Factor analysis improves the selection of prescribing indicators

Abstract

To test a method for improving the selection of indicators of general practitioners' prescribing. We conducted a prescription database study including all 180 general practices in the County of Funen, Denmark, approximately 472,000 inhabitants. Principal factor analysis was used to model correlation between 19 register-based indicators for the quality of non-steroidal anti-inflammatory drug (NSAID) prescribing. The correlation between indicators ranged widely from 0 to 0.93. Factor analysis revealed three dimensions of quality: (1) "Coxib preference", comprising indicators directly quantifying choice of coxibs, indicators measuring expenditure per Defined Daily Dose, and indicators taking risk aspects into account, (2) "Frequent NSAID prescribing", comprising indicators quantifying prevalence or amount of NSAID prescribing, and (3) "Diverse NSAID choice", comprising indicators focusing on the width of GPs' formularies. The number of indicators for measuring the important aspects of quality in prescribing of NSAIDs could be reduced substantially by selecting the indicator in each dimension with the highest factor loading. A high preference for coxibs indicated both appropriate and inappropriate prescribing, as revealed by the correlation of the indicators in the first factor. Correlation and factor analysis is a feasible method that assists the selection of indicators and gives better insight into prescribing patterns.