Diabetes Mellitus (DM): GBD 2017¶
WHO defines Diabetes Mellitus (DM) as a chronic, metabolic disease characterized by elevated levels of blood glucose (or blood sugar), which leads over time to serious damage to the heart, blood vessels, eyes, kidneys, and nerves. The most common is type 2 diabetes, usually in adults, which occurs when the body becomes resistant to insulin or doesn’t make enough insulin. In the past three decades the prevalence of type 2 diabetes has risen dramatically in countries of all income levels. Type 1 diabetes, once known as juvenile diabetes or insulin-dependent diabetes, is a chronic condition in which the pancreas produces little or no insulin by itself. [WHO-Diabetes-Definition]
GBD 2017 Modeling Strategy¶
According to GBD 2017, the case definitions and diagnostic criteria for overall diabetes mellitus, type 1 diabetes mellitus, and type 2 diabetes mellitus are presented differently. The overall diabetes mellitus model is defined as fasting plasma glucose (FPG) > 126 mg/dL (7 mmol/L) or being on treatment for diabetes. The overall type 1 diabetes mellitus model is defined as cases of DM that are on insulin or diagnosed with a biomarker (eg, c-peptide levels) that is not fasting plasma glucose. Type 2 diabetes mellitus cases are those that are not reported as type 1 diabetes mellitus. [GBD-2017-YLD-Capstone-Appendix-1-Diabetes-Mellitus]
Fasting Plasma Glucose (FPG) in GBD 2017¶
GBD 2017 predicted mean FPG from diabetes prevalence using an ensemble distribution. GBD 2017 also used the ensemble distribution methodology to estimate the prevalence of diabetes based on mean FPG in locations where data on prevalence of diabetes were not available.
Other measures of blood sugar to estimate diabetes in GBD 2017¶
GBD 2017 incorporated all available data related to population-representative estimates of diabetes, so other measures of blood sugar (glycated hemoglobin A1c, oral glucose tolerance test, post-prandial glucose test) were used to define diabetes and mean fasting plasma glucose (FPG) in a population when data on diabetes were not available as data inputs.
Cause Hierarchy¶
Restrictions¶
The following table describes any restrictions in GBD 2017 on the effects of this cause (such as being only fatal or only nonfatal), as well as restrictions on the ages and sexes to which the cause applies. If sub cause restrictions vary, then the conflicting restrictions are noted below.
Restriction Type |
Value |
Notes |
|---|---|---|
Male only |
False |
|
Female only |
False |
|
YLL only |
False |
|
YLD only |
False |
|
YLL age group start (DM Type 1) |
Early Neonatal |
(0, 6 days], age_group_id = 2 |
YLL age group start (DM Type 2) |
15 to 19 |
(15, 19], age_group_id = 8 |
YLL age group end |
95 plus |
(95, 125], age_group_id = 235 |
YLD age group start |
Early Neonatal |
(0, 6 days], age_group_id = 2 |
YLD age group end |
95 Plus |
(95, 125], age_group_id = 235 |
Vivarium Modeling Strategy¶
Scope¶
The aspects of the disease this cause model is designed to simulate is the basic structure of the disease, its sub causes, associated measures (deaths, prevalence, incidence, emr), associated sequelae, and associated disability weights. This cause model is designed differently, where simulants are selected and categorized as either ‘moderate’ or ‘severe’ diabetes state. Across the 2 diabetes sub causes, some of the associated sequelae will either be grouped into the ‘Moderate’ or ‘Severe’ diabetes state. The ‘uncomplicated’ sequelae for diabetes mellitus type 1 & type 2 are included in the ‘Moderate’ diabetes state, which are designated as non-fatal only and include only YLDs. The sequelae which map to ‘Severe’ diabetes state include all other sequelae. These sequelae are fatal and include YLLs and YLDs. The associated sequelae in each state can be found below in the ‘State Severity Split Definitions’ table.
Vivarium Modeling Strategy for Risk Factor FPG¶
This cause model is designed to simulate the basic structure of the risk factor (FPG) continuous exposure ensemble distribution model. The FPG distribution will range, starting at the theoretical minimum-risk exposure level (TMREL) of 4.5-5.4 mmol/L to the maximum FPG value for each location/sex/age group. For simulants that are in ‘Susceptible’ state in the vivarium model, the model will randomly draw a value of FPG that is equal to or less than the TMREL and less than 7.0 mmol/L (case definition for ‘With Condition’ of overall diabetes mellitus). For simulants that are included in the ‘Moderate’ or ‘Severe’ diabetes states, the model will randomly draw a value of FPG that is equal to or greater than 7.0 mmol/L.
Assumptions and Limitations¶
Assumptions¶
In vivarium, ‘uncomplicated DM Type 1 and Type 2’ is ‘Moderate’, which is different from how GBD 2017 is modelling it. In the future, severity splits will be revisited using disability weights.
This model has durations for moderate and severe DM that are too long because they arrive to that disease state immediately. Simulants in this model will not progress from moderate to severe.
This model assumes that EMR from ‘moderate’ and ‘severe’ disease states are equal.
This model assumes that Remission from ‘moderate’ to ‘susceptible’ and ‘severe’ to ‘susceptible’ are the same value.
Case definition cross-walks on FPG and HbA1c: GBD 2017 assumed that HbA1c >6.5% was equivalent to FPG >126 mg/dL.
The project requires that prevalence for each severity (mild/moderate v. severe) be correct, since this is a condition for treatment of LDL-C. DM is not a cause of interest for the current project beyond its impact on treatment. We will use prevalence weighted incidence to add new patients to both “moderate” and “severe” states as the simulation runs, and simulants may exit based on the CSMR (detailed below).
There is incidence into both “moderate” and “severe” (prevalence weighted incidence (parent cause incidence)) and remission (equal for both “moderate” and “severe”). There is no state-to-state transition for the reasons cited above.
Limitations¶
These limitations will impact the DM model in a couple ways:
The prevalence of each severity should be appropriate, and will be validated against GBD.
The rate of mortality from “moderate” will be high, since we are assuming EMR (calculated from CSMR) is the same for both “moderate” and “severe”. It is unclear how this will impact prevalence, since patients in “moderate” will be dying faster (which implies too few simulants in “moderate”), but not transitioning into “severe” (which implies too many simulants in “moderate”).
This may also impact DM related burden, since progression into more severe states of DM is not possible. In aggregate, if prevalence is right, the morbidity should be close to GBD estimates, but at the simulant level, we will underestimate burden.
Again, the impact of the assumptions stated regarding remission is unclear - simulants exiting from “severe” back to “susceptible” is inconsistent with GBD and should drive down burden, but the absence of progression from “moderate” to “severe” mitigates this. It is not clear yet which will have a greater impact. (Recall that our primary concern is getting prevalence correct for the current project’s treatment algorithm.)
Cause Model Diagram¶
Data Description¶
State and Transition Data Tables¶
State |
State Name |
Definition |
|---|---|---|
S |
Susceptible |
Susceptible to Diabetes Mellitus |
M |
Moderate |
Simulant is with condition of Uncomplicated Diabetes Mellitus, based on ‘uncomplicated’ sequelae of Diabetes Mellitus Type 1 and Type 2 |
Sev |
Severe |
Simulant is with condition of Severe Diabetes Mellitus, based on all other sequelae of Diabetes Mellitus Type 1 and Type 2 |
State |
State Name |
Definition |
|---|---|---|
S |
Susceptible |
Susceptible to Diabetes Mellitus |
M |
Moderate |
sequelae_mod = [s_5441, s_5465] |
Sev |
Severe |
sequelae_sev = [s_5429, s_5432, s_s5435, s_5438, s_5444, s_5447, s_5450, s_5453, s_5456, s_5459, s_5462, s_5468, s_5471, s_5474] |
State |
Measure |
Value |
Notes |
|---|---|---|---|
S |
simulants not prevalent with overall Diabetes Mellitus |
1 - prevalence_c587 |
|
M |
prevalence |
\({\sum_{s\in \text{sequelae_mod}}} \scriptstyle{\text{prevalence}_s}\) |
= (prevalence of Diabetes Mellitus Type 1 uncomplicated sequelae + prevalence of Diabetes Mellitus Type 2 uncomplicated sequelae |
Sev |
prevalence |
\({\sum_{s\in \text{sequelae_sev}}} \scriptstyle{\text{prevalence}_s}\) |
= (prevalence of Diabetes Mellitus Type 1 all other sequelae (not including uncomplicated) + prevalence of Diabetes Mellitus Type 2 all other sequelae (not including uncomplicated) |
Sev |
excess mortality rate (EMR) for severe DM |
\(\frac{\text{CSMR*_c587}}{\text{prevalence_c587}}\) |
cause-specific mortality rate of DM (*indicates calculated measure) / prevalence of DM |
M |
excess mortality rate (EMR) of moderate DM |
\(\frac{\text{CSMR*_c587}}{\text{prevalence_c587}}\) |
cause-specific mortality rate of DM (*indicates calculated measure) / prevalence of DM |
M |
disability_weight |
\(\frac{{\sum_{s\in \text{sequelae_mod}}} \scriptstyle{\text{disability_weight}_s \times\ \text{prevalence}_s}}{\text{prevalence_c587}}\) |
|
Sev |
disability_weight |
\(\frac{{\sum_{s\in \text{sequelae_sev}}} \scriptstyle{\text{disability_weight}_s \times\ \text{prevalence}_s}}{\text{prevalence_c587}}\) |
|
All |
cause-specific mortality rate (csmr) |
\(\frac{\text{deaths_c587}}{\text{population}}\) |
calculated, not a direct GBD 2017 data input |
Transition |
Source State |
Sink State |
Value |
Notes |
|---|---|---|---|---|
1 |
S |
M |
\(\frac{\sum_{s\in \text{prevalence_sequelae_mod.sub_causes.c587}}}{\text{prevalence_c587}} \times\ {\text{incidence_c587}}\) |
= weighted prevalence of moderate DM * incidence of DM |
2 |
S |
Sev |
\(\frac{\sum_{s\in \text{prevalence_sequelae_sev.sub_causes.c587}}}{\text{prevalence_c587}} \times\ {\text{incidence_c587}}\) |
= weighted prevalence of severe DM * incidence of DM |
3 |
M |
S |
remission_modelable_entity_id_2005 |
= remission from moderate DM to Susceptible |
4 |
Sev |
S |
remission_modelable_entity_id_2005 |
= remission from severe DM to Susceptible |
Variable |
Source |
Description |
Notes |
|---|---|---|---|
prevalence_c587 |
como |
prevalence of overall Diabetes Mellitus |
|
deaths_c587 |
codcorrect |
Count of deaths due to overall Diabetes Mellitus |
|
population |
demography |
Mid-year population for given sex/age/year/location |
|
prevalence_s{sid} |
como |
Prevalence of sequela with id {id} |
|
disability_weight_s{sid} |
YLD appendix |
Disability weight of sequela with id {id} |
|
remission_modelable_entity_id_2005 |
epi |
remission of overall Diabetes Mellitus from epi database |
|
incidence_c587 |
como |
incidence of overall Diabetes Mellitus |
Validation Criteria¶
Model Validation¶
Check the logical structure and input data for concept model, make sure that
the theories and assumptions underlying the conceptual model are correct
the data to build, evaluate, and test the model are correct
Logic¶
Parent cause is the sum of child causes and/or the sum of sequela
Fatal: Deaths (CSMR, Excess MR), YLLs
Non-fatal: YLDs, Prevalence, Incidence
DALYS = YLLs + YLDs
By location-/age-/sex-
Prevalence will be validate against GBD, as will morbidity and mortality. Given the assumptions described above, we prioritize validation of prevalence.
References¶
- WHO-Diabetes-Definition
Retrieved 30 Jan 2020. https://www.who.int/health-topics/diabetes
- GBD-2017-YLD-Capstone-Appendix-1-Diabetes-Mellitus
Supplement to: GBD 2017 Disease and Injury Incidence and Prevalence Collaborators. Global, regional, and national incidence, prevalence, and years lived with disability for 354 diseases and injuries for 195 countries and territories, 1990–2017: a systematic analysis for the Global Burden of Disease Study 2017. Lancet 2018; 392: 1789–858 (pp. 559-572)