The Seattle Heart Failure Model (SHFM) is 1 of the most widely used tools to predict survival in patients with heart failure. However, it does not accommodate very elderly patients. We decided to assess the applicability of the SHFM in patients >80 years old enrolled in a tertiary care heart failure clinic. We evaluated the difference between observed survival and mean life expectancy as predicted by the SHFM on 261 patients >80 years old enrolled in a heart failure clinic at the Jewish General Hospital, Montreal, Quebec, Canada from January 2002 through March 2010. Average age of the patient population was 85 ± 4 years (range 80 to 105). Sixty-two percent of the population consisted of men, 63% had ischemic cardiomyopathy (ICM), and average ejection fraction was 36 ± 18%. Median observed survival was 1.91 years (interquartile range 0.68 to 5.53) for the total population (n = 261). The SHFM (predicted median survival 6.7 years, interquartile range 3.8 to 11.2) overestimated life expectancy by an average of 4.79 years. For patients with ICM (n = 164) versus non-ICM (n = 97), the score overestimated survival by 4.29 versus 5.69 years, respectively. In conclusion, the SHFM overestimates life expectancy in elderly patients followed in a tertiary care heart failure clinic. Further studies are needed to more accurately estimate prognosis in this patient population.
Many clinical decision tools have been developed to estimate the prognosis of patients with heart failure. These tools include the Enhanced Feedback For Effective Cardiac Treatment model, which was intended to be used in patients hospitalized with heart failure, and the Heart Failure Survival Score, a prospectively validated model developed for patients with advanced heart failure (New York Heart Association class III or IV). The Seattle Heart Failure Model (SHFM) is another risk score that incorporates easily obtainable clinical variables including medications, laboratory information, and device therapy and predicts 1-, 2-, and 5-year survival and mean life expectancy (MLE) in patients with heart failure. This score was developed and validated in an outpatient population with heart failure from 4 clinical trials and 2 observational studies. The oldest mean age was 71 ± 7 years in 1 of the clinical trials used for score development. The SHFM was subsequently validated in different populations in which the mean age was 52 to 55 years. It was also validated in a hospital-based subgroup of patients >75 years of age. The SHFM calculator does not allow for the incorporation of ages >85 years, making it impossible to calculate the MLE for patients in this age group. The aim of the present study was to assess the applicability of the SHFM in patients >80 years old who are actively followed in an outpatient tertiary care heart failure clinic.
Methods
We retrospectively collected data necessary for the SHFM in patients >80 years old enrolled in the tertiary care heart failure clinic at the Jewish General Hospital, Montreal, Quebec, Canada from January 2002 through March 17, 2010. Patients eligible for entry into the clinic were referred by a cardiologist or general internist for multidisciplinary specialized care. Diagnosis of heart failure was based on history, physical examination, chest x-ray findings, laboratory information, and 2-dimensional Doppler echocardiographic results.
In total 271 patients >80 years old were enrolled in the heart failure clinic during the study period. Ten patients were excluded from the study because of multiple missing data, resulting in a total study population of 261 subjects. All patients entered the cohort on the date of their first visit to the heart failure clinic. They were followed until death (n = 175) or censoring owing to loss to follow-up (n = 10) or administrative censoring at the end of the study period (March 17, 2011; n = 76). Dates of death were determined using a heart failure database (VisionC, Vision Cardiology 2.0, Computerized Heart Failure Record, ©2000, Montreal, Quebec, Canada), hospital records, and the Quebec obituaries.
Data were collected from the first clinic visit and included patient age, gender, New York Heart Association class, weight, ejection fraction by 2-dimensional echocardiography, systolic blood pressure, cause of heart failure (i.e., ischemic cardiomyopathy [ICM] vs non-ICM), medications used (angiotensin-converting enzyme inhibitor, angiotensin receptor blocker, β blockers, statin, allopurinol, aldosterone blocker, and diuretic dose), and any device installed (implantable cardioverter–defibrillator, biventricular pacemaker, or biventricular implantable cardioverter–defibrillator). Specific laboratory parameters were also collected on entry into the heart failure clinic including hemoglobin, percent lymphocytes, uric acid level, total cholesterol, and serum sodium. Data regarding co-morbid conditions were also gathered based on chart review where these diagnoses were documented. Missing data at the time of the initial clinic visit were replaced by data collected within 12 months. Uric acid (n = 11, 4%) and cholesterol (n = 8, 3%) levels were missing for a small number of patients. For these patients only, mean overall cholesterol and uric acid levels were used. This method of dealing with missing data has been used previously. All heart failure patient data were stored in the VisionC database. The study was approved by the hospital ethics board.
The SHFM calculator does not allow for estimation of MLE in patients >85 years old; the maximum age accepted by the tool is 85 years. We therefore constructed a graph of age versus MLE for each patient >85 years of age (n = 133, 51%). This graph plotted theoretical patients with the same SHFM variables as the patient of interest except for age. Age was increased in a stepwise fashion starting with 40 years of age and ending with 85 years of age. The relation obtained was nearly linear with an r 2 value ≥0.95 in all cases. We used the equation of this regression line to extrapolate the MLE for our cohort of patients >85 years of age.
Demographic and clinical characteristics are presented as mean ± SD or count data with corresponding proportions. Observed survival was calculated for all patients by subtracting the date of the cohort entry from each patient’s last recorded date (date of death or date of censoring). Predicted survival was determined using the MLE from the SHFM. Kaplan–Meier survival curves were constructed with 95% confidence intervals to describe observed and predicted survival while accounting for censoring, and we calculated the median observed and predicted survival times with corresponding interquartile ranges (IQRs). Analyses were then repeated stratified by ICM status. Survival analyses were conducted using STATA 11.2 (STATA Corp, College Station, Texas).
Results
Baseline characteristics of all patients are listed in Table 1 . The average age of our patient population was 85 ± 4.0 years (n = 261). Fifty-one percent of the total population was >85 years old.
| Age (years), mean ± SD (range) | 85 ± 4 (80–105) |
| Survival (years), median (interquartile range) | 1.91 (0.68–5.53) |
| Men | 161 (62%) |
| New York Heart Association class | |
| I | 25 (9.6%) |
| II | 121 (46%) |
| III | 111 (43%) |
| IV | 4 (1.5%) |
| Ischemic cause | 164 (63%) |
| Ejection fraction (%), mean ± SD | 36 ± 18 |
| Average weight (kg), mean ± SD | 68 ± 12 |
| Systolic blood pressure (mm Hg), mean ± SD | 125 ± 24 |
| Diabetes mellitus | 97 (37%) |
| Hypertension | 169 (65%) |
| Dyslipidemia | 90 (34%) |
| Chronic kidney disease | 65 (25%) |
| Atrial fibrillation | 113 (43%) |
| Cancer ⁎ | 38 (15%) |
| Peripheral vascular disease and cerebrovascular accident/transient ischemic attack | 83 (32%) |
| Implantable cardioverter–defibrillator | 4 (1.5%) |
| Cardiac resynchronization therapy | 2 (0.8%) |
| Medications | |
| Loop diuretic | 239 (92%) |
| Hydrochlorothiazide | 0 (0%) |
| Metolazone | 15 (5.7%) |
| Aldosterone blocker | 30 (11.5%) |
| Angiotensin-converting enzyme inhibitor | 148 (56.7%) |
| β Blocker | 173 (66%) |
| Angiotensin receptor blocker | 62 (24%) |
| Statin | 137 (53%) |
| Allopurinol | 26 (10%) |
| Laboratory data | |
| Sodium (mEq/L), mean ± SD | 140.9 ± 4 |
| Hemoglobin (g/dl), mean ± SD | 11.9 ± 17.05 |
| Lymphocytes (%), mean ± SD | 20.76 ± 8.8 |
| Uric acid (mg/dl), mean ± SD | 8.56 ± 2.55 |
| Total cholesterol (mg/dl), mean ± SD | 152.2 ± 42.14 |
| Creatinine (mg/dl), mean ± SD | 161.7 ± 93.8 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
