|Home | About | Journals | Submit | Contact Us | Français|
Purpose:To evaluate the impact of a multicomponent health promotion and disease self-management intervention on physical function and health care expenditures among Medicare beneficiaries. To determine if these outcomes vary by urban or rural residence.Design and Methods:We analyzed data from a 22-month randomized controlled trial of a health promotion/disease self-management program that included 766 elderly Medicare beneficiaries from western New York, West Virginia, and Ohio. Physical function was measured by changes in self-reported dependencies in activities of daily living over the study period. Total health care expenditures were measured by aggregating expenditures from major sources (acute, postacute, and long-term care). We used ordinary least squares models to examine the effects of the intervention on both physical function and cost outcomes during the 22-month period.Results:The results indicated that the intervention reduced physical functional decline by 54% (p = .03) in the study sample. Stratified analyses showed that the intervention effect was much stronger in the rural sample. Mean total health care expenditures were 11% ($3,100, p = .30) lower in the intervention group. The effects of the intervention on average health care expenditures were similar among urban and rural participants.Implications:The intervention offered a promising strategy for reducing decline in physical function and potentially lowering total health care expenditures for high-risk Medicare beneficiaries, especially for those in rural areas. Future studies need to investigate whether the findings can be replicated in other types of rural areas through a refined intervention and better targeting of the study population.
Functional decline in activities of daily living (ADLs) represents a major risk factor for institutionalization and mortality among older adults (Spillman, 2004). Despite the noted reduction in the prevalence of disability in the population during the past three decades, the number of older Americans with disability continues to exceed 7 million (Manton & Gu, 2001). Projections estimate the absolute size of the disabled population is expected to exceed 12 million by 2030 (Manton, Corder, & Stallard, 1997; Manton, Gu, & Lamb, 2006). Of particular concern are aging adults who live in rural areas. Compared with their urban counterparts, rural Americans are more likely to engage in poor health behaviors (such as smoking and physical inactivity), suffer from chronic illnesses, and experience limitations in functional status (National Center for Health Statistics, 2001). Therefore, there exists an urgent need to develop and test the effectiveness of different strategies designed to improve physical functioning, or to at least delay physical functional decline, among elderly populations and to determine how well these strategies work in both urban and rural areas. Medicare and other health insurance agencies have begun to emphasize health promotion, especially among elderly populations, due to the noted improved overall health status and reduced disability in later life, with an associated potential for reduced health care expenditures (Ackermann et al., 2003; Fried, Bradley, Williams, & Tinetti, 2001; Woolf & Coffield, 2005).
The Medicare Primary and Consumer-Directed Care Demonstration (hereafter referred to as the Medicare demonstration), funded by the Centers for Medicare and Medicaid Services (CMS), tested the effects of a multicomponent nurse home visit intervention on functional and cost outcomes. The conceptual framework for the intervention was based on the logic of empowering, teaching, and coaching chronically ill Medicare beneficiaries to manage their own health and to interact more effectively with their primary care physicians. The health promotion aspect of the intervention was based on a combination of approaches derived from the Social Cognitive Theory (Bandura, 1997) dealing with predisposing, enabling, and reinforcing factors of human behavior; the Transtheoretical Model regarding the stages of intentional behavior change (Prochaska & DiClemente, 1983); and the Health Belief Model (Becker, 1974).
Disease self-management has been defined as the management of chronic illness using disease-specific protocols that enhance self-management and promote patient empowerment (Bodenheimer, Lorig, Holman, & Grumbach, 2002; Newman, Steed, & Mulligan, 2004; Reuben, 2002; Warsi, LaValley, Wang, Avorn, & Soloman, 2003; Warsi, Wang, LaValley, Avorn, & Soloman, 2004). Although disease self-management interventions have proven successful when targeted to specific diseases (e.g., arthritis and diabetes; Bodenheimer et al., 2002; Newman et al., 2004), a paucity of research has examined interventions that utilized a comprehensive approach, along with a core generic set of skills designed to effect change in self-care management behaviors (Warsi et al., 2004). Our intervention provided a unique opportunity to test the impact of a nurse home visit model coupled with an improved partnership with primary care physicians.
Rural residents face important barriers with respect to obtaining medical care and health-related information (Coburn, 2002; Rosenthal & Fox, 2000), and they tend to live more years with impairment than their urban counterparts (Laditka, Laditka, Olatosi, & Elder, 2007). These aspects of rural residence suggest that a health promotion and disease self-management intervention may have a greater impact on rural beneficiaries compared with urban beneficiaries because it addresses an important gap in the provision of health care services in rural areas.
The present investigation evaluated the impact of a health promotion and disease self-management intervention on physical function and health care expenditures among high-risk elderly Medicare beneficiaries over a 22-month period. We hypothesized that participants in the intervention group would demonstrate better physical function and associated decreased average total health care expenditures when compared with individuals in the control group. We also hypothesized that the impact of the intervention would produce different effects for study participants who had resided in rural versus urban geographical settings.
The Medicare demonstration was a 22-month randomized controlled trial that was designed to test the effects of the following three innovative models on health and cost outcomes: (a) a multicomponent nurse home visit intervention that individualized health promotion and chronic disease self-management, (b) a consumer-directed home care voucher model, and (c) a nurse plus voucher model. The analyses reported here are limited to the nurse model versus the control.
The study protocol was approved by the Institutional Review Board at the University of Rochester. Informed consent was obtained from the participants prior to study enrollment. Three hundred and seven primary care physicians referred patients who met the following criteria: (a) were enrolled in Medicare Parts A and B; (b) demonstrated impairment in physical functioning, with at least two limitations with ADLs or at least three limitations with instrumental activities of daily living (IADLs); and (c) had been either hospitalized or had been a patient in a nursing home or had received Medicare home health care within the past 12 months or had two or more emergency room (ER) visits in the past 6 months.
Seven hundred and sixty-six participants were randomized to the intervention group (443) and the control group (460) via a computer-generated assignment table. Enrollment began in July 1998 and ended in June 2002. The current investigation included the 452 participants (234 in the control group and 218 in the intervention group) who completed the 22-month follow-up assessment.
The Medicare demonstration was conducted in two geographic areas: 8 counties in upstate New York and 11 counties along the West Virginia–Ohio border. Of the total population of 1,528,207 in the study area, 1,144,897 (75%) residents lived in 6 urban counties and 383,310 (25%) lived in 13 rural counties. Rural counties were those located outside a Metropolitan Statistical Area (MSA). Rural counties had a higher proportion of adults older than 65 years compared with urban counties (14.1% and 13.2%, respectively). Only a small proportion of African Americans (3.3%) and Hispanics (0.9%) older than 65 years resided in the study area, and the proportion of African Americans was considerably higher in urban counties (4.2%) than in rural counties (0.34%).
An analysis of the study environment (Wamsley & Eggert, 2003) showed rural–urban differences in health care systems. For example, hospitals in urban counties had twice the available beds per 1,000 residents, had higher occupancy rates, longer lengths of stay, higher standard charges, and greater total revenues than rural hospitals. In contrast, Medicare hospital admissions were 28% higher in rural counties than in urban counties (380 and 295 per 1,000 beneficiaries, respectively). Nursing homes in urban counties had 55.86 beds per 1,000 residents older than 65 years compared to 50.2 beds in rural counties. Urban counties had more rehabilitation beds (hospitals and nursing homes) at 5.8 per 1,000 residents older than 65 years compared to 5.1 rehabilitation beds per 1,000 residents in rural counties. In addition, urban counties had three times more home health agencies and service providers than rural counties. Finally, urban counties had one physician for every 367 residents, 1.7 times more than that of rural counties (one physician for every 1,003 residents). The rural area also had a total of five primary care shortage areas, whereas the urban area had none.
The intervention included the following three major components: (a) Patient education: nurses completed monthly home visits that were designed to provide participants and/or their informal caregivers knowledge and skills relevant to successful disease self-management using Consumer Self-Care Strategies and “Healthwise for Life” handbooks (Mettler, Kemper, & Stilwell, 1996); (b) Individualized health promotion and disease self-management coaching: during home visits and telephone communications, the nurses used the PRECEDE (Predisposing, Reinforcing, and Enabling Constructs in Educational Diagnosis and Evaluation) health education planning model as the organizing framework for the application of health behavior change strategies (Green & Kreuter, 1991). The model provided a systematic framework to implement a planning process to empower individuals to engage in behavior change (diet, exercise, etc.), develop and sustain motivation, develop behavioral skills, and participate in community activities. As the majority of patients had reported multiple chronic illnesses, individualized disease self-management protocols were developed and applied based on the needs of the individual participants; (c) Physician care management: Medicare provided experimental benefit payments ($60 per conference, twice a year) to primary care physicians for conducting physician/nurse–patient/family conferences with the purpose of facilitating communication, supporting lifestyle behavior change, and enhancing treatment adherence. In addition, nurses sent periodic reports about patients’ medical conditions to their primary care physicians.
On average, the health promotion nurse completed 25 home visits per patient and organized two conferences with each patient’s primary care physicians during the 22-month intervention. Participants in the control group received their usual care via their Medicare benefits, and they were also compensated with $10 per month for completion of health care services utilization diaries.
Data collection interviews were completed at participants’ homes by trained interviewers at baseline and the 22-month follow-up. Questionnaires were used to collect information regarding demographics, health and physical functional status, and health care services utilization. Interviewers were trained by the research team and completed two reliability assessments of videotaped standardized interview sessions, one at the end of the training session and another during Year 2 of the project. The interrater agreement was 0.97 for all 27 interviewers (Meng, Friedman, Wamsley, Mukamel, & Eggert, 2005). Interviewers were blinded to participants’ group assignments.
Data on health care service use were collected via a weekly diary by the participant or his/her caregiver. Both participants and their caregivers were trained by project staff on how to complete the diaries. Utilization data have undergone extensive quality control procedures. For example, all data on hospital, nursing home, and ER use have been verified by comparing participants’ diary entries with service provider records.
The main outcome variables of interest for this paper included changes in physical function and total health care expenditures. Physical function was measured by the total number of ADL dependencies reported by the participant. Participants were asked: “Tell me how you bathe/dress/eat/use the toilet/walk/transfer from bed to chair? How much assistance is needed?” Participants’ responses to each question were classified as either “no dependency” or “some dependency.” ADL scores were then calculated by summing the total number of domains that the participant reported as having any dependence. Thus, changes in functional status (ranges from −4 to 6) were defined as the difference between the 22-month ADL score and the baseline ADL score. Therefore, negative change scores indicated improvement in physical function, whereas positive change scores indicated deterioration in function. (For example, a “change score” of 6 could have represented going from independence on all six domains of the ADL scale to dependence on all six domains during the study period.)
To calculate total health care expenditures, we multiplied utilization of each health care service by its unit price. Thus, total health care expenditures were obtained by summing expenditures on each service over the course of the study period. We used the following sources for price: local Medicare reimbursement rates (for ER, hospital, and postacute care), local health care provider charges (home health agencies), and actual costs reported by the participants (for out-of-pocket costs). Total health care expenditures included costs by various payers (Medicare, Medicaid, Veterans Affairs, and self-pay). Prescription drug costs were not collected due to expected high data collection costs. It is worth noting that such cost data allow us to estimate potential savings on total health care costs but prevent any comparison between urban and rural areas because of the effect of local economic conditions.
The intervention variable was coded as “1” if the participant had been assigned to the intervention group and “0” for everyone else. Rural status was defined as residing outside an MSA. We included three groups of baseline individual characteristics as independent variables based on factors associated with functional status reported by previous studies: (a) sociodemographic variables (age, gender, ethnicity, living arrangement, education, and income), (b) health and functional status variables (number of chronic conditions, Cognitive Performance Scale [CPS; Morris et al., 1994], and number of ADL dependencies at baseline), and (c) health services use prior to project enrollment (skilled home health care, hospital, ER, and nursing home). We also included an indicator site variable that identified whether the participant was from the New York study site.
We compared baseline characteristics of the intervention group with those of the control group by using the χ2 test for categorical variables and the t test for continuous variables. In order to assess the impact of excluding participants who did not complete the 22-month assessment, we also compared the baseline characteristics of those who were included in this analysis versus those who were not included. We then compared the mean ADL score at baseline and 22-month by urban–rural status and by treatment. We used ordinary least squares (OLS) regression models to estimate the effect of the intervention on changes in physical function while controlling for other covariates. We estimated a model that predicted the changes in ADL score as a function of the intervention variable and other covariates. We then performed stratified analysis by urban–rural status. We also calculated the adjusted R2 to examine the goodness of fit of the models as well as the Pregibon’s link test to examine the model specification (Pregibon & DiClemente, 1980). Finally, we examined the residual plots of the OLS models to ensure that model assumptions were met.
For the health care expenditures outcome, we performed unadjusted analysis by comparing the average total expenditures by intervention group for the full sample and by urban–rural status. We used t tests to examine statistical significance. Total health care expenditures were log-transformed to alleviate the skewness of its distribution. We then used multivariate linear regression to estimate the effect of the intervention on total health care expenditures while controlling for covariates. We used the same set of control variables as those used in the ADL models. We calculated the adjusted R2 to examine the goodness of fit of the models as well as the link test to examine the model specification (Pregibon & DiClemente, 1980).
Figure 1 depicts the flow of participants through different stages of the study. During the 22-month follow-up, of the 766 participants (384 in the control group and 382 in the intervention group) who received interventions, 271 (123 + 148) were excluded from the analysis because they discontinued the intervention (due to death, no longer interested, or other reasons). An additional 43 (27 + 16) participants were excluded from the analysis because their 22-month follow-up assessments were completed by proxies. The mortality rates of the intervention group and the control group were the same (18% in both groups, p = .90). The dropout rates for reasons other than death were higher in the intervention group (20% vs. 14% in the control group, p = .02), primarily due to the fact that more participants in the intervention group were “no longer interested” in the program (41 versus 18). Despite this difference, further analysis showed that those who dropped out of the study had similar mortality rates (28% in the control group and 27% in the intervention group, p = .91).
We compared those who were included in this analysis (n = 452) with those who were excluded (n = 314). We found that participants who were excluded were older; were more likely to have an informal caregiver; had more IADL disability and cognitive impairment; were more likely to have congestive heart failure (CHF), stroke, and chronic obstructive pulmonary disease; were less likely to be overweight or obese; and were more likely to have past health services use (results not shown).
Table 1 shows sociodemographics, health and physical function, and health services use prior to enrollment of participants at baseline for the study sample and by intervention and urban–rural status. The mean age of the sample was 76 years, and more than two thirds (71%) were female, 37% lived alone, and 70% had informal caregivers. Participants in the control and the intervention group had similar characteristics, with the exception of two variables: percent with CHF and stroke. In the urban sample, the only significant difference between the intervention and the control group was that more participants reported having CHF. In the rural sample, fewer participants in the intervention group reported being married. Urban and rural participants were similar with respect to sociodemographics, most health status measures, and past health services use, with the exception of the following: rural participants reported having fewer ADL dependencies, more myocardial infarction, and better self-rated health compared with their urban counterparts.
Figure 2 shows the unadjusted results of functional change for the study sample. The average participant demonstrated functional decline between the baseline and the 22-month follow-up, as evident by the increased mean number of ADL dependencies over time. The decline in function was much smaller in the intervention group (.23) when compared with that of the control group (.50), indicating a 54% ([.50 − .23] × 100/.50) reduction in functional decline among the intervention group.
Figure 3 illustrates functional change over 22 months by intervention and urban–rural status, indicating that the intervention is much more effective among rural participants. Rural participants who received the intervention had an 81% ([.62 − .12] × 100/.62) reduction in mean number of ADL dependencies compared with a 36% ([.45 − .29] × 100/.45) reduction among urban participants.
Table 2 shows the results from the OLS regression analyses, including the stratified analysis by urban–rural status. Coefficient estimates from the OLS models confirmed our finding of the intervention effect in reducing functional decline (coefficient estimate −0.27, p = .03) in the unadjusted analysis for the overall sample. When the same model was estimated for urban and rural participants separately, the intervention effect was no longer statistically significant among urban participants (−0.09, p = .56). However, the intervention effect among rural participants was more than double (−0.57, p = .01) the overall effect. Other significant predictors of functional change include age, baseline functional status, CPS, stroke, overweight or obese, self-rated health, and skilled home health care use during the year prior to enrollment.
Average total health care expenditures were 11% ($3,100, p=.30) lower in the intervention group during the 22-month study period ($26,100 in the intervention group and $29,200 in the control group). Table 3 shows the adjusted results of the impact of intervention on average total health care expenditures for the study sample and by urban–rural status. The result suggested that the intervention is cost neutral even factoring in the substantial cost of delivering the intervention ($3,500 per participant). Other significant predictors of total health care expenditures include baseline ADL, self-rated health, and skilled home health care use.
The results suggest that the intervention proved effective in reducing decline in physical functional status among community-dwelling Medicare beneficiaries who completed the intervention. The positive effect of the intervention was particularly strong among participants who lived in rural areas. To our knowledge, this represents the first study to show that a cost-neutral community-based health promotion and disease self-management intervention was effective in reducing ADL decline among the traditional fee-for-service Medicare beneficiaries over a 2-year period.
Our findings remained consistent with results from previous studies of various types of interventions among similar populations. For example, Gill and colleagues (2002) found that a 6-month home-based physical therapy intervention resulted in a 37% reduction in functional decline among a sample of frail older adults aged 75 and older. More recently, Gitlin and colleagues (2006) reported that a multicomponent home intervention led to fewer declines in ADL limitations over 6 months among 319 adults aged 70 years and older. Among enrollees of Medicare Health Maintenance Organizations Leveille and colleagues (1998) found that a health promotion chronic disease self-management intervention increased the likelihood of ADL improvement by 84% during a 12-month period and may have reduced inpatient hospital days. These findings suggest that it is possible to slow the decline in physical function for some subsets of the older population. Further studies are needed to determine if subpopulations of elderly Medicare beneficiaries with particular health conditions, such as heart disease, diabetes, and obesity, benefit more than others and to determine if these treatment strategies can be fine-tuned to achieve even better results.
Since 2002, the CMS has been testing various care management and disease management models in the traditional fee-for-service Medicare population (Guterman, 2007). Although these pilot programs have varied widely in their organizational structures, target populations, and interventions, all but two programs have used registered nurses as care coordinators, and most programs have conducted the intervention by telephone (one to three times per month). A recent report of the results of the first 2 years of 15 such programs indicated that none of these programs had achieved improvement in functional status or a reduction in Medicare expenditures (Brown et al., 2007).
We believe that the comprehensive nature of our intervention rather than any one component was associated with the treatment effect. Evidence from randomized controlled trials of disease self-management suggest that programs teaching self-management skills are more effective than patient education alone in improving clinical outcomes and that a patient–professional partnership involving collaborative care is also important (Bodenheimer et al., 2002). Although our results are consistent with these findings, a separate investigation is currently under way to more closely examine the structure and process of the intervention (Liebel, 2007).
A stronger intervention effect in rural areas may be explained in part by fewer functional limitations among rural participants at baseline (as illustrated in Table 1). Gill and colleagues (2002) also found a greater treatment effect for people with moderate frailty compared with those with severe frailty. Another possible explanation is that rural primary care providers may have been more likely than their urban counterparts to refer patients in earlier stages of functional impairment. Patients themselves may have been more motivated to accept self-management approaches in resource poor environments where treatment options were limited (Wamsley & Eggert, 2003).
One limitation of this study was the relatively high attrition rate (41%) over 2 years. However, this level of attrition seems to be reasonable, given the lack of exclusion of people with comorbidities. For example, Gill and colleagues (2002) reported that only 65% of participants completed their 12-month intervention. The attrition rate calls for better targeting in future studies so that people who are at exceedingly high risk of dying before the end of the program may be identified and referred to appropriate care settings.
Another limitation of this study was the small proportion of minorities in the demonstration catchment area and the study sample. Thus, the ability to generalize these findings to other populations of Medicare beneficiaries may be limited. The intervention may work differently in other regions of the country and with minority populations. The study is also limited by the operational definition of rural (non-MSA) used in the analyses. Recent developments in rural definitions incorporate gradients of rurality, each with strengths and weaknesses (Hart, Larson, & Lishner, 2005). However, the sample size of this study prevents an investigation of the effects of different levels of rurality.
Despite these limitations, this study maintained several strengths. The sample size was larger, and the follow-up period was longer than that of similar studies reported in the literature (Gill et al., 2002; Gitlin et al., 2006; Leveille et al., 1998; Phelan, Williams, Penninx, LoGerfo, & Leveille, 2004). In addition, the design of the intervention was unique in an attempt to improve chronic care in three important aspects: health promotion, disease self-management education, and reimbursed physician care coordination. Finally, the analyses attempted to address the issue of potential differences in delivering the intervention in urban versus rural areas.
The results of this study have important implications for public policy. The Office of Rural Health Policy (2006), in consultation with the Office of Management and Budget, has developed the goal of reducing the proportion of rural residents of all ages with limitation of activities caused by chronic conditions to 13.9% by the year 2010. This intervention of patient education, health promotion, and primary care coordination has great potential in achieving the goal in a cost-neutral way. Efforts are under way to determine if any medical criteria can help target to the subgroup of rural elderly who might benefit the most from the intervention. Future studies will examine the characteristics of primary care physicians and their practice settings to identify factors associated with improved functional status. In addition, rural areas are heterogeneous in that those near urban areas usually have a richer mix of health services and more physicians and nurses per thousand populations than more remote areas. A demonstration of a refined intervention across different types of rural areas could provide empirical evidence to adjust the intervention across the rural continuum from densely to sparsely populated places.