Symptom Burden Predicts Nursing Home Admissions Among Older Adults

Abstract

Context

Symptom burden has been associated with functional decline in community-dwelling older adults and may be responsive to interventions. Known predictors of nursing home (NH) admission are often nonmodifiable.

Objectives

To determine if symptom burden independently predicted NH admission among community-dwelling older adults over an 8½ year follow-up period.

Methods

A random sample of community-dwelling Medicare beneficiaries in Alabama, stratified by race, gender, and rural/urban residence, had baseline in-home assessments of sociodemographic measurements, Charlson comorbidity count, and symptoms. Symptom burden was derived from a count of 10 patient-reported symptoms. Nursing home admissions were determined from telephone interviews conducted every six months over the 8½ years of the study. Cox proportional hazard modeling was used to examine the significance of symptom burden as a predictor for NH admission after adjusting for other variables.

Results

The mean±SD age of the sample (N=999) was 75.3±6.7 years, and the sample was 51% rural, 50% African American, and 50% male. Thirty-eight percent (n=380) had symptom burden scores ≥ 2. Seventy-five participants (7.5%) had confirmed dates for NH admission during the 8½ years of follow-up. Using Cox proportional hazard modeling, symptom burden remained an independent predictor of time to NH placement (HR=1.11, P=0.02), even after adjustment for comorbidity count, race, sex and age.

Conclusion

Symptom burden is an independent risk factor for NH admission. Aggressive management of symptoms in older adults may reduce or delay NH admission.

Introduction

A growing number of older adults are living with multiple chronic conditions. From 1998 to 2008, the proportion of older adults reporting one or more chronic diseases increased from 86.9% to 92.2% (1). People with co-occurring diseases, also known as multimorbidity, often experience an array of symptoms that may go unrelieved. Symptoms are defined as the subjective evidence of disease or physical disturbance experienced by a patient (2). Disease-specific symptom burden has been described as the sum of the severity and impact of symptoms reported by a significant proportion of patients with a given disease or treatment (3). Symptom burden has been studied in cancer populations and other chronically ill populations defined by specific diseases, such as acquired immune deficiency syndrome (AIDS) and congestive heart failure (CHF) (4-6); however, there is minimal research examining the impact of symptom burden among community-dwelling older adults.

Previous studies of older adults with heart failure, chronic obstructive pulmonary disease (COPD) or cancer have shown the association between symptoms and poor health care outcomes, such as lower self-rated health, quality of life and functional disability (7). No studies to date have looked at the role of symptom burden in nursing home (NH) admission.

The commonly recognized risk factors for NH admission include advanced age, activities of daily living (ADL) dependency, cognitive impairment, and prior NH residency (8-11). In a 2009 meta-analysis, the strongest predictors of NH admission were the presence of three or more ADL dependencies, cognitive impairment, and prior NH residency (10). Most studies that evaluate risk factors do so to identify targets to reduce or delay NH admission. However, many of the previously identified risk factors for NH admission are not modifiable (11). Symptom burden is potentially modifiable and could be a novel target for intervention, particularly to improve outcomes related to quality of life. Therefore, the objective of this study was to assess whether symptom burden independently predicted NH admission in community-dwelling older adults over an 8½ year period of follow-up.

Methods

Setting and Participants

The University of Alabama at Birmingham (UAB) Study of Aging was designed to understand participant-specific factors predisposing older adults to mobility decline and racial differences in mobility changes associated with aging. The UAB Study of Aging is a prospective, observational study of 1000 participants recruited from community-dwelling Medicare beneficiaries aged 65 years or older, living in central Alabama. The sample was stratified by race, sex and county, with recruitment set to achieve a balanced sample in terms of race, sex, and rural/urban residence (12). Counties were classified as urban or rural based on population at the time of baseline interviews (13). After obtaining informed consent, trained interviewers conducted baseline in-home interviews between November 1999 and February 2001. Telephone follow-up interviews to assess NH admission and vital status were conducted at six-month intervals. During the follow-up interviews, information related to the outcome was provided by proxy response if the participant was unable to speak with the interviewer or was unavailable. Potential participants for the present study included all participants in the UAB Study of Aging who completed the baseline assessment and at least one follow-up interview during the 8½ years of the study. The UAB Institutional Review Board approved the study protocol.

Study Variables

Nursing Home Admissions

Data from telephone follow-up interviews completed every six months for up to 8½ years of follow-up were used for this analysis. Nursing home admission was defined by the confirmed admission date of a NH admission during the follow-up period of study. During each follow-up telephone interview, participants or a contact person identified by the participant at baseline were asked if they had been admitted to a NH during the previous six months. If there was a positive response to the question during any follow-up telephone call, a participant was defined as having had a NH admission. If a participant or proxy did not complete a telephone interview and had undetermined vital status, the interview was coded as missing. If vital status was undetermined at the end of the study, the participant was defined as unknown. Reported deaths were confirmed using the Social Security Death Index.

Symptom Burden

The symptom burden measure consisted of a sum of 10 indicators for the following symptoms: shortness of breath, feeling tired or fatigued, problems with balance or dizziness, perceived weakness in legs, constipation, poor appetite, pain, stiffness, anxiety, and anhedonia. These symptoms were assessed in the baseline interview. Six of the symptoms (shortness of breath, feeling tired or fatigued, stiffness, balance or dizziness, weakness in legs, and constipation) were asked with yes/no responses without regard to time. For example, for the symptom constipation, it was asked: “Do you have any constipation?” For the other five symptoms, it was asked, “Do you have problems with any of the following conditions: shortness of breath, feeling tired or fatigued, stiffness, balance or dizziness, weakness in legs?” Four symptoms (poor appetite, pain, anxiety, anhedonia) used a Likert-type response scale; dichotomous categorizations were later applied to the responses. For appetite, participants were asked “Would you say your appetite is usually very good, good, fair, or poor?” Those who answered “fair” or “poor” were characterized as having “poor appetite,” and those who said “very good’ or “good” were characterized as having “good appetite.” For pain, participants were asked “How frequently over the past four weeks have you experienced pain?” Response choices were: daily, 4-6 times per week, 1-3 times per week, less than once a week, or not at all. Those who said “daily” were characterized as having regular pain as a symptom. For anxiety, participants were asked, “During the past four weeks, how often have you been bothered by nervousness or your nerves?” For anhedonia, participants were asked, “In the past … how often have you had little interest or pleasure in doing things?” Responses for these two items were recorded on a five-point scale (always, very often, sometimes, almost never, or never); responses of “very often” or ”always” were coded as reporting the symptom and responses of “never,” “almost never,” and “sometimes” were coded as not reporting the symptom. The sum of these 10 dichotomous symptom items was used as an overall measure of symptom burden. A Cronbach's alpha of 0.76 indicated sufficient internal consistency for the total count. Symptom burden also was evaluated by grouping the number of symptoms into categories of low (0-2), medium (3-5), or high (≥6).

Comorbidity Count

A comorbidity count was calculated by assigning one point for each disease category of the Charlson Comorbidity Index without consideration of severity (15). Self-reported comorbid conditions were verified through prescription medications, physician or clinic questionnaires, and hospital discharge summaries.

Sociodemographic Variables

Race, sex, and age were self-reported at baseline.

Statistical Analysis

Descriptive characteristics included frequencies, proportions, means, standard deviations, and medians. Correlations between symptoms were evaluated by Pearson correlation coefficients. Survival analysis using NH admission as the event was modeled using Cox regression. Cox proportional hazard modeling was performed to examine the independence and significance of symptom burden as a predictor of time to NH admission after adjusting for comorbidity count, race, sex and age (16). Statistical analysis was done using SAS statistical software version 9.2 (SAS Institute Inc., Cary, NC).

Results

The analysis included 999 participants with known vital status and at least one telephone interview over 8½ years of follow-up. Follow-up ranged from six months to 8½ years. Seventyfive participants were admitted to a NH over the 8½ year study period. Vital status at 8½ years was determined for 492 participants who were living and 393 who were deceased. Vital status was unavailable for 114 participants. The 114 participants lost to follow-up were significantly more likely to be African American or have less than a high school education.

Table 1 describes the baseline demographic characteristics of the sample stratified by NH status (admitted and not admitted). Participants admitted to the NH were older and more likely to have lived in a rural county. Those admitted to a NH had a mean symptom burden score of 4.2, whereas those who were not admitted to a NH had a mean symptom burden of 3.4 (P-value 0.006). The residents admitted to a NH experienced a mean comorbidity count of 2.8, and the participants not admitted to a NH had a mean comorbidity count of 2.2 (P-value 0.002). Nursing home admission rates increased as more symptoms were endorsed at baseline. Approximately 5% with low symptom burden at baseline (≤2 symptoms) were admitted to a NH, whereas among those with high symptom burden (≥ 6 symptoms), the NH admission rate was doubled to10.5% (P-value 0.02).

Table 1. Baseline Demographic Characteristics of Population (N=999).

Characteristic, n (%)	Nursing Home Admission (n=75)	No Nursing Home Admission (n=924)	P-value
High-school graduate (Y)	32 (42.7)	469 (50.8)	0.178
Gender (Male)	33 (44)	468 (50.7)	0.268
Race (African American)	32 (42.7)	468 (50.6)	0.184
Rural residence	46 (61.3)	467 (50.5)	0.072
Mean age (SD)	79.4 (7.2)	74.9(6.55)	<0.001
Mean symptom burden (SD)	4.2 (2.6)	3.4 (2.5)	0.006
Mean comorbidity count (SD)	2.8 (1.7)	2.2 (1.6)	0.002
Symptom burden
Low (≤2)	22 (29.3)	396 (42.9)
Medium (3-5)	29 (38.7)	325 (35.2)
High (≥6)	24 (32.0)	203 (22)	0.020

SD = standard deviation.

Table 2 shows the proportion of participants reporting individual symptoms stratified by NH status. Among those admitted to a NH, problems with balance or dizziness, weakness, constipation, and poor appetite were more likely to be reported.

Table 2. Self-Reported Symptoms at Baseline (N=999).

Symptom, n (%)	Nursing Home Admission (n=75)	No Nursing Home Admission (n=924)	P-value
Shortness of breath	27 (36)	323 (35)	0.855
Tired	36 (48)	442 (47.8)	0.978
Dizzy	39 (52)	310 (33.5)	0.001
Weak	44 (58.7)	344 (37.2)	<0.001
Constipation	36 (48)	326 (35.3)	0.027
Poor Appetite	22 (29.3)	150 (16.2)	0.004
Daily pain	32 (42.7))	353 (38.2)	0.445
Stiffness	40 (53.3)	449 (48.6)	0.430
Anxiety	33 (44)	327 (35.4)	0.135
Anhedonia	8 (10.7)	108 (11.7)	0.791

AU: Please Define Bolded Categories/Numbers

Table 3 shows the regression analysis evaluating the role of symptom burden in future NH admission. For every one-point increase in baseline symptom burden score, the odds of being admitted to a NH were 10% higher. Fig. 1 shows the survival analysis of time to NH, which was shortest for those categorized as having high symptom burden at baseline compared with subjects in the low or medium symptom burden groups. When specific symptoms were individually added to the hazard model, feeling tired and fatigued or perceived weakness in legs remained independent predictors of NH admission (data not shown).

Table 3. Factors Independently Predicting Time to Nursing Home Admission (N=75).

Risk Factor	Hazard Ratio	95% Confidence Interval	P-value
Comorbidity Count	1.28	1.12, 1.46	0.0002
Symptom burden score	1.10	1.00, 1.20	0.0511
African American	0.55	0.34, 0.88	0.0123
Male	0.94	0.60, 1.50	0.8084
Age	1.12	1.08, 1.15	<0.0001

Fig. 1.

Time to nursing home admission by symptom burden group.

Discussion

We found that symptom burden independently predicted NH admission over 8½ years of follow-up among community-dwelling older adults. For each additional symptom reported at baseline, the hazard ratio for a NH admission increased by 10%. The prevalence of individual symptoms for those subsequently admitted to a NH ranged from 29.3% for poor appetite to 58.7% for perceived weakness. This is the first known study that evaluated the impact of symptoms on NH admission in a cohort of community-dwelling older adults. Symptom burden and management are often associated with palliative care and advanced chronic disease management (17-19). Symptom burden also has been used to evaluate time to death (20). However, we sought to explore symptom burden, not in the context of terminal disease, but as it occurred earlier in the course of a disease. This perspective adds to the literature showing that symptom burden independently affects patient outcomes, such as functional status (21-22), quality of life (7), and in this case, NH admission. Clinically, this provides impetus for health care professionals to assess symptoms and incorporate these findings into a comprehensive care plan.

A few limitations need to be mentioned. Selection bias for participation in the overall study may have excluded persons already experiencing symptoms or diseases predictive of NH admission. In this analysis, only a simple count of symptoms was used. This analysis only included symptoms from baseline and did not take into account how symptoms might have changed over time. This was consistent with the purpose of the analysis—to define symptom burden at one point in time and to see if the cumulative burden has predictive value. Symptom burden was used as a summative measure and subsequently, symptom severity was not addressed. Further work needs to be done to evaluate the impact of type, number, and severity of symptoms on NH admission. We did not include cognition in this assessment for several reasons. First, dementia was included in the Charlson comorbidity count. Second, our aim was not to create a prediction rule; therefore, we did not add known predictors of NH admission to the model. We controlled for sociodemographic factors and comorbidity because we wanted to distinguish symptom burden from disease diagnoses.

This paper explores the use of symptom burden as a summative assessment, which in our simple model, predicted NH admission. Many symptoms are potentially modifiable, particularly using palliative care interventions. These interventions have been traditionally used for end-of-life care and cancer populations but are now being used more frequently in advanced chronic disease management (4-6, 17-18).

Clinically, early recognition and mitigation of symptom burden could reduce risk for NH admission. It has been previously highlighted that older adults typically underreport symptoms, particularly when symptoms are considered a part of normal aging (23). Routine systematic assessment by clinicians could identify symptom burden, which could then be used to develop interdisciplinary care plans for symptom management including traditional measures of support such as referral for home-based services.

Placing symptom assessments routinely into primary care practice and identifying symptom burden could facilitate more aggressive symptom management. Because many effective modalities have been identified for symptom management regardless of underlying chronic disease conditions, consistent assessment of symptoms may provide opportunities to decrease symptom burden, with a resulting impact on health care utilization and the ability of older adults to remain in the community. Incorporating symptom assessment into routine clinical practice and managing the symptoms identified may lead to improved quality of life, functional status and maintenance of independence. This study suggests that symptom assessment and management may impact NH utilization. Future research should address if managing older adults' symptom burden delays or limits transitions to alternative care settings such as a NH.