Background Exacerbation of COPD is the second commonest reason for hospital admission, with high subsequent readmission and mortality rates among those who survive to discharge. Risk stratification would inform efficient use of services.
Methods Consecutive patients admitted with an exacerbation of COPD who survived to discharge were recruited by screening admissions units and searching coding records. Six UK hospitals took part: the derivation and internal validation cohort involved the same two hospitals at different time periods, and the external validation involved four hospitals.
Clinical data, and 90-day death and readmission rates were recorded. Multivariate logistic regression analysis was used to develop a tool to predict 90-day readmission, or death without readmission. Performance was assessed by the area under the receiver operator characteristic (AUROC) curve.
Results 2,417 patients were analysed (derivation 824, internal validation 824, external validation 791). Female 54.0%, mean (SD) age 72.6 (10.2) years, FEV1 45.3 (18.2) %predicted, 90-day readmission or death 38.7%.
In the derivation cohort, the five strongest predictors (odds ratio, 95% confidence interval given for whole population) were: two or more Previous admissions in the preceding year (OR 3.17, 95% CI 2.59–3.87), stable-state dyspnoea assessed by the Extended MRCD score (eMRCD 4 OR 1.46, 95% CI 1.12–1.90; eMRCD 5a OR 2.35 95% CI 1.79–3.08; eMRCD 5b OR 3.00 95% CI 2.19–4.11), Age 80 or more (OR 1.48, 95% CI 1.22–1.81), cor-pulmonale “Right heart failure” (OR 1.93, 95% CI 1.41–2.66), Left heart failure (OR 1.45, 95% CI 1.07–1.97). Two or more previous admissions and eMRCD 5b were assigned a score of 3, eMRCD 5a scored 2, while eMRCD 4 and remaining indices scored 1. The risk of readmission and/or death is shown in Table 1.
The AUROC was: derivation 0.73 (95% CI 0.69–0.77); internal validation 0.68 (95% CI 0.64–0.72); and external validation 0.70 (95% CI 0.66–0.73).
Discussion In patients hospitalised with an exacerbation of COPD the PEARL score is a robust predictor of readmission and death and may be used to inform efficient use of resources according to risk.