Simulation modeling of neighborhood-level differences in life course health outcomes: a proof of concept

Adults in low-resource neighborhoods face greater burden of preventable diseases than the general population. Methodologies to measure the impact of population-level efforts to reduce neighborhood-level disparities in health outcomes are lacking. We introduce a simulation modeling approach for incident disease and mortality risk over the life course and apply the model to assess the impact of a hypothetical population health initiative to reduce neighborhood-level disparities.

Using electronic health records, we constructed a probabilistic dynamic systems model that simulated the life course of each patient in a large primary care population, taking into account specific relationships by sex and Area Deprivation Index (ADI) quintile. The model predicted long-term incidence of 10 chronic conditions and all-cause mortality.

The model was reliable, with strong discrimination (C-statistic for 10-year mortality: 0.871) and calibration. Predicted median life expectancy was nearly a decade lower for patients ages 40-45 who resided in the highest (ADI quintile 5) vs the lowest (quintile 1) deprivation neighborhoods (difference [95% CI]; women, 8.6 [8.2, 9.0] years; men, 10.0 [9.6, 10.3]. A hypothetical initiative to reduce the number of smokers by 10% had minimal effect on disparities in life expectancy when implemented in the general patient population (women, -0.4 years; men, -0.2), but meaningfully narrowed disparities when focused on high-deprivation neighborhoods (women, -0.8 years; men, -0.6).

In a broad patient population, it was feasible to measure socioeconomic disparities in health outcomes and life expectancy and evaluate interventions to improve health equity.