Abstract
Objective
Our objective was to develop models for short- (30-day) and long- (5-year) term mortality after heart failure (HF) hospitalization that include geriatric conditions, specifically dementia and mobility disability, to determine whether these conditions emerge as strong and independent risk factors.
Background
Although 80% of patients with HF are 65 years or older, no large studies have focused on the prognostic importance of geriatric conditions.
Methods
We analyzed medical record data from a national sample of Medicare beneficiaries hospitalized for HF. To identify independent predictors of mortality, we performed stepwise selection in multivariable logistic regression models. We used net reclassification improvement to assess the incremental benefit of adding geriatric conditions to a model containing traditional risk factors for mortality.
Results
The mean age of patients included in the analysis was 80 years; 59% were female, 13% were non-white, 10% had dementia and 39% had mobility disability. Mortality rates were 9.8% at 30 days and 74.7% at 5 years. Twenty-one variables were considered for inclusion in the final multivariable model. Dementia and mobility disability were among the top predictors of short- and long-term mortality, with among the top 6 largest absolute standardized estimates (SE) in the final model for 30-day mortality, and among the top seven largest SE for 5-year mortality. The net reclassification improvement when geriatric conditions were added to traditional factors was 5.1% at 30 days and 4.2% at 5 years.
Conclusions
Geriatric conditions are strongly and independently associated with short- and long-term mortality among older patients with HF.
Keywords: Heart Failure, Aging, Mobility, Dementia, Prognosis
INTRODUCTION
Prognostic information about patients with heart failure (HF) can be used to assist in clinical decision-making and identify potential targets for intervention. Published tools to stratify the risk of HF patients have traditionally focused on demographic factors, HF severity, medically diagnosed comorbid diseases, physical examination findings, and laboratory values (1–10). Although 80% of patients with HF are 65 years or older, no large studies have focused on the prognostic importance of geriatric conditions, including mobility disability and dementia. These conditions fall outside of the traditional disease model, which permeates HF research and clinical care, and thus may be overlooked in the care of older patients with HF.
To address this issue, we evaluated data from a nationally representative sample of Medicare patients hospitalized with HF in the United States that included information on geriatric conditions, cardiac function, comorbid diseases, and laboratory values. Our objective was to develop models for short- (30-day) and long- (5 years) term mortality after hospitalization for HF that included both traditional factors and geriatric conditions to determine whether geriatric conditions would emerge as strong and independent risk factors and to determine the incremental benefit of including geriatric conditions in models predicting mortality.
METHODS
National Heart Care Project
The Centers for Medicare & Medicaid Services National Heart Care Project is an ongoing quality of care initiative for Medicare beneficiaries hospitalized with cardiovascular diseases, including HF. As part of the project, a cohort of fee-for-service Medicare beneficiaries hospitalized with a principal discharge diagnosis of HF (International Classification of Diseases, Ninth Revision, Clinical Modification codes 402.01, 402.11, 402.91, 404.01, 404.91, or 428) (11) between April 1998 and March 1999 or July 2000 and June 2001 inclusive, was identified to assess quality of medical care. In each sampling period, hospital medical records were grouped by state, and then sorted by patients’ age, sex, and race, as well as the treating hospital. Up to 800 records were then randomly selected from each state. All records were included if fewer than 800 hospitalizations occurred in a state during a sampling period (Alaska, Hawaii, Idaho, Utah, Vermont, and Wyoming in both samples). Medical records of selected patients were obtained from the treating hospital and underwent detailed review by trained data abstractors in central data abstraction centers.
The reliability of data collected for the National Heart Care Project was formally assessed by comparing results obtained by abstractors with results obtained by a clinically experienced panel reviewing the same medical records (n=154) to determine assessment of ejection fraction and prescription of an ACE inhibitor to patients with reduced systolic function. Kappa values for these measures were 0.95 and 0.88, respectively. Similarly, inter-rater reliability was assessed by having two abstractors independently abstract medical records (n=560). For the thirteen measures assessed (pertaining to prescription of ACE-I, measurement of ejection fraction, cigarette smoking, discharge instructions, and use of beta-blockers), six had a Kappa > 0.8 denoting almost perfect agreement, and five had a Kappa of 0.61 to 0.80 denoting substantial agreement.
The reliability of data collected to assess geriatric conditions was not assessed with the same rigor as the quality measures, not surprising given the National Heart Care Project’s focus on quality measures for HF hospitalization. However, while preparing for the second phase of data collection (July 2000 through June 2001), the agreement of assessments by abstractors and a clinically experienced panel was measured for mobility and dementia. One hundred charts were reviewed, and agreement was found in 85% of mobility assessments and 99% of dementia assessments.
Study cohort
The National Heart Care cohort included 78,882 records, of which 39,477 were from 1998–99 and 39,405 from 2000–01. Of the 78,882 initially sampled hospitalizations, 6558 patients younger than 65 years were excluded because they are not representative of the majority of Medicare patients. Patients who arrived by inter-hospital transfer (n = 2419) and those without chart documentation of HF on admission (n = 5003) were excluded to ensure a cohort of patients presenting with HF. After chart abstraction identified one of the National Heart Care Project’s exclusion criteria (chronic hemodialysis, discharge against medical advice, and repeat hospital admission) (12), 4340 patients were excluded. Patients were also excluded if their vital status could not be verified (n=48). In total, 16,552 records met one or more of these exclusions, resulting in a cohort of 62,330 patients.
Geriatric Conditions
For the assessment of geriatric conditions, abstractors were trained to exclude conditions that were not present before the onset of the current illness that led to hospitalization. For example, if the patient was unable to walk due to increasing shortness of breath over the 2 days before hospitalization, but did not have limitation in mobility before this illness, the patient’s mobility would be categorized as not limited.
Mobility was assessed using information from physician, nursing, and physical therapy notes, as well as the discharge summary. Mobility disability was defined as requiring assistance (from a device, such as a cane or a walker, or from another person) or being unable to walk. If there was no explicit documentation in the medical record regarding mobility, this was coded as “missing.” Abstractors were instructed not to infer status based on presence of other conditions.
The presence of dementia (defined as a chronic loss of mental function or slowly progressive mental decline) was determined using the same sources as used for the assessment of mobility (excluding physical therapy notes). Acute confusional states and delirium were not included. Abstractors were trained to search for documentation of a history of dementia. If no documentation was found, dementia was considered not present.
Patient Outcomes
Mortality at 30 days and 5 years after admission was determined using the Medicare Enrollment Database and the Death Master File of the Social Security Administration (13). While the greatest (incident) mortality risk occurs in the month after HF hospitalization (14), mortality at 5 years was also examined to determine the long-term importance of geriatric conditions.
Statistical Analysis
For descriptive purposes, we compared patient characteristics (demographic and clinical) between those with at least one geriatric condition and those with none using the chi-square or t-test. Thirty-one percent of patients were missing data on mobility disability, while 37% were missing data on left ventricular ejection fraction; all other variables were missing < 5% of values. We did not impute missing data, nor did we exclude subjects with missing data. Rather, we created “dummy variables” for missing values. Because some variables would be expected to have non-linear relationships with the outcome, we examined the bivariate relationship of each predictor with five-year mortality. For variables where the strength of the linear association varied across the range of values observed (including serum sodium and potassium, systolic blood pressure, and body temperature), splines were considered. However the final results were minimally affected, so for enhanced clarity these variables were considered in a continuous fashion.
Non-parametric Spearman rank-order correlations between variables were evaluated, and when the correlation coefficient was greater than 0.4, denoting potential collinearity, a single factor was chosen based on clinical judgment and the strength of the association with the outcome in bivariate analyses. The only correlation coefficients that exceeded 0.4 were those between systolic and diastolic blood pressure (0.62) and BUN and creatinine (0.72). Systolic blood pressure and serum creatinine were then selected for inclusion in the final model based on strength of bivariate association
To identify independent predictors of mortality for each time point (30-days and 5 years), we derived 1000 data samples using bootstrapping simulation and performed stepwise selection in logistic regression models (P value for entry=0.01 and for exit=0.001) for each data sample. Variables retained in more than 80% (more than 800 times in the 1000 iterations) of simulated stepwise selection models at either time point were included in the final, multivariable logistic regression models for both time points. In addition to odds ratios, standardized estimates (SEs) were derived from the multivariable logistic regression models to compare the effect size of each variable on mortality. The SE is the correlation coefficient determined after standardizing each variable to have a mean of 0 and a SD of 1, thereby minimizing the effect of unit size on observed associations. It represents the change in response for a change of 1 SD in a predictor.
We also utilized a newly described reclassification approach to assess the incremental benefit of adding geriatric conditions to a multivariable model containing traditional risk factors (age, sex, race, body temperature, systolic blood pressure, heart rate, respiratory rate, serum sodium, potassium, creatinine, white blood cell count, hematocrit, prior HF, aortic stenosis, coronary artery disease, cerebrovascular disease, cancer, hypertension, diabetes mellitus, COPD, and left ventricular ejection fraction) to predict mortality (15). This approach expands upon previously published reclassification methods (16, 17). For the purposes of risk reclassification, patients were classified in quartiles of mortality risk. The net reclassification improvement (NRI) focuses on the reclassification of persons who developed events and those who did not after new information is added to the predictive model. Among persons who experienced an event (i.e., mortality), reclassification to a higher risk group was considered upward movement/improvement in classification. Reclassification downward was considered worsening in classification for those who developed an event. Conversely, among persons who did not experience an event, reclassification upward was considered worsening and reclassification downward was considered improvement in classification. Reclassification improvement was then estimated by taking the sum of differences in proportions of individuals reclassified upward minus the proportion reclassified downward for people who developed events, and the proportion of individuals moving downward minus the proportion moving upward for those who did not develop events. Using these methods, the overall reclassification sum is the NRI, and the statistical significance of the overall improvement is assessed with an asymptotic test.
All analyses were conducted using SAS software version 9.1 (SAS Institute, Cary, North Carolina). The authors had full access to and take full responsibility for the integrity of the data. All authors have read and agree to the manuscript as written.
RESULTS
As shown in Table 1, the mean age of the study population was 79.6 years; 58.5% were female and 13.4% were non-white. Most patients (71.6%) had a known diagnosis of HF before hospitalization. Coronary artery disease (57.2%), hypertension (63.7%), diabetes mellitus (39.7%), and COPD (34.0%) were common comorbidities. Overall, 43.8% of patients had at least one geriatric condition (39.2% had mobility disability and 9.7% of patients had dementia). Compared with patients without a geriatric condition, those with at least one geriatric condition were older and more likely to be female. Patients with a geriatric condition were more likely to have pre-existing HF, aortic stenosis, cerebrovascular disease, and diabetes mellitus but less likely to have coronary artery disease or depressed ejection fraction. They also had lower systolic and diastolic blood pressure, higher serum BUN and lower hematocrit compared with patients without a geriatric condition.
Table 1.
Baseline Characteristics of study population
| All patients (N=62,330) | No Geriatric Condition (n=35,002) | ≥ 1 Geriatric Condition (n=27,328) | P value* | |
|---|---|---|---|---|
| Demographics | ||||
| Mean age, years (SD) | 79.6 (7.8) | 77.9 (7.5) | 81.8 (7.6) | <0.001 |
| Female, % | 58.5 | 53.5 | 64.9 | <0.001 |
| Non-white race, % | 13.4 | 13.4 | 13.5 | 0.90 |
| Medical history, % | ||||
| Prior HF | 71.6 | 68.6 | 75.6 | <0.001 |
| Aortic stenosis | 7.5 | 6.8 | 8.3 | <0.001 |
| Coronary artery disease | 57.2 | 58.8 | 55.2 | <0.001 |
| Cerebrovascular disease | 18.8 | 14.2 | 24.6 | <0.001 |
| Cancer | 2.4 | 2.4 | 2.5 | 0.98 |
| Hypertension | 63.7 | 63.3 | 64.2 | 0.03 |
| Diabetes mellitus | 39.7 | 38.9 | 40.8 | <0.001 |
| COPD | 34.0 | 33.8 | 34.3 | 0.14 |
| Admission physical exam findings | ||||
| Mean body temperature, degrees Fahrenheit (SD) | 97.7 (1.5) | 97.6 (1.6) | 97.7 (1.5) | <0.001 |
| Mean systolic blood pressure, mm Hg (SD) | 146.5 (31.7) | 147.8 (32.2) | 144.8 (31.1) | <0.001 |
| Mean diastolic blood pressure, mm Hg (SD) | 77.9 (19.0) | 79.0 (19.2) | 76.4 (18.6) | <0.001 |
| Mean heart rate, beats/min (SD) | 90.0 (22.5) | 90.6 (22.9) | 89.1 (21.9) | <0.001 |
| Mean respiratory rate, breaths/min (SD) | 24.6 (6.7) | 24.4 (6.7) | 24.7 (6.8) | <0.001 |
| Peripheral edema, % | 74.6 | 72.7 | 77.0 | <0.001 |
| Admission laboratory findings, mean (SD) | ||||
| Sodium | 138.4 (5.1) | 138.5 (4.9) | 138.3 (5.3) | <0.001 |
| Potassium | 4.3 (0.7) | 4.3 (0.6) | 4.3 (0.7) | <0.001 |
| BUN | 30.5 (19.3) | 28.9 (18.2) | 32.5 (20.5) | <0.001 |
| Creatinine | 1.5 (0.8) | 1.4 (0.7) | 1.5 (0.8) | <0.001 |
| Admission laboratory findings, mean (SD) | ||||
| Glucose | 152.0 (71.3) | 153.3 (71.7) | 150.3 (70.8) | <0.001 |
| White blood cell count | 9.2 (4.1) | 9.1 (4.0) | 9.2 (4.1) | 0.11 |
| Hematocrit | 37.0 (5.9) | 37.5 (6.0) | 36.3 (5.8) | <0.001 |
| Left ventricular ejection fraction <40%, % | 29.8 | 32.6 | 26.2 | <0.001 |
| Geriatric Conditions, % | ||||
| Mobility disability | 39.2 | N/A | N/A | |
| Dementia | 9.7 | N/A | N/A | |
P Value for comparison between patients with no geriatric condition and those with at least 1 geriatric condition
31% of patients were missing data on mobility disability, while 37% were missing data on left ventricular ejection fraction; all other variables were missing < 5% of values
The 30-day mortality rate was 9.8% (n=6124). Variables shown in Table 1 were considered for inclusion in the final multivariable models, which had good performance (ROC = 0.73 for the 30 day model and ROC = 0.76 for the 5 year model). As shown in Table 2, mobility disability and dementia were among the top seven predictors of 30 day mortality in bivariate analyses, with among the largest absolute standardized estimates. In multivariable analyses, mobility disability and dementia were among the top six predictors of 30 day mortality, with 100% occurrence in the simulated regression models, and among the six largest absolute standardized estimates in the final model. The highest odds ratios were seen with serum creatinine (OR 1.43, 95% CI 1.39–1.48), cancer (OR 1.89, 95% CI 1.64–2.18), mobility disability (OR 1.96, 95% CI 1.81–2.12), and dementia (OR 1.86, 95% CI 1.73–2.01).
Table 2.
Short Term Mortality Models: Bivariate and Multivariable Results
| Bivariate Results |
Multivariable Results |
|||
|---|---|---|---|---|
| Standardized Estimate | Odds Ratio (95% CI) | Standardized Estimate | Odds Ratio (95% CI) | |
| Demographics | ||||
| Age (5 year increments) | 0.20 | 1.26 (1.23–1.28) | 0.15 | 1.19 (1.17–1.21) |
| Female | −0.05 | 0.82 (0.78–0.87) | −0.03 | 0.89 (0.84–0.95) |
| Non-white race | −0.09 | 0.61 (0.56–0.67) | −0.05 | 0.78 (0.70–0.86) |
| Admission physical exam findings | ||||
| Body temperature | −0.04 | 0.95 (0.94–0.96) | −0.03 | 0.97 (0.95–0.98) |
| Systolic blood pressure (10 mm Hg increments) | −0.35 | 0.82 (0.81–0.83) | −0.31 | 0.84 (0.83–0.85) |
| Heart rate | 0.03 | 1.00 (1.00–1.00) | 0.04 | 1.00 (1.00–1.01) |
| Respiratory rate | 0.08 | 1.02 (1.02–1.02) | 0.08 | 1.02 (1.02–1.03) |
| Admission laboratory findings | ||||
| Sodium | −0.11 | 0.96 (0.96–0.97) | −0.07 | 0.98 (0.97–0.98) |
| Potassium | 0.17 | 1.60 (1.54–1.66) | 0.07 | 1.20(1.16–1.25) |
| Creatinine | 0.17 | 1.51 (1.47–1.56) | 0.15 | 1.43(1.39–1.48) |
| White blood cell count | 0.13 | 1.06 (1.05–1.07) | 0.11 | 1.05 (1.04–1.05) |
| Hematocrit | −0.07 | 0.98 (0.97–0.98) | −0.01 | 1.00 (0.99–1.00) |
| Medical history | ||||
| Prior HF | 0.11 | 1.55 (1.45–1.65) | 0.04 | 1.17 (1.09–1.26) |
| Aortic stenosis | 0.05 | 1.43 (1.31–1.57) | 0.03 | 1.26 (1.15–1.39) |
| Coronary artery disease | −0.02 | 0.94 (0.90–1.00) | −0.01 | 0.95 (0.89–1.01) |
| Cerebrovascular disease | 0.06 | 1.32 (1.24–1.41) | 0.05 | 1.25 (1.16–1.33) |
| Cancer | 0.07 | 2.22 (1.95–2.53) | 0.05 | 1.89 (1.64–2.18) |
| Hypertension | −0.10 | 0.68 (0.64–0.71) | −0.03 | 0.88 (0.83–0.93) |
| Diabetes mellitus | −0.06 | 0.81 (0.77–0.86) | −0.02 | 0.92 (0.86–0.97) |
| COPD | 0.03 | 1.14 (1.08–1.20) | 0.03 | 1.13 (1.07–1.20) |
| Left Ventricular Ejection Fraction < 40 | 0.16 | 1.83 (1.71–1.95) | 0.06 | 1.26 (1.17–1.36) |
| Geriatric Conditions | ||||
| Mobility disability | 0.27 | 2.67 (2.48–2.88) | 0.18 | 1.96 (1.81–2.12) |
| Dementia | 0.15 | 2.45 (2.28–2.63) | 0.10 | 1.86 (1.73–2.01) |
At 5 years, the mortality rate was 74.7% (N = 46, 570). As shown in Table 3, mobility disability and dementia were among the top six predictors of 5 year mortality in bivariate analyses, with among the largest absolute standardized estimates. Similarly, in multivariable analyses, mobility disability and dementia were among the top seven predictors of 5 year mortality, with 100% occurrence in the simulated regression models and with among the top seven largest absolute standardized estimates in the final model. The highest odds ratios were seen with serum creatinine (OR 1.68, 95% CI 1.61–1.74), prior HF (OR 1.66, 95% CI 1.59–1.73), aortic stenosis (OR 1.69, 95% CI 1.55–1.85), cancer (OR 3.02, 95% CI 2.55–3.59), mobility disability (OR 1.78, 95% CI 1.70–1.87), and dementia (OR 2.01, 95% CI 1.84–2.19).
Table 3.
Long Term Mortality Models: Bivariate and Multivariable Results
| Bivariate |
Multivariable |
|||
|---|---|---|---|---|
| Standardized Estimate | Odds Ratio (95% CI) | Standardized Estimate | Odds Ratio (95% CI) | |
| Demographics | ||||
| Age (5 year increments) | 0.28 | 1.38 (1.36–1.40) | 0.27 | 1.37 (1.35–1.39) |
| Female | −0.06 | 0.80 (0.78–0.84) | −0.06 | 0.81 (0.78–0.85) |
| Non-white race | −0.06 | 0.72 (0.68–0.75) | −0.03 | 0.87 (0.82–0.92) |
| Admission physical exam findings | ||||
| Body temperature | −0.06 | 0.94 (0.92–0.95) | −0.03 | 0.96 (0.95–0.98) |
| Systolic blood pressure (10 mm Hg increments) | −0.19 | 0.90 (0.90–0.91) | −0.13 | 0.93 (0.92–0.95) |
| Heart rate | −0.02 | 1.00 (1.00–1.00) | 0.01 | 1.00 (1.00–1.00) |
| Respiratory rate | 0.05 | 1.02 (1.01–1.02) | 0.06 | 1.02 (1.01–1.02) |
| Admission laboratory findings | ||||
| Sodium | −0.08 | 0.97 (0.97–0.98) | −0.05 | 0.98 (0.98–0.99) |
| Potassium | 0.15 | 1.50 (1.46–1.55) | 0.04 | 1.12(1.09–1.16) |
| Creatinine | 0.30 | 2.03 (1.96–2.10) | 0.22 | 1.68(1.61–1.74) |
| White blood cell count | 0.05 | 1.02 (1.02–1.03) | 0.03 | 1.02 (1.01–1.02) |
| Hematocrit | −0.13 | 0.96 (0.96–0.96) | −0.06 | 0.98 (0.98–0.98) |
| Medical history | ||||
| Prior HF | 0.20 | 2.19 (2.11–2.28) | 0.13 | 1.66 (1.59–1.73) |
| Aortic stenosis | 0.10 | 1.97 (1.81–2.14) | 0.08 | 1.70 (1.55–1.85) |
| Coronary artery disease | 0.06 | 1.26 (1.21–1.30) | 0.03 | 1.13 (1.08–1.18) |
| Cerebrovascular disease | 0.09 | 1.53 (1.45–1.61) | 0.06 | 1.34 (1.27–1.41) |
| Cancer | 0.09 | 2.85 (2.42–3.35) | 0.09 | 3.02 (2.55–3.59) |
| Hypertension | −0.09 | 0.71 (0.69–0.74) | −0.07 | 0.78 (0.75–0.82) |
| Diabetes mellitus | 0.03 | 1.11 (1.06–1.14) | 0.07 | 1.28 (1.23–1.34) |
| COPD | 0.09 | 1.40 (1.35–1.46) | 0.11 | 1.51 (1.44–1.57) |
| Left Ventricular Ejection Fraction < 40 | 0.15 | 1.78 (1.70–1.86) | 0.08 | 1.40 (1.33–1.47) |
| Geriatric Conditions | ||||
| Mobility disability | 0.26 | 2.59 (2.48–2.71) | 0.16 | 1.78 (1.70–1.87) |
| Dementia | 0.18 | 2.92 (2.69–3.17) | 0.11 | 2.01 (1.84–2.19) |
At 30 days, the C-statistic before adding geriatric conditions to the multivariable model was 0.75, and this increased to 0.76 after the conditions were added. At five years, the C-statistic before adding geriatric conditions to the multivariable model was 0.75, and this increased to 0.76 after the conditions were added. For comparison, the C-statistic before adding serum creatinine to the 30-day multivariable model was 0.75, increasing to 0.76 in the full model. Of note, information about mobility was missing in 31% of records. At 30 days, the standardized estimate for “missing mobility” was 0.16, with OR (95% CI) of 1.88 (1.73–2.04). At five years, the standardized estimate for “missing mobility” was 0.05, with OR (95% CI) of 1.23 (1.17–1.29).
As shown in Table 4, at 30 days, most patients remained at the same level (i.e., quartile) of risk (along the shaded diagonal, from upper left to lower right) after information about mobility disability and dementia were included as additional variables in the prediction model containing variables shown in Tables 2 and 3 (age, sex, race, body temperature, systolic blood pressure, heart rate, respiratory rate, serum sodium, potassium, creatinine, white blood cell count, hematocrit, prior HF, aortic stenosis, coronary artery disease, cerebrovascular disease, cancer, hypertension, diabetes mellitus, COPD, and left ventricular ejection fraction). Considering those patients with an event, 767 were reclassified upward (above the diagonal) and 504 were reclassified downward (below the diagonal) after the geriatric conditions were added to the prediction model. For those without an event, 8469 were reclassified downwards and 8138 were reclassified upwards. Using previously reported methods (15), the NRI at 30 days was calculated to be 5.1%, P Value <0.001.
Table 4.
Reclassification of 30 day mortality risk with incorporation of geriatric conditions
| 30 Day probability of mortality with geriatric conditions | ||||||
|---|---|---|---|---|---|---|
| Q1 (<3.5%) | Q2 (3.5 – 6.8%) | Q3 (6.8 – 12.7%) | Q4 (≥ 12.7%) | TOTAL | ||
| 30 Day probability of mortality without geriatric conditions | Q1 (<3.9%) | E = 237 | E = 124 | E = 11 | E = 0 | E = 372 |
| N = 12108 | N = 2976 | N = 126 | N = 0 | N = 15210 | ||
| Q2 (3.9 – 7.2%) | E = 71 | E = 441 | E = 273 | E = 10 | E = 795 | |
| N = 3161 | N = 8583 | N = 2968 | N = 76 | N = 14788 | ||
| Q3 (7.2 – 12.7%) | E = 2 | E = 188 | E = 971 | E = 349 | E = 1510 | |
| N = 3 | N = 3246 | N = 8932 | N = 1892 | N = 14073 | ||
| Q4 (≥ 12.7%) | E = 0 | E = 0 | E = 243 | E = 3204 | E = 3447 | |
| N = 0 | N = 25 | N = 2059 | N = 10051 | N = 12135 | ||
| TOTAL | E = 310 | E = 753 | E = 1498 | E = 3563 | E = 6124 | |
| N = 15272 | N = 14830 | N = 14085 | N = 12019 | N = 56206 | ||
Q refers to quartile of risk
Estimates of mortality risk using traditional risk factors (vertical axis) and with traditional risk factors plus geriatric conditions (horizontal axis) are shown. (Traditional risk factors include age, sex, race, body temperature, systolic blood pressure, heart rate, respiratory rate, serum sodium, potassium, creatinine, white blood cell count, hematocrit, prior HF, aortic stenosis, coronary artery disease, cerebrovascular disease, cancer, hypertension, diabetes mellitus, COPD, and left ventricular ejection fraction).
Each cell includes number of events (E) and non-events (N).
Shaded cells represent patients who remained at the same level of risk in both models.
Similarly, Table 5 shows the reclassification of mortality risk at 5 years. Again, cells in the shaded diagonal represent those patients who remained in the same quartile of risk after the geriatric conditions were added to the model. For those with an event, 6150 were reclassified upward and 5883 were reclassified downward when the geriatric conditions were added to the prediction model. Considering those without an event, 1933 were reclassified downward and 1360 were reclassified upward. The NRI was 4.2%, P Value <0.001.
Table 5.
Reclassification of 5 year mortality risk with incorporation of geriatric conditions
| 5 year probability of mortality with geriatric conditions | ||||||
|---|---|---|---|---|---|---|
| Q1 (< 64.2%) | Q2 (64.2–78.8%) | Q3 (78.8 – 88.6%) | Q4 (> 88.6%) | TOTAL | ||
| 5 year mortality risk without geriatric conditions | Q1 (< 64.8%) | E = 6364 | E = 1426 | E = 108 | E = 0 | E = 7898 |
| N = 7054 | N = 603 | N = 27 | N = 0 | N = 7684 | ||
| Q2 (64.8 – 78.4%) | E = 1282 | E = 7483 | E = 2325 | E = 153 | E = 11243 | |
| N = 882 | N = 2958 | N = 482 | N = 18 | N = 4340 | ||
| Q3 (78.4 – 87.9%) | E = 0 | E = 2395 | E = 8509 | E = 2138 | E = 13042 | |
| N = 0 | N = 718 | N = 1593 | N = 230 | N = 2541 | ||
| Q4 (> 87.9%) | E = 0 | E = 0 | E = 2206 | E = 12181 | E = 14387 | |
| N = 0 | N = 0 | N = 333 | N = 862 | N = 1195 | ||
| TOTAL | E = 7646 | E = 11304 | E = 13148 | E = 14472 | E = 46570 | |
| N = 7936 | N = 4279 | N = 2435 | N = 1110 | N = 15760 | ||
Q refers to quartile of risk
Estimates of mortality risk using traditional risk factors (vertical axis) and with traditional risk factors plus geriatric conditions (horizontal axis) are shown. (Traditional risk factors include age, sex, race, body temperature, systolic blood pressure, heart rate, respiratory rate, serum sodium, potassium, creatinine, white blood cell count, hematocrit, prior HF, aortic stenosis, coronary artery disease, cerebrovascular disease, cancer, hypertension, diabetes mellitus, COPD, and left ventricular ejection fraction).
Each cell includes number of events (E) and non-events (N).
Shaded cells represent patients who remained at the same level of risk in both models.
DISCUSSION
Our study demonstrates that geriatric conditions, specifically mobility disability and dementia, are strongly and independently associated with short- and long-term mortality among older persons hospitalized with HF. The demographic and clinical characteristics of patients with at least one geriatric condition differed from those of patients without a geriatric condition, but the relationship between the geriatric conditions and mortality persisted even after adjustment for these factors. Moreover, reclassification methods that assessed geriatric conditions and mortality demonstrated significant improvement in risk classification when geriatric conditions were added to multivariable models containing traditional risk factors. The magnitude of this improvement (5.1% at 30 days and 4.2% at 5 years) was similar to that seen when information about levels of high-sensitivity C-Reactive Protein was added to models of cardiovascular disease risk (5.6%) (17).
These findings expand our current knowledge of prognosis after HF hospitalization. Previous studies have not examined the independent association of geriatric conditions with mortality after HF hospitalization. Geriatric conditions fall outside of the traditional medical model that underlies most HF research and clinical care, yet the significant reclassification effects of mobility disability and dementia underscore the importance of incorporating information about geriatric conditions into prognostic models and clinical decision making.
Our results demonstrate the importance of geriatric conditions for both short- and long-term mortality outcomes. The immediate post-hospitalization period (i.e., 30 days after hospitalization) is a time when care must be transitioned from the acute care hospital setting to the outpatient (or short or long term care facility) setting, often with changes in medications and numerous follow-up appointments. These issues enhance the relevance of cognitive and physical disabilities, which are also critical for ongoing, long-term HF self-care. For instance, although daily weighing is a cornerstone of HF self-care recommendations, patients with limited mobility may be unable to complete this assessment. Similarly, patients with dementia may have difficulty remembering to take medications and understanding sodium content of foods.
Although they may not be “curable”, geriatric conditions can be addressed in a variety of ways. For example, a course of physical therapy and exercise may improve mobility while increased caregiver and nursing support can be implemented to help patients with dementia adhere to medications. The benefits of interventions to address mobility and dementia are likely to extend beyond HF self-care. These interventions may also enhance patients’ abilities to avoid or cope with other medical problems, including infections, falls, and a number of chronic diseases.
This study used nationally representative data on over 62,000 patients hospitalized with HF followed for five years. However, our findings must be interpreted within the context of several potential limitations. First, because our sample included fee-for-service Medicare patients, our findings may not be applicable to elderly patients enrolled in Medicare managed care plans. While the extensive self-care required for optimal management of HF (including daily measurement of body weight and ongoing vigilance regarding dietary sodium intake) may enhance the relevance of mobility disability and dementia, it is possible that these geriatric conditions would hold the same prognostic value among patients with other chronic diseases, such as chronic obstructive pulmonary disease. Future work should assess the prognostic importance of these geriatric conditions in non-HF populations. Also, our analysis of geriatric conditions was limited to mobility disability and dementia. We chose mobility disability and dementia because of: 1) their prevalence among general populations of older patients; 2) their potential influence on HF self-care, including medication and dietary adherence, as well as daily self weighing and; 3) their robust assessment (i.e., through detailed medical record review) within the National Heart Care Project. Other potentially relevant geriatric conditions, including urinary incontinence, frailty, and falls, were considered but were not included because they were not assessed. Future research should examine the prognostic importance of a broader array of geriatric conditions. We do not have information about DNR status of patients, and it may be that these patients received less intensive, “life-prolonging” medical care. Of note, only 5% of hospitalized patients with severe heart failure have a do-not-resuscitate order. (18) Outcomes were available only for mortality; hence, we were unable to evaluate other important endpoints, including health status. Finally, information about mobility was missing in 31% of records. This likely reflects clinical practice, in which patients’ physical capabilities are often not assessed or recorded by clinicians. Of note, our results indicate that having missing information on mobility confers poorer prognosis.
Despite the aging of the population and the fact that HF primarily affects older persons in whom many complex conditions co-exist, current studies and guidelines have not incorporated routine assessment or management of geriatric conditions. There is increasing recognition of the need to move away from considering diseases as solitary entities, and to consider illness in the context of the patient (19). HF is the number one cause of hospitalization among persons age 65 years or older (20), placing tremendous burden on the health care system. Recent innovations in HF management have not yet translated into better outcomes in this population, underscoring the need for a new, more comprehensive approach to the care of this population (21). By addressing geriatric conditions, it may be possible to improve prognosis.
Acknowledgments
Dr. Chaudhry is supported by a Beeson Career Development Award (K23 AG030986). Drs. Chaudhry and Gill are supported by the Claude D. Pepper Older Americans Independence Center at Yale (P30-AG21342). Dr. Gill is the recipient of a Midcareer Investigator Award in Patient-Oriented Research (K24AG021507) from the National Institute on Aging.
Abbreviations List
- HF
heart failure
- ACE-I
angiotensin converting enzyme inhibitor
- SE
standardized estimate
- SD
standard deviation
- COPD
chronic obstructive pulmonary disease
- NRI
net reclassification improvement
- BUN
blood urea nitrogen
- ROC
receiver operating curve
- OR
odds ratio
- CI
confidence interval
Footnotes
Relationship with industry: Dr. Krumholz participates in a Scientific Advisory Board for United Healthcare. No disclosures for other authors.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
References
- 1.Fonarow GC, Peacock WF, Phillips CO, Givertz MM, Lopatin M. Admission B-type natriuretic peptide levels and in-hospital mortality in acute decompensated heart failure. J Am Coll Cardiol. 2007;49:1943–1950. doi: 10.1016/j.jacc.2007.02.037. [DOI] [PubMed] [Google Scholar]
- 2.Fonarow GC, Abraham WT, Albert NM, Stough WG, Gheorghiade M, Greenberg BH, O’Connor CM, Pieper K, Sun JL, Yancy CW, Young JB. Factors identified as precipitating hospital admissions for heart failure and clinical outcomes: findings from OPTIMIZE-HF. Arch Intern Med. 2008;168:847–854. doi: 10.1001/archinte.168.8.847. [DOI] [PubMed] [Google Scholar]
- 3.Januzzi JL, van Kimmenade R, Lainchbury J, Bayes-Genis A, Ordonez-Llanos J, Santalo-Bel M, Pinto YM, Richards M. NT-proBNP testing for diagnosis and short-term prognosis in acute destabilized heart failure: an international pooled analysis of 1256 patients: the International Collaborative of NT-proBNP Study. Eur Heart J. 2006;27:330–337. doi: 10.1093/eurheartj/ehi631. [DOI] [PubMed] [Google Scholar]
- 4.Lee DS, Austin PC, Rouleau JL, Liu PP, Naimark D, Tu JV. Predicting mortality among patients hospitalized for heart failure: derivation and validation of a clinical model. JAMA. 2003;290:2581–2587. doi: 10.1001/jama.290.19.2581. [DOI] [PubMed] [Google Scholar]
- 5.Kerzner R, Gage BF, Freedland KE, Rich MW. Predictors of mortality in younger and older patients with heart failure and preserved or reduced left ventricular ejection fraction. Am Heart J. 2003;146:286–290. doi: 10.1016/S0002-8703(03)00151-0. [DOI] [PubMed] [Google Scholar]
- 6.Jong P, Vowinckel E, Liu PP, Gong Y, Tu JV. Prognosis and determinants of survival in patients newly hospitalized for heart failure: a population-based study. Arch Intern Med. 2002;162:1689–1694. doi: 10.1001/archinte.162.15.1689. [DOI] [PubMed] [Google Scholar]
- 7.Blackledge HM, Tomlinson J, Squire IB. Prognosis for patients newly admitted to hospital with heart failure: survival trends in 12 220 index admissions in Leicestershire 1993–2001. Heart. 2003;89:615–620. doi: 10.1136/heart.89.6.615. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Kosiborod M, Smith GL, Radford MJ, Foody JM, Krumholz HM. The prognostic importance of anemia in patients with heart failure. Am J Med. 2003;114:112–119. doi: 10.1016/s0002-9343(02)01498-5. [DOI] [PubMed] [Google Scholar]
- 9.Fonarow GC, Adams KF, Jr, Abraham WT, Yancy CW, Boscardin WJ. Risk stratification for in-hospital mortality in acutely decompensated heart failure: classification and regression tree analysis. JAMA. 2005;293:572–580. doi: 10.1001/jama.293.5.572. [DOI] [PubMed] [Google Scholar]
- 10.Smith GL, Vaccarino V, Kosiborod M, Lichtman JH, Cheng S, Watnick SG, Krumholz HM. Worsening renal function: what is a clinically meaningful change in creatinine during hospitalization with heart failure? J Card Fail. 2003;9:13–25. doi: 10.1054/jcaf.2003.3. [DOI] [PubMed] [Google Scholar]
- 11.International Classification of Diseases, 9’th revision clinical modification. Los Angeles: Practice Management Information Corporation; 1998. [Google Scholar]
- 12.Masoudi FA, Ordin DL, Delaney RJ, Krumholz HM, Havranek EP. The National Heart Failure Project: a health care financing administration initiative to improve the care of Medicare beneficiaries with heart failure. Congest Heart Fail. 2000;6:337–339. doi: 10.1111/j.1527-5299.2000.80175.x. [DOI] [PubMed] [Google Scholar]
- 13.Department of Health and Human Services. Social Security Administration’s Death Master File. Washington D.C: 1999. NTS Order No. SUB-5251NQ. [Google Scholar]
- 14.Solomon SD, Dobson J, Pocock S, Skali H, McMurray JJ, Granger CB, Yusuf S, Swedberg K, Young JB, Michelson EL, Pfeffer MA. Influence of nonfatal hospitalization for heart failure on subsequent mortality in patients with chronic heart failure. Circulation. 2007;116:1482–1487. doi: 10.1161/CIRCULATIONAHA.107.696906. [DOI] [PubMed] [Google Scholar]
- 15.Pencina MJ, D’Agostino RB, Sr, D’Agostino RB, Jr, Vasan RS. Evaluating the added predictive ability of a new marker: from area under the ROC curve to reclassification and beyond. Stat Med. 2008;27:157–172. doi: 10.1002/sim.2929. discussion 207–112. [DOI] [PubMed] [Google Scholar]
- 16.Ridker PM, Buring JE, Rifai N, Cook NR. Development and validation of improved algorithms for the assessment of global cardiovascular risk in women: the Reynolds Risk Score. JAMA. 2007;297:611–619. doi: 10.1001/jama.297.6.611. [DOI] [PubMed] [Google Scholar]
- 17.Wilson P, Pencina M, Jacques P, Selhub J, D’Agostino R, O’Donnell C. C-Reactive Protein and Reclassification of Cardiovascular Risk in the Framingham Heart Study. Circulation Cardiovascular Quality Outcomes. 2008;1:92–97. doi: 10.1161/CIRCOUTCOMES.108.831198. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Ko DT, Alter DA, Austin PC, You JJ, Lee DS, Qiu F, Stukel TA, Tu JV. Life expectancy after an index hospitalization for patients with heart failure: a population-based study. Am Heart J. 2008;155:324–331. doi: 10.1016/j.ahj.2007.08.036. [DOI] [PubMed] [Google Scholar]
- 19.Tinetti ME, Fried T. The end of the disease era. Am J Med. 2004;116:179–185. doi: 10.1016/j.amjmed.2003.09.031. [DOI] [PubMed] [Google Scholar]
- 20.Koelling TM, Chen RS, Lubwama RN, L’Italien GJ, Eagle KA. The expanding national burden of heart failure in the United States: the influence of heart failure in women. Am Heart J. 2004;147:74–78. doi: 10.1016/j.ahj.2003.07.021. [DOI] [PubMed] [Google Scholar]
- 21.Kosiborod M, Lichtman JH, Heidenreich PA, Normand SL, Wang Y, Brass LM, Krumholz HM. National trends in outcomes among elderly patients with heart failure. Am J Med. 2006;119:616, e611–617. doi: 10.1016/j.amjmed.2005.11.019. [DOI] [PubMed] [Google Scholar]