Objectives To develop a mortality-predictive model for correct identification of patients with non-cancer multiple chronic conditions who would benefit from palliative care, recognise predictive indicators of death and provide with tools for individual risk score calculation.

Design Retrospective observational study with multivariate logistic regression models.

Participants All patients with high-risk multiple chronic conditions incorporated into an integrated care strategy that fulfil two conditions: (1) they belong to the top 5% of the programme’s risk pyramid according to the adjusted morbidity groups stratification tool and (2) they suffer simultaneously at least three selected chronic non-cancer pathologies (n=591).

Main outcome measure 1 year mortality since patient inclusion in the programme.

Results Among study participants, 201 (34%) died within the 1 year follow-up. Variables found to be independently associated to 1 year mortality were the Barthel Scale (p<0.001), creatinine value (p=0.032), existence of pressure ulcers (p=0.029) and patient global status (p<0.001). The area under the curve (AUC) for our model was 0.751, which was validated using bootstrapping (AUC=0.751) and k-fold cross-validation (10 folds; AUC=0.744). The Hosmer-Lemeshow test (p=0.761) showed good calibration.

Conclusions This study develops and validates a mortality prediction model that will guide transitions of care to non-cancer palliative care services. The model determines prognostic indicators of death and provides tools for the estimation of individual death risk scores for each patient. We present a nomogram, a graphical risk calculation instrument, that favours a practical and easy use of the model within clinical practices.