Spencer J Keene 1 , 2 , 3 , Rachel E Jordan 1 , Frits ME Franssen 3 , 4 , Frank de Vries 2 , 5 , James Martin 1 , Alice Sitch 1 , 6 , Alice Margaret Turner 1 , Andrew P Dickens 1 , David Fitzmaurice 7 , Peymane Adab 1
24 October 2019
Reviews suggest that the ADO score is the most discriminatory prognostic score for predicting mortality among chronic obstructive pulmonary disease (COPD) patients, but a full evaluation and external validation within primary care settings is critical before implementation.
To validate the ADO score in prevalent and screen-detected primary care COPD cases at 3 years and at shorter time periods.
One thousand eight hundred and ninety-two COPD cases were recruited between 2012 and 2014 from 71 United Kingdom general practices as part of the Birmingham COPD Cohort study. Cases were either on the practice COPD register or screen-detected. We validated the ADO score for predicting 3-year mortality with 1-year and 2-year mortality as secondary endpoints using discrimination (area-under-the-curve (AUC)) and calibration plots.
One hundred and fifty-four deaths occurred within 3 years. The ADO score was discriminatory for predicting 3-year mortality (AUC= 0.74; 95% CI: 0.69–0.79). Similar performance was found for 1- (AUC= 0.73; 0.66–0.80) and 2-year mortality (0.72; 0.67–0.76). The ADO score showed reasonable calibration for predicting 3-year mortality (calibration slope 0.95; 0.70–1.19) but over-predicted in cases with higher predicted risks of mortality at 1 (0.79; 0.45–1.13) and 2-year (0.79; 0.57–1.01) mortality.
The ADO score showed promising discrimination in predicting 3-year mortality in a primary care population including screen-detected cases. It may need to be recalibrated if it is used to provide risk predictions for 1- or 2-year mortality since, in these time-periods, over-prediction was evident, especially in cases with higher predicted mortality risks.