• Aorta
• great vessels and trauma
• arrhythmias
• cardiac function
• cardiac surgery
• heart failure

Heart failure (HF) is a significant public health issue with high morbidity and mortality despite persistent advances in therapeutic strategies. HF is physically and psychologically burdensome for patients and their families and costly to the NHS; the latter driven primarily by repeated prolonged hospital admissions.1 The greatest incidence and prevalence of HF is in older patients. The multiple and often complex comorbidities associated with this cohort impact on management strategies and prognosis.2 The 1-year survival following a new diagnosis of HF is invariably poor, but the decline towards death is often unpredictable. This chaotic illness trajectory contributes to what has previously been termed, ‘prognostic paralysis’.3 Consequently, anticipatory management is extremely difficult and so a ‘revolving door’ clinical pattern often ensues. It is now accepted that the morbidity and mortality for patients with HF is similar to or worse than many cancers with the first HF hospitalisation in particular heralding a very poor prognosis.4 Despite this, end of life care in HF remains uncoordinated and inadequate for many. As such the development and validation of a HF score to permit accurate identification and prognostication in this group is an attractive prospect clinically and in terms of resource allocation.

In this issue of Heart, Haga et al5 identified two scores for comparison: the qualitative Gold Standards Framework Prognostic Indicator Guide (GSF)6 and the better established quantitative Seattle Heart Failure Model (SHF).7 The latter has been extensively validated and is considered ‘gold standard’ for routine prognostication in ambulant HF populations, despite being derived exclusively from highly selected clinical trial patient data. The former, while advocated by the Clinical Standards for Heart Disease for identification of patients with HF approaching end of life,8 has no published data supporting its accuracy.

Haga et al recruited a total of 138 patients with New York Heart Association NYHA III-IV HF, predominantly older Caucasian men, from within their active HF liaison nurse service. The majority had moderate to severe left ventricular systolic dysfunction with an ischaemic aetiology and more than 75% were identified as having significant renal impairment. Almost one-third (43 patients) had died by the end of a 12-month follow-up period. This is in line with recently published large epidemiological studies.4 9 The majority of patients died in hospital, and death was most commonly the result of progressive HF. Haga et al demonstrated clearly that neither the GSF nor the SHF accurately predicted death in their cohort of ambulatory patients with NYHA III–IV HF. Furthermore, there was striking discordance between the scores with the GSF predicting death within 12 months for 86% of patients while the SHF predicted death in only 4% of the same HF population. Notably elevated serum creatinine alone demonstrated an overall accuracy for predicting 1-year mortality of 72%. The accuracy of this single factor for mortality prognostication was identical to the overall of accuracy of the quantitative SHF and clearly compared favourably to the overall accuracy of the GSF (41%).

While a number of HF prognostication models have been developed and validated in a range of patient groups, the accuracy of risk quantification in patient populations outwith the model derivation sample specifications is unclear.10 The deficiencies of these HF prognostication tools in routine practice are increasingly recognised and Haga et al elucidate this further. It has become apparent that existing HF scores do not accurately predict time to death or, arguably more importantly, those who may benefit most from palliative care intervention. Further research is needed in this area. Careful consideration of the HF cohorts from which to derive data to base novel scoring tools on will be imperative to ensure effective translation of prognostic scores into routine clinical practice within unselected HF communities.

If it is possible to develop an accurate HF prognostic score, then this will most likely involve a composite of clinical, echocardiographic and laboratory data. Haga et al allude to the potential prognostic power of an elevated creatinine level in patients with HF. Similarly it has previously been demonstrated that the inclusion of B-type natriuretic peptide (BNP) levels improves mortality prognostication by the SHF.11 One should also consider, however, that prognostication in HF may be too complex to be defined by a single score. Furthermore, those patients with the greatest symptom burden and most in need of palliative care may not necessarily be those approaching end of life. As such, accurate symptom assessment tools specific to HF, perhaps in conjunction with prognostic scores, may better identify those with the greatest need for palliative care and with the most to gain from any intervention.

Improving identification and prognostication in end-stage HF is clearly a valuable pursuit worthy of further time and resource, as it would facilitate more structured and anticipatory management strategies. The marked disparity between the coordinated multidisciplinary care received by patients with terminal cancer and the care of patients with end-stage HF is well recognised. The lack of structured end of life care in HF is reflected by the fact, as demonstrated by Haga et al, that many patients die in hospital. Indeed, a large Scottish epidemiological study based on a national database of over 12 000 patients comprising 4877 deaths over 3 years demonstrated that 73% of all HF deaths occurred in hospital.12 Hospital deaths are costly, and many patients would choose to die at home if possible.13 Several large surveys of the public have been undertaken in recent years to ascertain people's preferences and priorities in relation to end of life care. While preferences and priorities may change as death approaches the main findings were that most people would prefer to be cared for at home provided high quality care can be assured and minimal burden is placed on their families and carers.14 Previous studies have demonstrated that matching end-stage HF patients' preferred place of death with actual place of death is possible with collaboration between cardiology, palliative care and primary care. This collaboration has facilitated patients dying in their preferred place of death and also helped in the prevention of hospital admissions and futile attempts at cardiopulmonary resuscitation.13

Difficult mortality prognostication in advanced HF is frequently sited as an obstacle inhibiting discussions regarding end of life and appropriate initiation of palliative care.3 Given that the prognosis for the vast majority is exceptionally poor, there is a clear argument for the direction of palliative care resources being symptom centred rather than prognosis driven, that is, the trigger for palliative care input should be the presence of unmanageable symptoms despite maximum tolerated therapy rather than the perception that ‘end of life’ may be approaching. This strategy moves away from the historical concept of palliative care as a destination exclusively for those imminently dying and towards a more holistic patient and problem centred model that anticipates, reacts to and manages symptoms. As such, perhaps future research in end-stage HF should focus not only on more accurate mortality prognostication but also on the development and validation of effective symptom assessment tools.