Investigating Alternative Ways of Estimating the Prevalence of Serious Mental Illness Using the National Survey on Drug Use and Health
Kott, P. S., Liao, D., & Aldworth, J. (2011, October). Investigating Alternative Ways of Estimating the Prevalence of Serious Mental Illness Using the National Survey on Drug Use and Health. Presented at APHA 2011, Washington, DC.
The National Survey on Drug Use and Health (NSDUH) uses a two-phase process to estimate the proportion of a population with serious mental illness (SMI). The first phase is the NSDUH itself, a large self-administered national survey containing a series of questions on personal drug use and mental health. A randomly chosen subsample of the annual NSDUH, the Mental Health Surveillance Survey (MHSS) is drawn, and respondents are clinically evaluated for SMI. A prediction model is fitted in this MHSS subsample with the clinical evaluations treated as the "gold standard" and then applied to the entire NSDUH sample. Currently, an unadjusted (model-based) cut-point estimator is computed by assigning an SMI status to everyone in the NSDUH based on a fitted logistic model with two covariates, one based on K6 psychological-distress scores and the other based on WHODAS functional-impairment scores.
We investigated several potential alternatives to the unadjusted cut-point estimator above based on the same logistic-model fit with 2008 and 2009 NSDUH/MHSS data. These include a bias-adjusted (model-free) cut-point estimator, an unadjusted (model-based) probability estimator, which assigns a probability of being SMI to everyone in the NSDUH, and a bias-adjusted (model-free) probability estimator.
We measured the standard errors of the competing estimators, including the error from estimating logistic-model parameters, using linearization and Fay's version of balanced repeated replication (Fay's BRR). Linerization was particularly effective in assessing the potential for bias in the unadjusted estimators.