Application of a Latin square experimental design in health services research : estimation of the effects of setting clinical standards and performance review on the process and outcome of care in general practice

Abstract

The North of England Study of Standards and Performance in General Practice was set up to investigate whether the setting of clinical standards was an effective way of improving clinical performance (North of England Study, 1991). Doctors from 60 training practices met in small groups to set standards of good clinical performance for five symptomatic conditions of childhood-acute cough; acute vomiting; itchy rash; bedwetting; and recurrent wheezy chest. Data on the process and outcome of care were collected both before and after standard setting process. Some of the baseline data was fed back to the doctors to enable them to evaluate their performance in the first phase of the study. The clinical standards and baseline data were disseminated to the small groups of doctors according to a Latin square design. By comparing responses obtained during the second phase of data collection (after standards had been set) with those obtained in the first, it was possible to estimate the effects of standard setting and other methods of performance review on what doctors did (the process of care) and on the resulting outcome of care for their patients. The general analytic approach adopted was to fit generalised linear models to try and explain the variation in the observed data. Within this general framework, methods were developed for coping with a wide range of statistical problems including: heteroscedasticity correlated binary responses loss of orthogonality arising because of the incompleteness of many of the data sets; and overdispersion. Abstract The setting of clinical standards was found to have influenced doctors' prescribing of drugs and was found to have had a beneficial influence on outcome of care for children suffering from recurrent wheezy chest. Implications for the design of future studies to evaluate this type of intervention in the health service are discussed.