Markov models of major depression for linking psychiatric epidemiology to clinical practice

Abstract

Background

Most epidemiological studies of major depression report estimates of period prevalence. Such estimates are useful for public health applications, but are not very helpful for informing clinical practice. Period prevalence is determined predominantly by incidence and episode duration, but it is difficult to connect these epidemiological concepts to clinical issues such as risk and prognosis. Incidence is important for primary and secondary prevention, and prognostic information is useful for clinical decision-making. The objective of this study was to decompose period prevalence data for major depression into its constituent elements, thereby enhancing the value of these estimates for clinical practice. Data from a series of population-based Canadian studies were used in the analysis. Markov models depicting incidence, prevalence and recovery from major depressive episodes were developed. Monte Carlo simulation was used to constrain model parameters to the epidemiological data.

Results

The association of sex with major depression was found to be due to a higher incidence in women. In distinction, the higher prevalence in unmarried subjects was mostly due to a different prognosis. Age-related changes in prevalence were influenced by both factors. Education, which was not found to be associated with major depression in the survey data, had no impact either on risk or prognosis.

Conclusion

The period prevalence of major depression is influenced both by incidence (risk) and episode duration (prognosis). Mathematical modeling of the underlying epidemiological relationships can make such data more readily interpretable in relation to clinical practice.