Survival prediction for terminally ill patients with cancer: Revision of palliative prognostic score with incorporation of delirium.

Abstract

e19552 Background: The validated Palliative Prognostic (PaP) score predicts survival in terminally-ill cancer patients and is calculated on the basis of dyspnea, anorexia, Karnofsky Performance Status, clinical prediction of survival, total white blood count and lymphocyte rate. The impact of delirium is known but was not incorporated in the original PaP score. Our aim was to incorporate information on delirium into this score. Methods: The study was based on a retrospective series of 361 terminally-ill cancer patients. The main endpoint was overall survival. We used the ‘validation by calibration’ approach proposed by van Houwelingen et al. and later adapted by Miceli et al. to obtain a score revision with inclusion of a new variable. A new model is fitted including the original score and the new variable as covariates; the model parameter estimates are then used to calculate the partial score for the new variable and the revised cutoffs to define the prognostic categories. Results: The prognostic contribution of delirium was confirmed as statistically significant (p<0.001) and the variable was incorporated into the PaP score (D-PaP score) accordingly. With such a revision, 30 day survival in the three risk groups (A>70%, B 30-70%, C<30%) was 83% (95% Confidence Interval CI 77-90), 50% (95% CI 42-58) and 9% (95% CI 3-15) vs 87% (95% CI 80-93), 51% (95% CI 44-59) and 16% (95% CI 8-24) obtained using the original PaP score. PaP and D-PaP score classifications were in accordance in 292 patients (80.9%). Seventeen (4.7%) patients were switched to a group with a less favorable prognosis and 52 (14.4%) were switched to a more favorable prognostic group. The frequency of patients with delirium, 14 (10.1%) and 37 (25.9%) in groups A and B, sharply increased to 48 (60.8%) in group C. The overall prognostic performance of the revised score was better than that obtained with the original one. Conclusions: The model validation and revision was carried out by modifying the cut-off values used for prognostic grouping without affecting the partial scores of the original PaP score. The revision improved the performance of the original PaP score whilst maintaining its key feature of simplicity.