Comparison of indicators assessing the quality of drug prescribing for asthma

Objective. To compare different indicators for assessing the quality of dru g prescribing and establish their agreement in identifying doctors who may not adhere to treatment guidelines. Data Sources/Study Setting. Data from 181 general practitioners (GPs) from The Netherlands. The case of asthma is used as an example because, in this area, different quality indicators exist whose validity is questioned. The study is part of the European Drug Education Project. Study Design. Spearman rank correlations were assessed among the GPs' score s on self-report instruments, aggregated prescribing indicators, and indivi dualized prescribing indicators. Kappa values were calculated as agreement measures for identifying low adherence to the guidelines. Data Collection. Prescribing data from GPs were collected through pharmacie s, public health insurance companies, or computerized GP databases. Two sel f-report instruments were mailed to the GPs. The GPs first received a quest ionnaire assessing their competence regarding the treatment of asthma patie nts. Three months later they received a series of 16 written asthma cases a sking for their intended treatment for each case. Principal Findings. Correlations between scores based on self-report instru ments and indicators based on actual prescribing data were mostly nonsignif icant and varied between 0 and 0.21. GPs identified as not adhering to the guidelines by the prescribing indicators often had high scores on the self- report instruments. Correlations between 0.20 and 0.55 were observed among indicators based on aggregated prescribing data and those based on individu alized data. The agreement for identifying low adherence was small, with ka ppa values ranging from 0.19 to 0.30. Conclusions. Indicators based on self-report instruments seem to overestima te guideline adherence. Indicators assessing prescribing quality at an aggr egated level give clearly different results, as compared to indicators eval uating prescribing data on an individual patient level. Caution is needed w hen using such prescribing indicators to identify low adherence to guidelin es. Further validation studies using a gold standard comparison are needed to define the best possible indicator.