Multivariate analysis of outcome of mental health care using graphical chain models The South-Verona Outcome Project 1

Abstract

Background. Short-term outcome of mental health care was assessed in a multidimensional perspective using graphical chain models, a new multivariate method that analyses the relationship between variables conditionally, i.e. taking into account the effect of antecedent and intervening variables.

Methods. GAF, BPRS, DAS (at baseline and after 6 months), LQL and VSSS (at follow-up only) were administered to 194 patients attending the South-Verona community-based mental health service. Direct costs in the interval were also calculated. Graphical chain models were used to analyse: (1) the associations between predictors (psychopathology, disability, functioning, assessed at baseline); (2) the effects of predictors on costs; and (3) the effect of predictors and costs on outcomes (psychopathology, disability, functioning, quality of life and service satisfaction) as well as their correlation.

Results. Psychopathology, disability and functioning scores at baseline predicted the corresponding scores at 6-month follow-up, with greater improvement in the more severely ill. Higher psychopathology and poorer functioning at baseline predicted higher costs and, in turn, costs predicted poorer functioning at follow-up. Outcome indicators polarized in two groups: psychopathology, disability and functioning, which were highly correlated; and the dyad service satisfaction and quality of life. Service satisfaction was highly related to quality of life and was predicted by low disability and high dysfunctioning. No predictors for quality of life were found.

Conclusions. Graphical chain models were demonstrated to be a useful methodology to analyse process and outcome data. The results of the present study help in formulating specific hypotheses for future studies on outcome.