Article Text
PDF
Abstract
Objectives To obtain national data on the clinical characteristics, investigation, management and outcome of patients hospitalised with a diagnosis of heart failure.
Method A survey was carried out of the first 10 patients hospitalised with a primary diagnosis of heart failure each month in 86 hospitals providing services for acute medical admissions in England and Wales from April 2008 until March 2009. The main outcome measures were rates of investigations, treatments and specialist management, length of hospital stay and mortality.
Results The 86 hospitals enrolled 6170 patients with a median age of 78 years (IQR 70–85 years), including 2639 (43%) women. At admission, only 30% of patients were breathless at rest, while 43% had peripheral oedema. Echocardiograms were recorded in 75% of patients and left ventricular ejection fraction (LVEF) was ≤40% in 78%. Natriuretic peptides were rarely measured. Allowing for missing data, >90% of patients were treated with loop diuretics at discharge, 80% with ACE inhibitors or angiotensin receptor blockers, 50% with β-blockers and 30% with aldosterone antagonists. Patients with an LVEF <40% were more likely to receive these agents. Median hospital stay was 9 days (IQR 5–17) and in-patient mortality was 12%. Patients admitted to general medicine rather than cardiology wards were more likely to die (HR=2.5, 95% CI 2.0 to 3.3, p<0.001) even after adjusting for differences (HR=1.9, 95% CI 1.5 to 2.5, p<0.001). Projected 1-year mortality below and above age 75 years was 26% and 56%, with higher rates if managed on general medicine rather than cardiology wards (HR=1.4, 95% CI 1.2 to 1.6, p<0.001).
Conclusion The prognosis of patients hospitalised with heart failure remains poor and investigation and treatment suboptimal. Specialist services are associated with higher rates of investigation and treatment and improved outcome.
- Heart failure
- delivery of care
- epidemiology
- statistics
Statistics from Altmetric.com
Introduction
Despite advances in diagnosis and management, heart failure remains a common but complex syndrome associated with high rates of hospitalisation and death. In England and Wales, with a combined population of ∼50 million in 2006–2007, more than 250,000 hospital deaths and discharges were coded for heart failure, more than 65,000 of these in the first diagnostic position.1 For an average hospital trust, which serves a community of ∼300 000 people, this represents about 1500 episodes each year, with about 400 in the first diagnostic position. Assuming an average length of stay in excess of 10 days, this reflects ∼2.5 million bed-days per year overall and about 15,000 bed-days per year for each hospital trust. Assuming a cost of at least £225 per bed-day,2 this part of the cost of managing heart failure could alone amount to £563 million nationally and £3.4 million per hospital trust per year. These data probably underestimate true activity.3
In order to gain greater insight into the contemporary epidemiology, diagnosis and management of patients hospitalised with heart failure, the British Society for Heart Failure, the HealthCare commission and the National Health Service (NHS) Information Centre joined forces to develop, pilot and deploy an audit of patients with a death or discharge diagnosis of heart failure in England and Wales. A similar audit is being conducted in Scotland. Following pilot studies using different entry criteria, a uniform set of criteria for patient inclusion was agreed and implemented in March 2008. This is a report of the first year's activity.
Methods
The NHS Information Centre developed a secure, encrypted, web-based database using Lotus Notes for the collection of data relating to patients with heart failure. The full data set includes 233 fields and can be used as a patient electronic record. For audit purposes, 21 fields were designated as core fields, of which seven comprised basic demographic information such as patient identifiers, sex and date of enrolment. A list of core fields is shown in figure 1.
Mandatory fields for completion of the 2008–2009 audit. Note that mandatory fields could be completed as unknown.
The purpose of the survey was to assess the quality of care of patients with heart failure by gathering information on the rate of diagnostic investigations, treatment with medicines at discharge, referral to specialist services for continuing care, treatment ward and length of stay. All-cause mortality during and subsequent to hospitalisation, with a maximum potential follow-up of 1 year, was recorded.
Various methods of patient identification were considered including enrolment from outpatient clinics and hospital wards of patients in whom a definite or provisional diagnosis of heart failure had been made, but such a strategy could not address any deficiency in the diagnostic process and might enrol patients with an unrepresentatively high quality of care. Inclusion of all patients receiving a loop diuretic, patients in whom heart failure should be excluded, was considered, but pilot studies suggested that this was likely to be a large number of patients and not readily feasible at this time.4 Another possibility was to enrol all patients with a death or discharge diagnosis of heart failure in any coding position (up to 14 diagnoses can be coded for each patient), since a diagnosis of heart failure in any position carries a similar prognosis and it is a matter of judgement whether to code heart failure in the first or lower position. This would require recording information on ∼250,000 episodes per year and was also considered not feasible. Ultimately, a pragmatic decision was taken to focus only on patients with heart failure coded in the first position at death or discharge. During 2008/2009, participating hospitals were asked to provide data on the first 10 patients with a death or discharge diagnosis of heart failure in each month. If all hospital trusts dealing with adult general or cardiology admissions (166) in England and Wales complied with this request, this would generate ∼20 000 submissions from 10 000–15 000 patients. The intention is to code all patients with a death or discharge code of heart failure in the first diagnostic coding position from April 2010 onwards.
This is a national survey initiated by the NHS Information Centre and therefore no specific request for ethical approval from regulators was deemed necessary nor was patient consent sought. The NHS Information Centre acts as guarantor of patient anonymity and data security.
Statistical methods
Data were summarised using the median and IQR (continuous data) or by percentages for categorical data. Median length of stay and IQR excluded patients with an admission of <24 h for whatever reason, including death, since it was unclear what proportion of these were elective admissions for planned procedures. Variables predicting the likelihood of being discharged on guideline treatment were modelled using logistic regression and backwards elimination. We are aware that automated selection methods are not optimal,5 6 though backwards elimination is preferable to forwards stepwise.7 Models were validated using re-sampling based on 10-fold cross-validation.8 The data were divided into 10 subsets of approximately equal size. For each subset we generated a logistic regression model leaving out one subset at a time. The omitted subset was used to calculate a misclassification rate per model. Missing values for categorical variables generated a separate category which was included in the modelling. Analyses were conducted both in the overall population surviving to discharge and in the subpopulation with a reported left ventricular ejection fraction (LVEF) <40%, the main focus of therapeutic guidelines.
All-cause mortality was analysed by Cox regression from which HRs with 95% CIs were estimated. Deaths were confirmed by the Office for National Statistics (ONS). The proportionality of hazards assumption was assessed by residual plotting.9 Two Cox models were undertaken. First, in patients who died on index admission (neither treatment on discharge nor postdischarge referral to cardiac/heart failure services were considered as covariates in this group). A second model was developed for patients surviving to discharge. Graphical presentation was by Kaplan–Meier curves. A nominal level of 5% statistical significance (two-tailed) was assumed. The Stata statistical computer package was used to analyse the data (StatCorp, Jacksonville, Florida, USA).
Results
Eighty-six hospitals participated in this survey, reporting data on a median of 44 (IQR 13–111) patients. In the 12 months after 31 March 2008, 6170 patients were enrolled (compared with ∼10 000 had each hospital provided the expected 120 patients), of whom 2639 (43%) were women (tables 1–4). The median age at first admission was 78 years (IQR 70–85 years) with 54% of men and 71% of women being aged >75 years. Ethnic origin was reported in 4063 patients (66%) of whom 3727 (92%) were Caucasian.
Characteristics of all patients by age, sex and admission ward
Characteristics of patients with LVEF <40% by age, sex and admission ward
Characteristics of patients with LVEF ≥40% by age, sex and admission ward
Characteristics of patients in whom LVEF was not reported by age, sex and admission ward
Almost 50% of patients were managed on cardiology wards. Compared with patients on general medical wards, those managed on cardiology wards were younger and more likely to be men, were more likely to have echocardiography and to have left ventricular systolic dysfunction (LVSD), but rates of co-morbidity including ischaemic heart disease, atrial fibrillation and diabetes were similar. The diagnosis of heart failure provided by discharge codes was verified by site investigators in 81% of cases, with little difference according to age, sex, LVEF or admission ward.
Most patients had an ECG done (96%). Echocardiograms were usually done in patients admitted to cardiology wards (88%) but often not done on general medical wards (65%). A higher proportion of echocardiograms from cardiology wards showed an LVEF <40% (84% vs 71%). Only 1% of patients had a measurement of a natriuretic peptide. Overall, only 48% of patients were reported to have ischaemic heart disease and only 33% were reported to have had a prior myocardial infarction, although proportions were higher among men and younger patients. Hypertension was reported in 47%, diabetes in 27% and valve disease in 19%, and were more common among older patients. At the time of admission, only 30% of patients were reported to be breathless at rest, while 40% had breathlessness limiting ordinary activity and 43% had moderate or severe peripheral oedema. Symptom presentation differed little by age, sex or admission ward.
LVEF was not recorded in 25%, was <40% in 58% of all cases and >40% in only 17%. Patients with LVEF >40% and those who had no reported LVEF had similar characteristics and were more likely to be women, older, managed on general medical wards and have atrial fibrillation and valve disease but less likely to have had a myocardial infarction compared with those with LVEF <40%. Symptoms at presentation differed little by LVEF group.
Treatment at discharge was reported in ∼90% of patients (table 5). Allowing for missing data, >90% of patients were treated with loop diuretics at discharge regardless of age, sex or treatment ward. ACE inhibitors or angiotensin II receptor blockers (A2RBs) were prescribed in ∼80% of patients aged <75 years and ∼70% aged ≥75 years without evidence of bias according to sex. β-Blockers were prescribed in only 50% of patients. Younger patients and men were more likely to receive them. Overall, ∼30% of patients were discharged on an aldosterone antagonist, with men and those admitted to cardiology wards more likely to receive them. Diuretics and A2RBs were prescribed in a similar proportion of patients regardless of whether LVEF was above or below 40% or not documented. Patients with a documented LVEF <40% were more likely to receive ACE inhibitors, β-blockers and aldosterone antagonists than other patients.
Treatment at discharge
The median stay in hospital when >24 h was 9 days (IQR 5–17 days), but was slightly longer in older patients and shorter in patients admitted to general medical compared with cardiology wards (tables 6 and 7). Patients without a reported LVEF tended to have shorter hospital stays, but length of stay was similar in those with LVEF above or below 40%.
Outcomes for all patients and of the subgroup with LVEF <40%
Outcomes for all patients with LVEF ≥40% or no reported LVEF
Factors associated with discharge medications
These were determined using a series of 10 logistic regression models and excluded patients who died during the index admission. Four variables (age <75 years, loop diuretics, history of hypertension and LVEF <40%) were independently associated with the prescription of either ACE inhibitors or A2RBs and were retained in all 10 models (table 8). Misclassification varied between 23% and 27%. In the subset of patients with LVSD, three variables, namely age <75 years, loop diuretics and hypertension, were in all 10 models.
For use of β-blockers, three variables (age <75 years, loop diuretics, and history of hypertension) were retained in all 10 models (table 9).
Factors associated with the use of ACEI/ARBs: logistic regression models using 10-fold cross-validation for patients surviving to discharge
Factors associated with the use of β-Blockers: logistic regression models using 10-fold cross-validation for patients surviving to discharge
Also frequently reported were male sex, history of myocardail infarction and having an echocardiogram whether or not it showed an LVEF <40%. Misclassification varied between 31% and 42%. In the subset with LVSD, three variables, namely age <75 years, loop diuretics and myocardail infarction, were in all 10 models.
For the use of aldosterone antagonists, six variables (age <75 years, loop diuretics, male sex, history of valve disease, echo with LVEF <40% and moderate/severe oedema) were retained in all 10 models (table 10). Misclassification varied between 26% and 30%. In the subset with LVSD, three variables, (age <75 years, male sex, loop diuretics) were in all 10 models.
Factors associated with the use of ARAs: logistic regression models using 10-fold cross-validation for patients surviving to discharge
Deaths during the index admission
Overall, the mortality during the index hospitalisation was 12%. Patients aged <75 years and those managed on cardiology wards had a mortality of ∼5% on the index admission, compared with a mortality of >15% in other groups. Mortality was lowest (8%) among those with LVEF <40%, higher (11%) among those with LVEF >40%, perhaps due to the greater age of these patients, and highest among those without a recorded LVEF (18%). In the 652 patients who died during the index admission, median time to death was 11 days (IQR 3–20) days. Patients admitted to a general medicine ward were more likely to die than those admitted to a cardiology ward (HR=2.5, 95% CI 2.0 to 3.3, p<0.001). This relationship remained significant in a multivariable model adjusting for age, aetiology, echocardiography, heart rhythm, sex and symptoms (HR=1.9, 95% CI=1.5 to 2.5, p<0.001). Treatment (mainly recorded at discharge) was not included in these models.
Deaths subsequent to discharge
Median follow-up, censoring for death, was only 158 days (IQR=70–260 days) and maximum follow-up 365 days. Overall, 22% of patients who survived to discharge subsequently died. (figure 2A–H). Younger patients and those who had been managed on cardiology wards fared better. The predicted annual mortality was close to 30%, ranging from 10% in those aged <65 years up to 40% in those aged >85 years. Patients who did not have LVEF measured or who had LVEF above or below 40% had similar postdischarge mortality. Age but not sex appeared an important determinant of outcome. Patients discharged from general medicine wards were more likely to die than those discharged from cardiology wards (HR=1.4, 95% CI 1.2 to 1.6, p<0.001). After adjustment for age, sex, aetiology, symptoms, treatment and investigation, a significant relationship still existed (HR=1.1, 95% CI 1.0 to 1.3, p=0.048). After further adjustment for follow-up by cardiology/heart failure services, this relationship was accounted for (HR=1.1, 95% CI 0.9 to 1.2, p=0.22). A model excluding treatment ward identified greater age, greater severity of oedema, lack of use of an ACE inhibitor or A2RB and lack of use of β-blockers as the strongest predictors of an adverse outcome (all p<0.001). LVEF did not predict outcome.
(A–H) A series of Kaplan–Meier curves showing mortality subsequent to discharge according to (A) age, (B) sex, (C) management ward, (D) left ventricular ejection fraction (LVEF), (E) use of ACE inhibitors or angiotensin II receptor blockers (A2RBs), (F) use of β-blockers, (G) use of aldosterone antagonists (ARAs) and (H) according to whether or not patients were referred into a specialist follow-up programme.
Follow-up arrangements
Most patients with LVEF <40% regardless of age or management ward had cardiology or heart failure specialist nurse follow-up. Patients who were younger and managed on cardiology wards were also likely to have specialist follow-up even if their LVEF was >40%. A minority of patients were reported to have follow-up with care of the elderly or palliative care teams.
Overall mortality
Overall mortality, including deaths during the index admission, was 34% and the estimated 1-year mortality was >40%. For those aged <75 years, the overall predicted annual mortality was 26% and for those aged ≥75 years it was 56%.
Discussion
The most striking finding from this analysis is the poor overall prognosis of patients who require admission to hospital with a primary diagnosis of heart failure. This is substantially worse than data from clinical trials suggest, which may reflect the exclusion of older, frail and multimorbid patients and/or the survival benefit that appears to accrue from participating in clinical trials.10 Overall, 61% of patients were aged ≥ 75 years. Age, rather than LVEF or sex, was the most important determinant of prognosis, although 1-year mortality was high (25%) even in those aged <75 years, rising to >50% in those aged ≥ 75 years. These poor outcomes occurred in patients with LVEF <40% despite fairly high uptake of guideline-indicated pharmacological treatment including ACE inhibitors or AR2Bs in 81%, β-blockers in 59% and aldosterone antagonists in 37%, although whether these were prescribed in appropriate doses is uncertain as this was not included in the analysis. There is room for improved implementation of existing pharmacological treatments, but this alone may not be enough to reduce annual mortality to a notional target of <10% even in those aged <75 years.
This is the third substantial survey of heart failure deaths and discharges conducted in the UK. The EuroHeart Failure survey enrolled 1700 patients in a 6-week snapshot of UK hospital activity in 2001.11 It included patients with confirmed or suspected heart failure. These patients were slightly younger (mean age 75 years) than in the current survey but with similar sex and LVEF distribution. Only 21% of patients were managed on cardiology wards. Inpatient mortality was 9.1% and 12 week mortality 15.5%. Use of ACE inhibitors or A2RBs was 55%, β-blockers 24% and aldosterone antagonists 13.9%. An NHS survey conducted in 2005–2006 enrolled 9387 patients with a mean age of 77 years.12 Inpatient mortality was 15%. Use of ACE inhibitors or A2RBs was 71%, β-blockers 39% and aldosterone antagonists 25%. Comparing the surveys, use of heart failure medications appears to be increasing but the prognosis remains poor, perhaps in part due to the increasing age of the patients admitted.
Only ∼60% of the patients who should have been reported by participating hospitals were enrolled in this audit. Considering that the annual audit is still a work in prognosis, this rate of enrolment may be considered fair and the analysis relevant to contemporary clinical practice. Enrolment may have been biased towards younger patients and those managed on cardiology wards, although the age and sex distribution of patients was similar to two previous UK surveys of heart failure.11 12 Lack of availability of natriuretic peptides and diagnostic scepticism may have accounted for the relatively low rate of heart failure with preserved LVEF.13–15 As younger age and specialist management are associated with an increased probability of investigation, guideline-indicated management and prognosis, the real clinical situation may be worse than suggested by this survey. From April 2010, the audit should include all relevant hospitals and each should include at least 20 patients per month with a primary diagnosis of heart failure at death or discharge.
LVEF was not a strong determinant of prognosis and had a relatively small impact on the pattern of medication. Indeed, LVEF <40% was associated with a somewhat better prognosis, at least in part because, as shown in other studies, it is more prevalent in younger patients and perhaps because these patients are more likely to receive specialist follow-up. The prognosis of older patients with a low LVEF may be so dismal that they are under-represented in prevalence statistics. Patients with LVEF <40% were more likely to be treated with each class of drugs known to improve prognosis in patients with heart failure and this may also account, in part, for the better outcome among these patients. Although the evidence that ACE inhibitors, A2RBs, β-blockers and aldosterone antagonists should be given to patients with heart failure who have LVEF >40% is not robust, they are often used to manage co-morbidities such as ischaemic heart disease and hypertension and were widely used among such patients in this survey.16
Many patients had no LVEF reported. This is likely to reflect a mixture of underinvestigation and inadequate documentation of tests that were done. Measurement of natriuretic peptides provides a simple method of identifying patients who need investigation for heart failure and is a much better guide to prognosis than is LVEF.17 Despite National Institute for Health and Clinical Excellence (NICE) recommendations,18 these tests are not yet widely used in the UK (<2% of patients in this survey), perhaps because of existing financial structures that often deal poorly with problems such as heart failure for which responsibility is often spread across multiple business units. It is also likely that the diagnosis of heart failure is often overlooked. Only about a quarter of patients discharged from hospital on a loop diuretic carry a discharge diagnosis of heart failure, and yet the prognosis of these patients is similarly poor whether or not heart failure is reported.4 Widespread use of natriuretic peptides would improve detection rates, which is a necessary first step towards improved management of heart failure.
Clearly, this analysis has many limitations including case ascertainment, robust data on doses of medications and use of devices. This reflects, in part, the limited resources available to complete the audit, which depended, in large part, on the good will and enthusiasm of existing health professionals rather than new funding. These issues will be addressed in the coming years. Currently, hospital provision of care is suboptimal and the outcome of patients poor. The same rules that apply to suspected cancer should pertain to a disease with such a malign prognosis as heart failure, including easy access to the first diagnostic step (natriuretic peptides), speedy referral to an appropriate specialist, with inpatient and postdischarge management guided by appropriately trained specialist staff. The current system of care already provides care to all patients with heart failure, but this is often haphazard and disorganised. Re-organisation of existing resources, rather than new funding, will be key to success. Development of dedicated inpatient resources to manage patients with heart failure, similar to the introduction of coronary care units in the previous century that have revolutionised the care of acute coronary syndromes, should be considered.
Clearly, more needs to be done to improve the prognosis of patients with heart failure. It is reasonable to assume that most deaths in younger patients and a substantial proportion in older patients were cardiovascular. There are many strategies to improve the outcome of patients with heart failure. For example, appropriate deployment of implantable cardiac defibrillator or cardiac resynchronisation devices has been associated with an annual mortality of <5% in younger patients even with very poor left ventricular systolic function.19 Although our analysis did not include devices, it is very likely that they were deployed in only a small minority of patients. Many new pharmacological treatments directed at improving myocardial function are also in development, while interventions directed at key co-morbidities such as anaemia and renal impairment may also improve outcomes. Just as for acute coronary syndromes, the development of heart failure units will improve the deployment of existing treatments and accelerate the identification of new ones.
Perhaps the most important finding in this audit is that specialist care, both in the hospital and subsequent to discharge, is associated with better outcomes. This was not explained by differences in age, sex, LVEF, co-morbidity or discharge medication. This is consistent with data from randomised controlled trials, suggesting that intensified long-term management has a powerful effect on mortality.20 21 This may reflect better patient education and greater care with the titration and monitoring of treatment. Current evidence suggests that higher rates of prescription of evidence-based medicines especially ACE inhibitors, β-blockers and aldosterone antagonists,18 22 together with, where appropriate, the use of cardiac resynchronisation therapy (CRT) and implantable cardioverter-defibrillators (ICDs) within a comprehensive multidisciplinary system of specialist care would substantially improve outcomes. Audits such as this one continue to draw attention to the deficiencies in the management of patients with heart failure. Hopefully, future iterations of this audit will demonstrate sustained improvements in the quality of care, resulting in improved survival.
Acknowledgments
We acknowledge the efforts made by the staff of participating hospitals that entered information into the national audit database and to all the staff of the NHS Information Centre who assisted with implementation.