Objective To develop and validate a prognostic model to assess mortality risk at 24 months in patients with advanced chronic conditions.

Methods Retrospective design based on a previous population cohort study with 789 adults who were identified with the surprise question and NECPAL tool from primary and intermediate care centres, nursing homes and one acute hospital of Spain. A Cox regression model was used to derive a mortality predictive model based on patients’ age and six previously selected NECPAL prognostic factors (palliative care need identified by healthcare professionals, functional decline, nutritional decline, multimorbidity, use of resources, disease-specific criteria of severity/progression). Patients were split into derivation/validation cohorts, and four steps were followed: descriptive analysis, predictors’ assessment, model estimation and model assessment.

Results All predictive variables were independently associated with increased risk of mortality at 24 months. Performance model including age was good; discrimination power by area under the curve (AUC) was 0.72/0.67 in the derivation/validation cohorts, and correlation between expected and observed (E/O) mortality ratio was 0.74/0.70. The model showed similar performance across settings (AUC 0.65–0.74, E/O 1.00–1.01), the best performance in oncological patients (AUC 0.78, E/O 0.76) and the worst in dementia patients (AUC 0.58, E/O 0.85). Based on the number of factors affected, three prognostic stages with significant differences and a median survival of 38, 17.2 and 3.6 months (p<0.001) were defined.

Conclusion The NECPAL prognostic tool is accurate and eventually useful at the clinical practice. Stratification in risk groups may enable early intervention and enhance policy-making and service planning.