Background Predicting prognosis accurately would help patients and clinicians to make informed decisions about treatment and referral to appropriate services. But user-friendly tools are lacking in clinical practice. The Palliative Prognostic Index(PPI), based on simple clinical indicators, has shown promise in several studies.

Aims Following a previous multicentre study in the UK, the current prospective study involved 10 centres, to check centre-to-centre variability, and included a second assessment of PPI score 3–5 days after admission to investigate whether incorporating changes in PPI lead to improve predictions.

Methods PPI score was calculated on admission to inpatient hospice, and again 3–5 days later. Kaplan-Meier curves were constructed and predicted survival based on PPI was compared to actual survival, using standard measures.

Results Initial PPI (PPI_1) was recorded for 1164 patients. Median survival for patients with PPI _1 score ≤4, 4–6 and >6 was 38, 17 and 5 days, but there was significant variation between centres. A second PPI value (PPI_2) was recorded for 962 patients. Most of the remaining patients had high PPI_1 and died before the second assessment. PPI_2 was a more reliable predictor of survival than PPI_1. Based on PPI_2, the sensitivity, specificity, positive predictive value and negative predictive value for prediction for <21 days were 64%, 85%, 84%, 65% and for ≥42 days were 64%, 77%, 53%, 84%. Again, there was considerable centre-to-centre variability. Kaplan-Meyer and ROC curves constructed (Survival <21 days: PPI1: 0.734; PPI2: 0.820. Survival ≥ 42 days: PPI1: 0.723; PPI2: 0.795). Further analysis of the effect of changes of score and comparison with recent studies from Taiwan and Japan showed similar results in many aspects. Possible causes of variability of results between centres are discussed.

Conclusion PPI score is more accurate if calculated twice and the rate of change of PPI is useful.