Customize your JAMA Network experience by selecting one or more topics from the list below.
Stuck AE, Minder CE, Peter-Wüest I, et al. A Randomized Trial of In-Home Visits for Disability Prevention in Community-Dwelling Older People at Low and High Risk for Nursing Home Admission. Arch Intern Med. 2000;160(7):977–986. doi:10.1001/archinte.160.7.977
In-home preventive visits with multidimensional geriatric assessments can delay the onset of disabilities in older people.
This was a stratified randomized trial. There were 791 participants, community-dwelling people in Bern, Switzerland, older than 75 years. The participants' risk status was based on 6 baseline predictors of functional deterioration. The intervention consisted of annual multidimensional assessments and quarterly follow-up in-home visits by 3 public health nurses (nurses A, B, and C), who, in collaboration with geriatricians, evaluated problems, gave recommendations, facilitated adherence with recommendations, and provided health education. Each nurse was responsible for conducting the home visits in 1 ZIP code area.
After 3 years, surviving participants at low baseline risk in the intervention group were less dependent in instrumental activities of daily living (ADL) compared with controls (odds ratio, 0.6; 95% confidence interval, 0.3-1.0; P=.04). Among subjects at high baseline risk, there were no favorable intervention effects on ADL and an unfavorable increase in nursing home admissions (P=.02). Despite the similar health status of subjects, nurse C identified fewer problems in the subjects who were visited compared with those assessed by nurses A and B. Subgroup analysis revealed that among low-risk subjects visited by nurses A and B, the intervention had favorable effects on instrumental ADL (P=.005) and basic ADL (P=.009), reduced nursing home admissions (P=.004), and resulted in net cost savings in the third year (US $1403 per person per year). Among low-risk subjects visited by nurse C, the intervention had no favorable effects.
These data suggest that this intervention can reduce disabilities among elderly people at low risk but not among those at high risk for functional impairment, and that these effects are likely related to the home visitor's performance in conducting the visits.
THE INCREASING health and social care services required by disabled older persons will be a growing burden and a major societal concern.1,2 One possible means of coping with this development would be to devise and implement strategies for detecting and modifying biological, psychological, social, and environmental risk factors for disabilities in order to prevent or delay their onset.3 Several investigators have developed and tested preventive home visitation programs in community-dwelling older people as a strategy for delaying the development of disabilities and reducing nursing home admissions in older people.4-12 Although some studies have shown favorable effects on health status and health care use in older people,4,6,8-10,12 others have not.5,7,11 Some of the discrepancies in these findings might be explained by differences in the design of the preventive interventions, such as the duration and intensity of the home visits.13 However, other factors might also explain the variable study findings. First, it is uncertain whether all or only subgroups of the older population benefit from preventive home visits. Exploratory subgroup analyses of one trial suggested that subjects with good baseline functional status benefit from preventive home visits.14 Second, we identified variations in the intervention approach between the nurses conducting the home visits in an earlier study.4 Although such variations in the intervention process might influence outcomes in any type of preventive home visitation or comprehensive geriatric assessment program, none of the previous studies addressed this issue.4-18
To address these uncertainties, we conducted a randomized controlled trial to test the hypothesis that preventive home visits with annual multidimensional assessments have more favorable effects on functional status and nursing home admissions in low-risk compared with high-risk older persons. In addition, we used a study design that would permit an analysis of the impact on patient outcomes of differences in the intervention process among nurses.
The study was approved by the institutional review committee at the University of Bern, Bern, Switzerland. The study population was drawn from a health insurance list of community-residing subjects aged 75 years and older living in 3 ZIP code areas in Bern. This approach allowed us to obtain a representative study sample (more than 95% of older people in Switzerland had health insurance coverage in 1993) and to have access to health care use and expenditure data. First, we took a random sample of 1998 subjects from this list (57% of total enrollment) and excluded 806 of them for the following reasons: 165 people had died, 595 were living in nursing homes or board and care facilities, 17 had moved away, 17 did not speak German, and 12 reported that they had a terminal disease. Before asking the remaining subjects if they would participate in the study, we decided that only 1 member per household should be represented, since the unit of randomization was the individual subject and randomizing 1 eligible household member to the intervention group and another to the control group would not have been feasible. Therefore, we excluded subjects whose household partner had already agreed to participate in the study and replaced these subjects with persons who were randomly selected from the original health insurance list but matched for household membership status. Overall, among the eligible 1192 subjects, 401 (34%) declined and 791 (66%) agreed to participate in the study (Figure 1). There were no significant differences among participants and refusers by sex (women, 73% vs 73%, age (81.6 vs 81.8 years, respectively), self-perceived health (fair or poor, 33% vs 37%), and number of physician consultations in the 6 months prior to baseline assessment (6.6 vs 6.6 consultations).
The baseline interview of the study participants was conducted at home. Trained interviewers using a structured interview form collected information on sociodemographic factors, basic activities of daily living (ADL) (bathing, dressing, feeding, transferring from bed to chair, and moving around inside the house),19 instrumental ADL (cooking, handling finances, handling medication, shopping, using public or private transportation, and using the telephone),19 self-perceived general health,20 cognitive function,21 depression,22 gait and balance performance,23 medication use,24 and self-reported chronic conditions. Information about the subjects' baseline economic situation was obtained from tax files.
Before randomization, subjects were classified according to their risk for future nursing home admissions, which we determined by developing a list of 6 criteria based on a literature search.25-27 A total of 347 subjects fulfilled one or more of these criteria: (1) need for assistance in at least one of the basic ADL19; (2) score higher than 5 on the Geriatric Depression Scale, Short Form22; (3) Mini–Mental State Examination score lower than 2421; (4) impaired gait and balance (score <19)23; (5) more than 3 self-reported chronic conditions; and/or (6) use of 6 or more medications (as a proxy for comorbid conditions).
Based on the prerandomization baseline interview, all subjects who gave written informed consent were assigned to the appropriate risk stratum and listed within each of the 2 risk strata separately in ascending sequence of preassigned random numbers. As soon as 3 participants were available in any 1 risk stratum, an independent test center calculated the treatment assignment based on a random letter table and informed the project team. One subject from each triplet was allocated to the intervention and 2 were allocated to the control group. Since randomization was conducted separately within each risk stratum, the ratio of the number of subjects in the intervention group to the number of subjects in the control group was 1 to 2 within each risk stratum.
The subjects in the intervention group underwent annual multidimensional geriatric assessments in their homes. Three health nurses (certified registered nurses with an additional degree in public health nursing based on an 8-month postgraduate course) obtained medical histories, gave physical examinations, and measured hematocrit and glucose levels in blood samples. In addition, they performed a comprehensive geriatric assessment by evaluating subjects for hearing, vision, nutritional status, oral health, appropriateness of medication use, safety in the home, ease of access to the external environment, and social support.28,29 Based on this in-home visit and on the information collected at the prerandomization interview, the nurses prepared a problem list and discussed each case with one of the project team's geriatricians, developed rank-ordered recommendations, and conducted in-home follow-up visits every 3 months to monitor the implementation of the recommendations, make additional recommendations if new problems were detected, and facilitate compliance. In exceptional circumstances, the nurse telephoned the participant or was available by telephone. Nurses also provided health education, encouraged subjects to participate in self-care, and attempted to improve their ability to discuss problems with their physicians. If participants gave their permission, their primary care physician received a letter from the project team with selected findings of the baseline geriatric assessment and selected recommendations to the physician; in urgent and complex situations, one of the project team's geriatricians called the primary care physician to discuss a problem or a recommendation. In addition, an interdisciplinary team (physical therapist, occupational therapist, dietitian, and social worker) was available to the nurse for discussing complex problems.
Because health nurses have little training in physical assessment and gerontology in the Swiss system, the nurses received additional training in these areas and in conducting preventive home visits before and during the project. The rate of identified problems in older persons was measured with a method that we developed in an earlier study.30 The nurses listed all identified problems of each older person and allocated these problems to a list of 24 predefined problem categories.
Primary outcome measures were the need for assistance in basic and instrumental ADL19 at 3 years and the number of permanent admissions to nursing homes during the 3-year follow-up. Trained telephone interviewers independent of the intervention and blinded to the treatment assignment of study subjects collected information on ADL at 3 years. Information about nursing homes and survival was collected from proxies by telephone and was verified based on records from government and health insurance registries. This follow-up was conducted at 3 years, since the nurses gave additional recommendations at the final 2-year intervention visit; thus, older people were expected to implement these recommendations during the third year. We did not offer additional intervention follow-up visits during the third study year because of financial constraints.
Secondary outcome measures were collected in the homes of all surviving study participants at the 2-year follow-up visit. For this purpose, nurses collected information on affect,22 cognitive function,23 gait and balance,24 general health,20 number of medications being taken, and influenza vaccination status.
Health care cost and utilization data covering 6 months before and 2 years after randomization of all participants were obtained from insurance files. Data from the second half of the third year were not available because of a reform of the Swiss health care system. For the cost analysis, third-year ambulatory care cost data were imputed from data for the first half of the third year. It is a peculiarity of the Swiss health system that the financing of institutional care is shared between insurance companies and the public sector. Since hospital cost insurance records reflect only part of total hospital cost, public sector subsidies had to be imputed. In 1995, the daily cost for nursing home care was SFr 222 (US $133), and the daily cost for hospital care ranged from SFr 475 (US $285) to SFr 1579 (US $947), depending on the type of hospital.31
We calculated the sample size needed to detect a relative 30% risk reduction for disability in the intervention group compared with controls. In addition, for budgetary reasons, we decided to use a 1:2 ratio of the number of people in intervention vs control groups, since research study costs per subject in the intervention group were much higher than those for controls. Based on a power of 0.8 and a prevalence of instrumental ADL limitation of 50% among controls, required sample sizes were 138 for the intervention group and 276 for controls. In order to be able to analyze the high and low baseline subgroups separately, we doubled these sample sizes to 276 and 552, respectively.
All analyses were by intention to treat, including subjects who refused intervention. Multivariable logistic regression analyses were employed to analyze dichotomous outcomes; multivariable regression analyses were used to analyze outcomes measured as scores (eg, cognitive performance); and multivariable Poisson regression analyses corrected for overdispersion32 to analyze the number of hospital admissions and physician visits. Sensitivity analyses were conducted by repeating the analyses without adjusting for baseline characteristics of the participants and by using nonparametric procedures. For analyzing differences of intervention effects between risk strata or between ZIP codes, interaction analyses were conducted. The statistical models predicted outcomes (eg, basic ADL and instrumental ADL) with the following independent variables: a term indicating intervention group assignment, the interaction term (eg, intervention in subjects at high risk), a term indicating the risk stratum (or the ZIP code), age, sex, baseline functional status, and baseline self-perceived health status.
The overall comparison between intervention and control groups showed comparable baseline characteristics, with the exception of a higher rate of people who depended on others to perform basic ADL in the intervention group compared with controls (12% vs 7%; P=.02) (Table 1). Of the 264 persons in the intervention group, 7 persons declined to be visited, and 8 persons died during the first year of the study. The remaining 249 persons received 8.5 ± 2.9 home visits (mean ± SD) during the 2 years of the intervention. A single visit lasted 74 ± 12 minutes (mean ± SD), with yearly assessments taking approximately 2 hours and quarterly visits approximately 1 hour. More than 90% of the participants' primary care physicians were contacted at least once during the project by one of the project team's geriatricians.
At the 3-year follow-up, the analysis of the entire study sample revealed surviving subjects in the intervention group were less dependent on others to perform instrumental ADL compared with controls (P=.03) (Table 2). The intervention effects on the other main outcomes were not statistically significant. Table 2 reveals that the favorable effect on instrumental ADL occurred among the subjects at low baseline risk (P=.04). In contrast, among subjects at high baseline risk, the intervention had no effect on basic and instrumental ADL and a significant unfavorable effect on nursing home admissions (P=.02). Sensitivity analyses based on alternate outcome definitions (eg, combining death with poor functional outcome, combining basic and instrumental ADL) did not change the conclusions.
Interaction analyses revealed that the odds ratio (OR) for the intervention effect on nursing home admissions among subjects at low risk (OR, 0.5) (Table 2) was significantly (P=.03) different from the OR for the intervention effect on nursing home admissions among subjects at high risk (OR, 2.1) (Table 2).
Table 3 depicts the secondary outcomes for subjects at low baseline risk. At the 2-year follow-up, the mean gait and balance score was significantly higher in intervention subjects compared with controls. Clinically, this difference corresponds to the difference of whether or not a person had to use his or her hands for standing up from a chair and is related to an overall increase in strength, particularly lower-extremity strength. There were no significant differences in the affective or cognitive scores. The influenza vaccination rate was higher among subjects in the intervention group compared with controls (33% vs 25%, P=.01) (Table 3). At 2 years, the mean number of medications taken by subjects in the intervention group was higher (mean difference, 0.6) than among controls. There were no significant differences for hospital care use. The number of visits to primary care providers but not the number of visits to specialist physicians was higher for the intervention group compared with controls (P=.05). There was no difference in traditional home care use between the intervention and control groups.
In contrast, the intervention had no effects on secondary outcomes (P≥.10 for all) among subjects at high baseline risk. (Tabulated data are available from the authors on request.)
After the first year of the intervention, despite similar training of the 3 nurses, nurse C identified significantly fewer problems compared with nurses A and B (P<.001) (Table 4). Differences in problem identification were found particularly for problems that required clinical judgment (eg, chronic obstructive pulmonary disease, medication management, social and environmental problems). Differences in the health status of people living in different ZIP code areas might account for minor differences among the nurses but not a variation of this kind. Over the 2-year period, the number of problems for which nurse C intervened (ie, gave recommendations) was significantly lower compared with the number of problems for which nurses A and B intervened (Table 4).
Because of this observation, at the end of the first study year, we hypothesized more favorable effects in persons visited by nurses A and B compared with those visited by nurse C. We therefore analyzed outcomes for subjects in the ZIP code areas visited by nurses A and B (ZIP codes A and B) and compared them with the outcomes for subjects in the ZIP code area visited by nurse C (ZIP code C). Because the primary analyses showed differences in intervention effects according to the risk status of the study populations, we conducted these subgroup analyses separately for subjects at low and high baseline risk.
Among subjects at low baseline risk in ZIP codes A and B, 2 persons (3% of 77 surviving subjects) from the intervention group compared with 22 (12% of 177 surviving subjects) from the control group were dependent on assistance for at least 1 of the ADL at the 3-year follow-up (Table 5). None of the 85 persons allocated to the intervention group, compared with 11 controls (6% of 191 subjects), was admitted to a nursing home during the study period. In contrast, there were no favorable effects for low-risk subjects in ZIP code C. Interaction analyses revealed that for subjects in ZIP codes A and B, the intervention effect on basic ADL was significantly (P=.047) more favorable compared with the intervention effect on basic ADL for subjects in ZIP code C. Among subjects at low risk in ZIP codes A and B, the intervention had a favorable effect on gait and balance (P=.005), but no effect in ZIP code C (P=.26). Among subjects at low risk in ZIP codes A and B, the intervention increased the influenza vaccination rate (39% vs 24%; P=.01), but among subjects at low risk in ZIP code C, the intervention did not affect the influenza vaccination rate (26% vs 29%, P=.37).
Among subjects at high baseline risk in ZIP code C, the intervention had a statistically significant unfavorable effect (P=.002), with a higher number of nursing home admissions in the intervention group compared with controls (Table 5). Among high-risk subjects in ZIP codes A and B, there was no significant difference in nursing home admissions between the intervention and control groups (P=.54). Interaction analyses revealed that the intervention effect on nursing home admissions for high-risk subjects in ZIP code C was significantly different (P=.003 for interaction term) from the effect on nursing home admissions for high-risk subjects in ZIP codes A and B. Among high-risk subjects in ZIP codes A and B, there was a trend toward higher mortality in the intervention group compared with controls (Table 5), but the intervention effects on mortality did not differ between the ZIP codes A and B and ZIP code C (P=.69 for interaction term). The intervention did not have significant effects on secondary outcomes (P≥.10 for all; data available from the authors on request).
Based on 3-year outcomes, the average yearly health care cost for people in the intervention group was approximately SFr 1500 (US $900) more than that for controls (a table with cost analyses for the entire sample is available from the authors on request). Table 6 presents the results in the subgroup of subjects with favorable results (ie, low baseline risk, ZIP codes A and B). Table 6 also shows the net cost difference per person and by year for subjects in the intervention group compared with controls. In the first and second years of the program, the annual cost of the intervention itself was SFr 460 (US $276) per subject allocated to the intervention group. Since there was no cost for the intervention among control subjects (those who did not receive the intervention), the net difference for the intervention was SFr 460 (US $276) per subject per year. At the beginning of the third year of the study, the final annual in-home visit was conducted, and there were no follow-up home visits in the third year of follow-up. Therefore, the intervention costs were only SFr 184 (US $110) in the third year.
In addition, the intervention resulted in additional net costs because it increased ambulatory care costs compared with those for controls. This increase of ambulatory care costs was a result of the fact that the intervention increased the number of visits to primary care physicians (Table 3). As depicted in Table 6, the increase in ambulatory care costs was most pronounced in the first 2 years of the project. Among subjects at low risk in ZIP codes A and B, the mean yearly acute-care hospital cost for the intervention group was SFr 16 (US $10) less than that for the control group. Since analyses of 2-year outcomes revealed that the intervention had no statistically significant effect on acute-care hospital use (number of admissions, length of stay), this difference was rounded to 0. Finally, the intervention resulted in a reduction of nursing home admissions in subjects at low risk in ZIP codes A and B.
Among subjects at low risk in ZIP codes A and B, 11 persons from the control group were permanently admitted to a nursing home compared with none from the intervention group (Table 5). In the control group, the net costs for nursing home care were SFr 636 (US $382) per person in the second year of follow-up and SFr 2683 (US $1610) per person in the third year of follow-up. Overall, in the first 2 years of follow-up, total health care costs for subjects in the intervention group were higher than those for controls. However, in the third year of the project, the prevention of nursing home admissions in the intervention group resulted in substantial savings that more than offset the additional costs for home visits and ambulatory care, which resulted in net savings of SFr 2336 (US $1403) per person per year.
At 1 year after baseline assessment, in all ZIP codes, over 90% of subjects said they liked the home visits, and over 70% said the visits had been helpful. At 3 years, an independent interview was conducted with surviving subjects who had participated in the intervention. There were no significant differences in the proportion of subjects from the 3 ZIP code areas who said that, as a result of the intervention, they knew more about their medications (38% of interviewed subjects), felt more confident discussing their problems with their physician (36%), or had increased their level of physical activity (29%). However, the proportion of subjects who said that they were sorry the visits had stopped was significantly higher in ZIP code C (69%) compared with ZIP codes A and B (52%) (P=.04).
In this trial, we used an intervention strategy similar to that used in an earlier trial in Santa Monica, Calif.4 Despite comparable study populations, the intervention effects differed between the 2 studies. The intervention effects of the Bern trial on functional status outcomes were weaker compared with those in the Santa Monica trial. It is likely that the low problem-detection rate in ZIP code C was a main factor contributing to the weaker intervention effects in the Bern trial. However, other factors must play a role, since the intervention effects were weaker in the Bern trial compared with the Santa Monica trial even if ZIP codes A and B were analyzed alone. The duration of the intervention in Bern was 2 years compared with 3 years in Santa Monica. The length of follow-up has been shown to be a key factor in determining the effects of geriatric assessment programs.13 Furthermore, physician adherence to preventive recommendations in Bern was lower than in Santa Monica. For example, the influenza vaccination rate was lower in Bern compared with the rate in Santa Monica (31% vs 62%).
The unfavorable intervention effects on nursing home admissions among subjects at high baseline risk were unexpected. This increase in nursing home admissions was caused by an increase of nursing home admissions in ZIP code C. Thus, the unfavorable effect of the intervention on nursing home admissions among high-risk subjects might be related to the intervention process of nurse C (eg, failure of the nurse to recognize early signs of catastrophic conditions). The trend toward a higher mortality rate in the intervention group among high-risk subjects in ZIP codes A and B compared with controls was probably a result of chance. A survival follow-up at 40 months in the high-risk subjects of ZIP codes A and B showed that the P value for the mortality difference between the intervention and control groups increased to .23. Overall, these results suggest that preventive home visits probably have no favorable effects on outcomes in high-risk older people, and, under certain circumstances, might even have unfavorable effects. We believe that for older persons at high risk, programs combining rehabilitation and care coordination should be used instead of preventive home visitation programs. For example, randomized studies have shown favorable effects of programs emphasizing tertiary prevention and rehabilitation in patients with recurrent congestive heart failure,15,16 in patients requiring home care services,17 or in patients with dementia.18
The finding of favorable intervention effects among low-risk subjects is consistent with our hypothesis that this type of intervention is more effective among well-functioning elderly people. In fact, secondary analyses of the Santa Monica study confirmed that favorable intervention effects were more pronounced in elderly people who were better functioning at baseline.14 The present study lends insight into possible mechanisms that contribute to the benefit of the intervention. The intervention resulted in better gait and balance function but had no relevant impact on cognitive function, suggesting that this intervention does not reduce nursing home admissions related to dementia. The increase in primary care physician visits among intervention subjects compared with controls underscores the fact that the recognition of problems that physicians were unaware of and referrals to patients' primary care physicians contribute to the benefit.34 While the intervention increased overall medication use among low-risk persons, data on medication categories suggested that this observed difference resulted from an increase in medications that were previously underused (eg, antihypertensive agents) and a decrease in previously overused medications (eg, benzodiazepines, nonsteroidal anti-inflammatory agents).
The results of this study demonstrate that the adequate detection of risk factors for functional status decline and the implementation of recommendations to address them are probably key factors contributing to the success of in-home prevention programs. The development of improved validated multidimensional assessment tools and guidelines for preventive recommendations could help to improve the intervention process. Based on a systematic literature review of risk factors for functional status decline,35 we are currently validating a new multidimensional preventive assessment tool.36 Computer technology might then be used to analyze the collected information and create checklists or individualized recommendations that could help the person performing the intervention to implement optimal individualized intervention plans. The results of the present study also suggest that home visits by volunteers (ie, performance of nurse C) may not be effective and that the intervention should be conducted by specially trained and selected professionals. To ensure the implementation of an adequate intervention process, comparative analyses of identified problems or the number and type of recommendations might help to identify potential deficits in the preventive interventions. In addition, use of "tracer" conditions (ie, the influenza vaccination rate, control of arterial hypertension, or level of patients' physical activity) and, in like fashion, ongoing monitoring of outcomes could be used for quality assurance. Although patient satisfaction with the preventive home care program is important for program success, measures of patient satisfaction are inadequate for judging the performance of the home visitors, and may, as shown in this study, even be misleading.
The cost analysis revealed that in the subgroup with favorable outcomes (ie, among low-risk subjects in ZIP codes A and B), there was a temporal pattern (Table 6). The program resulted in additional costs in the early phase of the intervention, but in the third year, the additional health care costs (for preventive home visits and additional physician visits) were more than offset by savings in nursing home costs, indicating that this intervention requires an initial investment, with savings occurring only after time.
The analysis of the impact of the intervention process on outcomes was based on the differences in performance among 3 nurses. Studies based on larger numbers of persons who perform interventions would be needed to analyze the impact of the intervention process on outcomes in more detail. Theoretically, factors other than the process of the intervention, such as differences in patient characteristics among those living in different ZIP code areas, might account for differences in outcomes between them. However, this is unlikely, since the results remained significant after controlling for baseline population factors. The classification into 2 risk strata was based on a combination of 6 individual risk factors, as used by others in the literature. Future studies with larger sample sizes would be needed to refine these criteria and identify those factors that best predict who might benefit from preventive home visits. With regard to the entire field of geriatric assessment, the findings regarding nurse C suggest that the intervention approach used by the professional conducting the assessment might be a key factor for determining effects and, together with risk status, might account for some of the previously unexplained discordant effects among earlier studies of comprehensive geriatric assessment.14
Limitations of the study include the following: (1) Although we cannot exclude the possibility that the significant prerandomization difference in basic ADL between the intervention and control groups may have influenced, in part, the outcomes among high-risk subjects, it is unlikely that that this had an impact on the findings because the results for high-risk persons remained significant after controlling for baseline differences. (2) Since some of the secondary outcomes (health status and medication use) were collected by the nurses, we cannot exclude a measurement bias for these variables. (3) The hypothesis of weaker intervention effects in ZIP code C compared with ZIP codes A and B was formulated after and not prior to the randomization. However, although these analyses must be classified as secondary analyses, they were conducted according to an analytic plan with hypotheses formulated before outcome data collection. Thus, the present study suggests that preventive home visitation programs are effective in preventing functional decline in low-risk older people and may result in long-term cost savings by reducing the number of nursing home admissions in this population, and that favorable effects are likely related to the home visitor's performance in conducting the visits.
Accepted for publication August 16, 1999.
This research was supported by grant 4032-35637 from the Swiss National Science Foundation, Bern, Switzerland, and by grants from the Cantonal Department of Health and Social Affairs, Bern; the W. K. Kellogg Foundation, Battle Creek, Mich; the Novartis Foundation for Gerontological Research, Basel, Switzerland; and the Visana Health Insurance Co, Bern.
We thank Robert E. Bjork, PhD, Arizona Center for Medieval and Renaissance Studies, Arizona State University, Tempe, for his editorial review of this manuscript.
Corresponding author (Europe): Andreas Stuck, MD, Geriatrics Research Unit, Zieglerspital, Morillonstrasse 75, CH-3001 Bern, Switzerland (e-mail: firstname.lastname@example.org). Corresponding author (United States): John C. Beck, MD, School of Medicine, University of California, Los Angeles, 10833 Le Conte, 32-144 CHS, Los Angeles, CA 90024-1687.
Create a personal account or sign in to: