1 Bernoulli 5 10 Compliance_10 COMMENT 1 1 pioen 1 Compliance_40 z i yan el Compliance_diab
9. Update Treatment Stage 4 Update Costs Cost Quality of Life update rules in the simulation Affected Parameter If In State PCI CABG YearlyEventCost Cost_Comment YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost VearlvFventC ost Occurrence Probability Or CVD_Procedure_ Ente 1 1 Or Proliferative_left_Ent ESRD_Transplant_Entered Amputation_Entered Angina_Entered Or CVD_Procedure_ Ente Or And MI_Entered 1 C Or And MI Entered CVD Or CHFwoML Entered E Stroke Fntered Function Bernoulli 0 67 0 1 YearlyEventCost 1101 YearlyEventCost 138071 YearlyEventCost 42929 YearlyEventCost 8282 YearlyEventCost Iif PC YearlyEventCost 41744 YearlyEventCost 60865 YearlyEventCost 34635 VearlvFventCost 55778 Notes S amp S Procedure 3 1 are CABG 3 2 ar E Set Yearly Cost to 0 ADD UP EVENT COST Add cost of amputation ESRD_transplant cost accordin Add cost of amputation Event Cost for Angina Procedure Cost 3 1 are CABG Mycardial Infarction without pr Mycardial Infarction with proc Add cost of CHF hospitalization Fvent cost for Stroke Sy Cost QoL parameter If in State aa ba 4 1 3 1 Defining cost values Occurrence Probability ta aza Function The MMD can calculate yearly and cumulative direct medical costs related to diabetes management and its co
10. superpack python2 7 exe download or http www scipy org Download and download the NumPy library Requires Python a version suitable for Python version 2 7 for Windows e Visit http sourcetorge net projects scipy tiles scipy 0 10 0 scipy 0 10 0 win32 superpack python2 7 exe download or http www scipy org Download and download the SciPy library Requires Python and NumPy a version suitable for Python version 2 7 e Visit http code google com p sympy downloads detail name sympy 0 7 1 win32 exe or http code qoogle com p sympy downloads list and download the Sympy library Requires Python Version 0 7 1 OS X installation e Python for OS X is included by default on all OS X installations e Install pip to assist with the installation of non standard Python modules used by the IEST software by visiting the following webpage http pip readthedocs org en latest installing html and downloading the get pip py file Save the file to your desktop e Open the application Terminal through Applications gt Utilities gt Terminal and issue the following commands o sudo python Desktop get pip py o sudo pip install numpy o sudo pip install scipy e Download wxPython2 8 12 ansi version NOT unicode like Windows from above by visiting the following webpage and install the subsequent dmg Michigan Model for Diabetes User Manual file http sourceforge net projects wxpython files wxPython 2 8 12 1 wxPython2 8
11. 396 282 in 2009 US for proteinuria and the event costs were assumed to be the same as the ongoing costs Nichols GA Vupputuri S Lau H Medical care costs associated with progression of diabetic nephropathy Diabetes Care 2011 34 2374 8 Based on Tables K7 K9 and K11 in the following report U S Renal Data System USRDS 2013 Annual Data Report Atlas of Chronic Kidney Disease and End Stage Renal Disease in the United States National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases Bethesda MD 2013 Based on Table 2 in the following study the event costs were derived from averaging total costs at 0 1 year in the incident heart failure cohort and the ongoing costs were derived from averaging each of yearly total costs between year 1 and year 5 in the incident heart failure cohort Liao L Jollis JG Anstrom Ku et al Costs for heart failure with normal vs reduced ejection fraction Arch Intern Med 2006 166 112 8 These data were from email consultation with Dr Christopher Hogan on March 19 2015 who Is the president of Direct Research LLC in Vienna VA These costs of death were the incremental per capita medical payments between the diabetes survivors in 2012 costs in the year of 2012 and the diabetes decedents in 2012 costs in the last 12 months of life who were Medicare fee for service beneficiaries with Part A and Part B enrollment and with any diagnosis of diabetes on any physician or hospit
12. Insert v Calibri X X ey EE General pE H 4 iB es nse Ay Ai Ea r a g Delete X l ic z o 59 09 Conditional Formatas Cell a Sort amp Find amp Formattingy Tabley Stylesy Ei Formate Filter Select Clipboard M x Alignment F Number x Styles Cells Editing Ea M12 A B C D E F G H I J K L M N O Diabetes_Type_2 Alive Age Duration_Of_Diabetes Male Race BMI Height SBP DBP Smoke HDLCholesterol LDLCholesterol Triglycerides TotalCholestel 1 1 The current version of the IEST software does not accept missing values When the data is ready save the file as a csv file and change the file name x WH ls MyPopulation csy Microsoft Excel f 52 Home Insert Page Layout Formulas Data Review View Developer a o ep 28 Fa fp e _ B HHA 3 Insert 2 ae Calibri y il Aa F yr General v zl E B Ba 5 oy 3 oerte E 2Y Paste B Z U Hr amp Ar E ZE HS l ar 9 0 00 Conditional Format Cell PS Sot amp Find amp v SA E ze a eee F Eas gt eta Formatting as Table Styles 9 Format lt 27 Filter Select i Clipboard A Font Alignment Number Ta Styles Cells i Editing i Al v fe Diabetes Type 2 B C D E F G H l J K L M N O Diabetes Type 2 Alive Age Duration_Of Diabetes Male Race BMI Height SBP DBP Smoke HDLCholesterol LDLCholesterol Triglycerides TotalCholester te 1 1 42 4 0 I 25 1 547 131 110 0 1 24 372 293 6 3 3 1
13. Simulation Name Created On 2015 07 28 12 32 34 591000 Notes How to simulate an Observational Study Last Modified 2015 07 28 12 50 55 266000 Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the simulation My First Diabetes Simulatic Derived From Qbervational Study Temple Primary Model No of Simulation Steps 49 Michigan Model For Diabetes 201 iu E Population Set My population No of Repetitions 4 Affected Parameter Function Notes Threshold_SBP 140 Threshold for increasing the le Threshold_Alc 75 Threshold for increasing the le Threshold_LDL 150 38 67 Threshold for increasing the le Max_Level_ACE 8 Highest level of treatment with On the Update Cost tab you can find a series of updating rules for calculating event costs ongoing costs and utility values Project Definition Simulation Name My First Diabetes Simulatic Derived From Qbervational Study Templ Primary Model Population Set My population Created On Michigan Model For Diabetes 201 u gt oJ 2015 07 28 12 32 34 591000 Last Modified 2015 07 28 12 50 55 266000 No of Simulation Steps 49 No of Repetitions 1 Notes How to simulate an Observational Study vr Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3
14. when applying the model to a population in a country with less access to revascularization procedures users can adjust the transition probabilities to match the revascularization procedure rates in their countries The current MMD software provides raw simulated data for all simulated individuals e g risk factors complications status yearly medical cost and utility score for each simulated year We provide SAS programs that can generate estimates of life expectancy quality adjusted life years and costs of complications for the working examples in Section 8 The provided SAS programs can also output longitudinal trajectories for important risk factors cumulative event rates and long term history rates Using the raw results users can also write their own programs to summarize other quantities of their own interest Michigan Model for Diabetes User Manual 2 Changes in Version 2 0 The MMD has been substantially revised since its original publication in 2005 Zhou et al 2005 and is implemented by using newly developed software that models chronic diseases New features of the MMD include 1 Modeling disease progression through evolution of multiple biomarkers and risk factors 2 An updated coronary heart disease sub model that incorporates the possibility of recurrence of myocardial infarction MI congestive heart failure and cardiac procedures either before or after MI 3 Modeling modern diabetes treatment regimens and managemen
15. 0 15 a Cost QoL Wizard Covariate If in State Occurrence Probability Function Notes b a NE vv u gi On this tag there are eight parameters that are used to set up treatment thresholds maximum levels of treatment allowed in the simulation and yearly rate of quitting smoking and starting aspirin See Appendix A2 for how treatments are specified in MMD The eight parameters are described in the following table Threshold_A1c At the end of each year if the HbA1c level is higher than the threshold level specified anti hyperglycemia treatment will be increased by 1 level for compliant patients Threshold_ SBP mmHg At the end of each year if the SBP level is higher than the threshold level specified treatment for hypertension treatment will be increased by 1 level for compliant patients Threshold LDL mmol L At the end of each year if the LDL level is higher than the threshold level specified treatment for dyslipidemia will be increased by 1 level for compliant patients Max_Level_Diabetes_Irt There are totally 6 levels of anti hyperglycemia treatment defined in the MMD 0 No treatment 1 Diet and exercise 2 One oral non insulin medication metformin 3 Two oral non insulin medications metformin 14 Michigan Model for Diabetes User Manual sulfonylureas 4 Basal insulin 5 Intensive bolus insulin You can set this parameter to any integer between 0 and 5 See Appendix A2 for the effect of
16. 1 13 2 0 Cardiovascular Death 15 1 26 7 Example 3 Users may want to simulate disease progression for a population with known distributions of characteristics instead of a single subject To undertake this type of simulation proceed as follows Step 1 Duplicate the project Interventional Study Template and rename it as Example 3 In the Population Set dropdown menu select Template for Specifying Distribution as shown in Section 5 2 54 Michigan Model for Diabetes User Manual Step 2 Set the No of Simulation Steps to 5 years the No of Repetition to 5000 Use the default setting of interventional study template Step 3 Run the model and then export the data to a csv file Use the included SAS program Example3_Summary sas to generate report of simulation results The QALE should be approximately 2 80 0 41 QALYs Total cost is approximately 31 768 Estimates may differ slightly between simulations as the MMD may have used a different set of random numbers To generate these estimates the model has simulated values for smoking status total LDL amp HDL cholesterol systolic amp diastolic blood pressure and HbA1c for each year based on the baseline risk factor values entered built in treatment regimens treatment thresholds specified and compliance rates The following figures show the individual and population average time paths for a few of these risk factors
17. 2 diabetes and BMI of 30 kg m who is treated with diet and exercise and has no microvascular neuropathic or cardiovascular complications Costs are expressed in year 2014 US dollars using the general Consumer Price Index to reflect inflation According to the statements in 2 JACC papers about one third of patients undergoing PCI in the US have diabetes see page e83 in the attached File 1 and about 35 of CABG patients have diabetes See page e167 in the attached File 2 Also according to a recent Circulation paper it was estimated that in 2010 in the US 492 000 patients underwent PCI while 219 000 underwent CABG see page e275 in the attached File 3 With calculations using these data what we could have is The estimated number of diabetic patients treated with PCI in 2010 in the US would be 164 000 492 000 1 3 while that treated with CABG would be 76 650 219 000 0 35 Thus based on these 2 calculated numbers we could get that about 68 of diabetic patients who need the coronary revascularization procedures may use PCI while 32 of them may get CABG 87 References Ts 2 Brandle M Zhou H Smith BR et al The direct medical cost of type 2 diabetes Diabetes Care 2003 26 2300 4 Based on Table 2 in the following study the ongoing costs for retinopathy related complications except blindness were assumed to be 75 in 2000 US and the event cost for nonproliferative retinopathy was assumed to be the same as the o
18. 50 and 40 comply with treatment for hyperglycemia beta blocker dyslipidemia hypertension and aspirin respectively This means 90 of patients are willing to comply with hyperglycemia treatment dyslipidemia treatment hypertension treatment and aspirin when there is a CVD event Among the above 90 of patients 8 out of 9 80 of the initial sample comply with treatment for hyperglycemia regardless of their CVD complication history among the 80 of compliers with treatment for hyperglycemia 7 out of 8 70 of the initial sample comply with the prescription of beta blocker etc among the total population 40 comply with all five treatments regardless of their CVD complication history To implement the above treatment and compliance rules the simulation program does the following Before the start of the simulation cycle each patient is assigned a treatment specific compliance profile that includes six variables one for compliance when there is a CVD event and five for treatment specific compliance rates i e one for each of five types of treatments To set up the simulation a user needs to specify the four following sets of parameters 1 Treatment threshold parameters 2 Parameters for maximum level of treatment 3 Yearly rates for starting aspirin and quitting smoking 4 Compliance rate parameters Next we will show how to specify treatment and compliance related parameters 13 Michigan Model for Diabetes User Manual
19. Basallnsulin Insulin Blind_One_Eye_Only Blind_Both_Eyes Orf Micro_Albuminuria Proteinuria 1 1 HealthUtilityScoreThis ear 0 038 HealthUtilityscoreThis ear 0 021 HealthUtilityScoreThisVear 0 023 HealthUtilityScoreThis ear 0 034 HealthUtilityScoreThis ear 0 045 HealthUtilityScoreThis ear 0 1 0 HealthUtilityScoreThis ear 0 011 HealthUtilityScoreThisVear 0 078 HealthUtilityscorel his ear 0 078 4 Cost QoL parameter If in State ally Occurrence Probability Function Cost QoL Wizard Notes ca Notes Calculati Set Healt Female p obese pe Penalty fi using ins Penalty fi Penalty fips 2 Change the event cost for amputation in the Function cell You can also modify the text in the Notes cell to keep notes of this change Cost Quality of Life update rules in the simulation Affected Parameter If In State Cost_Comment HealthUtilityScoreThis ear HealthuUtilityScoreThisY ear HealthuUtilityScoreThisY ear HealthUtilityScoreThisY ear HealthuUtilityScoreThis Year HealthutilityScoreThisY ear HealthUtilityScore This Year HealthutilityScoreThisYear ESRD_Dialysis HealthUtilityScoreThis ear ESRD_Transplant HealthUtilitvyScoreThisYear Clinical Neuropathy Occurrence Probability 1 1 Female Ge BMI 30 Or Metformin OtherOralMedication Or Basallnsulin Insulin Blind_One_Eye_Only Or Micro_Albuminuria Proteinuria 1 1 1 Function 0 689 HealthuUtilityScoreThisYear 0 038
20. Function Notes oi H a H Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Covariate update rules in the simulation Affected Parameter If In State Occurrence Probability Function Notes Was_OtherOralMedication Eq Time 1 OtherOralMedication Was_BasalInsulin Eq Time 1 BasalInsulin E Was_Insulin Eq Time 1 Insulin a Started_Diabetic Eq Time 1 Sets to 1 if diabetic at start eii COMMENT 1 Define Intervention For Year 1 Add1_Diabetes Eq Time 1 Eq Insulin 0 Compliance_diabetes Ge A ic Threshold_A ic Add1_ Statin Eq Time 1 Compliance_statin Ge LDLCholesterol Threshold Intervention on dyslipidemia in Add1_Aspirin Eq Time 1 Eq Aspirin 0 Compliance_Aspirin Bernoulli YearlyRateOfStartA Taking Aspirin in Year 1 Smoke Ge Time 1 Eq Smoke 1 Bernoulli 1 YearlyRateOfQuittingSmoking smoking cessation in Year 1 COMMENT 1 Update risk factors Age Ge Time 2 Age 1 Age increase Temp_x1 1 CappedGaussian3 Generate random number used 4 1 Covariate If in State Occurrence Probability Function Notes Add1_ACE Eq Time 1 Coming ACE Ge SEP a Intervention on hypertension in Year 1 29 Michigan Model for Diabetes User Manual 6 You now can modify the function in the editing window For example below we modify the function so that the treatment threshold for hypertension at baseline
21. HealthUtilityScoreThis ear 0 021 HealthUtilityScoreThisYear 0 023 HealthutilityScoreThisYear 0 034 HealthutilityScoreThis ear 0 043 HealthUtilityScoreThis ear 0 011 HealthutilityScoreThisYear 0 078 HealthUtilityScoreThis ear 0 078 HealthUtilityScoreThisYear 0 065 4 SS e e Cost QoL parameter HealthutilityScoreThisY If in State rai Occurrence Probability Blind_Both_Eyes Cost QoL Wizard Notes DEG Penalty for both eyes blind Notes Calculati Set Healt Female p obese pe Penalty fi using ins Penalty fi Penalty fi Penalty l Penalty fi Penalty fi 7 j m 25 Michigan Model for Diabetes User Manual 3 When you are done with modifying click on the Up Arrow and bring back the parameter to the cost utility window Cost Quality of Life update rules in the simulation Affected Parameter If In State Cost_Comment HealthUtilityScore This Year HealthUtilityscoreThisYear HealthUtilityScore This Year HealthUtilityScoreThis Year HealthUtilityScore This Year HealthUtilityScoreThis ear HealthUtilityScore This Year HealthUtilityScoreThisVear ESRD _ Dialysis HealthUtility coreThisYear ESRD _ Transplant HealthUtilityScoreThisYear Clinical Neuropathy 4 Cost QoL pararneter If in State HealthUtilityScoreThisY Occurrence Probability 1 1 Female Ge BMI 30 Or Metformin OtherOralMedication Or Basallnsulin Insulin Blind_One_Eye_Only Or Micro_Albumin
22. Max_Level_Diabetes_Trt Max_Level_Statin Highest level of treatment with Max_Level_Aspirin YearlyRateOfQuittingSm YearlyRateOfStartAspirin Equ_CompetingDeath 10 Use modified death rates f a x conto If in State Occurrence Probability Function e 6 1 Select the population set and set number of subjects Use the dropdown menu to select the Population Set you would like to conduct the simulation on oH ROJECT File Help Project Definition Simulation Name My First Diabetes Simulatic Created On 2015 07 19 19 21 38 652000 Notes How to simulate an Observational Study Derived From Example Project 1 Last Modified 2015 07 19 19 21 58 203000 Primary Model Michigan Model For Diabetes 201 No of Simulation Steps 15 Run Simulation Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the simulation Affected Parameter Function Notes Threshold_SBP 150 Threshold for increasing the le Threshold_Alc 11 Threshold for increasing the le 44 Michigan Model for Diabetes User Manual lf you are using a population set defined by distributions to set the number of subjects to be included in the simulation write down the number of subjects in the small window of No of Repetitions PROJECT DEFINITION Ecc Ea File Help Project Definition S
23. O O QO 31 32 35 38 44 44 45 45 47 49 57 87 89 90 Michigan Model for Diabetes User Manual 1 Introduction and Background The Michigan Model for Diabetes MMD is a computerized disease model that enables the users to simulate the progression of diabetes over time its complications retinopathy neuropathy and nephropathy and its major comorbidities cardiovascular and cerebrovascular disease and death Transition probabilities can be a function of individual characteristics current disease states or treatment status The model also estimates the medical costs of diabetes and its comorbidities as well as the quality of life related to the current health state of the subject In contrast to other proposed models the transition probabilities implemented in the MMD were obtained by synthesizing the published literature Specifically transition probabilities in the newly updated coronary heart disease sub model that reflects the direct effects of medical therapies on outcomes were derived from the literature and calibrated to recently published population based epidemiologic studies and randomized controlled clinical trials This method not only allowed us to build a model without access to individual level data from a long term prospective study but allowed us to update the model by incorporating data from new studies as they become available In addition different from other proposed models our model allows a user to contro
24. Penalty fi using ins Penalty fi Penalty fi z Penalty fi Penalty f Penalty fi t Cost QoL parameter If in State HealthUtilityScoreThisY v is a Occurrence Probability v Blind_Both_Eyes Cost QoL Wizard Function v HealthUtilityScoreThisY v Penalty for both eyes blind Notes 4 1 3 3 Discount rates The MMD allows the users to set the annual discount rate to be applied to life expectancy quality adjusted life expectancy and medical cost estimates Two different discount rates can be applied for example a discount rate of 0 03 8 can be specified for the first 10 years and then 0 015 1 5 for all subsequent years If discounting is not required enter O To modify the discount rates click on the Stage 0 Initialization tab and use the Down Arrow and Up Arrow at the bottom of the tab 26 Michigan Model for Diabetes User Manual E PROJECT DEFINITION babe File Help Project Definition Simulation E R Unda Name My First Diabetes Simulatic Created On 2015 07 28 12 32 34 591000 Motes How to simulate an a f ye Observational Stu Derived From Qbervational Study Temple Last Modified 2015 07 28 12 50 55 266000 ty x Primary Model Michigan Model For Diabetes 201 m No of Simulation Steps 10 Run Simulation Population Set jy population No of Repetitions 41 Vien Result Delete Results Initialize the simulation Aff
25. SD of change 3 5kg SD of change 0 3kg year UKPDS 13 1995 exercise weight loss Metformin one Mean change 2kg Mean change 0 3kg year Kahn et al 2006 OAD non insulin SD of change 0 3kg SD of change 0 3kg year med Metformin Mean change 2kg Mean change 0 kg year Phung et al 2010 Sulfonylureas two SD of change 1kg SD of change 0 3 kg year OADs non insulin meds Add Basal insulin to Mean change 1 9kg Mean change 0 8kg year Holman et al 2009 OAD non insulin SD of change 4 2kg SD of change 0 5kg med Intensive insulin Mean change 1 2kg Mean change 0 8kg year Rosenstock et al 2009 therapy SD of change 0 5kg SD of change 0 5kg year A2 2 Changes in HbA1c There are 6 levels in glycemic control treatment 77 0 No treatment 1 Diet and exercise 2 Oral non insulin medication metformin 3 Two oral non insulin medications metformin sulfonylureas 4 Basal insulin 5 Intensive bolus insulin Changes of HbA1c for patients under each treatment is described in Table A11 Patient will transition to next stage when HbA1c level becomes 2 7 Table A11 Changes of HbAic under different anti hyperglycemia treatment scenarios Anti Initial effect first year change Changes after one year hyperglycemia treatment Treatment Level 0 Mean change 0 35 year No treatment SD of change abs mean change 3 Treatment Level 1 Mean change 1 9 Mean change 0 2 year Intensive lifestyle 0 5 currentHbA1c 9 1
26. SD of change abs mean diet and SD of change abs mean change 3_ change 3 exercise weight loss Treatment Level 2 Mean change 1 0 Mean change 0 14 year Metformin one 0 5 currentHbA1c 8 3 SD of change abs mean OAD non insulin SD of change abs mean change 3_ change 3 med Treatment Level 3 Mean change 0 8 Mean change 0 2 year Metformin 0 5 currentHbA1c 8 3 SD of change abs mean Sulfonylureas two SD of change abs mean change 3 change 3 Comments This way HbA1c will increase about 2 in 6 years on average for diabetics who are not appropriately treated UKPDS Group 1998 Figure 2 showed 1 5 increase in 6 years It was arbitrarily increased to reflect faster increase without any treatment An arbitrary variation was added to allow the change to be between zero and twice the value calculated from the references UKPDS 13 1995 UKPDS 33 1998 Sherifali et al 2010 Kahn et al 2006 Phung etal 2010 Charbonnel et al 2005 78 OADs non insulin meds Treatment Level 4 Mean change 0 8 Add Basal insulin 0 5 currentHbA1c 8 4 to OAD non insulin SD of change abs mean change 3 med Treatment Level 5 Mean change 1 2 CurrentHbA1c Intensive insulin 8 2 0 5 therapy SD of change 0 326 Reference for initial change Reference for change after one year A2 3 Changes in lipids Mean change 0 2 year SD of change abs mean change 3 No change H
27. User Manual es a retinopath Right eye retinopathy Nonproliferative_right Right eye has non sub model 1 Yes O No proliferative retinopathy Proliferative_right Right eye has 1 Yes O No proliferative ue retinopath Blind Eye_right No _Macular_edema_left Left eye does not Left eye retinopathy 1 Yes O No have macular sub model edema If left eye is blind Macular edema left both variables should macular edema be set to be 0 No_Macular_edema_right Right eye does not Right eye retinopathy 1 Yes O No have macular sub model edema If right eye is blind Macular _edema_right Right eye has both variables should 1 Yes O No macular edema be set to be 0 Medication There are five stages for anti OtherOralMedication Two or more oral non insulin hyperglycemia treatment in MMD These five stages are mutually exclusive of each other At most Basallnsulin only one of them can 1 Yes 0 No 1 Yes O0O No medications e g metformin sulfonylureas be set to 1 and the rest of them need to be set to zero If a subject is on both insulin and metformin s he should be considered as at the 5 stage treatment for hyperglycemia and therefore only the variable Insulin is set to be 1 Beta_Blocker Whether a subject is taking beta blocker es O No Ace_Inhibitor Whether a subject is taking any es O No hypertension medication that is no beta blocker medication for dyslipidemia Whethe
28. also modify the text in the Notes cell A Initialize the simulation Affected Parameter Function Notes Threshold_SBP 140 Threshold for increasing the le Threshold_LDL 150 38 67 Threshold for increasing the le Max_Level_ACE Highest level of treatment with Max_Level_Diabetes_Trt Max_Level_Statin Highest level of treatment with Max_Level_Aspirin YearlyRateOfQuittingSm YearlyRateOfStartAspirin Equ_CompetingDeath 10 Use modified death rates f Discount_FirstLOYears Discount_Afterl0Years asr wa Covariate F in State Occurrence Probability action Threshold_Alc v Threshold for increasing the level of 16 Michigan Model for Diabetes User Manual 3 Click on the Up Arrow Stage Q Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the simulation Affected Pararneter Threshold_SBP Threshold LDL Max_Level_ ACE Max_Level_Diabetes_Trt Max_Level_ Statin Max_Level_Aspirin YearlyRateOfQuittingSm YearlyRateOfStartAspirin Equ_CompetingDeath Discount_FirstLOV ears Discount_Afterl0Vears Covariate Function 140 150 38 67 Threshold _Alc 7 v ewe Wi Occurrence Probabwity Function Notes Threshold for increasing the le Threshold for increasing the le Highest level of treatment with Highest level of treatment with 10 Use modified dea
29. can be saved and would need to be re done This is a problem the future version of IEST will fix 48 Michigan Model for Diabetes User Manual 8 Worked Examples Example 1 To determine the likely impact of a difference in HbA1c values at the time of diagnosed type 2 diabetes say 11 0 versus 7 0 on Life Expectancy and Quality Adjusted Life Expectancy for a fifty year old white male patient proceed as follows Step 1 Using the nout Population Template csv file enter characteristics for two patients that have identical risk factor levels except for their HbA1c level Variable Name Diabetes Type 2 Alive O1 30 kg m 3 0 mmol L 1 6 mmol L 4 9 mmol L 7 for subject one and 11 for subject two Oe a Disease Status Within the same sub model defined below one and only one variable should be set to one No cerebrovascular disease No _Cerebrovascular_ Disease 1 Survive _Stroke No CVD No coronary heart disease Angina CHFwoMl CADwProc Survive_ Ml No _ Nephropath Micro Albuminuria Proteinuria ESRD_ Dialysis ESRD_ Transplant No_Neuropath 49 No nephropathy L TI Michigan Model for Diabetes User Manual zeo a O Amputation oo No left eye retinopathy Nonproliferative_left JO Proliferative left O Blind_Eye_left O No right eye retinopathy Nonproliferative right 0 Proliferative right Blind _Eye_right oOo o No left eye retinopathy Macular_edema left O
30. false Or x1 x2 x3 will perform a Boolean OR operation on two or more inputs And x1 x2 x3 will perform a Boolean AND operation on two or more inputs Not x will perform a Boolean Not operation on a single input Is True x will return 1 for a numeric x that is not 0 Will return O otherwise D5 Mathematical functions Exp x exponential Log x n logarithm of base n Ln x natural logarithm Log10 x decimal logarithm 90 Pow x n power operator similar to Sqrt x square root operator similar to 0 5 Pi the mathematical constant approximately equal to 3 14159 Mod x n Modulus of base n Abs x Absolute value of x Floor x closest integer equal to or below x Ceil x closest integer equal to or above x Max a1 a2 a3 the maximum value in the list Min b1 b2 b3 the minimum value in the list D6 Random number generators These random functions can be used to define the distribution of parameters Bernoulli p Binomial n p Geometric p Uniform a b the arguments a and b define the lower and upper limits of the interval Gaussian mean std D7 Cumulative distribution functions The last argument x represents a number for quantiles Bernoulli p x Binomial n p x Geometric p x Uniform a b x the arguments a and b define the lower and upper limits of the interval Gaussian mean std x D8 Control lif Statement TrueResult FalseResult Returns TrueResult if Statement is not 0 Fals
31. for treatment Compliance _Aspirin 1 Compliance _100 Compliance levels for aspirin Compliance_beta 1 Compliance_100 Compliance levels for beta_blocke Compliance_diabetes 1 Compliance _100 Compliance levels for treatment Compliance_statin 1 Compliance_100 Compliance levels for treatment Compliance_CVD 1 Compliance _100 Compliance levels for treatment 50 Michigan Model for Diabetes User Manual Step 3 Run the model and then export the data to a csv file Use the included SAS program Example1 amp 2_Summary sas to summarize the simulation results The default setting in this program summarizes the results for subject one To get Summaries on subject two change the P statement in the first data step in the program The quality adjusted life expectancy for subject one should be approximately 18 6 3 9 years 11 0 2 3 QALYs and for subject two with the higher HbA1c at the beginning slightly smaller at approximately years 18 5 3 9 10 6 2 3 QALYs Total cost is approximately 108 024 for subject one and 129 549 for subject two Estimates may differ slightly between simulations as the MMD may have used a different set of random numbers To generate these estimates the model has simulated values for smoking status total LDL amp HDL cholesterol systolic amp diastolic blood pressure and HbA1c for each year based on the baseline risk factor values entered built in treatment regimens treatment threshold specifie
32. of Franklin et al MI gt Procedure after MI Repeat MI 63 Jensen et al 2011 2004 Deedwania 2011 Jensen et al 2011 MI 25 xP CHF t Deedwania 2011 i gt CHF after MI Repeat MI 37 xP CHF t Jensen et al 2011 Deedwania E 1 h Short term survival of A 7 2 oa after MI gt Deedwania 2 f Procedure after MI year of MI Deedwania a oT 64 Michigan Model for Diabetes User Manual CHF after MI For subject has CHF before MI 78 75 as For subject has no CHF before MI 78 75 xP CHF T Repeat MI Jensen et al 2011 For subject has CHF before repeat MI 81 xP CHF t Deedwania 2011 For subject has no CHF before repeat MI 81 xP CHF tT j Procedure after MI gt Hx MI Jensen et al 2011 Franklin et al of MI For subject has CHF before MI 0 2004 For subject has no CHF before MI 78 75 x 1 Jensen et al 2011 P CHF T Repeat MI Jensen et al 2011 For subject has CHF before repeat MI 0 Deedwania 2011 For subject has no CHF before repeat MI 78 75 x 1 P CHF t year of MI gt CHD death year of MI gt CHF after MI year of MI gt Hx of MI tP CHF 0 13 Age _Modifier Gender _Modifier 0 45 Medication _Modifier for MI module P CHF 0 13 Age_Modifier Gender_Modifier Medication Modifier for repeat MI module 65 Michigan Model for Diabetes User Manual The age and gender modifier in the P CHF equations in Table A2 are shown in Table A3 Table A3 Age and Gender Modifier in Table
33. time has elapsed since the simulation starts 46 Michigan Model for Diabetes User Manual 7 Outputs When simulation is completed click OK on the pop up window that informs you the completion of the simulation To view results click on the View Result button 5 PROJECT DEFINITION o e File Help Project Definition Simulation aun Name Example Project 1 Created On 3014 11 14 12 30 36 266000 Notes How to simulate an Observational Derived From Last Modified 2915 07 20 13 24 47 788000 PAR ii Primary Model Michigan Model For Diabetes 2015 a uw No of Simulation Steps 4 Run Simuiation Population Set Example Population 1 ssa No of Repetitions 4 Delete Results Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs On the pop up window select the number of rows you would like to view in IEST and then click OK Rows to show Define number or records to show The result grid is very large 818 rows x 514 columns and itis probably not practical to show it on the screen Please decide how many rows you wish to view Note that you can later export the results to view them in full in another application Pressing Cancel will show all rows and in some cases may overwhelm the system cancel The following window shows the simulate
34. 0 20 Atrial fibrillation 1 Yes O No Disease Status Within each sub model defined below one and only one variable should be set to one No Cerebrovascular _ Disease No Cerebrovascular cerebrovascular disease sub model disease Survive _ Stroke histor coronary heart disease sub model yee T wU U U 1 Yes O0O No 1 Yes 0 N o Michigan Model for Diabetes User Manual J isles Angina Coronary artery disease without 1 Yes O No history of MI or heart failure CHFwoMl History of heart 1 Yes O No failure but not MI CADwProc History of 1 Yes 0O No revascularization procedure with no history of MI Survive_ Ml History of MI can 1 Yes O No be more than once with no history of heart failure failure and history of MI model Micro Albuminuria Microalbuminuria 1 Yes 0O No is defined as 30 bit lt ACR lt 300 ESRD_ Dialysis End stage renal disease with need of dialysis but no history of transplant 1 Yes O0O No ESRD_ Transplant End stage renal 1 Yes O No disease with history of transplant Neuropathy sub sensory neuropathy Amputation History of 1 Yes O No amputation due to diabetic neuropathy No_ Proliferative Retinopathy left Normal left eye Left eye retinopathy 1 Yes 0 No Nonproliferative_left Left eye has non sub model 1 Yes 0O No proliferative retinopathy Proliferative_left Left eye has 1 Yes 0 No 223 Michigan Model for Diabetes
35. 0648 x3 Diff_Ln_LDL 0 0738 Age 0 00412 Age Age 0000463 Ln_ Triglycerides 0 0114 Ln_LDL 138 Ln_HDL 0 00620 Diff_Ln_FastingGlucose 0 0821 Diff_ BMI 0 00906 Female 0 00600 0 111 x2 0 00206 x3 Diff_Ln_Triglyceride 157 Age 0 00728 Age Age 0000660 Triglycerides_Ln 112 Ln_LDL 0 0189 Ln_HDL 0496 Diff_Ln_FastingGlucose 0 268 Diff_BMI 0 0275 Female 0 021540 1359 x1 0 00734 x2 0 0189 x3 Diff_Ln_HDL future change in Ln_HDL Diff_ Ln_LDL future change in Ln_LDL Diff_Ln_triglyceride future change in Ln_triglyceride Ln_HDL logarithm e based transformed current HDL Ln_LDL logarithm e based transformed current LDL Ln_ Triglycerides logarithm e based transformed current triglyceride Diff_Ln_FastingGlucose future change in logarithm e based transformed current fasting glucose mmol L Diff_BMI future change in BMI A2 4 Changes in blood pressure Drug effect We assume a patient can go through a maximum of 9 levels of anti hypertensive treatments including no treatment 0 No anti hypertensive treatment one drug half dose one drug full dose two drugs half dose two drugs full dose three drugs half dose three drugs full dose four drugs half dose four drugs full dose COON OOF WN ACE inhibitor ARB will be the first drug to be added regardless of whether a patient is receiving B blocker or not 80 Table A12 Effect of anti hypertensive treatment Drug effect Comments treatment ch
36. 1 Bernoulli 0 244 0 10 1 0 10 This variable is an indicator for the state CAD w o MI as in shown in Appendix A For historical reason this variable name for this state was name as Angina in the software This variable is an indicator for the state CHF after MI as in shown in Appendix A For historical reason this variable name for this state was name as CHF in the software 43 Michigan Model for Diabetes User Manual 6 Running the Model To run the model use the project window to set the following parameters and then to start the simulation a fon Ex File Help Project Definition Simulation Ce e OOOO O O OoOO N U O Name My First Diabetes Simulatic Created On 2015 07 19 19 21 38 652000 Notes How to simulate an a Observational St Derived From Example Project 1 Last Modified 2015 07 19 19 21 58 203000 pune Primary Model Michigan Model For Diabetes 201 v No of Simulation Steps 15 Population Set Example Population 1 v ise No of Repetitions 4 Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the simulation Affected Parameter Function Notes Threshold_SBP Threshold for increasing the le Threshold_Alc Threshold for increasing the le Threshold_LDL Threshold for increasing the le Max_Level_ACE Highest level of treatment with
37. 1 1 Initialization Age Ge Time 2 Age 1 Age increase 4 m Covariate Ifin State Occurrence Probability Function Notes z 2 Change the value for this parameter in the Function cell You can also modify the text in the Notes cell Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Covariate update rules in the simulation Affected Parameter If In State Occurrence Probability Function Notes Compliance_15 Eq Time 1 Bernoulli 15 20 Compliance_20 Compliance _10 Eq Time 1 Bernoulli 10 15 Compliance_15 Compliance_5 Eq Time 1 Bernoulli 5 10 Compliance_10 COMMENT 1 1 Compliance levels for diseases are defined Compliance_ACE 1 Compliance _30 Compliance levels for treatment for hypertension Compliance_beta 1 Compliance _30 Compliance levels for beta_blocker Compliance_diabetes 1 Compliance _30 Compliance levels for treatment for dysglycemia Compliance_statin 1 Compliance _30 Compliance levels for treatment for dyslipidemia Compliance_CVD 1 Compliance_90 Compliance levels for treatment for dysglycemia dysli COMMENT 1 1 Initialization Age Ge Time 2 Age 1 Age increase Generate random number used for correlations If in State Compliance _Aspirin x 3 Click on the Up Arrow Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4
38. 1 2 This is an assumption that will be kept until a reference with more information is introduced Note that for non insulin takers the number actually Originates from the non proliferative to Blindness transition since the proliferative to Blindness transition inherits this number 75 A1 6 Other death Table A9 Transition probabilities for death due to non diabetic causes Death probability Alive to Other Death 0 0006 to 0 0546 The data was retrieved from http www cdc gov nchs hdi htm in which the depends on table topic of Mortality and life expectancy was selected and then the table of age gender race Mortality by underlying and multiple cause ages 18 US 1981 2013 Source E NVSS was selected Rates underl was selected in the Measure section and the data of year 2013 was used The rates of death due to non diabetic causes were calculated as a summary of all death rates with a given cause selected as the underlying cause of death except for diabetes major cardiovascular diseases and kidney diseases Thus these data would represent deaths from causes other than those that have been already counted in other sub models and these data depended on age gender and race ethnicity A2 Cardiovascular risk factors and related treatments Besides glycemia level we also model weight BMI lipid profiles and systolic and diastolic blood pressures SBP and DBP Each year the model updates glycaemia level and o
39. 1 39 3 0 2 31 1 52 150 90 0 1 22 4 59 2 55 6 98 4 1 1 61 10 0 23 1 52 140 90 0 123 2 99 Ppl 5 18 5 1 1 60 9 0 2 33 1 62 140 90 0 1 08 5 69 yy ak 8 01 6 1 1 64 10 0 1 32 1 51 131 90 0 0 92 2 56 0 8 3 68 E 1 1 26 0 0 1 34 1 513 155 90 1 0 8 1 86 1 91 3 53 8 ut 1 64 15 1 1 31 1 78 138 89 ut 1 12 FAL 2 49 5 37 9 1 1 50 12 0 1 28 1 65 121 68 0 1 16 6 6 2 8 82 To read in the population data do the following steps 1 Click on the Populations button on the left side of the main window to open the population sets window If you have your Project Definition window open you need to first close it to have access to the main window 235a Michigan Model for Diabetes User Manual File Forms Help Indirect Estimation and Simulation Tool Project Narne Project Type Notes Add New Project Obervational Study Template Simulation How te simulate an Observational Study Interventional Study Ternplate Simulation How to simulate an interventional study My First Diabetes Simulation Simulation How te simulate an interventional study Study Model Transitions Probabilities Populations Parameters 2 Click the Add button on the Population Sets window to start creating a new population set File Help B Mame Definition Type Created On Notes Source Dered From Last Modified Template for Specifying Di Distribution based 2015 07 23 12 12 57 3 2015 07 27 14 26 08 1 3 Name your population data
40. 9 5 629 35 Kahn SE Haffner SM Heise MA Herman WH Holman RR Jones NP Kravitz BG Lachin JM O Neill C Zinman B and Viberti G for the ADOPT Study Group Glycemic Durability of Rosiglitazone Metformin or Glyburide Monotherapy N Eng J Med 2006 355 2427 2443 Kiadaliri AA Gerdtham U Nilsson P Eliasson B Gudbjornsdottir S Carlsson KS Towards Renewed Health Economic Simulation of Type 2 Diabetes Risk Equations for First and Second Cardiovascular Events from Swedish Register Data 2013 PLOS ONE 8 5 e62650 Klein R Klein BE Moss SE Cruickshanks KJ The Wisconsin Epidemiologic Study of diabetic retinopathy XIV Ten year incidence and progression of diabetic retinopathy Arch Ophthalmol 112 1217 1228 1994 Klein R Klein BE Moss SE Cruickshanks KJ The Wisconsin Epidemiologic Study of Diabetic Retinopathy XV The long term incidence of macular edema Ophthalmology 102 7 16 1995 22 Law MR Wald NJ Morris JK Jordan RE Value of low dose combination treatment with blood pressure lowering drugs analysis of 354 randomised trials BMJ 2003 Jun 28 326 7404 1427 Law MR Morris JK Wald NJ Use of blood pressure lowering drugs in the prevention of cardiovascular disease meta analysis of 147 randomised trials in the context of expectations from prospective epidemiological studies BMJ 2009 338 01665 Mellbin LG Malmberg K Norhammar A Wedel H Ryd n L DIGAMI 2 Investigators Prognostic implication of glucose lowering treat
41. A2 Franklin et al 2004 Factor Category Modifier Age For example for a 60 years old male subject not on beta blocker or ACE Inhibitor P CHF for the MI module 0 13 0 87 0 86 0 45 Medication Modifier is as described in the main text 66 Michigan Model for Diabetes User Manual A1 1 2 Prediction model for the risk of congestive heart failure CHF in type 2 diabetes T2DM based on the Cardiovascular Health Study Data source The Cardiovascular Health Study CHS was a study of risk factors for the development and progression of CHD and stroke in people aged 65 years of age and older The 2 962 women and 2 239 men were recruited and examined yearly from 1989 through 1999 The added minority cohort of 256 men and 431 women was examined from 1992 to 1999 Examination components included medical history questionnaires echocardiograms ambulatory electrocardiograms cerebral magnetic resonance imaging abdominal and carotid ultrasound studies measurement of ankle brachial index spirometry and retinal photographs CHS has undertaken extensive follow up for ascertainment of cardiovascular events including myocardial infarction MI CHF stroke claudication and death Our goal was to develop a long term prediction model for CHF in T2D conditional on the subject s history of angina and MI In the original CHS cohort 862 subjects had diabetes at the baseline visit without history of CHF including 416 who had newly diagnosed diabetes i
42. BMI centered at 28 2 BMI Plus function BMI 33 Gender Male vs Female 1 39 1 08 1 79 AF Yes vs No 2 45 1 56 3 85 Age at diabetes onset centered at 65 1 05 1 02 1 07 C index at 10 year 0 699 BMI 33 BMI 33 when BMI 33 gt 0 otherwise 0 70 A1 2 Cerebrovascular disease sub model 1 No 3 Survived 2 Stroke 4 Stroke Death Cerebrovascular Stroke Lo Figure A5 Structure of cerebrovascular disease sub model Disease Table A5 Transition probabilities in cerebrovascular disease sub model Transition Transition probability Comments Clarke et al 2004 outcomes model modified by direct medication effect 2 to 3 This is the complementary for the transition from Stroke to Stroke Death Death Changes in that transition should be reflected in this probability 2 to 4 Fatality equation from UKPDS 68 Clarke et al 2004 3 to 2 If had stroke last year 30 x transition The calibration factor was influenced by numbers in table 2 in Sacco et al probability of 1 to 2 1994 If had stroke before last year 10 x transition probability of 1 to 2 0 5 0 1064 Table 2 in Sacco et al 1994 Similar to the existing diabetes formula that distinguishes the first year from subsequent years combine the following numbers in first year 0 201 and other years 0 0738 1 1 0 412 1 0 201 1 0 4 0 0738 The above probability was multiplied by a calibration factor of 0 5 to reflect the advance in healthcare si
43. INITIO File Help Project Definition Simulation fae eae al Mame Dered From Primary Model Population Set Stage 0 Initialization My First Diabetes Simulatic Created On 2015 07 28 12 32 34 591000 Notes How to simulate an Sees Observational Stu Obervational Study Templ Last Modified 2015 07 28 12 50 55 266000 sf Michigan Model For Diabetes 201 No of Simulation Steps 10 No of Repetitions 4 Run Simulation My population View Result Delete Results Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Cost Quality of Life update rules in the simulation Affected Parameter PC CABG YearlyEventCost Cost_Comment YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlwFvent ost Cost QoL parameter YearlyEventCost If In State Occurrence Probability Occurrence Probability Orf CVD_Procedure_Ente 1 1 Or Proliferative_left_Ent ESRD_Transplant_Entered Angina_Entered Orf CVD_Procedure_Ente OrlAnd ML Entered l E OrlAnd ML Entered CVD Or CHFwoML Entered E Stroke_Entered Hyooohwremia oe Amputation_Entered fearlyEventCost 4 The modified numbers is back in the list Affected Parameter PC CABG YearlyEventCost Cost_Comment YearlyEventCost YearlyEventCost YearlyEventCost
44. MI Charlton Menys V Fuller JH CARDS investigators Primary prevention of cardiovascular disease with atovarstatin in type 2 diabetes in the collaborative Atorvastati Diabetes Study CARDS multicentre randomized placebo controlled trial Lancet 2004 364 685 696 Cole JH Jones EL Craver JM Guyton RA Morris DC Douglas JS Ghazzal Z Weintraub WS Outcomes of repeat revascularization in diabetic patients with prior coronary surgery J Am Coll Cardiol 2002 40 1968 1975 Deedwania PC Ahmed MI Feller MA Aban IB Love TE Pitt B Anmed A Impact of diabetes mellitus on outcomes in patients with acute myocardial infarction and systolic heart failure Eur J heart Fail 2011 12 551 559 Franklin K Goldberg RJ Spencer F Klein W Budaj A Brieger D Marre M Steg PG Gowda N Gore JM GRACE Investigators Implications of diabetes in patients with acute coronary syndromes The global registry of acute coronary events Arch of Intern Med 2004 164 1457 1463 Fried LP Borhani NO Enright P Furberg CD Gardin JM Kronmal RA Kuller LH Manolio TA Mittelmark MB Newman A The Cardiovascular Health Study design and rationale Ann Epidemiol 1991 1 3 263 76 83 12 13 14 15 16 17 18 19 20 21 22 23 24 20 Gall MA Hougaard P Borch Johnsen K Parving HH Risk factors for development of incipient and overt diabetic nephropathy in patients with non insulin dependent diabetes mellitus prospective o
45. Modifying the default MMD For advanced users only lf your project needs additional changes which was not mentioned in the instructions above please contact us at help MichiganModelForDiabetes umich edu Me Michigan Model for Diabetes User Manual 5 Entering Population Information Populations can either be inputted as data to be used in a Simulation or an Estimation or set by specifying a distribution to be used in Estimation or for randomly generating population sets It is the responsibility of the users of MMD to ensure that only valid values are entered as the software applies a few data entry checks The items needed for each subject are listed in the following table Variable Name Legal Range System Variables Diabetes Type 2 State indicator for having type 2 diabetes State indicator for being alive Demographics Characteristics Age 1 100 diabetes 1 Male 2 Black BMI Weight Height 10 50 Weight in kilograms 1 0 kg 2 2 pounds Height in meters 1 0 meter 39 inches Height Height in meters 1 0 meter 39 inches 2 NO Ol Current Risk Factors Systolic blood pressure mmHa 60 280 Diastolic blood pressure mmHg 20 140 Smoke Smoking status smoker 1 Smoker HDLCholesterol High density lipoprotein cholesterol in 0 3 5 mmol L 38 dl Z O 7 mmol L 1 mmol L 38 6mq dl mmol L 38 6mq dl TotalCholesterol 0 6 25 12 mmol L 38 6mq dl HbAic Hemoglobin Alc
46. No right eye retinopathy Macular_edema righ 0 Medication Currently use intensive life style for controlling Metformin O glucose level _OtherOralMedication O Basallnsulin 0O Insulin S 0 Import this population sheet following instructions in section 5 1 Step 2 Follow instructions in section 4 1 1 to create a new observational project On the project window in the Population Set manual select the population you have just created and read in Set the No of Simulation Steps to 20 years the No of Repetition to 1000 To see how diabetes progresses in these two patients in the scenario that they both comply with all treatments use the setup in the observational study template change the compliance rate for all treatments to 100 following instructions in Section 4 1 2 For all other parameters use the default setting PROJECT DEFINITION o eE File Help Project Definition Simulation ee ere Name My Example 1 Created On 2015 07 20 16 15 11 623000 Notes How to simulate an Observational b Study Derived From Qbervational Study Template LastModified 2915 07 21 13 28 01 844000 Primary Model Michigan Model For Diabetes 2015 w No of Simulation Steps 0 Run Simulation na es ies Ce Cs Covariate update rules in the simulation Affected Parameter If In State Occurrence Probability Function Notes z Compliance _ACE 1 Compliance _100 Compliance levels
47. White and Black in the race columns was selected and 4 the data was divided by 1 000 to represent the yearly transition probability A1 4 Neuropathy sub model 1 No 2 Clinical 3 Amputation Neuropathy Neuropathy Figure A7 Structure of neuropathy sub model Table A7 Transition probabilities in neuropathy sub model Transition probability 1 to2 0 0518 Sands et al 1997 Table 1 first line Note that in the future it may be possible to use sex or age covariates using the same table data 0 0113 Adler et al 1999 Table 4 last row Note that the table considers only men in the future other data may be considered 73 A1 5 Retinopathy sub model Two eyes are modeled separately and assume to be independent Retinopathy macular edema are two parallel sub sub processes 3 Proliferative 2 Non 4 Blindness Proliferative Caused by DR Retinopathy or Retinopathy Macular Edema 1 No 7 Blind Retinopathy 5 Proliferative gt Retinopathy or Macular Edema 6 Blindness Caused by Macular Edema Figure A8 Structure of retinopathy sub model Table A8 Transition probabilities in retinopathy sub model Transition probability Comments 1 to2 0 0653 for diabetics Klein 1994 Table 8 70 2 10 yr progression rate was used for insulin taking group and who do not need 49 1 10 yr progression rate was used for non insulin taking group The first row and the I
48. YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost YearlyEventCost VYearwFventt ost 4 1 3 2 Defining utility scores If In State Occurrence Probability OrfCVD_Procedure_Ente 1 1 Or Proliferatre_left_Ent ESRD_Transplant_Entered Amputation_Entered Angina_Entered Or CVD_Procedure_Ente OrfAnd MLEntered 1 C OrfAnd MIEntered CVD Orf CHFwoML Entered E Stroke Frtered ata axa Function Bernoulli 0 67 0 1 YearlyEventCost 1101 YearlyEventCost 138071 YearlyEventCost 8282 YearlyEventCost IPC YearlyEventCost 41744 YearlyEventCost 60865 YearlyEventCost 34635 YearlyEventCost 55278 VearwFyventtost 1h991 Function Function Bernoulli 0 67 0 1 YearlyEventCost 1101 YearlyEventCost 1380 1 YearlyEventCost 60000 YearlyEventCost 6282 YearlyEventCost Iif PC YearlyEventCost 41744 YearlyEventCost 60865 YearlyEventCost 34635 VYearwFvent ost 55778 Cost QoL Wizard Notes Procedure 3 1 are CABG 3 2 ar Set Yearly Cost to 0 ADD UP EVENT COST Add cost of amputation ESRD transplant cost accordin Event Cost for Angina Procedure Cost 3 1 are CABG Mycardial Infarction without pr Mycardial Infarction with proc Add cost of CHF hospitalization Event cost for Stroke Confirmed Clinical Neuronathw Add cost of amputation Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications
49. a From the menu bar at the top of the main form select File b From the File menu select Open 5 JEST Indirect Estimation and Simu Forms Help Indirect Estimation and Simulation Tool Project Name Project Type Notes Single Report Report All Exit c Select the requested filename path of the zip file of MMD from the new window that appeared and press the Open Button Organize New folder E Videos J Mame di Documentation di Temp EI MichiganModelForDiabetes2 0 zip a Testing zip 4 TestingEstimation zip 4 TestingEstimationWithResults zip JE Computer amp Local Disk C EE DVD RW Drive O DATCARD G SPH U E wye s iPhone No preview available hy Network d InstalllEST_0 85 0 0 di Documentation oal Th Filename MichiganModelForDiabetes2 0 zip d The label at the top of the windows should show the path of the file and the project list should show projects held within the loaded file lice ed A ed bd el U O File Forms Help Indirect Estimation and Simulation Tool Project Name Project Type Notes Add New Project Obervational Study Template Simulation How to simulate an Observational Study Interventional Study Ternplate Simulation How to simulate an interventional study Michigan Model for Diabetes User Manual 4 Implementation of the Michigan Model for Diabetes in IEST For each subject the model software re
50. act of microvascular complications treatment and geographic location Diabetes Care 2007 30 1241 1247 Bretzel RG Nuber U Landgraf W Owens DR Bradley C Linn T Once daily basal insulin glargine versus thrice daily prandial insulin lispro in people with type 2 diabetes on oral hypoglycaemic agents APOLLO an open randomised controlled trial Lancet 2008 Mar 29 371 9618 1073 84 Charbonnel B Schernthaner G Brunetti P Matthews DR Urquhart R Tan MH Hanefeld M Long term efficacy and tolerability of add on pioglitazone therapy to failing monotherapy compared with addition of gliclazide or metformin in patients with type 2 diabete Diabetologia 2005 48 6 1093 104 Chaitman BR Hardison RM Adler D Gebhart S Grogan M Ocampo S Sopko G Ramires JA Schneider D Frye RL Bypass Angioplasty Revascularization Investigation 2 Diabetes BARI 2D Study Group The bypass angioplasty revascularization investigation 2 diabetes randomized trial of different treatment strategies in Type 2 diabetes mellitus with stable ischemic heart disease Circulation 2009 120 2529 2540 Clarke PM Gray AM Briggs A et al UK Prospective Diabetes Study UKDPS Group A model to estimate the life time health outcomes of patients with type 2 diabetes the United Kingdom Prospective Diabetes Study UKPDS Outcomes Model UKPDS no 68 Diabetologia 2004 47 1747 1759 Colhoun HM Betteridge DJ Durrington PN Hitman GA Neil HA Livingstone SJ Thomason MJ Mackness
51. ads in or simulates the subject s baseline characteristics and then advances the subject through a specific number of years or until death Each year the model updates in the four stages as indicated by blue blocks in the following figure including 1 Update risk factors i e weight BMI HbA1c fasting glucose systolic blood pressure SBP diastolic blood pressure DBP lipids according to treatment status and natural history of changes in glycaemia blood pressure and lipids See Appendix A1 for details of model specification 2 Update disease states and complications based on transition probabilities which can be functions of individual characteristics current disease states or treatment status See Appendix A1 for details of model specification 3 Update treatments when certain risk factor passes pre specified threshold or subject experiences a major complication event taking account of pre specified compliance parameters 4 Assign cost and utility values for the specific year according to complication experiences Input Initiation Population parameters for Baseline treatment and information compliance 3 Update Examination 4 Update Compliance Cost QoL Treatment 1 Update 2 Update Risk Disease States Factors Complications The first year of this process differs for observational studies and intervention studies For an observational study the first step updating risk factors is skipped during the first year c
52. al inpatient or outpatient claims in 2011 and 2012 88 Appendix C Michigan Model for Diabetes Utility Model Table C1 Penalty functions for QWB SA health utility scores Disease status Complication Level QWB SA Penalty Ooo intercept Ci a Diabetes Intervention Macular edema or proliferative 0 000 neLNOpAtY retinopathy Blind in one eye 0 043 Blind in two eyes 0 170 No nephropath Microalbuminuria or proteinuria 0 011 Nephropathy Microalbuminuria or proteinuria 0 011 Neuropathy Cerebrovascular disease EE maki Coffey et al 2002 did not provide a penalty for having history of Angina or MI PTCA CABG In Zhang et al 2012 the penalty for other heart disease is approximately half of the penalty for CHF We therefore imputed the penalty for Angina and MI PTCA CABG as half of the penalty for CHF Reference 1 Coffey JT Brandle M Zhou H Marriott D Burke R Tabaei BP Engelgau MM Kaplan RM Herman WH Valuing health related quality of life in diabetes Diabetes Care 25 2238 2243 2002 2 Zhang P Brown MB Bilik D Ackermann RT Li R Herman WH Health Utility Scores for People With Type 2 Diabetes in U S Managed Care Health Plans Diabetes Care 35 2250 2256 2012 89 Appendix D Python Expressions Used in IEST Expressions include mathematical and logical formulas Expressions can be as simple as 1 2 they can use another parameter as in Age 1 They can be complex expressions using mathemat
53. an DE Predictors of mortality and recurrence after hospitalized cerebral infarction in an urban community the Northern Manhattan Stroke Study Neurology 44 626 634 1994 sands ML Shetterly SM Franklin GM Hamman RF Incidence of distal symmetric Sensory neuropathy in NIDDM The San Luis Valley Diabetes Study Diabetes Care 1997 20 322 329 UK Prospective Diabetes Study UKPDS Group Relative efficacy of randomly allocated diet sulohonylurea insulin or metformin in patients with newly diagnosed non insulin dependent diabetes followed for three years UKPDS 13 BMJ 1995 310 383 UK Prospective Diabetes Study UKPDS Group Intensive blood glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes UKPDS 33 Lancet 1998 Sep 12 352 9131 837 53 U S Renal Data System USRDS 2002 Annual Data Report Atlas of End Stage Renal Disease in the United States National Institutes of Health National Institute of Diabetes and Digestive and Kidney Diseases Bethesda MD 2002 Table F 20 in Section F 441 2002 Available at http www usrds org 2002 pdf F pdf cited 08 25 2015 Saran R Li Y Robinson B et al US Renal Data System 2014 annual data report epidemiology of kidney disease in the United States Am J Kidney Dis 2015 66 1 suppl 1 S1 S306 Table H 4 1 in Section H Available at http www usrds org reference aspx cited 08 25 2015 Saran R Li Y Rob
54. and click on the Data button on the right to open the data window T Population Sets Ea File Help Population Sets Mame Definition Type Created On Notes Source Derived From Last Modified 2015 07 27 14 28 08 1 x Template for Specifying Di Distribution based 2015 07 23 12 12 57 3 Dai 36 Michigan Model for Diabetes User Manual 4 On the data window click on the Import button to read in your population data set Parameter Notes Alc Short for HbA ic in 3 Alc_Diff The difference in Alc Alc Difi The difference in Aic lagged 1 time ste Alc Diff2 The difference in Aic lagged 2 time st Alc Prev Alc from previous year ACE Level acne Albha mim be Menschen esha 5 If the data is successfully read in you can see it on the Data tab Click OK and close the Population Sets window to save the this population set 297 Michigan Model for Diabetes User Manual 5 2 Specify a distribution An alternative to inputting a data set with individual information is to simulate a baseline population using population level summary statistics To do so you can use the template for specifying a distribution that we included in the default MMD 1 Click on the Populations button on the left side of the main form to open the population sets window 2 TEST C Users wye Desktop MichiganModelForDiabetes2 0_07282015 zip 3 File Forms Help Indirect Estimation and Simulation Tool Pr
55. ange No treatment PP No drug gt one drug half If the first drug is ACE inhibitor ARB Law et al 2009 standard dose Mean change of SBP 6 9mhg 0 08 SBP 150 Wald et al 2009 Mean change of DBP 3 7mhg 0 09 DBP 90 Law et al 2003 If the first drug is B blocker Mean change of SBP 7 4mhg 0 08 SBP 150 Mean change of DBP 5 6mhg 0 09 DBP 90 Already on drug gt Mean change of SBP nx3 4mhg nx0 04 SBP 150 receive an increase of Mean change of DBP nx1 8mhg nx0 04 DBP 90 treatment of n levels No drug gt treatment lf the first drug is ACE inhibitor ARB level n n gt 1 Mean change of SBP 6 9 nx3 4mhg 0 08 nx0 04 x SBP 150 Mean change of DBP 3 7 nx1 8mhg 0 09 nx0 04 x DBP 90 lf the first drug is B blocker Mean change of SBP 7 4mhg nx3 4mhg 0 08 nx0 04 x SBP 150 Mean change of DBP 5 6mhg nx 1 8mhg 0 09 nx0 04 x DBP 90 Aging effect x4 x5 are two randomly drawn independently distributed standard normal variables They are re drawn each year The two following equations calculate the change in SBP and DBP based on the current value of SBP DBP change in BMI gender and race DBP_diff 0 2 Age 0 282913980 DBP 0 031328327 SBP 0 030871363 Age SBP 0 000770741 Age DBP 0 003093990 BMI_Diff 0 372137437 Female 0 379980806 IsAfricanAmerican 0 567931842 2 5848 Temp_x5 SBP_diff 34 7 Age 1 02313914 DBP 0 13180962 SBP 0 18569020 Age SBP 0 00590678 Age DBP 0 00268753 BMI_Diff 1 79346394 Female 0 52748318
56. appedGaussian3 8 6 53 4 Duration_Of Diabetes Max 0 CappedGaussian3 2 7 Bernoulli 0 56 Bernoulli 0 18 1 Notes bir Short for HbA 1c in Notes The difference in Aic The difference in Aic lagged 1 time ste The difference in Alc lagged 2 time st Alc from previous year A Ih ome fe Pem m immm e a Pr 3 Below we use the Age variable as an example to show you how to modify the distribution Click and highlight the line of the variable you would like to modify and click the Down Arrow button File Help Population Set Tem Farameter Corr_SBP_DBP Diabetes Type 2 Alive g CappedGaussian3 8 6 53 4 Duration Of Diabetes Max 0 CappedGaussian3 2 7 Male Bernoulli 0 56 Race Bernoulli 0 18 1 Parameter Notes Alc Short for HbA 1c in 3 Alc Diff The difference in Alc Alc_Diffi The difference in Alc lagged 1 time ste Alc _Diff2 The difference in Alc lagged 2 time st Alc Prev Alc from previous year ACE Level A bha raim be en hmm e prr 39 Michigan Model for Diabetes User Manual 4 The original distribution for the variable age disappears from the top list and appears in the narrow window in the middle CappedGaussian3 is a system function that generates a standard normal random number with all numbers lt 3 or gt 3 truncated i e any randomly drawn numbers lt 3 are set to be 3 any randomly drawn numbers gt 3 are set to be 3 8 6 is the standard deviation and 53 4 is the mean
57. benefit and by additionally adjusting the Avogaro et al 2007 race smoking al 2004 hazard by a factor 0 7 men and women HbA1c SBP Avogaro et al B No CHD gt UKPDS IHD equation adjusted for medication separately lipid ratio and 2007 CAD w o Ml benefit and by additionally adjusting the hazard function by a factor of 3 O No CHD gt UKPDS MI equation IHD 0 CHF 0 adjusted for CHD death medication benefit and by additionally adjusting the hazard by a factor 0 091 AA No CHD gt CHF w o MI CHS risk equation Section C in this document Age at diabetes Fried LP et al Angina 0 Ml 0 adjusted for medication benefit onset sex SBP 1991 DBP lipid ratio BMI history of angina history of MI AF and medications K CAD w o MI The UKPDS MI equation IHD 1 CHF 0 adjusted Calibrated to Age sex race Clarke et gt CHD death for medication benefit and by additionally adjusting Colhoun et al 2004 smoking HbA1c al 2004 the hazard by a factor 0 668 placebo groups SBP lipid ratio Colhoun et al CAD w o MI gt The UKPDS MI equation IHD 1 CHF 0 adjusted and 2004 MI for medication benefit and by additionally adjusting medications I the hazard by a factor 1 68 H CAD w o MI The UKPDS MI equation IHD 1 CHF 0 adjusted gt CHD for medication benefit and by additionally adjusting procedure the hazard by a factor 7 62 BB CAD w o MI CHS risk equation Sec
58. betes Simulatic Created On 2015 07 19 19 21 38 652000 Notes How to simulate an 2 ae Ob tional St Derived From Example Project 1 Last Modified 3915 07 19 19 21 58 203000 SS y Primary Model Michigan Model For Diabetes 201 v mm No of Simulation Steps Run Simulation Population Set My population v am No of Repetitions 1000 View Result Delete Results Stage0 Initialization Stage Update Covariates Stage Update Complications tage Update Treatment Staged Update Costs Initialize the simulation Affected Parameter Function Notes Threshold_SBP 150 Threshold for increasing the le TL i At 11 Ti sacl old Lan io anan nie am l a la 6 3 Run simulation Save all the changes before running a simulation Otherwise if the program is aborted all the changes will be lost 45 Michigan Model for Diabetes User Manual Click on the Run Simulation button to start the simulation Project Definition Simulation Name My First Diabetes Simulation Created On 2015 07 20 10 58 38 090000 Notes How to simulate an Observational Derived From Example Project 1 Last Modified 2915 07 20 11 06 50 038000 i Primary Model Michigan Model For Diabetes 2015 w No of Simulation Steps 5 l My population x be No of Repetitions 1000 View Result Delete Results Stage 0 Initaizaton Once you start the simulation a small window pops up to show how much
59. bservational study BMJ 314 783 788 March 15 1997 Harrell FE Jr Lee KL Califf RM Pryor DB Rosati RA Regression modeling strategies for improving prognostic prediction Stat Med 1984 3 2 1438 52 Hayes AJ Leal J Gray AM et al UKPDS outcomes model 2 a new version of a model to simulate lifetime health outcomes of patients with type 2 diabetes mellitus using data from the 30 year United Kingdom Prospective Diabetes Study UKPDS 82 Diabetologia 2013 56 1925 1933 Holman RR Thorne KI Farmer AJ Davies MJ Keenan JF Paul S Levy JC 4 T Study Group Addition of biphasic prandial or basal insulin to oral therapy in type 2 diabetes N Engl J Med 2007 357 17 1716 30 Holman RR Farmer AJ Davies MJ Levy JC Darbyshire JL Keenan JF Paul SK 4 T Study Group hree year efficacy of complex insulin regimens in type 2 diabetes N Engl J Med 2009 Oct 29 361 18 1736 47 Humphrey LL Ballard DJ Frohnert PP Chu CP O Fallon WM Palumbo PJ Chronic renal failure in non insulin dependent diabetes mellitus A population based study in Rochester Minnesota 1 Ann Intern Med 1989 Nov 15 111 10 788 96 Jensen LO Maeng M Thayssen P Tilsted HH Terkelsen CJ Kaltoft A Lassen JF Hansen KN Ravkilde J Christiansen EH Madsen M S rensen HT Thuesen L Influence of diabetes mellitus on clinical outcomes following primary percutaneous coronary intervention in patients with ST segment elevation myocardial infarction Am J Cardiol 2012 10
60. cemia is set to be 6 5 a patient whose HbA1c value is larger than 6 5 at baseline will receive treatment enhancement right after the simulation starts Their HoA1c and weight values will change accordingly To modify the rules for the first year risk factors and treatment changes do the following steps 1 Follow instruction in 4 1 1 to set up your own simulation project by copying Interventional Study Template a D7 Michigan Model for Diabetes User Manual 2 On the Project Definition window click on the tab Stage1 Update Covariates A File Help Project Definition Simulation Name My Interventional Study Derived From Interventional Study Temp Last Modified 2015 07 28 13 09 21 585000 Primary Model Michigan Model For Diabetes 201 m Population Set Template for Specifying Distributi v ums ae Created On 2015 07 28 13 09 21 585000 No of Simulation Steps 5 No of Repetitions 5000 Notes How to simulate an a interventional study vr Run Simulation Initialize the simulation Affected Parameter Threshold_SBP Threshold_Alc Threshold_LDL Max_Level_ACE Max_Level_Diabetes Trt Max_Level_Statin Max_Level_Aspirin YearlyRateOfQuittingSm YearlyRateOfStartAspirin Equ_CompetingDeath Discount_Afterl0Years Discount First 0Vears Function 135 0 03 a x Cawan Occurrence Probability Function Notes Threshold for increasing the
61. d and compliance rates The following figures show the time paths for a few of these risk factors in subject one and subject two respectively Subject One Subject Two Alc over time Alc over time Alc Alc Time Time Time Time Eia Michigan Model for Diabetes User Manual Subject One Subject Two Total Cholesteroal over time Total Cholesteroal over time TC Time Time It is also possible to examine cumulative event rates adjusted for death as a competing risk over the years specified in the simulation The following table shows the simulated incidence rate for subject one and subject two per 1000 person years PYs For example for subject one the estimated incidence rate of experiencing first MI is 5 7 1000 PYs in 20 years the probability for subject one to experience MI is 10 2 Complication Subject One Subject Two Incidence rate Cumulative Incidence rate Cumulative 1000 PY Incidence 1000 PY Incidence MI 5 7 10 2 6 0 CHF 4 7 8 4 7 0 Stroke 1 5 2 8 2 0 Revascularization 8 2 14 5 8 4 Amputation 3 7 6 7 4 6 Blind In Both Eyes 0 38 0 7 ESRD 1 0 1 9 Cardiovascular Death 3 7 6 9 Example 2 We may also want to undertake a simulation based on no compliance to any treatment at all To study this proceed as follows Step 1 Use the project window in Example 1 change the compliance rate for all treatments to 0 following instructions
62. d to start or intensify treatment and a diagnosis of CAD they will comply with the treatment change with 50 probability The remaining patients will never start or intensify these treatments Beta blocker is started For compliers when there is a CVD event or the patient is diagnosed with CAD the treatment will be started For non compliers treatment will never start 249 Michigan Model for Diabetes User Manual Aspirin is started Among subjects who are not currently on aspirin For compliers after a new CVD event or the patient is diagnosed with CAD aspirin will be started The remaining compliers are randomly assigned to start aspirin each year at a user specified rate For the non compliers who become willing to comply with treatment when there is a CVD event aspirin is started when there is a CVD event For the same group of patients if they are diagnosed with CAD they will comply with the treatment change with 50 probability The remaining patients will never start or intensify this treatment Smoking cessation When there is a new CVD event a current smoker quits smoking When CAD is diagnosed a current smoker quits smoking with 50 probability The remaining smokers quit smoking each year at a user specified rate We further assume a hierarchical structure of patients for compliance For ease of exposition lets assume 90 of patients comply with all treatments when there is a CVD event 80 70 60
63. d yearly results for all the simulated individuals The current IEST software only provides limited results summaries We suggest that users export the individual results to csv files and calculate summary statistics and perform additional analyses using other software In the Worked Example section we provide a few SAS programs for summarizing simulation results To export results click on the Export To File button and follow the steps to select the desired path to save the results as a CSV file SIMULATION RESULT 3 fmezel File Help Results for Project Example Project 1 Simulation ID 1 S Sars Delete Delete All IndividualID Repetition Time Male Age Alc Weight BMI ition_Of Diab Sn 1 1 0 0 0 0 42 0 12 8 84 4 35 25646 4 0 0 0 2 1 0 1 0 0 42 0 13 0910517934 84 49956 79784 35 3080604256 4 0 0 0 3 2 0 0 0 39 0 8 656477 73 4 31L 76939 3 0 0 0 4 2 0 1 0 0 39 0 8 79613204496 75 347323857 31 7454693498 35 0 0 0 5 3 o 0 0 0 61 0 15 94194 78 0 33 76039 10 0 0 0 6 3 o 1 0 0 61 0 13 9539358335 78 46 16616751 33 96020675 10 0 0 0 7 4 o 0 0 0 60 0 13 98495 59 0 27 83103 9 0 1 0 A Once you have exported the results it is a good practice to delete all the results using the Delete All button before you make further modifications to any parameters 47 Michigan Model for Diabetes User Manual under the project window including steps in the Sections 4 1 2 4 1 4 Otherwise no modifications on the project
64. djusted for thes medication benefit and by additionally adjusting the procedure hazard by a factor by 3 074 G Hx of MI gt CHS risk equation Section C in this document Age at diabetes Fried LP el al CHF after MI Angina 1 Ml 1 adjusted for medication benefit onset sex SBP 1991 DBP lipid ratio BMI history of angina history of MI AF and medications Repeat MI gt See details in the Ml repeat MI module Table A2 See Table A2 See Table A2 See Table A2 a MI Repeat MI gt See details in the a MI module Table A2 cH after MI Repeat i gt Dadane details in the uh chi a MI module Table A2 cH death The inca ancihitanisiatill MI equation IHD 1 CHF 1 adjusted Calibrated to Age gender Clarke et 7S Repeat MI for medication benefit and by additionally adjusting Deedwania 2011 race smoking al 2004 the hazard by a factor 1 088 and Mellbin a al HbA1c SBP Deedwania et T CHF after The UKPDS MI equation IHD 1 CHF 1 adjusted 2011 lipid ratio and al 2011 for medication benefit and by additionally adjusting medications Mellbin et al the hazard by a factor 0 489 2011 X CHF after The UKPDS MI equation IHD 1 CHF 1 adjusted vn Repeat for medication benefit and by additionally adjusting procedure the hazard by a factor 6 201 V Repeat 95 if subject does not have CHF None None Cole et al procedure gt Hx 0 if subject have CHF 2002 of MI 62 Michigan Mode
65. do the following steps using threshold for HbA1c as an example 1 Highlight the parameter you would like to modify and click on the Down Arrow at the bottom of the window to bring down the parameter line to the editing cell 15 Michigan Model for Diabetes User Manual File Help Project Definition Simulation Name My First Diabetes Simulatic Created On 2015 07 28 12 32 34 591000 Notes How to simulate an yl Observational St Derived From Obervational Study Temple Last Modified 2015 07 28 12 32 34 591000 Pa ay Primary Model Michigan Model For Diabetes 201 v No of Simulation Steps 10 Population Set My population gt oJ werent Delete Results Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the simulation Affected Parameter Notes Threshold_SBP Threshold Alc Threshold_LDL Threshold for increasing the le Max_Level_ACE Highest level of treatment with Max_Level_Diabetes_Trt Threshold for increasing the le Threshold for increasing the le Max_Level_Statin Highest level of treatment with Max_Level_Aspirin YearlyRateOfQuittingSm YearlyRateOfStartAspirin Equ_CompetingDeath t 10 Use modified death rates f Discount_Firstl0Years Discount Afterl0VYears oa ume eS eee eee 2 Change the value for this parameter in the Function cell You can
66. e Ml repeat MI module Table A2 See Table A2 death M MI gt CHF See details in the Ml repeat MI module Table A2 after MI Age gender race smoking HbA1c SBP lipid ratio and medications Age at diabetes onset sex SBP DBP lipid ratio BMI history of angina history of MI AF and medications Age at diabetes onset sex SBP DBP lipid ratio BMI history of angina history of MI AF and medications Age at diabetes onset sex SBP DBP lipid ratio BMI history of angina history of MI AF and medications See Table A2 Clarke et al 2004 Chaitman et al 2009 Fried LP el al 1991 Clarke et al 2004 Deedwania 2011 Mellbin et al 2011 Clarke et al 2004 Deedwania 2011 Mellbin et al 2011 See Table A2 61 Michigan Model for Diabetes User Manual a MI gt Hx of See details in the Ml repeat MI module Table A2 UKPDS MI equation IHD 1 CHF 0 adjusted for Calibrated to Jensen Age gender Clarke et naa death medication benefit and by additionally adjusting the et al 2011 and race smoking al 2004 hazard function by a factor 0 232 Mellbin et 2011 HbAic SBP Mellbin et al Hx of MI gt UKPDS MI equation IHD 1 CHF 0 adjusted for lipid ratio and 2011 Jensen sae MI medication benefit and by additionally adjusting the medications et al 2011 hazard by a factor by 1 247 W Hx of MI gt UKPDS MI equation IHD 1 CHF 0 a
67. e simulation Eq Time 1 Affected Parameter If In Occurrence Probability Function Started_Diabetic Eq Time 1 1 COMMENT 1 Addl_Diabetes Eq Time 1 Eq Insulin 0 Compliance_diabetes Ge Alc Threshold_Alc Add1_Statin Eq Time 1 Compliance_statin Ge LDLCholesterol Threshold_LDL Add1_Aspirin Eq Time 1 Eq Aspirin 0 Compliance_Aspirin Bernoulli YearlyRateOfStartAspirin Compliance_ACE Ge SBP Threshold_SBP 10 Notes a Sets to 1 if diabetic at start er Define Intervention For Year1 Intervention on dyslipidemia i Taking Aspirin in Year1 Intervention on hypertension i Smoke Ge Time 1 Eq Smoke 1 Bernoulli 1 YearlyRateOfQuittingSmoking smoking cessation in Year1 COMMENT 1 Update risk factors Age Ge Time 2 Age 1 Age increase Temp_xl 1 CappedGaussian3 Generate random number use Temp x2 1 CappedGaussian3 Generate random number use 4 mW r Covariate If in State Occurrence iy Function Notes Add1_ACE v v Eq Time 1 v Compliance ACE Ge SE Intervention on hypertension in Year A Treatment changes not only happen to subjects enrolled in an active treatment arm but also mostly happen to subjects enrolled in placebo arms as well When simulating disease progression for subjects in a placebo arm of an interventional study one should not use the template for an observational study to simulate a placebo arm in an interventional study 30 Michigan Model for Diabetes User Manual 4 2
68. eResult if Statement is 0 D9 Table Table TableParameters A multi dimensional table TableParameters are provided as a string of comma separated values The Table input argument pattern is D Ny ND V4 Voutena tnp M1 Rio Rint Mp Ropo Ronp D number of dimensions N Np dimension size for dimension 1 to D V1 Wontene ND table values M Mp dimension names for dimension 1 to D Rio Rini lf the dimension is discrete define Riop NaN O O O O 91 Ri Rii values for each level in the ith dimension If the levels dimension is continuous the levels of each dimension are defined by cutpoints which represent the lower and upper bounds for each interval Rio the lower bound of the first interval R the upper bound of the first interval and the lower bound of the second interval Rini the upper bound of the Nth interval An example The following table can be stored in the system with the expression Table 2 2 3 1 2 3 4 5 6 Gender NaN 0 1 Age 0 30 60 120 0 lt Age lt 30 30 lt Age lt 60 60 lt Age lt 120 a ae r Gender O _ _ 1 Gender 1 Ea ee C A A E A G D 2 this is a 2 dimensional table N1 2 the dimension size is 2 for the first dimension N2 3 the dimension size is 3 for the second dimension M1 Gender the dimension name is Gender for the first dimension M2 Age the dimension name is Age for the second dimension Rio NaN the Gender dimen
69. eatment Stage 4 Update Costs latn Covariate update rules in the simu Affected Parameter If In State Occurrence Probability Function Notes COMMENT 1 1 Compliance levels are defined Compliance_100 1 1 Compliance_95 Eq Time 1 Bernoulli 0 95 Compliance_90 Eq Time 1 Bernoulli 90 95 Compli Compliance_85 Eq Time 1 Bernoulli 85 90 Compii Compliance_80 Eq Time 1 Bernoulli 80 85 Compii Compliance_75 Eq Time 1 Bernoulli 75 80 Compii Compliance_70 Eq Time 1 Bernoulli 70 75 Compii Compliance_65 Eq Time 1 Bernoulli 65 70 Compii Compliance_60 Eq Time 1 Bernoulli 60 65 Compii Compliance_55 Eq Time 1 Bernoulli 55 60 Compii Compliance 50 FafTime 1 Rernoullif50 55 Comoill reir Occurrence Probability Function Use the Scrollbar on the right to scroll down the page and find the section where the compliance levels for treatments are defined Occurrence Probability Function Eq Time 1 Bernoulli 15 20 Compliance_20 Eq Time 1 Bernoulli 10 15 Compliance_15 Eq Time 1 Bernoulli 5 10 Compliance_10 Compliance levels for diseases are defined Compliance levels for treatment for hypertension Compliance levels for aspirin Compliance levels for beta_blocker Compliance levels for treatment for dysalycemia Compliance levels for treatment for dyslipidemia Compliance levels for treatment for dysglycemia dysili 1 1 1 1 1 1 1 1 If in State
70. ected Parameter Function Notes Thresheld_Alc 5 Threshold for increasing the le Threshold _LDL 150 38 67 Threshold for increasing the le Max_Level_ ACE 8 Highest lewel of treatment with Max_Level_Diabetes_Trt Max_Level_ Statin 2 Highest level of treatment with Max_Level_Aspirin YearlyRateOfQuittingSm YearlyRateOfStartAspirin Equ_CompetingDeath 10 75 10 Use modified death rates f Discount_FirstlOYears pE Discount_AfterlOVears Covariate If in State Occurrence Probability Function r r 4 1 4 Defining the first year treatment parameters when simulating an intervention study When setting up a simulation the most important difference between an observational study and an interventional study is how to set up the first year In an observational study the transition probabilities for disease progression are calculated based on the baseline parameters In contrast in an interventional study since patients receive an intervention right after they are enrolled in the study risk factors often change largely after they started due to changes in treatment Therefore when setting up an interventional study in the first year of the simulation MMD allows user to model the change of treatments which consequently changes the risk factor levels before calculating transition probabilities In the default model the first year changes follow the same rule as other years For example if the treatment threshold for hypergly
71. en if it has been or hereafter advised of the possibility of such damages Michigan Model for Diabetes User Manual List of Abbreviations BMI CAD CVD MI CHD CHF MMD SBP DBP ACR PTCA CABG ACE ARB QALE QALYs Michigan Model for Diabetes User Manual Table of Contents 1 Introduction and Background 2 Changes in Version 2 0 3 Download and Installation 3 1 Download the disease modeling software IEST and Michigan Model for Diabetes 3 1 1 Installation of Python environment 3 1 2 IEST software and MMD installation 3 1 3 Running the IEST software 3 2 Loading the Michigan Model for Diabetes in the IEST software 4 Implementation of the Michigan Model for Diabetes in IEST 4 1 Running simulation using the default MMD 4 1 1 Start your own project 4 1 2 Defining general treatment parameters and compliance rates 4 1 3 Defining cost values and utility scores 4 1 4 Defining first year treatment parameters when simulating an intervention study 4 2 Modifying the default MMD For advanced users only 5 Entering Population Information 5 1 Input as data 5 2 Specify a distribution 6 Running the Model 6 1 Select the population set and set number of subjects 6 2 Number of years simulated 6 3 Run simulation 7 Outputs 8 Worked Examples Appendix A Disease Model Appendix B Cost Model Appendix C Utility Model Appendix D Python Expressions Used in IEST D Q CON N O O DO OO f O DYN NDN N
72. entional study Copy Record Study Model From the dropdown menu select Copy Record You should see a new project added to the list named as Observational Study Template_0 IEST C Users wye Desktop MMD MichiganModelForDiabetes2 0_beta zip File Forms Help Indirect Estimation and Simulation Tool Project Name Project Type Notes Add New Project Observational Study Template Simulation How to simulate an Observational Study Interventional Study Template Simulation How to simulate an interventional study Observational Study Template_0 Simulation How to simulate an Observational Study Chih Madal 2 Change the name of the new project to your own Double click on the line of the new project to open the popup window for PROJECT DEFINITION On the upper left corner of the PROJECT DEFINITION window change the project name to your own On the upper right corner change the notes to your own if desired A IEST C Users wye Desktop MMD MichiganModelForDiabetes2 0_beta zip File Forms Help Indirect Estimation and Simulation Tool Project Name Project Type Notes Add New Project Observational Study Template Simulation How to simulate an Observational Study Interventional Study Template Simulation How to simulate an interventional study Observational Study Template _0 Simulation How to simulate an Observational Study 10 Michigan Model for Diabetes Us
73. er Manual gt E PROJECT DEFINITION File Help Project Definition Simulation bop Name observational Study Templatef Created On 2015 08 19 16 57 16 174000 Notes Derived From Observational Study Template LastModified 2915 08 19 16 57 16 174000 Primary Model Michigan Model For Diabetes 2015 No of Simulation Steps 10 Run Simulation Population Set Template for Specifying Distribution am No of Repetitions 4 View Result f Delete Results Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the simulation Affected Parameter j Notes Threshold_SBP Threshold for increasing the level Threshold_Aic Threshold for increasing the level Threshold_LDL Threshold for increasing the level Max_Level_ACE Highest level of treatment with A Max_Level_Diabetes_ Trt Max_Level_Statin Highest level of treatment with St Max_Level_Aspirin YearlyRateOfQuittingSmoking YearlyRateOfStartAspirin Equ_CompetingDeath 10 Use modified death rates fr Discount_Firsti0Years Discount_After 10Years A Before modifying any parameters under the project window including steps in 4 1 2 4 1 4 one needs to delete existing results using the Delete Results button Otherwise no modifications on the project can be saved and need to be redone This is a problem the future vers
74. etes 1 Compliance _30 Compliance_statin 1 Compliance _30 Compliance_CVD 1 Compliance _90 COMMENT 1 1 Age Ge Time 2 Age 1 Notes Compliance levels for diseases are defined Compliance levels for treatment for hypertension Compliance levels for aspirin Compliance levels for beta_blocker Compliance levels for treatment for dysglycemia Compliance levels for treatment for dyslipidemia Compliance levels for treatment for dysglycemia dysili Initialization Age increase X 4 1 3 Defining cost values and utility scores The MMD provides a cost module and a utility score module To access these modules following the two steps below 1 Inthe main window click on the project name you are working on ee eee ee TAER Tee rey ee ee eee E TEST C Users wye Desktop MichiganModelForDiabetes2 0_07282015 zip File Forms Help f a alee Indirect Estimation and Simulation Tool Project Name Project Type Add New Project Obervational Study Ternplate Simulation Interventional Study Ternplate Simulation My First Diabagg Simulation Simulation Study Model Transitions Probabilities Parameters Notes How to simulate an Observational Study How to simulate an interventional study How to simulate an Observational Study 21 Michigan Model for Diabetes User Manual 2 Inthe project window click on the tab Stage 4 Update Costs File Help Project Definition
75. for the normal distributed age variable in the template We use CappedGaussian instead of the standard normal random number to avoid extreme values POPULATION DATA File Help Population Set Template for specifyin Parameter Distribution Corr_SBP_DBP 0 82 Diabetes _Type_ 1 Alive 1 Duration Of Diabetes Max 0 CappedGaussian3 2 7 Male Bernoulli 0 56 Race Bernoulli 0 18 1 BMI Max 0 Min 45 Gaussian 31 98 5 16 Parameter Notes E Alc Short for HbA 1c in LS Distribution Alc Diff The difference in Aic Alc Difi The difference in Alc lagged 1 time ste Alc Diff The difference in Alc lagged 2 time st Alc Prev Alc from previous year ACE Level acm 5 Type in the narrow window to modify the distribution and click the Up Arrow button to send the distribution definition for Age back to the upper list E POPULATION DATA File Help Population Set Template for specifying distribution Parameter Distribution Corr_SBP_DBP 0 82 Diabetes Type_ 1 Alive 1 Duration_Of Diabetes Max 0 CappedGaussian3 2 7 Male Bernoulli 0 56 Bernoulli 0 18 1 Max 0 Min 45 Gaussian 31 98 5 16 Parameter 40 Michigan Model for Diabetes User Manual 6 The updated list looks like this 5 POPULATION DATA co fee File Help Population Set Template for specifying distribution Columns Parameter Distribution Corr_SBP_DEP Diabetes Type 2 Race
76. ical functions as in Exp Age They can also use if statements as in lif Gr Age 1 50 1 0 These expressions can also represent tables as in Table 1 3 0 0 5 1 Age NaN 20 30 40 These formulas may contain as literals parameter names including parameters that hold values parameters that specify user defined functions state indicator names and some reserved words mathematical operators system built in functions Below is a list of allowed operators D1 Supported arithmetic functions Addition operator negative subtraction operator multiplication operator division operator note that integers will be treated as floating point numbers power operator D2 Other supported literals e Parenthesis to determine the order of the calculation e brackets enclosing comma separated values describe vectors and matrices Note that this type of expression is limited to defined vectors and matrices D3 Comparison operators Eq x1 x2 will return 1 if x1 x2 and 0 otherwise Ne x1 x2 will return 1 if x1 lt gt x2 and 0 otherwise Gr x1 x2 will return 1 if x1 gt x2 and 0 otherwise Ge x1 x2 will return 1 if x1 gt x2 and 0 otherwise Ls x1 x2 will return 1 if x1 lt x2 and 0 otherwise Le x1 x2 will return 1 if x1 lt x2 and 0 otherwise D4 A list of Boolean operators In the following Boolean operators the results are either 1 or 0 Any argument that not zero is considered be true and zero is treated as
77. imulation oe a Name My First Diabetes Simulatic Created On 2015 07 19 19 21 38 652000 Notes How to simulate an Observational St Derived From Example Project 1 Last Modified 2015 07 19 19 21 58 203000 aeee Primary Model Michigan Model For Diabetes 201 v No of Simulation Steps 5 ate Population Set My population F No of Repetitions View Result Delete Results Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the simulation Affected Parameter Function Notes Threshold_SBP 150 Threshold for increasing the le i Tho ld At a1 Th secl ola Low ia ewww wie l a la If you are using a population set with individual data the number in the small window of No of Repetitions tells the computer how many repetitions for each subject in your population set will be simulated For example if you have 100 subjects in the population set and you set No of Repetitions to be 2 the program will simulate 200 subjects in total 6 2 Number of years simulated To set the length of the simulation fill in the number of years to simulate in the small window of No of Simulation Steps In the following example the length of simulation is set to be 5 years PROJECT DEFINITION f Se File Help Project Definition Simulation Save Copy Undo Name My First Dia
78. in Section 4 1 2 52 Michigan Model for Diabetes User Manual Step 2 Run the model and then export the data to a csv file Use the SAS program Example1 amp 2_Summary sas to generate reports on the simulation results The quality adjusted life expectancy for subject one should be approximately 17 6 4 5 years 10 7 2 8 QALYs and for Subject two with the higher HbA1c at the beginning somewhat smaller at approximately 16 3 5 0 years 9 9 3 0 QALYs Total cost is approximately 142 266 for Subject one and 170 612 for Subject two The following figures show the time paths for a few of these risk factors in Subject one and Subject two respectivel Subject One Subject Two Alc over time Alc over time Time Time SBP over time SBP over time SBP SBP SBP SBP _53 Michigan Model for Diabetes User Manual Subject One Subject Two Total Cholesteroal over time Total Cholesteroal over time TC Time Time The following table shows the simulated incidence rate for subject one and subject one if neither of them complies with any treatment Complication Subject One Subject Two Incidence rate Cumulative Incidence rate Cumulative 1000 PY Incidence 1000 PY Incidence MI 18 9 31 0 30 3 44 3 CHF 12 5 20 5 21 6 Stroke 3 7 6 4 10 4 Revascularization 28 1 43 0 Amputation 4 0 7 0 Blind In Both Eyes 0 23 0 4 ESRD
79. in the simulation Affected Parameter Occurrence Probability Function Notes i PCL CABG Or CVD_Procedure_ Ente Bernoulli 0 67 Procedure 3 1 are CABG 3 2 ar z YearlyEventCost 1 0 Set Yearly Cost to 0 Cost_Comment 1 1 ADD UP EVENT COST YearlyEventCost Or Proliferative_left_Ent YearlyEventCost 1101 Add cost of amputation YearlyEventCost ESRD_Transplant_Entered YearlyEventCost 138071 ESRD_transplant cost accordin YearlyEventCost Angina_Entered YearlyEventCost 8282 Event Cost for Angina YearlyEventCost Or CVD_Procedure_Ente YearlyEventCost Iif PC Procedure Cost 3 1 are CABG YearlyEventCost Or And MLEntered 1 C YearlyEventCost 41744 Mycardial Infarction without pr YearlyEventCost Or And MLEntered CVD YearlyEventCost 60865 Mycardial Infarction with proc YearlyEventCost Or CHFwoML Entered E YearlyEventCost 34635 Add cost of CHF hospitalization YearlyEventCost Stroke_Entered YearlyEventCost 55278 Event cost for Stroke YearlvFventC nst Hvynonlvcemia VearlvFventCost 14991 Confirmed Clinical Neurnnathy Cost QoL Wizard Cost QoL parameter Occurrence Probability action Notes YearlyEventCost T v Amputation_Entered v fYearlyEventCost 4292 Add cost of amputation Michigan Model for Diabetes User Manual 3 When you are done with modifying click on the Up Arrow and bring back the parameter to the cost utility window Pea ee eee toe ae Cer ee a HABENG SS On eR ly ata NN 5 PROJECT DEF
80. inson B et al US Renal Data System 2014 annual data report epidemiology of kidney disease in the United States Am J Kidney Dis 2015 66 1 suppl 1 S1 S306 Table H 10 1 in Section H Available at http www usrds org reference aspx cited 08 25 2015 Rosenstock J Fonseca V McGill JB Riddle M Hall JP Hramiak I Johnston P Davis M Similar progression of diabetic retinopathy with insulin glargine and neutral protamine Hagedorn NPH insulin in patients with type 2 diabetes a long term randomised open label study Diabetologia 2009 Sep 52 9 1778 88 85 39 AO 41 42 43 44 45 Sherifali D Nerenberg K Pullenayegum E Cheng JE Gerstein HC The Effect of Oral Antidiabetic Agents on HbA1c Levels A systematic review and meta analysis Diabetes Care August 2010 33 1859 1864 UK Prospective Diabetes Study UKPDS Group Intensive blood glucose control with sulphonylureas or insulin compared with conventional treatment and risk of complications in patients with type 2 diabetes UKPDS 33 Lancet 1998 352 837 853 doi 10 1016 S0140 6736 98 07019 6 Wald DS Law M Morris JK Bestwick JP Wald NJ Combination therapy versus monotherapy in reducing blood pressure meta analysis on 11 000 participants from 42 trials Am J Med 2009 Mar 122 3 290 300 Yki Jarvinen H Kauppinen Makelin R Tiikkainen M Vahatalo M Virtamo H Nikkil K Tulokas T Hulme S Hardy K McNulty S H nninen J Lev nen H Lahdenper S Lehtonen R R
81. ion of IEST will fix 4 1 2 Defining general treatment parameters and compliance rate There are five types of treatments and one behavior change modeled in MMD 1 Treatment for hyperglycemia Treatment for hypertension Treatment for dyslipidemia Beta blocker Aspirin therapy Smoking cessation W N O In MMD the change of treatment depends on four factors levels of risk factors disease history or diagnosis the maximum level of treatment available and patients compliance characteristics 1 The need for change of treatment or behavior The need for starting or intensifying treatments for hyperglycemia hypertension and dyslipidemia are triggered by a relevant risk factor passing the specific treatment threshold The need for starting beta blocker is triggered by a CVD event CVD myocardial infarction MI revascularization procedure stroke or heart failure and 11 Michigan Model for Diabetes User Manual diagnosis of coronary artery disease CAD Aspirin and smoking cessation are recommended for all patients especially subjects with CVD or CAD 2 Compliance characteristics We assume each person has a fixed compliance profile for all the five types treatments e g for each type of treatment a patient either complies all the time or never complies with any prescriptions For current smoker the model does not assign a compliance status i e all current smokers can potentially quit 3 History of disease o
82. is 10 units higher than the usual treatment threshold Ce te ke Primary Model Michigan Model For Dij Function Population Set Example Population 1 Stage 0 Initialization Stage 1 Update Cowari Covariate update rules in the simulation Affected Parameter Started_Diabetic COMMENT Addl_Diabetes Addl_Statin Add1_Aspirin Smoke COMMENT Age Temp xd Temp _x2 Temp x4 a Covariate If in State Addl_ACE Modify Text and Press OK or Press Cancel to ignore changes Compliance_ACE Ge SBP Threshold_SBP 10 Occurrence Probability Eq Time 1 ae we LUF Te ee i ge CappedGaussian3 CappedGaussian3 CappedGaussian3 m aca a Function Notes Run Simulation Delete Results Stage 4 Update Costs Notes Sets to 1 if diabetic at start er Define Intervention For Year 1 E Intervention on dyslipidemia ii Taking Aspirin in Year 1 smoking cessation in Year 1 Update risk factors Age increase Generate random number use Generate random number use Generate random number use t Compliance_ACE Ge SE Intervention on hypertension in Year 7 Close the editing window by clicking OK and then click on the Up Arrow to bring the modified line back to the Stage1 Update Covariate tab window Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Stage 0 Initialization Covariate update rules in th
83. l for Diabetes User Manual Y Repeat 95 if subject have CHF procedure gt 0 if subject does not have CHF CHF Z Repeat 5 procedure gt CHD death Medications in this table refer to aspirin lipid drug ACE inhibitor and beta blocker 63 Michigan Model for Diabetes User Manual Table A2 Calibration and references for transition probabilities in Ml repeat MI module Figure A3 Transition Transition Probability Calibration Reference a MI gt CHD death fatal MI Modified the UKDPS fatality equation by add Calibrated to10 fatal Clarke et al 2004 MI gender effect The new odds of death is MI for men and 15 Colhoun et al 3 2514 2 772 Ln Age 52 59 HbA1c fatal MI among all first 2004 Roffi et al 7 09 0 114 2 640 Female Ln 3 5 MI events in Colhoun et 2013 We then calculate the probability of death using the al 2004 study These odds and adjusted by a factor 0 18 disregard fatality rate is based on whether a patient has CHF or not information in Roffi et al 2013 Repeat MI Calibrated to Jensen et Clarke et al 2004 For subjects with CHF Using the probability from the al 2011 Jensen et al 2011 modified odds as described above For subjects without CHF Using the probability from the modified odds further adjusted by a factor 0 53 b MI gt Short term 1 transition probability in a Clarke et al 2004 survival of MI Colhoun et al 2004 Roffi et al 2013 c Short term survival
84. l risk factor changes by defining treatment thresholds and compliance rates for hyperglycemia dyslipidemia and hypertension and compliance to quitting smoking and taking aspirin Given the fact that modern medicines have largely decreased the complication rate in type 2 diabetes through management of these risk factors it is important to explicitly model these management strategies and allow users to modify them to match the specific scenarios that they are simulating Most of the risk equations adapted in the coronary heart disease sub model and cerebrovascular disease sub model are from the UKPDS Outcomes Model 1 Appendix A Reference 5 which was based on a population of newly diagnosed diabetics between 25 and 65 years of age that were followed for 14 years These equations model race with only two categories Caucasians and Blacks In light of this and recognizing that the other data sources for our model are studies that were conducted in the United States and Western Europe and considering the difference in medical practice across countries caution should be applied when model results are extrapolated to populations that differ significantly from the model target population relatively young 25 79 years of age Caucasians or Black populations with type 2 diabetes in the United States and Western Europe Despite this the IEST software which houses our model allows users to adjust parameters to better suit their own situations For example
85. le Threshold for increasing the le Threshold for increasing the le Highest level of treatment with Highest level of treatment with 10 Use modified death rates f 3 Scroll down on this tab you can find the section for defining treatment changes at the beginning of Year 1 a S 2 eee File Help Pe Denman iie sve Gor ne Name My Interventional Study Created On 2015 07 28 13 09 21 585000 Notes How to simulate an a z int tional st Derived From Interventional Study Temp Last Modified 2015 07 28 13 09 21 585000 imei 7 Primary Model Michigan Model For Diabetes 201 Population Set Template for Specifying Distributi v m No of Simulation Steps 5 No of Repetitions 5000 A on pape Covariate update rules in the simulation Affected Parameter Was _BasalInsulin Was Insulin Started_Diabetic COMMENT Addi_Diabetes Addi_Statin Addi_Aspirin Addi_ACE Occurrence Probability Eq Time 1 Eq Time 1 Eq Time1 Eq Time 1 Eq Insulin 0 Eq Time1 Eq Time 1 Eq Aspirin 0 Eq Time 1 Ge Time 1 Eq Smoke 1 Ge Time 2 1 Covariate If in State Occurrence Probability Function BasalInsulin Insulin 1 1 Compliance_diabetes Ge Alc Threshold_Alc Compliance_statin Ge LDLCholesterol Thres Compliance_Aspirin Bernoulli YearlyRateOfs Compliance_ACE Ge SBP Threshold_SBP Bernoulli 1 YearlyRateOfQuittingSmoking Function Sets t
86. lsAfricanAmerican 0 96 762149 7 300000 Temp_ x4 2 505755 Temp_ x5 SBP_diff change in SBP DBP_diff change in DBP Age current age 81 SBP current SBP DBP current DBP BMI_ diff future change in BMI A3 Hypoglycemia severe Anti hyperglycemia treatment Intensive lifestyle None diet and exercise weight loss Metformin one OAD non insulin med Metformin Sulfonylureas two 0 004 event per person per year Spangas scat an et al 2010 OADs non insulin meds Add Basal insulin to OAD non insulin 0 02 event per person per year 1 Event per patient per year median 0 4 events med in 243 patients 1 7 Holman et al 2007 2 0 severe event in LANMET study Yki Jarvinen et al 2006 3 0 03 event per patient per year Bretzel et al 2008 Intensive insulin therapy 0 12 event per person per year 0 02 0 35 event per patient per year Zammitt and Frier 2005 82 References 10 11 Adler Al Boyko EJ Ahroni JH Smith DG Lower extremity amputation in diabetes The independent effects of peripheral vascular disease sensory neuropathy and foot ulcers Diabetes Care 22 1029 1035 1999 Avogaro A Giorda C Maggini M Mannucci E Raschetti R Lombardo F Spila Alegiani S Turco S Velussi M Ferrannini E Diabetes and Informatics Study Group Association of Clinical Diabetologists Istituto Superiore di Sanita Incidence of coronary heart disease in type 2 diabetic men and women imp
87. ltiplying Incidence with Number of risk at each row Both rounded and not rounded incident counts were close The rounded calculation was selected The sum of incidences was divided by the total number at risk to obtain the 10 year probability The 1 year equivalent transition probabilities were calculated Since there were no incidences of Blindness for non taking Insulin at this age group an assumption is made The assumption is that the chance of blindness from Proliferative is the same as the probability from Non Proliferative These numbers are temporary and require modification Klein et al 1995 Table 3 Numbers were calculated by summing all the incidents from all rows in the table except the first and last rows Only older onset numbers were used Incidences were calculated from multiplying Incidence with Number of risk at each row Both rounded and not rounded incident counts were close The rounded calculation was selected The sum of incidences was divided by the total number at risk to obtain the 10 year probability The 1 year equivalent transition probabilities were calculated See the XL spreadsheet for detailed calculations lt was decided to use progression probabilities similar to the transition from Proliferative to blindness The reason these were used is that Moss et al 1994 Table 3 shows Macular Edema has similar loss in the visual angle to Proliferative retinopathy in the taking insulin column 60 7 vs 52 0 69 2 50 0 8
88. ment in patients with acute myocardial infarction and diabetes experiences from an extended follow up of the diabetes mellitus insulin glucose infusion in acute myocardial infarction DIGAMI 2 study Diabetologia 2011 54 1308 1317 84 26 27 28 29 30 31 32 33 34 35 36 37 38 Moss SE Klein R Klein BE Ten year incidence of visual loss in a diabetic population Ophthalmology 101 1061 1070 1994 Phung OJ Scholle JM Talwar M Coleman Cl PharmD Effect of Noninsulin Antidiabetic Drugs Added to Metformin Therapy on Glycemic Control Weight Gain and Hypoglycemia in Type 2 Diabetes JAMA 2010 303 14 1410 1418 Ravid M Savin H Jutrin Bental T Katz B Lishner M Long term stabilizing effect of angiotensin converting enzyme inhibition on plasma creatinine and on proteinuria in normotensive type II diabetic patients Ann Intern Med 118 577 581 1993 Rhoads GG Dain MP Zhang Q Kennedy L Two year glycaemic control and healthcare expenditures following initiation of insulin glargine versus neutral protamine Hagedorn insulin in type 2 diabetes Diabetes Obes Metab 2011 Aug 13 8 711 7 Roffi M Radovanovic D Erne P Urban P Windecker S Eberli FR for the AMIS Plus Investigator Gender related mortality trends among diabetic patients with ST segment elevation myocardial infarction insights from a nationwide registry 1997 2010 Eur Heart J 2013 2 4 342 349 Sacco RL Shi T Zamanillo MC Kargm
89. mplications We divided disease specific costs into two categories 1 event costs that are the one time costs and accrue within the year in which a complication first occurs and 2 state costs that are intended to reflect the ongoing costs in subsequent years that are associated with the management of the complications Table B1 22 Michigan Model for Diabetes User Manual in Appendix B shows the detailed costs of complications for MMD All default costs are expressed in 2014 US dollars Users can modify costs following the steps below using the cost of amputation as an example 1 Highlight the cost you would like to modify and click on the Down Arrow at the bottom of the window to bring down the parameter line to the editing cells File Help Project Definition Simulation Notes How to simulate an Observational Study Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Cost Quality of Life update rules in the simulation Name My First Diabetes Simulatic Created On 2015 07 28 12 32 34 591000 Derived From Qbervational Study Temple Last Modified 2915 07 28 12 50 55 266000 Primary Model Michigan Model For Diabetes 201 No of Simulation Steps 10 Population Set My population v aa No of Repetitions 1 Function Notes If In State Ta Procedure 3 1 are CABG 3 2 ar E Affected Parameter PCL CABG
90. nce 1994 in this scope The multiplier is somewhat an arbitrary assumption and should be improved in the future with concrete evidence 71 A1 3 Nephropathy sub model pO 1 No 2 Micro 4 ESRD with 5 ESRD with 3 Proteinuria 6 ESRD Death Nephropathy Albuminuria Dialysis Transplant Figure A6 Structure of nephropathy sub model Table A6 Transition probability in nephropathy sub model FP hin I Lil Transition Gall et al 1997 number for 5 year progression in key messages p 787 is 0 23 Adjusted for 1 year from 5 years 1 1 0 23 1 5 2to3 0 1032 Ravid et al 1993 the risk for developing this degree of proteinuria within 5 years of follow up was 19 45 42 in the placebo group Number adjusted for 1 year from 5 years 0 1032 1 1 0 42 1 5 3 to 4 0 0082 Humphrey et al 1989 page 791 page 791 after 5 year 7 0 8 4 developed it by 10 years and 11 6 by 15 years the 15 year number was selected Number adjusted for 1 year from 15 years 0 0082 1 1 0 116 1 15 4to5 0 006 to 0 084 This data of the renal transplant rates in dialysis patients in year 2013 was provided depends on by KECC at the University of Michigan The data was processed using the age gender and race following criteria 1 only the data for diabetes as ESRD cause was selected 2 the data depended on age gender and race 3 the data for White and Black was selected 4 the data was divided by 100 to represent the yearly t
91. ncident cohort and 446 had previously diagnosed diabetes prevalent cohort Duration of diabetes of the prevalent cohort is unknown During the median follow up 10 years 308 subjects in the prevalent cohort and 134 subjects in the incident cohort developed CHF Predictors Selection of potential predictors was informed by characteristics included in the UKPDS Outcome Models I amp Il Clarke et al 2004 Hayes et al 2013 and Risk Equations for First and Second Cardiovascular Events from Swedish Register Data Kiadaliri et al 2013 Initially 15 risk factors were selected as candidate predictors for the regression model including history of angina history of MI history of angioplasty history of bypass surgery Atrial fibrillation AF most recent value of fasting glucose LDL lipid ratio total cholesterol HDL SBP DBP BMI sex race smoking status and age at CHS study baseline visit Of these 15 risk factors sex race smoking status and age at baseline are time independent covariates The other nine risk factors are time dependent covariates Data analysis and model selection Given that duration of diabetes is a very important risk factor for CHF Kiadaliri et al 2013 one would typically use the incident cohort only to derive the CHF prediction model However the smaller number of events in the incidence cohort limited the statistical power for model development At least 10 20 events per candidate predictor have been pr
92. ngoing cost O Brien JA Patrick AR Caro J Estimates of direct medical costs for microvascular and macrovascular complications resulting from type 2 diabetes mellitus in the United States in 2000 Clin Ther 2003 25 1017 38 Ward A Alvarez P Vo L Martin S Direct medical costs of complications of diabetes in the United States estimates for event year and annual state costs USD 2012 J Med Econ 2014 17 176 83 Ward et al followed the same approach as O Brien et al s article published in Clin Ther 2003 25 101 7 38 but the reasons for the higher cost of hypoglycemia requiring hospitalization reported by Ward et al may include 1 Ward et al s data was from year 2010 while O Brien et al used 1998 data and thus inflation for these many years won t reflect the increment on hospitalization cost for these many years 2 inpatient physician resource use profile inpatient physician fee and post acute care cost may have been different between 1998 and 2010 for example the physician fees applied in the O Brien et al s analysis were much more lower and definitely the hospital stay costs have changed a lot between these two analyses 3 cost to charge ratios were different 0 61 used by O Brien et al vs 0 231 0 767 used by Ward et al and 4 there has been a minor change in the definition of the cases Based on Table 3 in the following study the ongoing costs were determined to be 396 in 2009 US for microalbuminuria and 678
93. nical Neuropath 1 No_Neuropath sub model Amputation K No_Proliferative_Retin Bernoulli 0 78 Left eye opathy_left retinopathy Nonproliferative_left lif No_ Proliferative Retinopathy sub model _left 0 Bernoulli 0 5 Proliferative left lif No_ Proliferative Retinopathy _left Nonproliferative_left 0 1 Blind Eye left No_Proliferative Retin Bernoulli 0 78 Right eye opathy_right retinopathy Nonproliferative_right lif No_ Proliferative Retinopathy _right 0 Bernoulli 0 5 sub model Proliferative_right lif No_ Proliferative Retinopathy _right Nonproliferative_right 0 1 OO 42 Blind Eye_right No _Macular_edema_le Bernoulli 0 90 Left eye retinopathy Macular_edema_left 1 No Macular _edema_left sub model No _Macular_edema_ri Bernoulli 0 90 Right eye ght retinopath F F Michigan Model for Diabetes User Manual Macular_edema_right 1 No_Macular_edema_right cain Medication lif IntensiveLifeStyle 0 Bernoulli 3 9 lif IntensiveLifeStyle Metformin 0 Bernoulli 2 6 Basallnsulin lif IntensiveLifeStyle Metformin OtherOralMedication O Bernoulli 1 4 lif IntensiveLifeStyle Metformin OtherOralMedication Basallnsul in O 1 lif Or Survive_MI Survive_ Stroke 1 Bernoulli 0 15 0 10 1 0 10 lif Or Survive_MI Survive_Stroke 1 Bernoulli 0 389 0 10 1 0 10 lif Or Survive_MI Survive_Stroke 1 Bernoulli 0 531 0 10 1 0 10 lif Or Survive_MI Survive_Stroke
94. nsulin treatment progression column for both categories were selected Numbers were adjusted for 1 year 0 1140 for diabetics progression 0 1140 0 114024676 1 1 0 702 1 10 0 0653 0 065301 1 1 who need Insulin 0 491 1 10 treatment 74 0 0390 for diabetics need Insulin treatment 0 0233 for diabetics who do not need Insulin treatment 0 0148 for diabetics need Insulin treatment 0 0166 for diabetics who do not need Insulin treatment 0 0148 for diabetics need Insulin treatment 0 0166 for diabetics who do not need Insulin treatment Klein et al 994 Table 8 70 2 10 yr progression rate was used for insulin taking group and 49 1 10 yr progression rate was used for non insulin taking group The first row and the progression column for both categories were selected Numbers were adjusted for 1 year progression 0 1140 0 114024676 1 1 0 702 1 10 0 0653 0 065301 1 1 0 491 1 10 For IGT the probability is from Ref F1 Table 3 The nondiabetic retinopathy incidence after 5 6 years is 24 out of 24 278 When this is converted to yearly probabilities we get 1 1 24 0 24 278 1 5 6 0 014677981118243144 0 0147 Retinopathy is assumed to be non proliferative for IGT since our model does not allow non diabetic proliferative retinopathy Moss et al 1994 Table 2 Only older onset numbers were used the last 4 rows were used Severity 60 85 PDR Incidences were calculated from mu
95. ntered BMI gt 5 Therefore we used linear splines with one knot at BMI 33 centered BMI 4 2 to model BMI effect y2 test showed that this transformed BMI variable provided a significantly better fit p 0 012 To select the best prediction model we used a stepwise selection procedure with higher than standard p value We used Akaike s Information Criterium AIC which implies a p value lt 0 157 for selection of predictions with 1 df Results The stepwise selection approach selected a model with 10 predictors Estimated regression confidents are reported in Table A4 C index for this model varies from 0 678 to 0 699 at 1 to 10 years indicating acceptable discrimination Using non linear regression analysis we fitted a Weibull baseline cumulative function to the estimated non parametric baseline function of the 68 Michigan Model for Diabetes User Manual incidence cohort strata Figure A4 The estimated Weibull function parameters p and A are also shown in Table A4 Figure A4 Weibull baseline cumulative hazard functions 005 0 10 015 020 Baseline Cumulative Hazard Function 0 00 0 2 4 6 6 10 12 14 Time Year 69 Table A4 Parameters in the prediction model for risk of congestive heart failure in T2DM Parameter Parameter Hazard Ratio 95 Cl Estimate P Value ooo e m T a Ln TC HDL centered at 4 62 0 782 0 00026 2 19 1 44 3 32 SBP centered at 136 9 e a DBP centered at 69 4 BMI
96. o 1 if diabetic at start eit Define Intervention For Year1 Intervention on dyslipidemia in Taking Aspirin in Year 1 Intervention on hypertension i smoking cessation in Year1 Age increase Generate random number use Notes 28 Michigan Model for Diabetes User Manual 4 To modify the treatment changing rules in year 1 highlight the treatment you would like to modify and click on the Down Arrow at the bottom of the window to bring down the parameter line to the editing cells Stage 1 Update Covariates Covariate update rules in the simulation Affected Parameter If In State Occurrence Probability Function Was_OtherOralMedication Eq Time 1 OtherOralMedication Was_Basallnsulin Eq Time 1 BasalInsulin Was_Insulin Eq Time 1 Insulin Started Diabetic Eq Time 1 1 Sets to 1 if diabetic at start eii COMMENT 1 Define Intervention For Year 1 Add 1_Diabetes Eq Time 1 Eq Insulin 0 Compliance_diabetes Ge Aic Threshold _A ic Add1_Statin Eq Time 1 Compliance_statin Ge LDLCholesterol Threshold Intervention on dyslipidemia in Eq Time 1 Eqg Aspirin 0 Compliance_Aspirin Bernoulli YearlyRateOfStartA Taking Aspirin in Year 1 Eq Time 1 Compliance _ACE Ge SBP Threshold SBP Intervention on hypertension i Ge Time 1 Eq Smoke 1 Bernoulli 1 YearlyRateOfQuittingSmoking smoking cessation in Year 1 1 Update risk factors Ge Time 2 Age 1 Age increase Covariate If in State Occurrence Probability
97. oject Name Project Type Notes Add New Project Obervational Study Template Simulation How to simulate an Observational Study Interventional Study Template Simulation How to simulate an interventional study My First Diabetes Simulation Simulation How to simulate an interventional study Study Model Populations Parameters 2 Click on the Data button on the right side of the Template for specifying distribution line to open the data window Population Sets ae n fm File Help Population Sets add Find Name Definition Type Created On Notes Source Derived From Last Modified x Example Population 1 Data based 2014 11 10 13 42 19 3 2014 11 10 13 44 36 9 x Template for specifying distrib Distribution based 2014 11 14 13 59 15 9 2014 11 17 16732 05 71 In the following data window you can see a list of distributions for all the required variables as listed in the table in Section 5 page 32 35 You can change the definition for any of these variables to suit your population You may use different type of expressions and functions to define you population See Appendix D for a list of Python expressions that are allowed in the IEST software It is important to keep the order of how these distributions are defined 38 Michigan Model for Diabetes User Manual File Help Population Set Template for specifying distribution Parameter Corr_SBP_DBP Diabetes Type 2 1 Alive 1 Age C
98. olman et al 2007 Rhoads et al 2011 Holman et al 2009 Since the individuals in the 4T study did receive intensive insulin therapy after one year of basal insulin most of them had already an HbA1c lt 8 0 Baseline HbA1c before initiation of intensive therapy was 7 6 and median HbAic after 2 years was 6 9 CI 6 6 to 7 1 Therefore we would change the decrease in HbA1c using intensive insulin for our model to 1 0 SD 0 1 Every year the change of lipid is calculated by adding initial change induced by treatment change if any and the change following that which can be attributed to aging or disease progression Drug effect Currently we model two levels of treatment for dyslipidemia For each of these two levels the drug induced change is 25 decrease 5 increase and 6 increase in LDL C HDL C and triglyceride respectively Aging effect x1 x2 x3 are three randomly drawn independently distributed standard normal variables They are redrawn each year The three following equations calculate the change in logarithm e based transformation of HDL LDL and triglyceride based on the current value of Ln HDL Ln_LDL Ln_triglyceride change in BMI change in logarithm e based transformed fasting glucose and gender 79 Diff_Ln_ HDL Change 0 0340 Age 00112 Age Age 0 0000117 Ln_Triglycerides 0145 Ln_LDL 000961 Ln_HDL 0844 Diff_Ln_FastingGlucose 0364 Diff_BMI 00414 Female 0 0147 0
99. oposed in previous guidelines for the development of prediction models Harrell et al 1984 In order to overcome the problem caused by missing duration of diabetes in the prevalent cohort and to make use of the information provided by this cohort we employed the following analysis strategy First we used a Cox proportional hazard regression model stratified by cohort types i e prevalent cohort and incident cohort This model allowed us to derive a non parametric estimation of baseline hazard function for each of the two cohorts separately while using data from both cohorts to select predictors and estimate corresponding risk coefficients By including data from both cohorts we had a total of 442 CHF events which provided 29 events per candidate predictor This was more powerful than lt 10 events per candidate predictor 67 Michigan Model for Diabetes User Manual which the incident cohort alone would have provided This model also allowed us to accommodate both time independent and time dependent predictors Second in order to use the model for long term prediction we used a non linear regression model to fit a Weibull cumulative hazard function to the estimated non parametric cumulative baseline hazard function of the incident cohort derived from the Cox proportional hazard model The Weibull model assumes a baseline hazard given by the function ho t pt exp A and the hazards model for the ith subject at time t is h t xi
100. or changes in different levels Max Level ACE There are 9 levels of anti hypertensive treatment defined in the MMD 0 No anti hypertensive treatment one drug half dose one drug full dose two drugs half dose two drugs full dose three drugs half dose three drugs full dose four drugs half dose four drugs full dose You can set this parameter to any integer between 0 and 8 See Appendix A2 for the effect of or change in different levels Max_Level_ Statin There are a totally of 2 level of anti dyslipidemia treatment defined in the MMD 0 No anti dyslipidemia treatment 1 one drug half dose 2 one drug full dose You can set this parameter to any integer between 0 and 2 See Appendix A2 for the effect of or change in different levels YearlyRateOfQuittingSmoking This parameter allows you to define the yearly rate of smoking cessation among current smokers who did not experience any major CVD nor was diagnosed with CAD This parameter can be any value from 0 to 1 YearlyRateofStartAspirin For patients who did not experience any major CVD and were not diagnosed with CAD you can define a compliant rate to aspirin therapy as shown in section 4 1 2 At the same time not all the compliant patients start taking aspirin at the beginning This parameter allows you to define the rate of starting aspirin among all aspirin compliant patients This oarameter can be any value from 0 to 1 To modify the above parameters
101. ortion of patients that comply with treatment for hyperglycemia regardless of CVD event history to 80 you should set Compliance_diabetes Comp liance_ 80 To modify the above parameters do the following steps using compliance rate for aspirin as an example 19 Michigan Model for Diabetes User Manual 1 Highlight the parameter you would like to modify and click on the Down Arrow at the bottom of the window to bring down the parameter line to the editing cells Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Covariate update rules in the simulation Affected Parameter If In State Occurrence Probability Function Notes Compliance_15 Eq Time 1 Bernoulli 15 20 Compliance_20 Compliance _10 Eq Time 1 Bernoulli 10 15 Compliance_15 E Compliance_5 Eq Time 1 Bernoulli 5 10 Compliance _10 COMMENT 1 1 Compliance levels for diseases are defined Compliance_ACE 1 Compliance _30 Compliance levels for treatment for hypertension 1 Compliance _30 Compliance levels for aspirin Compliance_beta 1 Compliance _30 Compliance levels for beta_blocker Compliance_diabetes 1 Compliance _30 Compliance levels for treatment for dysglycemia Compliance_statin 1 Compliance _30 Compliance levels for treatment for dyslipidemia Compliance_CVD 1 Compliance_90 Compliance levels for treatment for dysalycemia dysli COMMENT
102. osx ansi 2 8 12 1 universal py2 7 dmg download 3 1 2 IEST software and MMD installation After Python environment has been properly installed Visit http AWwww med umich edu mdrtc cores DiseaseModel model htm to download the package that includes both IEST software and MMD Downloading the file requires compliance to its license and registration e Extract the downloaded zip file archive to a directory of your choice This will be your working directory e lf using OS X or Linux unzip the IEST software and issue the following command in the unzipped IEST working directory o python Main py 3 1 3 Running the IEST software Open the working directory created during installation and double click Main py The main form of the system titled Indirect Estimation and Simulation Tool will open A IEST Indirect Estimation and Simulation Tool E File Forms Help Indirect Estimation and Simulation Tool Project Name Project Type Notes States Study Model Transitions Probabilities A As the User Manual for MMD this document does not include detailed information on IEST To access the help system for IEST click on the Help menu or click here For a set of videos tutorials for IEST please click here Michigan Model for Diabetes User Manual 3 2 Loading the Michigan Model for Diabetes in the IEST software To load the MMD in the IEST software follow the steps below
103. r a subject is taking aspirin 1 Yes 0 No 34 Insulin Intensive bolus 1 Yes 0O No insulin p p lt Michigan Model for Diabetes User Manual This variable is an indicator for the state CAD w o MI as in shown in Appendix A For historical reason this variable name for this state was name as Angina in the software This variable is an indicator for the state CHF after MI as in shown in Appendix A For historical reason this variable name for this state was name as CHF in the software Additional instructions to set up five variables of medications for anti hyperglycemia treatment 1 If a subject is on insulin therapy in which only basal insulin or only premixed insulin is used s he should be considered at the 4 stage treatment for hyperglycemia and therefore only the variable Basallnsulin is set to be 1 2 If a subject is on insulin therapy in which any of rapid acting insulin short acting insulin or intermediate acting insulin is used s he should be considered at the 5 stage treatment for hyperglycemia and therefore only the variable Insulin is set to be 1 5 1 Input as data In the download folder the users can find an Excel file that provides a template for creating an input population labeled Input Population Template csv Input PopulationTemplate csv Excel t O X HOME INSERT PAGE LAYOUT FORMULAS DATA REVIEW VIEW DEVELOPER POWERPIVOT SAS Ye Wen a v HH oy
104. r diagnosis For the first three treatments i e treatments for hyperglycemia hypertension and dyslipidemia we also assume most patients are willing to comply with the need of treatment when they experience a CVD event Among the subjects who are non compliers but become willing to comply when they experience a CVD event when diagnosed with a CAD they comply with 50 probability 4 Maximum level of treatment There are a maximum of 5 2 and 8 treatment levels available for hyperglycemia treatment hypertension treatment and dyslipidemia treatment respectively When the maximum level of treatment has been reached no further intensification is available even if there is a need for that The following table shows the rule for each treatment behavior change See Appendix A2 for details on treatment regimens Treatment behavior Start or intensification rules change Hyperglycemia For each of these treatments if a compliers relevant risk factor i e HbA1c for hyperglycemia SBP for Hypertension hypertension LDL cholesterol for dyslipidemia passes a ACE I or ARB is started or user specified threshold the treatment will be started or intensified intensified Dyslipidemia Statin is started or For patients who are non compliant but become intensified compliant when there is a CVD event the treatment is started or intensified when the risk factor is higher than the threshold For the same group of patients if there is a nee
105. ransition probability and 5 the case counts for 0 21 age groups were probably too low to report the rates appropriately and thus the transplant rates in 22 44 age groups were used for 0 21 age groups 4to6 0 0434 to 0 5472 depends on Saran R Li Y Robinson B et al US Renal Data System 2014 annual data report gender age race Hypertension epidemiology of kidney disease in the United States Am J Kidney Dis 2015 66 1 adjusted by other death suppl 1 S1 S306 Table H 4 1 in Section H Available at causes http www usrds org reference aspx cited 08 25 2015 The data from the USRDS 72 table was processed using the following criteria 1 only the data for diabetes was selected 2 the data depended on age gender and race 3 the data for non Hispanic White and Black in the race columns was selected and 4 the data was divided by 1 000 to represent the yearly transition probability 0 0081 to 0 245 Saran R Li Y Robinson B et al US Renal Data System 2014 annual data report depends on epidemiology of kidney disease in the United States Am J Kidney Dis patton non igs oo 2015 66 1 suppl 1 S1 S306 Table H 10 1 in Section H Available at adjusted by other deat http www usrds org reference aspx cited 08 25 2015 The data from the USRDS causes Ha table was processed using the following criteria 1 only the data for diabetes was selected 2 the data depended on age gender and race 3 the data for non Hispanic
106. sion is nominal R 0 the value for the first level in the Gender dimension is 0 R 2 1 the value for the first level in the Gender dimension is 1 Roo 0 the lower bound of the first interval in the Age dimension is 0 R 30 the upper bound of the first interval and the lower bound of the second interval in the Age dimension is 30 R 22 60 the upper bound of the second interval and the lower bound of the third interval in the Age dimension is 60 Ro3 120 the upper bound for the third and last interval in the Age dimension is 60 D10 Special math symbols Note that these may be platform dependent Boolean operators treat NaN Not a Number as false as well as any other non number type such as a vector matrix Inf inf will be recognized by the system as infinite This symbol is not to be used in mathematical calculations as it may generate error It can be used for bound checks for parameters NaN nan will be recognized by the system as not a number Note that comparison of NaN to any number including NaN will return False Arithmetic operations using NaN produce NaN and may raise errors and therefore should be avoided Note that missing values are not supported by the system An exception is population data upload in which case missing data values are ignored by default in simulation 92
107. t ho t exp Bxi t pt exp A B xi t where x t is a vector of the risk factors for subject i at time t This two step strategy allowed us to derive a Weibull proportional hazard model with time dependent and time independent predictors Ideally a one step analysis to fit a Weibull proportional hazard model is preferred However such a model requires modeling the multiple longitudinal factors simultaneously and no existing software is available Figure S4 compares the non parametric cumulative baseline hazard from the Cox proportional hazard model and the fitted Weibull function The Weibull function fits the non parametric function very well Before any modeling was performed the distributions of all potential predictors were carefully examined for extreme values Biologically implausible values were set to missing values and the remaining extreme values were truncated by shifting the values below 1 centile and above 99 centile to truncated points Such truncation may prevent distortion of the relationship between predictor and outcome due to high leverage of the extreme values To define appropriate transformation of continuous variables we used p spline functions to explore the potential nonlinear effect of potential continuous predictors The only continuous predictor that has a non linear function form is BMI Based visual inspection we assumed no BMI effect until centered BMI centered at 28 2 5 and a linear effect for ce
108. t for hyperglycemia dyslipidemia and hypertension Modeling direct benefits of medications and compliance Updated transition probability tables for end stage renal disease Updated competing death table 4 5 6 7 Updated cost and utility models Michigan Model for Diabetes User Manual 3 Download and Installation In order to run the MMD one has to download both the MMD files and a disease modeling software the Indirect Estimation and Simulation Tool IEST 3 1 Download the disease modeling software IEST and Michigan Model for Diabetes 3 1 1 Installation of Python environment The IEST software is written using Python language It requires installation of Python version 2 7 and a few Python libraries as follows NOTE This software has been tested on Microsoft Windows XP Windows 7 and Linux Note that other operating systems such as OS X and other Windows versions may work yet were not fully tested Windows installation e Visit http python org ftp python 2 7 2 python 2 7 2 msi or http python org download releases 2 7 2 and download Python version 2 7 for Windows e Visit http downloads sourceforge net wxpython wxPython2 8 win32 unicode 2 8 12 1 py27 exe or htitp Awww wxpython org download php stable and download wxPython Requires Python a Unicode version suitable for Python version 2 7 for Windows 32 bit e Visit http sourceforge net projects numpy files NumPy 1 6 1 numpy 1 6 1 win32
109. th rates f kd Threshold for increasing the level of 4 You should see that the parameter is back in the list of parameters with the new value Stage 0 Initialization Stage 1 Update Covariates Stage 2 Update Complications Stage 3 Update Treatment Stage 4 Update Costs Initialize the simulation Affected Parameter Threshold_SBP Threshold Alc Threshold_LDL Max_Level_ACE Max_Level_Diabetes Trt Max_Level_Statin Max_Level_Aspirin YearlyRateOfQuittingSm YearlyRateOfStartAspirin Function Notes Threshold for increasing the le Threshold for increasing the le Threshold for increasing the le Highest level of treatment with Highest level of treatment with gt 17 Michigan Model for Diabetes User Manual Compliance Parameters To change specify treatment compliance rates click on Stage 1 Update Covariates to bring this tag to the front Project Definition Simulation Name My First Diabetes Simulatic Created On 2015 07 28 12 32 34 591000 Notes How to simulate an a Observational St Derived From Qbervational Study Templ Last Modified 2015 07 28 12 32 34 591000 mia Primary Model Michigan Model For Diabetes 201 v No of Simulation Steps 10 f Population Set My population No of Repetitions 4 Stage 0 Initialization Stage 1 Update kx Stage 2 Update Complications Stage 3 Update Tr
110. ther cardiovascular risk factors before calculating transition probabilities for each of the six sub models In order to correctly model the casual relationships between these risk factors we update them in the following order 1 Weight 2 HbAtc 3 Lipids 4 SBP and DBP The changes in these risk factors are determined by both treatment statues and aging disease progression When a patient is on lifestyle intervention only changes in BMI drives the changes in HbA1c When a patient is on oral non insulin glucose control drug s or insulin the drug affects the changes in HbA1c and weight independently which might not be the case but we do not have data 76 including the changes in the first year when the new treatment is initiated and the following years before next step of intensification of the treatment This set of models also models a causal relationship between different types of biomarkers For example the prediction models for lipids changes include both BMI and HbA1c changes as predictors thus allow changes in BMI and HbA1c drive the changes in lipids The other example is that changes in BMI drive the changes in DBP and SBP A2 1 Changes in Weight and BMI BMI changes is derived from weight changes Table A10 Changes of body weight under different anti hyperglycemia treatment treatment pinn p O ea eo SD of change 0 3kg year Intensive lifestyle Mean change 3 7kg Mean change 1 kg year Baseline 80 4kg SD 15 6 kg diet and
111. tion C in this document Age at diabetes Fried LP elal gt CHF w o MI Angina 1 Ml 0 adjusted for medication benefit onset sex SBP 1991 DBP lipid ratio BMI history of angina history of MI AF and medications L Immediate Cole 2002 60 Michigan Model for Diabetes User Manual fprcedu Toney atonal C CAD with procedure gt Ml UKPDS MI equation IHD 1 CHF 0 adjusted for medication benefit and by additionally adjusting the hazard function by a factor 1 387 UKPDS MI equation IHD 1 CHF 0 adjusted for medication benefit and by additionally adjusting the hazard function by a factor 0 37 based on calibration CHS risk equation Section C in this document Angina 1 Ml 0 adjusted for medication benefit Calibrated to the prompt group in Chaitman et al D CAD with 2009 procedure gt CHD death CC CAD with procedure gt CHF w o MI DD CHF w o MI gt Ml UKPDS MI equation IHD 1 if subjects had history of angina CHF 1 adjusted for medication benefit and by additionally adjusting the hazard function by a factor 0 07 Calibrated to Deedwania 2011 and Mellbin et al 2011 EE CHF w o MI gt CHD death UKPDS MI equation IHD 1 if subjects had history of angina CHF 1 adjusted for medication benefit and by additionally adjusting the hazard function by a factor 0 43 Calibrated to Deedwania 2011 and Mellbin et al 2011 E MI gt CHD See details in th
112. tion initiated from the initial state but the sub process can be ended due to another sub process reaches the terminal state i Nested parallel sub processes 57 Michigan Model for Diabetes User Manual A1 1 Coronary heart disease CHD sub model A1 1 2 Structure and transition probabilities for CHD sub model CHF w o MI Repeat MI AA CHD Procedure CAD with Procedure D Keys Regular State C Event State C gt Module Transition Figure A2 Coronary heart disease states and progression CHD coronary heart disease CAD coronary artery disease CHF w o Ml congestive heart failure without MI Ml myocardial Infarction CHF after Ml congestive heart failure after experience of MI Hx history w o without CHD procedure revascularization procedure 58 Michigan Model for Diabetes User Manual Ml repeat MI Module Re infarction within 1 yr of MI Procedure after MI MI Short term survival of MI Figure A3 Myocardial infarction module Ovals indicate instant states To Hx of MI To CHF after MI To CHD death 59 Michigan Model for Diabetes User Manual Table A1 Calibration and references for transition probabilities in the main CHD sub model Figure A2 Transition Transition Probability Calibration Risk factors Reference medications UKPDS MI equation IHD 0 CHF 0 adjusted for Calibrated to Age gender Clarke et medication
113. uria Proteinuria 1 1 1 Occurrence Probability Blind_Both_Eyes Function ilityScoreThisY ea Function 1 0 689 HealthUtilityScoreThis ear 0 038 HealthUtilityScoreThisYear 0 021 HealthUtilityScoreThis ear 0 025 HealthUtilityScoreThisVear 0 034 HealthUtilityScoreThis ear 0 043 HealthUtilityScoreThis ear 0 011 HealthUtilityScoreThisVear 0 078 HealthUtilityScoreThisVear 0 078 HealthUtilityscoreThis ear 0 065 Cost QoL Wizard Notes Penalty for both eyes blind Notes Calculati Set Healt Female p obese pe Penalty fi using ins Penalty fi Penalty fiz Penalty Ae Penalty fi Penalty fi j 4 The modified numbers is back in the list Cost Quality of Life update rules in the simulation Affected Parameter If In State Cost_Comment HealthUtilityScoreThisY ear HealthUtilityScoreThisY ear HealthUtilityScoreThisY ear HealthUtilityScoreThisY ear HealthUtilityScoreThisY ear HealthUtilityScoreThisYear HealthUtilityScoreThisYear Occurrence Probability 1 1 Female Ge BMI 30 Or Metformin OtherOralMedication Or BasalInsulin Insulin Blind_One_Eye_Only Blind Both Eyes Or Micro_Albuminuria Proteinuria 1 Function 1 0 689 HealthUtilityScoreThisY ear 0 038 HealthUtilityScoreThisYear 0 021 HealthUtilityScoreThisYear 0 023 HealthUtilityScoreThisYear 0 034 HealthUtilityScoreThisYear ESRD Transplant 4 i m Notes Calculatii Set Healt Female p obese pe
114. ycle so that all transition probabilities are calculated based on baseline characteristics For an intervention study risk factors will be changed according to treatment regimen used in the study to reflect the immediate intervention or on trial effect lf you wish to use the default MMD model parameters you only need to specify population baseline information and initial parameters i e treatment threshold maximum treatment level and compliance rate as model inputs Please read section 4 1 for instructions If you wish to further modify the MMD model parameters to suit your own situation please contact us at help MichiganModelForDiabetes umich edu Michigan Model for Diabetes User Manual 4 1 Running simulation using the default MMD 4 1 1 Start your own project The MMD zip file includes two example projects one observational study and one intervention study To start your own project do the following 1 Make a copy of the example that matches your project For example if you wish to simulate an observational study on the project list right click the line for Observational Study Template J EST CAUsers wye Desktop MMD MichiganModelForDiabetes2 0 betazip File Forms Help Indirect Estimation and Simulation Tool Project Name Project Type Notes Add New Project Observational Study Temolate imulati How to simulate an Observational Study Interventional Study Te Delete Record How to simulate an interv
115. yysy L Insulin glargine or NPH combined with metformin in type 2 diabetes the LANMET study Diabetologia 2006 Mar 49 3 442 51 Zammitt NN Frier BM Hypoglycemia in type 2 diabetes pathophysiology frequency and effects of different treatment modalities Diabetes Care 2005 28 2948 Zhou H Isaman DJ Messinger S Brown MB Klein R Brandle M Herman WH A computer simulation model of diabetes progression quality of life and cost Diabetes Care 2005 28 12 2856 63 Zoungas S Patel A Chalmers J de Galan BE Li Q Billot L Woodward M Ninomiya T Neal B MacMahon S Grobbee DE Kengne AP Marre M Heller S ADVANCE Collaborative Group Severe hypoglycemia and risks of vascular events and death NEJM 2010 363 1410 8 86 Appendix B Michigan Model for Diabetes Cost Model Table B1 Costs of complications for Michigan Model for Diabetes Event and ongoing costs of complications for 2014 US dollars Michigan Model for Diabetes Event Ongoing Retinopathy es ee Nephropathy SoS o d O Microalbuminuria ooo 48487 748 4 511 iN oO NI inuri Oo 748 5 5 Neuropathy o S S O Clinical neuropathy HFT 2 Cardiovascular disease DO a 2 2 Percutaneous transluminal coronary angioplasty 2 2 Congestive heart failure 34635 7620 6 Acute metabolic complication oo o Doo doooSo Death by age in years 2 o d l SO NA not applicable The baseline cost is the annual direct medical cost for a white man with type

The Michigan Model for Diabetes User Manual

Contents

Download Pdf Manuals

Related Search

Related Contents