Dynamic Modelling of Health and its application to the large scale analysis of Body Mass Index, using data from consecutive set of surveys

TL;DR

This paper introduces a causality-based dynamic model for analyzing population health changes over time, applied to BMI data from multiple surveys to assess health trends and the impact of health interventions.

Contribution

It presents a novel dynamic modeling approach that incorporates causality directly, enabling analysis of health trends using cross-sectional survey data with an iterative regression algorithm.

Findings

Analyzed BMI trends in men from 1972-2002 using 7 surveys.

Provided unbiased insights into health promotion effectiveness.

Demonstrated model's ability to handle surveys with different age ranges.

Abstract

The methods used so far for the analysis of time changes in population health suffer from the lack of causality in their design. This results in problems with their implementation and interpretation. Here the method is presented with causality directly implemented in the design. This is done by, first, building up a dynamic model of population, postulating existence of Driving Force acting at subjects, while they move along their cohort lines, causing the changes of their substantial health indicators , State Variables, at rate proportional to this Force. The correspondent rates , named Cohort Trends, or C-trends, describe health history in each birth cohort. Having initial value and C-trends , the model allows to calculate health level (the means of State Variables) in each birth cohort, and thus, in the whole population. The task for statistical method is to identify the dynamic model (evaluate C-trends and Initial values) using data from a set of consecutive independent cross-sectional surveys. This is done by an iterative algorithm, running multiple regression procedure at each step, until the specified smoothing conditions are fulfilled. The algorithm can operate with surveys having different age ranges. The illustrative example shows the results of analysis of Body Mass Index for men , using 7 surveys in period 1972-2002 with age ranges 25-64 and 25-74. Since C-Trend is proxy for Driving Force, the year - age pattern of C-Trends provides unbiased information for health authorities on efficiency of health promotion actions or negative effects of uncontrolled harmful factors.