- Email: [email protected]
Associations Among Processes and Outcomes of Care for Medicare Nursing Home Residents with Acute Heart Failure
Associations Among Processes and Outcomes of Care for Medicare Nursing Home Residents with Acute Heart Failure Evelyn Hutt, MD, Elizabeth Frederickson, BA, Mary Ecord, RN, and Andrew M. Kramer, MD Objective: To characterize Medicare skilled nursing facility (SNF) residents who become acutely ill with heart failure (HF) and assess the association between the outcomes of rehospitalization and mortality, and severity of the acute exacerbation, comorbidity, and processes of care. Design: SNF medical record review of Medicare patients who developed an acute exacerbation of heart failure (HF) during the 90 days following nursing home admission. Setting: A total of 58 SNFs in 5 states during 1994 and 1997. Participants: Patients with 156 episodes of acute HF among 4693 random Medicare nursing home admissions. Measurements: Demographic variables, symptoms, signs, comorbidity, nursing home characteristics, nurse staffing ratios, and processes of care were compared between acute HF subjects transferred to hospital and those not transferred; and between subjects who died
Heart failure (HF) is the leading primary diagnosis among hospitalized older adults; 12% of men and 21% of women with heart failure are discharged from hospital to nursing home.1 Of the admissions to skilled nursing facilities from acute care hospitals, 18% have a HF diagnosis.2 Rehospitalization occurs frequently,3,4 is often for relatively low-risk symptoms5 and is a major contributor to the cost of managing this disease. The increasing prevalence of HF, with its associated morbidity and cost, has been described as an epidemic.6 In spite of these facts, information about the treatment of Division of Health Care Policy and Research, University of Colorado Health Sciences Center. Address correspondence to Evelyn Hutt, MD, Division of Health Care Policy and Research, University of Colorado Health Sciences Center, 3570 East 12th Avenue, Suite 300, Denver, CO 80206. Email: [email protected].
Copyright ©2003 American Medical Directors Association DOI: 10.1097/01.JAM.0000073964.19754.C0 ORIGINAL STUDIES
within 30 days of an acute exacerbation and those who survived. Results: After adjusting for age, disease severity, and comorbidity, residents whose change in condition was evaluated during the night shift were more likely to be hospitalized (OR 4.20, 95%CI 1.01–17.50). Residents who were prescribed an angiotensin-converting enzyme inhibitor or who received an order for skilled nursing observation more often than once a shift were 1/3 as likely to die as those who did not (OR 0.303, 95%CI 0.11– 0.82), after adjusting for hypotension, delirium, do not resuscitate orders, and prior hospital length of stay. Conclusion: For residents who develop an acute exacerbation of HF during a SNF stay, there is an association between attributes of nursing home care and the outcomes of rehospitalization and mortality. (J Am Med Dir Assoc 2003; 4: 195–199) Keywords: congestive heart failure; long-term care; nursing home
HF in nursing homes is limited. Studies to date have focused on the use and withdrawal of digoxin,7–10 and more recently, on the underuse11,12 and benefit2 of angiotensin-converting enzyme inhibitors (ACE-I) in this setting. No study to date has reported on acute HF presentation and treatment in nursing homes. The objectives of this study, therefore, were to characterize Medicare skilled nursing facility (SNF) residents with an incident episode of acute HF, and describe the care they received. We hypothesized that rehospitalization and mortality each would be associated with severity of the acute exacerbation, comorbidity, and processes of care. METHODS Setting Skilled nursing facilities were recruited from 5 states that participated in the Multistate Nursing Home Case Mix and Quality Demonstration Evaluation as previously deHutt et al 195
scribed.13,14 In each state, the facilities were randomly selected such that, when pooled from all 5 states, the sample would include a representative proportion of profit versus nonprofit and hospital-based versus freestanding facilities. Fifty-eight skilled nursing facilities, evenly distributed among Kansas, Maine, South Dakota, Texas, and New York were included in the sample. Residents A random sample of Medicare SNF admissions during 1994 and 1997 (the baseline and intervention years of the Demonstration) was drawn from the Medicare Provider Analysis and Review (MEDPAR) file. To be eligible, a resident was required to be in the first 90 days of his or her SNF stay, and to have had a hospitalization of at least 3 days within 30 days before the SNF admission. The first 90 days of a stay are a critical time for nursing home residents. Nearly 40% of rehospitalizations from nursing home occur within 90 days of admission,15 and residents are very unlikely to return to more independent living after 90 days in a SNF.16 The number of residents selected from each nursing home ranged from 15 to 40 based on the number of admissions to that SNF. SNF charts of 4693 random Medicare nursing home admissions to 58 skilled nursing facilities were reviewed retrospectively. Using the definition outlined below, and excluding residents whose onset of acute illness was less than 4 days from admission, 156 episodes of HF exacerbation were identified for analysis, 88 in 1994 and 68 in 1997. Data collection and analysis were institutional review board exempt because the facilities and residents were enrolled in a Centers for Medicare and Medicaid Services (CMS, then called HCFA) payment demonstration, and the results were pertinent to the demonstration evaluation. Confidentiality of the subjects was assured. Measures and Data Data Collection and Definitions Using a systematic instrument and protocol to review records, a trained nurse researcher identified incident exacerbation or a first episode of HF in the nursing home by presence of either documented pulmonary edema on chest X-ray in the nursing home or a hospital discharge diagnosis of HF, and two or more of the following clinical signs/symptoms: increased dyspnea on exertion, paroxysmal nocturnal dyspnea, orthopnea, massive edema, or altered sensorium. A single nurse data collector for each state conducted the chart review on site. A research nurse, analyst, and the project manager used actual SNF records to train the data collectors over a 2- to 3-day period at the University of Colorado. The items used in these analyses include only those with a Cohen’s kappa on both interrater and test/retest reliability of 0.70 or greater. Detailed information including demographics, advance care directives, vital signs, comorbidity and disease severity, use of an ACE-I, notification and response of physician, and whether and to where the resident was transferred were collected for each event. Advance care directives included “do 196 Hutt et al
not hospitalize” orders, “do not resuscitate (DNR)” or “full resuscitation” orders, or more specific restrictions such as those related to intubation or feeding tubes. To obtain information on comorbid disease, the following chronic diagnoses were obtained from the MEDPAR data using any of the 10 available diagnoses on the nursing home claims: chronic respiratory diseases (COPD), dementia, depression, Parkinson’s disease, quadriplegia, paraplegia, hemiplegia, arthritis, stroke, and hip and pelvis fractures. Facility characteristics were obtained from the Online Survey Certification and Reporting System (OSCAR) data and disaggregated to the resident level. Two outcomes during the 60 days from symptom onset were studied: transfer to hospital for treatment in the emergency department or admission, and mortality. Symptom onset was defined as the first mention in the chart of a significant change such as markedly abnormal vital signs, severe dyspnea, sudden decline in function, falls, acute confusion, decreased level of consciousness, new onset peripheral edema, or markedly abnormal laboratory results. Transfer to the hospital or emergency department was defined as the nurses’ notes or physician order stating that the transfer was for treatment of the acute event. Date of death was obtained from MEDPAR data and confirmed by the Social Security Death Index.17 Analysis We compared acutely ill residents who were transferred to the hospital for admission or treatment in the emergency department with acutely ill residents who were not transferred, on demographic variables, symptoms, signs, comorbidity, nursing home characteristics, nurse staffing ratios, time of symptom onset and processes of care. The Fisher exact test was used for dichotomous and Student t test for continuous independent variables. Residents with orders not to be rehospitalized were excluded from this analysis. Using logistic regression in SAS, we fit a multivariate model predicting rehospitalization. The following risk factors were selected based on the two-group comparisons, review of correlations between characteristics and rehospitalization, clinical considerations, and step-wise procedures: age, hypotension (systolic blood pressure ⬍90mmHg), dementia diagnosis, residence in a nursing home before the current Medicare SNF admission, length of preceding hospitalization, and symptom onset during the night shift (defined as 11:00 PM to 7:00 AM, with allowance made for how the facility defined its shifts). Using the riskadjusted model, we tested the association between process variables and rehospitalization, controlling for risk factors. We also compared residents acutely ill with an exacerbation of HF who had died by 60 days after symptom onset with those alive at 60 days, on demographic variables, symptoms, signs, comorbidity, nursing home characteristics, nurse staffing ratios, time of symptom onset and processes of care. The Fisher exact test was used for dichotomous and Student t test for continuous independent variables. Using logistic regression in SAS, we fit a multivariate model predicting mortality at 60 days. The following risk factors were selected based on the two-group comparisons, review of correlations between characteristics and mortality, clinical considerations, and JAMDA – July/August 2003
Table 1. Characteristics of Patients Who Became Acutely Ill with Congestive Heart Failure (N ⫽ 156) Characteristic Demographics Female White Average age (years) Previously resided in nursing home Average hospital length of stay prior to NH (days) Has do not resuscitate order (DNR) Has do not hospitalize order Signs and Symptoms Systolic blood pressure ⬍90 Pulse 具60典 100 beats per minute Altered sensorium Oxygen saturation ⬍90% Massive edema Comorbidity Depression Chronic obstructive pulmonary disease Known coronary artery disease Dementia Facility Characteristics Nonprofit Hospital based Urban Average ratio of licensed nurse hours/res/day Outcomes Transferred to hospital or emergency room Alive 30 days after symptom onset
% (n)* 70.0 (109) 90.4 (141) 83.2 ⫾ 8.99 34.4 (52) 20.3 ⫾ 23.3 66.7 (104) 1.3 (2) 18.9 (28) 44.5 (65) 35.3 (55) 68.5 (37) 19.9 (31) 8.5 (13) 28.8 (44) 24.6 (49) 15.0 (23) 54.3 (83) 17.7 (27) 62.8 (96) 0.23 ⫾ 0.197 56.6 (81) 78.9 (123)
NH ⫽ nursing home. * Percent, unless otherwise indicated. 4693 charts were reviewed identifying 156 cases of acute HF (3.3%).
step-wise procedures: hypotension, altered sensorium, order for skilled nursing observation more frequently than once per 8-hour shift, length of the preceding hospitalization, and prescription of an ACE-I. Using the risk-adjusted model, we tested the association between process variables and mortality, controlling for risk factors.
Table 2. Transfer for Rehospitalization or Emergency Department Evaluation for Heart Failure When the Resident’s Change of Condition was Evaluated During the Night Shift (n ⫽ 119) Characteristic
Odds ratio
95% CI for odds ratio
Age Hypotension Previous residence in NH Dementia Night shift
0.96 0.32 3.70
(0.91, 1.01) (0.11, 0.92) (1.44, 9.54)
0.32 4.20
(0.10, 1.04) (1.01, 17.50)
NH ⫽ nursing home. Chi square ⫽ 16.7. c statistic ⫽ .75.
groups were treated with an ACE-I); the episodes were pooled for the current analysis. Rehospitalized residents were significantly younger (81.6 vs. 85.1, P ⫽ 0.007), less likely to be hypotensive with a systolic blood pressure ⬍90 mmHg (10% vs. 29%, P ⫽ 0.003), and less likely to be demented (10% vs. 22%, P ⫽ 0.04), in bivariate comparisons. In multivariate analysis (Table 2), the sample size decreased to 119, because of missing data on the shift when symptoms were first noted. Previous residence in a nursing home was a significant positive predictor of rehospitalization, whereas hypotension and dementia were negative predictors. After adjusting for these variables and age, the odds of being rehospitalized increased more than four-fold if symptoms were evaluated at night (OR 4.20, 95%CI 1.01–17.50). Patients who died were more likely to have an altered sensorium (64% vs. 28%, P ⬍0.001) and to be hypotensive (52% vs. 10%, P ⬍0.001) in bivariate comparisons. They did not, however, have greater comorbidity. In multivariate analysis, shown in Table 3, the sample size decreased to 137, because of missing data on systolic blood pressure and on whether or not frequent skilled nursing observation had been ordered. Hypotension and an altered sensorium were predictive of death. After adjusting for these signs, treatment with
RESULTS Patients who became acutely ill with HF during the first 90 days of their nursing home admission were typically women with lengthy prior hospital stays (average 20.3 days), considerable comorbidity, and DNR orders (Table 1). They were similar in age and gender to residents without HF, but were more likely to be black than residents without HF (8% vs. 4%), had longer preceding hospital lengths of stay (20 vs. 15 days), and were more likely to have resided in a SNF before the incident admission (34% vs. 26%). The most prevalent presenting signs were hypoxemia, bradycardia or tachycardia, and altered sensorium. Because residents with an acute HF exacerbation in 1994 and 1997 were similar in age, number, and type of comorbidity; percentage who presented with unstable vital signs; and processes of care (eg, about 40% of both ORIGINAL STUDIES
Table 3. Mortality at 60 Days When the Resident was Treated with an ACE-I or Received Skilled Nursing Observation More Frequently than Once per Shift (n ⫽ 137) Characteristic
Odds Ratio
95% CI for Odds Ratio
Hypotension Altered sensorium Order not to resuscitate Preceding hospital length of stay Treatment with ACE Inhibitor Frequent skilled nursing observation
9.30 3.09 2.24 0.98 0.31 0.30
(3.20, 27.02) (1.18, 8.12) (0.78, 6.43) (0.95, 1.00) (0.11, 0.89) (0.11, 0.82)
ACE-I ⫽ angiotensin-converting enzyme inhibitor. Chi square ⫽ 28.1. c statistic ⫽ .85. Hutt et al 197
an ACE-I and skilled nursing observation more often than once per shift in the nursing facility decreased the odds of dying by 70%. DISCUSSION This study is the first to report on the clinical and demographic characteristics of acute HF exacerbations in recently admitted nursing home residents. At presentation, these residents were very ill: two-thirds were hypoxic, nearly half had abnormal pulse rates; one-third were delirious. More than half were transferred to the hospital or emergency department for treatment; one-fifth died. As the number and proportion of nursing home residents with HF increases over the coming years,6 it will become increasingly important to understand what processes of care in SNFs are effective in decreasing rehospitalization and preserving life and function. A unifying theory of the clinical and demographic characteristics predicting hospitalization of nursing home residents for acute congestive heart failure cannot be derived from the current analysis. Rehospitalized subjects were less likely to have dementia, and more likely to have been living in a nursing home before the incident admission, suggesting that chronic physical frailty without dementia predisposes to rehospitalization. The acute HF episodes that resulted in rehospitalization, however, were less severe than those managed in the nursing facility. The incongruence of these findings may reflect our lack of data about the functional and hemodynamic characteristics of HF in nursing home residents. Unfortunately, physicians are currently making rehospitalization decisions for these patients in a knowledge vacuum. There are no published studies that consider which HF exacerbations can be managed safely and effectively in nursing facilities, although at least one study suggests that many hospitalizations for HF are for low-risk symptoms.5 After adjusting for these risk factors, we found that symptom presentation and evaluation at night increased the odds of rehospitalization four-fold. Fried et al.18 found a similar increase in odds of rehospitalization for pneumonia when residents required evaluation for an acute illness in the evening. Staffing, and thus the nursing facility’s ability to manage acutely ill residents, is markedly decreased at night. This implies that what transpires in the nursing home can have a marked effect on important outcomes, such as rehospitalization. The clinical and demographic characteristics of residents who died within 60 days after an exacerbation of HF were not surprising. Hypotension and delirium predicted death. Like Gambassi and colleagues,2 we found a significant difference in the use of ACE-I between NH residents with HF who died and those who survived. After adjusting for these variables, residents who received skilled nursing observation more frequently than once per shift were one-third as likely to die as those who did not. This suggests that processes of care in the nursing home for residents with an acute HF exacerbation may impact their survival. A similar association between processes of care in the nursing home and survival from an episode of nursing home acquired pneumonia was recently described.19 198 Hutt et al
Our study has several important limitations. Most importantly, we had no assessment of the subjects’ functional status, either by an activities of daily living (ADL) score or by New York Heart Association (NYHA) classification. Future studies of the outcome of acute HF exacerbations in nursing homes will require this information, because it has such a large effect on hospitalization and mortality.20 Studies of HF in nursing home residents are necessarily limited by the fact that few patients in this setting have current echocardiograms defining their left-ventricular function or the presence of diastolic dysfunction. Second, more detailed information about the kinds of treatment ordered in the nursing facility would have been useful. How many patients were treated with new or increased potency diuretics? How many patients had monitoring of their hydration status by daily weights or recorded input and output? How many patients had a proximate change in the use of ACE-I, anti-arrhythmic medications, digoxin, or beta-blockers? Data like these might allow future studies to begin defining which exacerbations can be safely managed in nursing homes, and which are best treated in an acute care hospital. Finally, the general application of this study is limited by the fact that the subjects came from facilities in the five states involved in the Multistate Nursing Home Case Mix and Quality Demonstration, and had been admitted to nursing facilities only a short time before becoming acutely ill. Clinical characteristics and management of HF exacerbations in the long-term care segment of the nursing home population are likely to be quite different from those newly admitted under Medicare payment. CONCLUSION In spite of its limitations, this study suggests that a relationship exists between processes of care, such as use of an ACE-I, time of day when a resident is evaluated, and frequency of skilled nursing observation, and the outcomes of rehospitalization and mortality. Further study is needed to better understand this relationship and to begin defining which processes of care for heart failure in nursing homes will be able to decrease rehospitalization rates, stabilize function, and decrease mortality. REFERENCES 1. Haldeman GA, Croft JB, Giles WH, Rashidee A. Hospitalization of patients with heart failure: National Hospital Discharge Survey, 1985 to 1995. Am Heart J 1999;137:352–360. 2. Gambassi G, Lapane KL, Sgadari A, et al. Effects of angiotensin-converting enzyme inhibitors and digoxin on health outcomes of very old patients with heart failure. Arch Intern Med 2000;160:53– 60. 3. Hutt E, Ecord M, Eilertsen TB, et al. Precipitants of emergency room visits and acute hospitalization among short-stay Medicare nursing home residents. J Am Geriatr Soc 2002;50:223–229. 4. Krumholz HM, Parent EM, Tu N, et al. Readmission after hospitalization for congestive heart failure among Medicare beneficiaries. Arch Intern Med 1997;157:99 –104. 5. Butler J, Hanumanthu S, Chomsky D, Wilson J. Frequency of low-risk hospital admissions for heart failure. Am J Cardiol 1998;81:41– 44. 6. US Department of Health. Congestive Heart Failure in the United States: A New Epidemic. NHLBI-NIH Data Fact Sheet, 1– 6. 1996. Washington DC. JAMDA – July/August 2003
7. Carter BL, Small RE, Garnett WR. Monitoring digoxin therapy in two long-term facilities. J Am Geriatr Soc 1981;29:263–268. 8. Wilkins CE, Khurana MS. Digitalis withdrawal in elderly nursing home patients. J Am Geriatr Soc 1985;33:850 – 851. 9. Macarthur C. Withdrawal of maintenance digoxin from institutionalized elderly. Postgrad Med J 1990;66:940 –942. 10. Forman DE, Coletta D, Kenny D, et al. Clinical issues related to discontinuing digoxin therapy in elderly nursing home patients. Arch Intern Med 1991;151:2194 –2198. 11. Forman DE, Chander RB, Lapane KL, et al. Evaluating the use of angiotensin-converting enzyme inhibitors for older nursing home residents with chronic heart failure. J Am Geriatr Soc 1998;46:1550 –1554. 12. Ruths S, Straand J, Nygaard HA, Hodneland F. Drug treatment of heart failure–Do nursing-home residents deserve better? Scand J Prim Health Care 2000;18:226 –231. 13. Burke RE, Feldman JI, Reilly K, et al. Multistate Nursing Home Case Mix and Quality Demonstration: Final Report. Baltimore: Health Care Financing Administration, 1994.
ORIGINAL STUDIES
14. Kramer AM, Frederickson EB, Ecord MK, et al. Nursing Home Case Mix and Quality Demonstration Evaluation, Final Report, Vol. 2: Effects on Outcomes and Quality. Denver: University of Colorado Health Sciences Center, 2001. 15. Barker WH, Zimmer JG, Hall WJ. Rates, patterns, causes, and costs of hospitalization of nursing home resident: A population-based study. Am J Public Health 1994;84:1615–1620. 16. Hutt E, Frederickson E, Eilertsen T, et al. Prospective payment for nursing homes increased therapy provision without community discharge rates. J Am Geriatr Soc 2001;49:1071–1079. 17. Ancestory.com. MyFamily.com Inc. 2001. Accessed May 15, 2003. 18. Fried TR, Gillick MR, Lipsitz LA. Whether to transfer? Factors associated with hospitalization and outcome of elderly long-term care patients with pneumonia. J Gen Intern Med 1995;10:246 –250. 19. Hutt E, Frederickson EB, Ecord M, Kramer AM. Processes of care predict survival following nursing home acquired pneumonia. J Clin Outcome Manage 2002;9:249 –256. 20. Reiley P, Howard E. Predicting hospital length of stay in elderly patients with congestive heart failure. Nurs Econ 1995;13:210 –216.
Hutt et al 199
Heart failure (HF) is the leading primary diagnosis among hospitalized older adults; 12% of men and 21% of women with heart failure are discharged from hospital to nursing home.1 Of the admissions to skilled nursing facilities from acute care hospitals, 18% have a HF diagnosis.2 Rehospitalization occurs frequently,3,4 is often for relatively low-risk symptoms5 and is a major contributor to the cost of managing this disease. The increasing prevalence of HF, with its associated morbidity and cost, has been described as an epidemic.6 In spite of these facts, information about the treatment of Division of Health Care Policy and Research, University of Colorado Health Sciences Center. Address correspondence to Evelyn Hutt, MD, Division of Health Care Policy and Research, University of Colorado Health Sciences Center, 3570 East 12th Avenue, Suite 300, Denver, CO 80206. Email: [email protected].
Copyright ©2003 American Medical Directors Association DOI: 10.1097/01.JAM.0000073964.19754.C0 ORIGINAL STUDIES
within 30 days of an acute exacerbation and those who survived. Results: After adjusting for age, disease severity, and comorbidity, residents whose change in condition was evaluated during the night shift were more likely to be hospitalized (OR 4.20, 95%CI 1.01–17.50). Residents who were prescribed an angiotensin-converting enzyme inhibitor or who received an order for skilled nursing observation more often than once a shift were 1/3 as likely to die as those who did not (OR 0.303, 95%CI 0.11– 0.82), after adjusting for hypotension, delirium, do not resuscitate orders, and prior hospital length of stay. Conclusion: For residents who develop an acute exacerbation of HF during a SNF stay, there is an association between attributes of nursing home care and the outcomes of rehospitalization and mortality. (J Am Med Dir Assoc 2003; 4: 195–199) Keywords: congestive heart failure; long-term care; nursing home
HF in nursing homes is limited. Studies to date have focused on the use and withdrawal of digoxin,7–10 and more recently, on the underuse11,12 and benefit2 of angiotensin-converting enzyme inhibitors (ACE-I) in this setting. No study to date has reported on acute HF presentation and treatment in nursing homes. The objectives of this study, therefore, were to characterize Medicare skilled nursing facility (SNF) residents with an incident episode of acute HF, and describe the care they received. We hypothesized that rehospitalization and mortality each would be associated with severity of the acute exacerbation, comorbidity, and processes of care. METHODS Setting Skilled nursing facilities were recruited from 5 states that participated in the Multistate Nursing Home Case Mix and Quality Demonstration Evaluation as previously deHutt et al 195
scribed.13,14 In each state, the facilities were randomly selected such that, when pooled from all 5 states, the sample would include a representative proportion of profit versus nonprofit and hospital-based versus freestanding facilities. Fifty-eight skilled nursing facilities, evenly distributed among Kansas, Maine, South Dakota, Texas, and New York were included in the sample. Residents A random sample of Medicare SNF admissions during 1994 and 1997 (the baseline and intervention years of the Demonstration) was drawn from the Medicare Provider Analysis and Review (MEDPAR) file. To be eligible, a resident was required to be in the first 90 days of his or her SNF stay, and to have had a hospitalization of at least 3 days within 30 days before the SNF admission. The first 90 days of a stay are a critical time for nursing home residents. Nearly 40% of rehospitalizations from nursing home occur within 90 days of admission,15 and residents are very unlikely to return to more independent living after 90 days in a SNF.16 The number of residents selected from each nursing home ranged from 15 to 40 based on the number of admissions to that SNF. SNF charts of 4693 random Medicare nursing home admissions to 58 skilled nursing facilities were reviewed retrospectively. Using the definition outlined below, and excluding residents whose onset of acute illness was less than 4 days from admission, 156 episodes of HF exacerbation were identified for analysis, 88 in 1994 and 68 in 1997. Data collection and analysis were institutional review board exempt because the facilities and residents were enrolled in a Centers for Medicare and Medicaid Services (CMS, then called HCFA) payment demonstration, and the results were pertinent to the demonstration evaluation. Confidentiality of the subjects was assured. Measures and Data Data Collection and Definitions Using a systematic instrument and protocol to review records, a trained nurse researcher identified incident exacerbation or a first episode of HF in the nursing home by presence of either documented pulmonary edema on chest X-ray in the nursing home or a hospital discharge diagnosis of HF, and two or more of the following clinical signs/symptoms: increased dyspnea on exertion, paroxysmal nocturnal dyspnea, orthopnea, massive edema, or altered sensorium. A single nurse data collector for each state conducted the chart review on site. A research nurse, analyst, and the project manager used actual SNF records to train the data collectors over a 2- to 3-day period at the University of Colorado. The items used in these analyses include only those with a Cohen’s kappa on both interrater and test/retest reliability of 0.70 or greater. Detailed information including demographics, advance care directives, vital signs, comorbidity and disease severity, use of an ACE-I, notification and response of physician, and whether and to where the resident was transferred were collected for each event. Advance care directives included “do 196 Hutt et al
not hospitalize” orders, “do not resuscitate (DNR)” or “full resuscitation” orders, or more specific restrictions such as those related to intubation or feeding tubes. To obtain information on comorbid disease, the following chronic diagnoses were obtained from the MEDPAR data using any of the 10 available diagnoses on the nursing home claims: chronic respiratory diseases (COPD), dementia, depression, Parkinson’s disease, quadriplegia, paraplegia, hemiplegia, arthritis, stroke, and hip and pelvis fractures. Facility characteristics were obtained from the Online Survey Certification and Reporting System (OSCAR) data and disaggregated to the resident level. Two outcomes during the 60 days from symptom onset were studied: transfer to hospital for treatment in the emergency department or admission, and mortality. Symptom onset was defined as the first mention in the chart of a significant change such as markedly abnormal vital signs, severe dyspnea, sudden decline in function, falls, acute confusion, decreased level of consciousness, new onset peripheral edema, or markedly abnormal laboratory results. Transfer to the hospital or emergency department was defined as the nurses’ notes or physician order stating that the transfer was for treatment of the acute event. Date of death was obtained from MEDPAR data and confirmed by the Social Security Death Index.17 Analysis We compared acutely ill residents who were transferred to the hospital for admission or treatment in the emergency department with acutely ill residents who were not transferred, on demographic variables, symptoms, signs, comorbidity, nursing home characteristics, nurse staffing ratios, time of symptom onset and processes of care. The Fisher exact test was used for dichotomous and Student t test for continuous independent variables. Residents with orders not to be rehospitalized were excluded from this analysis. Using logistic regression in SAS, we fit a multivariate model predicting rehospitalization. The following risk factors were selected based on the two-group comparisons, review of correlations between characteristics and rehospitalization, clinical considerations, and step-wise procedures: age, hypotension (systolic blood pressure ⬍90mmHg), dementia diagnosis, residence in a nursing home before the current Medicare SNF admission, length of preceding hospitalization, and symptom onset during the night shift (defined as 11:00 PM to 7:00 AM, with allowance made for how the facility defined its shifts). Using the riskadjusted model, we tested the association between process variables and rehospitalization, controlling for risk factors. We also compared residents acutely ill with an exacerbation of HF who had died by 60 days after symptom onset with those alive at 60 days, on demographic variables, symptoms, signs, comorbidity, nursing home characteristics, nurse staffing ratios, time of symptom onset and processes of care. The Fisher exact test was used for dichotomous and Student t test for continuous independent variables. Using logistic regression in SAS, we fit a multivariate model predicting mortality at 60 days. The following risk factors were selected based on the two-group comparisons, review of correlations between characteristics and mortality, clinical considerations, and JAMDA – July/August 2003
Table 1. Characteristics of Patients Who Became Acutely Ill with Congestive Heart Failure (N ⫽ 156) Characteristic Demographics Female White Average age (years) Previously resided in nursing home Average hospital length of stay prior to NH (days) Has do not resuscitate order (DNR) Has do not hospitalize order Signs and Symptoms Systolic blood pressure ⬍90 Pulse 具60典 100 beats per minute Altered sensorium Oxygen saturation ⬍90% Massive edema Comorbidity Depression Chronic obstructive pulmonary disease Known coronary artery disease Dementia Facility Characteristics Nonprofit Hospital based Urban Average ratio of licensed nurse hours/res/day Outcomes Transferred to hospital or emergency room Alive 30 days after symptom onset
% (n)* 70.0 (109) 90.4 (141) 83.2 ⫾ 8.99 34.4 (52) 20.3 ⫾ 23.3 66.7 (104) 1.3 (2) 18.9 (28) 44.5 (65) 35.3 (55) 68.5 (37) 19.9 (31) 8.5 (13) 28.8 (44) 24.6 (49) 15.0 (23) 54.3 (83) 17.7 (27) 62.8 (96) 0.23 ⫾ 0.197 56.6 (81) 78.9 (123)
NH ⫽ nursing home. * Percent, unless otherwise indicated. 4693 charts were reviewed identifying 156 cases of acute HF (3.3%).
step-wise procedures: hypotension, altered sensorium, order for skilled nursing observation more frequently than once per 8-hour shift, length of the preceding hospitalization, and prescription of an ACE-I. Using the risk-adjusted model, we tested the association between process variables and mortality, controlling for risk factors.
Table 2. Transfer for Rehospitalization or Emergency Department Evaluation for Heart Failure When the Resident’s Change of Condition was Evaluated During the Night Shift (n ⫽ 119) Characteristic
Odds ratio
95% CI for odds ratio
Age Hypotension Previous residence in NH Dementia Night shift
0.96 0.32 3.70
(0.91, 1.01) (0.11, 0.92) (1.44, 9.54)
0.32 4.20
(0.10, 1.04) (1.01, 17.50)
NH ⫽ nursing home. Chi square ⫽ 16.7. c statistic ⫽ .75.
groups were treated with an ACE-I); the episodes were pooled for the current analysis. Rehospitalized residents were significantly younger (81.6 vs. 85.1, P ⫽ 0.007), less likely to be hypotensive with a systolic blood pressure ⬍90 mmHg (10% vs. 29%, P ⫽ 0.003), and less likely to be demented (10% vs. 22%, P ⫽ 0.04), in bivariate comparisons. In multivariate analysis (Table 2), the sample size decreased to 119, because of missing data on the shift when symptoms were first noted. Previous residence in a nursing home was a significant positive predictor of rehospitalization, whereas hypotension and dementia were negative predictors. After adjusting for these variables and age, the odds of being rehospitalized increased more than four-fold if symptoms were evaluated at night (OR 4.20, 95%CI 1.01–17.50). Patients who died were more likely to have an altered sensorium (64% vs. 28%, P ⬍0.001) and to be hypotensive (52% vs. 10%, P ⬍0.001) in bivariate comparisons. They did not, however, have greater comorbidity. In multivariate analysis, shown in Table 3, the sample size decreased to 137, because of missing data on systolic blood pressure and on whether or not frequent skilled nursing observation had been ordered. Hypotension and an altered sensorium were predictive of death. After adjusting for these signs, treatment with
RESULTS Patients who became acutely ill with HF during the first 90 days of their nursing home admission were typically women with lengthy prior hospital stays (average 20.3 days), considerable comorbidity, and DNR orders (Table 1). They were similar in age and gender to residents without HF, but were more likely to be black than residents without HF (8% vs. 4%), had longer preceding hospital lengths of stay (20 vs. 15 days), and were more likely to have resided in a SNF before the incident admission (34% vs. 26%). The most prevalent presenting signs were hypoxemia, bradycardia or tachycardia, and altered sensorium. Because residents with an acute HF exacerbation in 1994 and 1997 were similar in age, number, and type of comorbidity; percentage who presented with unstable vital signs; and processes of care (eg, about 40% of both ORIGINAL STUDIES
Table 3. Mortality at 60 Days When the Resident was Treated with an ACE-I or Received Skilled Nursing Observation More Frequently than Once per Shift (n ⫽ 137) Characteristic
Odds Ratio
95% CI for Odds Ratio
Hypotension Altered sensorium Order not to resuscitate Preceding hospital length of stay Treatment with ACE Inhibitor Frequent skilled nursing observation
9.30 3.09 2.24 0.98 0.31 0.30
(3.20, 27.02) (1.18, 8.12) (0.78, 6.43) (0.95, 1.00) (0.11, 0.89) (0.11, 0.82)
ACE-I ⫽ angiotensin-converting enzyme inhibitor. Chi square ⫽ 28.1. c statistic ⫽ .85. Hutt et al 197
an ACE-I and skilled nursing observation more often than once per shift in the nursing facility decreased the odds of dying by 70%. DISCUSSION This study is the first to report on the clinical and demographic characteristics of acute HF exacerbations in recently admitted nursing home residents. At presentation, these residents were very ill: two-thirds were hypoxic, nearly half had abnormal pulse rates; one-third were delirious. More than half were transferred to the hospital or emergency department for treatment; one-fifth died. As the number and proportion of nursing home residents with HF increases over the coming years,6 it will become increasingly important to understand what processes of care in SNFs are effective in decreasing rehospitalization and preserving life and function. A unifying theory of the clinical and demographic characteristics predicting hospitalization of nursing home residents for acute congestive heart failure cannot be derived from the current analysis. Rehospitalized subjects were less likely to have dementia, and more likely to have been living in a nursing home before the incident admission, suggesting that chronic physical frailty without dementia predisposes to rehospitalization. The acute HF episodes that resulted in rehospitalization, however, were less severe than those managed in the nursing facility. The incongruence of these findings may reflect our lack of data about the functional and hemodynamic characteristics of HF in nursing home residents. Unfortunately, physicians are currently making rehospitalization decisions for these patients in a knowledge vacuum. There are no published studies that consider which HF exacerbations can be managed safely and effectively in nursing facilities, although at least one study suggests that many hospitalizations for HF are for low-risk symptoms.5 After adjusting for these risk factors, we found that symptom presentation and evaluation at night increased the odds of rehospitalization four-fold. Fried et al.18 found a similar increase in odds of rehospitalization for pneumonia when residents required evaluation for an acute illness in the evening. Staffing, and thus the nursing facility’s ability to manage acutely ill residents, is markedly decreased at night. This implies that what transpires in the nursing home can have a marked effect on important outcomes, such as rehospitalization. The clinical and demographic characteristics of residents who died within 60 days after an exacerbation of HF were not surprising. Hypotension and delirium predicted death. Like Gambassi and colleagues,2 we found a significant difference in the use of ACE-I between NH residents with HF who died and those who survived. After adjusting for these variables, residents who received skilled nursing observation more frequently than once per shift were one-third as likely to die as those who did not. This suggests that processes of care in the nursing home for residents with an acute HF exacerbation may impact their survival. A similar association between processes of care in the nursing home and survival from an episode of nursing home acquired pneumonia was recently described.19 198 Hutt et al
Our study has several important limitations. Most importantly, we had no assessment of the subjects’ functional status, either by an activities of daily living (ADL) score or by New York Heart Association (NYHA) classification. Future studies of the outcome of acute HF exacerbations in nursing homes will require this information, because it has such a large effect on hospitalization and mortality.20 Studies of HF in nursing home residents are necessarily limited by the fact that few patients in this setting have current echocardiograms defining their left-ventricular function or the presence of diastolic dysfunction. Second, more detailed information about the kinds of treatment ordered in the nursing facility would have been useful. How many patients were treated with new or increased potency diuretics? How many patients had monitoring of their hydration status by daily weights or recorded input and output? How many patients had a proximate change in the use of ACE-I, anti-arrhythmic medications, digoxin, or beta-blockers? Data like these might allow future studies to begin defining which exacerbations can be safely managed in nursing homes, and which are best treated in an acute care hospital. Finally, the general application of this study is limited by the fact that the subjects came from facilities in the five states involved in the Multistate Nursing Home Case Mix and Quality Demonstration, and had been admitted to nursing facilities only a short time before becoming acutely ill. Clinical characteristics and management of HF exacerbations in the long-term care segment of the nursing home population are likely to be quite different from those newly admitted under Medicare payment. CONCLUSION In spite of its limitations, this study suggests that a relationship exists between processes of care, such as use of an ACE-I, time of day when a resident is evaluated, and frequency of skilled nursing observation, and the outcomes of rehospitalization and mortality. Further study is needed to better understand this relationship and to begin defining which processes of care for heart failure in nursing homes will be able to decrease rehospitalization rates, stabilize function, and decrease mortality. REFERENCES 1. Haldeman GA, Croft JB, Giles WH, Rashidee A. Hospitalization of patients with heart failure: National Hospital Discharge Survey, 1985 to 1995. Am Heart J 1999;137:352–360. 2. Gambassi G, Lapane KL, Sgadari A, et al. Effects of angiotensin-converting enzyme inhibitors and digoxin on health outcomes of very old patients with heart failure. Arch Intern Med 2000;160:53– 60. 3. Hutt E, Ecord M, Eilertsen TB, et al. Precipitants of emergency room visits and acute hospitalization among short-stay Medicare nursing home residents. J Am Geriatr Soc 2002;50:223–229. 4. Krumholz HM, Parent EM, Tu N, et al. Readmission after hospitalization for congestive heart failure among Medicare beneficiaries. Arch Intern Med 1997;157:99 –104. 5. Butler J, Hanumanthu S, Chomsky D, Wilson J. Frequency of low-risk hospital admissions for heart failure. Am J Cardiol 1998;81:41– 44. 6. US Department of Health. Congestive Heart Failure in the United States: A New Epidemic. NHLBI-NIH Data Fact Sheet, 1– 6. 1996. Washington DC. JAMDA – July/August 2003
7. Carter BL, Small RE, Garnett WR. Monitoring digoxin therapy in two long-term facilities. J Am Geriatr Soc 1981;29:263–268. 8. Wilkins CE, Khurana MS. Digitalis withdrawal in elderly nursing home patients. J Am Geriatr Soc 1985;33:850 – 851. 9. Macarthur C. Withdrawal of maintenance digoxin from institutionalized elderly. Postgrad Med J 1990;66:940 –942. 10. Forman DE, Coletta D, Kenny D, et al. Clinical issues related to discontinuing digoxin therapy in elderly nursing home patients. Arch Intern Med 1991;151:2194 –2198. 11. Forman DE, Chander RB, Lapane KL, et al. Evaluating the use of angiotensin-converting enzyme inhibitors for older nursing home residents with chronic heart failure. J Am Geriatr Soc 1998;46:1550 –1554. 12. Ruths S, Straand J, Nygaard HA, Hodneland F. Drug treatment of heart failure–Do nursing-home residents deserve better? Scand J Prim Health Care 2000;18:226 –231. 13. Burke RE, Feldman JI, Reilly K, et al. Multistate Nursing Home Case Mix and Quality Demonstration: Final Report. Baltimore: Health Care Financing Administration, 1994.
ORIGINAL STUDIES
14. Kramer AM, Frederickson EB, Ecord MK, et al. Nursing Home Case Mix and Quality Demonstration Evaluation, Final Report, Vol. 2: Effects on Outcomes and Quality. Denver: University of Colorado Health Sciences Center, 2001. 15. Barker WH, Zimmer JG, Hall WJ. Rates, patterns, causes, and costs of hospitalization of nursing home resident: A population-based study. Am J Public Health 1994;84:1615–1620. 16. Hutt E, Frederickson E, Eilertsen T, et al. Prospective payment for nursing homes increased therapy provision without community discharge rates. J Am Geriatr Soc 2001;49:1071–1079. 17. Ancestory.com. MyFamily.com Inc. 2001. Accessed May 15, 2003. 18. Fried TR, Gillick MR, Lipsitz LA. Whether to transfer? Factors associated with hospitalization and outcome of elderly long-term care patients with pneumonia. J Gen Intern Med 1995;10:246 –250. 19. Hutt E, Frederickson EB, Ecord M, Kramer AM. Processes of care predict survival following nursing home acquired pneumonia. J Clin Outcome Manage 2002;9:249 –256. 20. Reiley P, Howard E. Predicting hospital length of stay in elderly patients with congestive heart failure. Nurs Econ 1995;13:210 –216.
Hutt et al 199