Accuracy of clinical coding of pleural empyema: A validation study

Routinely recorded coding data are increasingly being used for clinical research, but the quality of that data is often variable. The clinical coding of pleural empyema has not been studied. If data are not accurately recorded, linking coding data through, for example, Hospital Episode Statistic data (HES), will lead to inappropriate conclusions.

We extracted the coding data for a year from three hospitals in the Southwest of England, UK: One large district general hospital, one tertiary pleural referral centre, and one thoracic surgical centre. We verified and validated this coding to ensure its accuracy. To assess the sensitivity of coding, we then verified the coded cases against a cohort of proven empyema patients from a recent observational study.

One hundred eighty-two patients with a coded diagnosis of pleural empyema were extracted. In 158 of these patients, the diagnosis was correct clinically, with a positive predictive value of 86%. In the cohort of 25 patients with a known clinical diagnosis of empyema, 18 were coded as empyema, a sensitivity of 72%. There was no significant difference in coding between hospitals.

In the centres studied, empyema coding was generally good (>85% accuracy), although some cases were miscoded. Using codes for elective admissions helped to differentiate post-operative from community acquired empyema. It is likely empyema is under rather than over-recorded in HES data.