Diagnostic accuracy in the Swedish national patient register: a review including diagnoses in the outpatient register

Background

The Swedish National Patient Register (NPR) is an important source of data for epidemiological research. A review in 2010 described the validity of recorded diagnoses for inpatient care, but did not include specialised outpatient care.

Method

Using systematic searches of medical literature databases (Embase, Medline), and reports from members of the Swedish Epidemiological Association, we aimed to identify all studies validating diagnoses and procedure codes in inpatient care since 2010 and all studies validating specialised outpatient care. In addition, we summarize findings from register validation work performed by the National Board of Health and Welfare.

Results

The literature search and personal reports generated 3990 non-duplicate original studies, of which 89 were deemed relevant. Compared to data in patient charts (reference), the median positive predictive value (PPV) for diagnostic codes in the NPR was 84% (interquartile range 72–93%), but with clear differences between types of diagnoses. The median PPV for surgical procedures was 97% (86–99%). The median sensitivity of diagnoses and procedures compared to other registers and cohorts was 73% (45–80%). The completeness of the register has improved over time. Missingness originates mainly from underreporting of procedures performed by private healthcare providers, and for certain variables, e.g. medication codes.

Conclusion

The NPR has good diagnostic accuracy for most diagnoses and very good for surgical procedures. The sensitivity is lower. Longitudinal comparisons of incidence or prevalence are affected by changes in completeness. Missingness is low, although it is higher among private healthcare providers and for specific variables such as drug administration.