Introduction Heart failure has a worse survival rate than many common cancers, yet few patients receive any palliative care input during the course of their illness. One of the main difficulties in providing palliative care for heart failure patients is the uncertainty around the course of the disease and the patient′s life expectancy. The aim of this study was to compare the “Gold Standards Framework” (GSF) criteria, which were developed to determine the need for palliative care in non-cancer patients, with the “Seattle Heart Failure (SHF) Model”, which provides a method of calculating a patient′s predicted mean life expectancy using physiological variables.

Methods Chronic heart failure patients, in NYHA class III or IV, who were being managed in the specialist, heart failure nursing service, were identified from a clinical heart failure database. GSF criteria were assessed by interviewing the specialist nurse responsible for each patient′s care. Clinical data required for the SHF model were obtained from two, online databases and were used to estimate mean life expectancy and predicted mortality at 1 year. Patients were then followed up, at 1 year, to evaluate; 1) all cause mortality, 2) place of death, and 3) the sensitivity and specificity of the GSF and SHF to predict death at 1 year.

Results 138 NYHA III-IV patients were identified from a total of 368 patients currently managed within the specialist nurse service; 66% were male, and the mean age was 77 years. GSF criteria, identified 119/138 (86%) patients that met the minimum requirement for palliative care input. However, the SHF model predicted that only 6/138 patients (4.3%) had a predicted life expectancy of less than 1 year. Patients who met GSF criteria for palliative care had significantly more hospital admissions (p=0.001) and had significantly lower predicted survival rates at 1 year (p=0.038) than those patients that did not meet GSF criteria. At follow-up, 43/138 patients had died (31%). Of these, 58% (25/43) died in hospital, following an acute admission. The sensitivity and specificity for the GSF and SHF model were 22%/83% and 98%/12% respectively. Overall, the patients renal function (eGFR<35 ml/min) was the best predictor of mortality, (sensitivity/specificity=82%/56%).

Discussion Neither the GSF nor the SHF were very accurate in predicting which patients were in the last year of life, in this selected sample. Despite the increasing drive towards palliation in heart failure, clinicians are still faced with a substantial prognostic barrier. Therefore, the progress of palliative care in heart failure patients may require a shift away from the traditional “end of life” model developed in cancer treatment, and focus instead on a patient′s increasing needs coupled with an understanding that death, itself, may remain unpredictable.