Adaptation of the Palliative Prognostic Index in patients with advanced medical conditions

Abstract

ObjectiveTo analyze the accuracy of the Palliative Prognostic Index (PPI) in patients with advanced medical diseases and to recalibrate it in order to adapt it to the profile of these patients. MethodsMulticenter, prospective, observational study that included patients with one or more advanced medical diseases. Calibration (Hosmer–Lemeshow goodness of fit) and discriminative power (ROC and area under the curve [AUC]) of PPI were analyzed in the prediction of mortality at 180 days. Recalibration was carried out by analyzing the scores on the PPI of each quartile upward of dying probability. Accuracy of PPI was compared with that obtained for the Charlson index. ResultsOverall mortality of the 1788 patients was 37.5%. Calibration in the prediction of mortality was good (goodness of fit with p=.21), the prognostic probabilities ranging from 0 to 0.25 in the first quartile of risk and from 0.48 to 0.8 in the last quartile. Discriminative power was acceptable (AUC=69; p=.0001). In recalibrated groups, mortality of patients with 0/1–2/2.5–9.5/≥10 points was 13, 23, 39 and 68%, respectively. Sensitivity (S) and negative predicative value (NPF) of the cutoff point above 0 points were 96 and 87%, respectively, while specificity (sp) and positive predictive value (PPV) of the cutoff point above 9.5 points were 95 and 68%. Calibration of the Charlson index was good (p=.2), and its discriminative power (AUC=.52; p=.06) was suboptimal. ConclusionsPPI can be a useful tool in predicting 6-month survival of patients with advanced medical conditions.