Epidemiological theory, decision theory and mental health services research

Abstract

Background: Mathematical models describing the epidemiology of major depression are potentially useful for epidemiological analyses, as decision support tools and in economic analyses. The objective of this project was to develop a Markov model based on epidemiological theory that may be useful for decision analysis and health services research.

Methods: Longitudinal data from a Canadian national survey, the National Population Health Survey (NPHS), were used. The NPHS has collected longitudinal data on a cohort of 17,262 subjects since 1994. The analysis employed a Markov tunnel in order to model the dependence of recovery probabilities on episode duration.

Results: Episode incidence ranged between 6.2 % per year in women under 35 to 0.26 % in men over the age of 65.A greater proportion of subjects over 35 years old reported episodes lasting more than 26 weeks. The probability of recovery declined with increasing episode duration, independently of sex. Under steady-state assumptions, a Markov model integrating these parameters predicted a point prevalence of approximately 2% in women and 1% in men under the age of 55. In older age groups, the predicted point prevalence declined in both sexes.

Conclusions: These models support the hypothesis that sex differences in major depression prevalence are due primarily to differences in incidence rather than episode length. These results also indicate that there is no meaningful "central tendency" describing the distribution of episode length in major depression episode. Estimates of mean episode duration represent an intermixing of frequent brief episodes with infrequent protracted episodes. This finding may have important policy implications.