Uncertainty Analysis in Population-Based Disease Microsimulation Models

Table 1

Components and parameters of a simplified version of POHEM-OA^∗ that was used for the uncertainty analysis example.

Demographic components
Age distribution in 2001^**
Sex distribution in 2001^**
Province of residence distribution in 2001^**
Birth and mortality rates by age and sex over time^#
Migration^##
Disease components
(i) Baseline parameters
BMI^¥ distribution in 2001^**
Incidence of OA in 2001 by sex and 5-year age groups^¥¥
Reference (“baseline”) hazard rates of OA^§
Prevalence of OA in 2001 by sex and 5-year age groups^§§
(ii) Risk factors model
Change in BMI over time^§*
(iii) Disease incidence
Effect of BMI on incident OA by sex^§#

^*POHEM-OA: Population Health Microsimulation model for osteoarthritis (OA) [46]. ^**Observed in Canadian Community Health Survey (CCHS-2001) [50]. ^#Projected mortality in Canada by age and sex from Statistic Canada. ^##Migration data obtained from Statistics Canada projections. ^¥BMI: weight/height². ^¥¥From administrative data in BC [52]; OA is defined as at least 2 visits to a health professional within 2 years or 1 hospitalization with the ICD-9 code 715. Incident cases in 2001 are identified after excluding prevalent cases prior to 2001 using a 10 year run-in period [46]. ^§Obtained numerically using calibration to match the marginal distribution of incidence in BC administrative data. ^§§Obtained as the final stable prevalence from a simulation of the Canadian population over a 50-year horizon, under constant age-specific incidence rates. ^§*Obtained from a linear regression model including age, sex, province, education, and prior BMI. ^§#BMI was categorized into 4 standard categories (see Table 3). The effect (hazard ratio) for each level of BMI is obtained from a survival regression model using longitudinal NPHS data (two cycles: 2000 and 2002) [49], separately for men and women.