A Bayesian multivariate approach to estimating the prevalence of a superordinate category of disorders

A Bayesian multivariate meta-analysis of prevalence data

When conducting a meta-analysis involving prevalence data for an outcome with several subtypes, each of them is typically analyzed separately using a univariate meta-analysis model. Recently, multivariate meta-analysis models have been shown to correspond to a decrease in bias and variance for multiple correlated outcomes compared with univariate meta-analysis, when some studies only report a subset of the outcomes. In this article, we propose a novel Bayesian multivariate random effects model to account for the natural constraint that the prevalence of any given subtype cannot be larger than that of the overall prevalence. Extensive simulation studies show that this new model can reduce bias and variance when estimating subtype prevalences in the presence of missing data, compared with standard univariate and multivariate random effects models. The data from a rapid review on occupation and lower urinary tract symptoms by the Prevention of Lower Urinary Tract Symptoms Research Consortium are analyzed as a case study to estimate the prevalence of urinary incontinence and several incontinence subtypes among women in suspected high risk work environments.

The Prevalence of Anxiety Disorders During Pregnancy and the Postpartum Period: A Multivariate Bayesian Meta-Analysis

OBJECTIVE

To estimate the prevalence of anxiety disorders in pregnant and postpartum women and identify predictors accounting for variability across estimates.

DATA SOURCES

An electronic search of PsycINFO and PubMed was conducted from inception until July 2016, without date or language restrictions, and supplemented by articles referenced in the obtained sources. A Boolean search phrase utilized a combination of keywords related to pregnancy, postpartum, prevalence, and specific anxiety disorders.

STUDY SELECTION

Articles reporting the prevalence of 1 or more of 8 common anxiety disorders in pregnant or postpartum women were included. A total of 2,613 records were retrieved, with 26 studies ultimately included.

DATA EXTRACTION

Anxiety disorder prevalence and potential predictor variables (eg, parity) were extracted from each study. A Bayesian multivariate modeling approach estimated the prevalence and between-study heterogeneity of each disorder and the prevalence of having 1 or more anxiety disorder.

RESULTS

Individual disorder prevalence estimates ranged from 1.1% for posttraumatic stress disorder to 4.8% for specific phobia, with the prevalence of having at least 1 or more anxiety disorder estimated to be 20.7% (95% highest density interval [16.7% to 25.4%]). Substantial between-study heterogeneity was observed, suggesting that "true" prevalence varies broadly across samples. There was evidence of a small (3.1%) tendency for pregnant women to be more susceptible to anxiety disorders than postpartum women.

CONCLUSIONS

Peripartum anxiety disorders are more prevalent than previously thought, with 1 in 5 women in a typical sample meeting diagnostic criteria for at least 1 disorder. These findings highlight the need for anxiety screening, education, and referral in obstetrics and gynecology settings.

A Bayesian multivariate approach to estimating the prevalence of a superordinate category of disorders

OBJECTIVE

Epidemiological research plays an important role in public health, facilitated by the meta-analytic aggregation of epidemiological trials into a single, more powerful estimate. This form of aggregation is complicated when estimating the prevalence of a superordinate category of disorders (e.g., "any anxiety disorder," "any cardiac disorder") because epidemiological studies rarely include all of the disorders selected to define the superordinate category. In this paper, we suggest that estimating the prevalence of a superordinate category based on studies with differing operationalization of that category (in the form of different disorders measured) is both common and ill-advised. Our objective is to provide a better approach.

METHODS

We propose a multivariate method using individual disorder prevalences to produce a fully Bayesian estimate of the probability of having one or more of those disorders. We validate this approach using a recent case study and parameter recovery simulations.

RESULTS

Our approach produced less biased and more reliable estimates than other common approaches, which were at times highly biased.

CONCLUSION

Although our approach entails additional effort (e.g., contacting authors for individual participant data), the improved accuracy of the prevalence estimates obtained is significant and therefore recommended.