Drug attitude and other predictors of medication adherence in schizophrenia: 12 months of electronic monitoring (MEMS(®)) in the Swedish COAST-study.

The aim was to investigate clinical predictors of adherence to antipsychotics. Medication use was electronically monitored with a Medication Event Monitoring System (MEMS(®)) for 12 months in 112 outpatients with schizophrenia and schizophrenia-like psychosis according to DSM-IV. Symptom burden, insight, psychosocial function (PSP) and side effects were rated at baseline. A comprehensive neuropsychological test battery was administered and a global composite score was calculated. The Drug Attitude Inventory (DAI-10) was filled in. A slightly modified DAI-10 version for informants was distributed as a postal questionnaire. Non-adherence (MEMS(®) adherence ≤0.80) was observed in 27%. In univariate regression models low scores on DAI-10 and DAI-10 informant, higher positive symptom burden, poor function, psychiatric side effects and lack of insight predicted non-adherence. No association was observed with global cognitive function. In multivariate regression models, low patient-rated DAI-10 and PSP scores emerged as predictors of non-adherence. A ROC analysis showed that DAI-10 had a moderate ability to correctly identify non-adherent patients (AUC=0.73, p<0.001). At the most "optimal" cut-off of 4, one-third of the adherent would falsely be identified as non-adherent. A somewhat larger AUC (0.78, p<0.001) was observed when the ROC procedure was applied to the final regression model including DAI-10 and PSP. For the subgroup with informant data, the AUC for the DAI-10 informant version was 0.68 (p=0.021). Non-adherence cannot be properly predicted in the clinical setting on the basis of these instruments alone. The DAI-10 informant questionnaire needs further testing.