an eight-step method for assessing diagnostic data quality in practice: chronic obstructive pulmonary disease as an exemplar

Abstract

Background Chronic obstructive pulmonary disease (COPD) is an important cause of mortality and morbidity. Its management is shifting from the secondary to the primary care setting. The quality of data is known to vary between practices, and individual practices need to be able to assess their data quality. Objectives To measure the quality of diagnostic data in COPD. Subjects 10 975 patients registered with a computerized general practice in the south of England, and 190 patients likely to have COPD. Methods An eight-step method was developed: (1) research the expected prevalence of the diagnosis and define audit criteria; (2) find out how the diagnosis might be coded – look at the terminology and the codes presented by the computer interface; (3) examine the characteristics of the practice population; (4) calculate the prevalence and infer its reliability; (5) investigate the completeness; (6) accuracy; (7) currency and consistency; and (8) calculate sensitivity and positive predictive value of the data. Results The prevalence of COPD in the literature ranges between 3% and 10%. The coding for bronchitis and COPD is complex and it is easy to select an incorrect code. The test population is younger but of similar social class to the national average. The prevalence of COPD in this study was 1.3%. The data were incomplete and some were inaccurate; patients with COPD had to be identified from additional searches. The sensitivity of the use of the diagnostic code was 79%, and the positive predictive value 75.3%. Conclusions The method provides a tool to help practices and localities assess their diagnostic data quality.