Adverse events recorded in English primary care: observational study using the General Practice Research Database

Background

More accurate and recent estimates of adverse events in primary care are necessary to assign resources for improvement of patient safety, while predictors must be identified to ameliorate patient risk.

Aim

To determine the incidence of recorded iatrogenic harm in general practice and identify risk factors for these adverse events.

Design and setting

Cross-sectional sample of 74 763 patients at 457 English general practices between 1 January 1999 and 31 December 2008, obtained from the General Practice Research Database.

Method

Patient age at study entry, sex, ethnicity, deprivation, practice region, duration registered at practice, continuity of care, comorbidities, and health service use were extracted from the data. Adverse events were defined by Read Codes for complications of care (Chapters S, T, and U). Crude and adjusted analyses were performed by Poisson regression, using generalised estimating equations.

Results

The incidence was 6.0 adverse events per 1000 person-years (95% confidence interval CI] = 5.74 to 6.27), equivalent to eight adverse events per 10 000 consultations (n = 2 540 877). After adjustment, patients aged 65–84 years (risk ratio RR] = 5.62, 95% CI = 4.58 to 6.91; P<0.001), with the most consultations (RR = 2.14, 95% CI = 1.60 to 2.86; P<0.001), five or more emergency admissions (RR = 2.08, 95% CI = 1.66 to 2.60; P<0.001), or the most diseases according to expanded diagnosis clusters (RR = 8.46, 95% CI = 5.68 to 12.6; P<0.001) were at greater risk of adverse events. Patients registered at their practice for the longest periods of time were less at risk of an adverse event (RR = 0.40, 95% CI = 0.35 to 0.47; P<0.001).

Conclusion

The low incidence of recorded adverse events is comparable with other studies. Temporal sequencing of risk factors and case ascertainment would benefit from data triangulation. Future studies may explore whether first adverse events predict future incidents.