The effect of interdisciplinary primary care teams on the use of health services by patients with multiple chronic conditions is uncertain. This study aimed to measure the effect of guided care teams on multimorbid older patients' use of health services.
Eligible patients from 3 health care systems in the Baltimore, Maryland–Washington, DC, area were cluster-randomized to receive guided care or usual care for 20 months between November 1, 2006, and June 30, 2008. Eight services of a guided care nurse working in partnership with patients' primary care physicians were provided: comprehensive assessment, evidence-based care planning, monthly monitoring of symptoms and adherence, transitional care, coordination of health care professionals, support for self-management, support for family caregivers, and enhanced access to community services. Outcome measures were frequency of use of emergency departments, hospitals, skilled nursing facilities, home health agencies, primary care physician services, and specialty physician services.
The study included 850 older patients at high risk for using health care heavily in the future. The only statistically significant overall effect of guided care in the whole sample was a reduction in episodes of home health care (odds ratio, 0.70; 95% confidence interval, 0.53-0.93). In a preplanned analysis, guided care also reduced skilled nursing facility admissions (odds ratio, 0.53; 95% confidence interval, 0.31-0.89) and days (0.48; 0.28-0.84) among Kaiser-Permanente patients.
Guided care reduces the use of home health care but has little effect on the use of other health services in the short run. Its positive effect on Kaiser-Permanente patients' use of skilled nursing facilities and other health services is intriguing.
clinicaltrials.gov Identifier: NCT00121940
The US health care system frequently fails to provide high-quality, well-coordinated, cost-effective health care to older Americans with multiple chronic conditions.1-3 Instead, care is fragmented, hospitals and emergency departments are overused, tests are duplicated, and patients and their families are disengaged from their care. As a result, the Medicare program's costs are unnecessarily and unsustainably high, and its chronically ill beneficiaries and their families are often dissatisfied.4
New models of interdisciplinary primary care have demonstrated the ability to improve multimorbid older patients' quality of care, quality of life, and satisfaction with care,5-8 but the cost of operating such teams is a barrier to their widespread implementation. Although better quality of outpatient care may reduce the need for acute and postacute care and thereby offset some of the cost of operating these teams, the degree to which this actually occurs is uncertain, ranging from negligible5-7 to modest.8,9
The model of interdisciplinary team care for multimorbid older patients tested most recently in a rigorous controlled trial is called guided care. This model builds on lessons from its predecessors in providing primary care that includes comprehensive geriatric assessment, evidence-based planning, case management, transitional care, self-management, and caregiver support. Guided care is provided by a team that includes a specially trained registered nurse, 2 to 5 physicians, and members of a primary care office staff. This team provides 8 clinical services to a panel of 50 to 60 of the practice's older patients at highest risk of using health care heavily during the following year. For each patient, the guided care nurse (1) performs a comprehensive assessment at home, (2) creates an evidence-based care guide and a patient-friendly version called an action plan, (3) monitors the patient on a monthly basis, (4) smooths the patient's transitions among sites of care, (5) coordinates the efforts of all the patient's providers of care, (6) uses motivational interviewing to promote patient self-management, (7) educates and supports family caregivers, and (8) facilitates access to appropriate community resources.10
Preliminary reports from a recent multisite cluster-randomized controlled trial (cRCT) suggest that guided care improves the quality of multimorbid older patients' chronic care,11,12 reduces family caregivers' strain,13,14 and increases primary care physicians' satisfaction with the care they provide to these patients.15 Data from the first 8 months of that cRCT also indicate that, on average, patients who received guided care had 24% fewer hospital days (odds ratio [OR], 0.76; 95% confidence interval [CI], 0.51-1.13), 37% fewer skilled nursing facility days (0.63; 0.35-1.15), 29% fewer episodes of home care (0.71; 0.47-1.08), and 15% fewer emergency department visits (0.85; 0.62-1.18) than the control patients, resulting in a net annual savings of $75 000 per nurse or $1364 (11%) per patient.16 Although these early reductions in use and costs of care were encouraging, they did not attain statistical significance.
In this article, we report the effects of guided care on multimorbid older patients' use of 6 health services through an additional 12 months of the aforementioned cRCT. Hypothesizing that this model of care may affect use differently among patients with differing insurance coverage and among those at different levels of risk, we also explore the effects of guided care on these same services among subgroups of patients defined by 2 characteristics: source of health insurance and baseline risk of using health services heavily in the future.
In 2006, we launched a cRCT of guided care in 14 primary care teams housed in 8 community-based primary care practices in urban and suburban neighborhoods in the Baltimore, Maryland–Washington, DC, metropolitan area. Six teams in 3 practices were operated by Kaiser-Permanente Mid-Atlantic States (KPMAS), a group-model managed care organization; 6 teams in 4 practices were operated by Johns Hopkins Community Physicians (JHCP), a statewide network of community-based practices; and 2 teams in 1 practice were operated by MedStar Physician Partners, a multisite group practice. The study population included older persons insured by Kaiser-Permanente (who received care at the KPMAS sites), traditional fee-for-service Medicare (who received care at the JHCP and MedStar sites), or the TRICARE/US Family Health Plan (USFHP), a federal health insurance program for retired military personnel and their dependents (who received care at the JHCP sites). A detailed description of the study design, which was approved by the 3 relevant institutional review boards, was published previously.11
Recruitment of participants
From April 1 through June 30, 2006, we screened the insurance claims of all patients of the 14 teams to identify those who were 65 years or older and at high risk of using health services heavily during the following year, as estimated by the claims-based hierarchical condition category (HCC) predictive model.17 High risk was equated with HCC scores in the highest quartile of the population of older patients covered by their primary health care insurer. Eligible high-risk patients who provided informed consent completed in-home baseline interviews.
Each of the 14 primary care teams included 2 to 5 primary care physicians, their office staff, and their consenting high-risk patients. The study's statistician, masked to the identities of the teams, randomly allocated 1 of each of the 7 pairs of teams to the guided care group; the other 7 teams constituted the usual care control group. Thus, clusters of patients, rather than individuals, were randomized according to the team of physicians who provided their primary care.
Registered nurses who had completed a course in guided care nursing18 joined their assigned primary care teams in May 2006. During the following 6 to 8 months, each guided care nurse was integrated into the practice and established a caseload of 50 to 60 guided care patients. The date on which each patient's care guide was created was set as his or her start date for receiving guided care. Patients in the usual care group continued to receive care from their established primary care physicians. Their start dates for receiving usual care were assigned to mirror the distribution of the guided care patients' start dates.
Information regarding all participants' baseline characteristics was obtained from prerandomization in-home interviews, and their baseline HCC scores were recalculated from insurance claims from the 12 months immediately preceding their start dates. Data regarding participants' use of health services during the cRCT were obtained from paid insurance claims. We computed each participant's annualized use of health services from his or her start date through June 30, 2008. An estimate of the use of long-term custodial nursing home days was obtained from participants' responses to a telephone interview question 8 and 20 months after their start dates: “How many days did you spend in a nursing home during the past year?” We counted as long-term custodial days only those self-reported nursing home days that exceeded a participant's days in nursing homes for postacute rehabilitation, as reflected by their insurance claims.
As described in detail previously,11 we imputed values for missing responses to baseline interview questions: 5 imputed data sets were generated and inferences were combined across data sets using Rubin's combining rules.19 Less than 1% of baseline responses were missing except for the question regarding finances (4% missing). In comparing the 2 study groups at baseline, we used site-stratified testing procedures to evaluate differences in all characteristics except health insurance because of its strong correlation with site.
We compared the guided care and usual care patients' use of health services during the intervals between their start dates and June 30, 2008. For each health service, we constructed a marginal regression model, which accounted for the correlation of multiple outcomes within individuals,20 to estimate the effect of guided care (vs usual care) on the mean units of service used per person per year. For each service, we constructed an a priori model of the logarithm of the mean rate as a linear function of treatment group, age, race, sex, educational level, finances, HCC score, self-rated health, activities of daily living, instrumental activities of daily living, and practice site, plus an offset term for exposure period. Regression parameters were estimated using generalized estimating equations with a working independence covariance structure. The estimated variance-covariance matrix of all the regression estimators was obtained using the sandwich variance technique.21,22 The adjusted treatment effect for each outcome is interpreted as the ratio of the mean units of service used per guided care recipient (vs usual care recipient) during a common exposure period.
We also estimated an approximate posterior (ie, Bayesian) probability that guided care reduced the use of a service by at least 10.0% by simulating 500 000 draws from a multivariate normal distribution in which the mean was set equal to the estimated regression parameters and the variance-covariance matrix was set equal to the estimated variance-covariance matrix.23 From these simulations, we computed, for each service, the proportion of effects that yielded greater than 10.0% improvement. These probabilities are presented to inform decision makers in health care organizations regarding the applicability of the study's findings to their daily operations.
We used these same methods to conduct the preplanned analyses of the effects of guided care within subgroups of study participants defined by 2 characteristics: health insurance plan and baseline HCC score (HCC score ≥1.6). All analyses were conducted using the statistical software R, version 2.9.1 (R Foundation for Statistical Computing, Vienna, Austria) and Stata, version 9.0 (StataCorp LP, College Station, Texas).
As shown in the Figure and described in detail previously,11 we screened 13 534 older patients for eligibility and offered 2391 eligible patients the option of participating in the study. Of these, 904 (37.8%) provided written informed consent and were randomized to receive guided care (n = 485) or usual care (n = 419). We excluded from the present analyses randomized participants who died before their start dates (2.4%), who did not have a start date before June 30, 2007 (1.5%), and whose insurance claims were unavailable (2.0%). We analyzed data regarding the remaining 850 participants (n = 446 in the guided care group, n = 404 in the usual care group). At the study baseline, there were no statistically significant differences between the guided care and usual care groups' sociodemographic characteristics or health status except that a higher percentage of the guided care patients reported having adequate money left at the end of each month (58.1% vs 51.2%, P = .01) (Table 1).
Table 2 gives the raw mean annual per capita use of health care services by patients in the guided care and usual care groups, as well as the adjusted ORs of service use (and 95% CIs) by guided care and usual care patients. Except for the 29.7% reduction in the use of home health care by guided care patients (OR, 0.70; 95% CI, 0.53-0.93), none of the differences represented by these ORs is statistically significant. The rightmost column shows the Bayesian probability that guided care reduced the use of each service by at least 10.0%.
Ten patients in the usual care group reported a total of 1928 long-term custodial days in nursing homes, whereas 4 guided care patients reported 610 such days. The small number of patients in both groups who received custodial care precluded meaningful testing of the statistical significance of this difference.
The results of the preplanned analysis of insurance subgroups indicated that guided care may have reduced the use of some health services among the study participants who were insured by Kaiser (and received their primary care from KPMAS clinics). As indicated in Table 3, guided care reduced the use of 7 services (mean annual per capita use). After adjustment, the differences between the groups' skilled nursing facility admissions (OR, 0.53; 95% CI, 0.31-0.89) and days (0.48; 0.28-0.84) were statistically significant. Given these data, as shown in the rightmost column, the probability of observing at least a 10.0% reduction in guided care patients' use of the 7 reduced services ranged from 39.8% (specialist visits) to 98.6% (skilled nursing facility days).
In the other 2 insurance subgroups (fee-for-service Medicare and TRICARE/USFHP), guided care had less effect on patients' use of health services. After adjustment, there were no significant differences in the use of any of the 9 services in the TRICARE/USFHP–insured group. In the fee-for-service Medicare-insured group, there were no significant differences between the 2 groups' use of 8 services, but the guided care patients had significantly more visits to specialist physicians (OR, 1.34; 95% CI, 1.06-1.70).
The other prespecified subgroup analysis evaluated the effect of guided care on health service use in subgroups defined by participants' baseline risk of generating high health care insurance expenditures in the future (ie, by HCC score of ≥1.6 or HCC score of <1.6). As indicated in Table 4, in the highest-risk subgroup, there were no statistically significant differences between the study groups' use of health services.
We undertook this analysis to determine whether the clinical improvements attributed to the guided care model (ie, enhanced quality of chronic care,11,12 greater physician satisfaction,15 and less family caregiver strain13,14) are associated with changes in patients' use of health services during the 20 months after their initial contact with a guided care nurse. Compared with the patients who received usual care, the patients who received guided care used similar amounts of care from hospitals, skilled nursing facilities, emergency departments, and physicians. The only statistically significant difference between the groups' use was a 29.7% reduction in home health care episodes by the guided care group. The Bayesian analysis indicates that the probability that guided care reduced home health episodes by at least 10.0% was 95.9%. The Bayesian analysis also suggests that the probability that guided care reduced 30-day readmissions by at least 10.0% was 74.9%. The guided care group also used less than one-third the number of long-term custodial days in nursing homes, but these days were associated with too few patients to allow meaningful statistical testing of the significance of this reduction.
Of note, in the sizable Kaiser-Permanente subgroup (n = 365, 42.9% of the entire sample), guided care appeared to reduce use to a greater extent than in the entire sample. Reductions in skilled nursing facility admissions (by 47.2%) and days (by 51.6%) were statistically significant. This subgroup's reductions in hospital admissions (15.0%), 30-day readmissions (48.7%), hospital days (20.7%), and emergency department visits (17.4%) were clinically significant because these services are so expensive, but they did not attain traditional levels of statistical significance.
These modest findings are similar to those reported from other studies of interdisciplinary team care for older persons with multiple chronic conditions. As noted previously, several such models have improved patients' quality of care and quality of life, but only one has shown significant reductions in patients' use of expensive hospital services. The GRACE (Geriatric Resources for Assessment and Care of Elders) model reduced high-risk patients' use of emergency departments and hospitals in the second year of an RCT.8
Why is it so difficult for interdisciplinary primary care teams to reduce multimorbid patients' use and cost of health services? The answer is probably multifactorial. First, chronic diseases are incurable and many are progressive. Thus, even high-quality, well-coordinated care cannot avert some of the exacerbations that require expensive acute care. Second, we are still learning how teams should be structured, organized, and operated to reduce unnecessary use and cost while preserving or improving patients' quality of care and quality of life. Third, there are few penalties or rewards that provide consistent incentives for teams to improve the quality and outcomes of their care. Value-based purchasing of health care may create such incentives in the future, but its actual effects are still uncertain. Fourth, we are still learning how to select the patients who are most likely to benefit from team care. Fifth, even when teams are successful in preventing the preventable, detecting statistically significant reductions in use is difficult in studies of relatively small samples of high-cost patients. Sixth, high-quality team care may appropriately increase the use of certain health services by some high-risk patients.
An interesting finding of this study is the apparent ability of the environment in which an intervention is implemented to affect its ability to achieve its desired outcomes. Guided care appears to have been more effective in reducing the use of skilled nursing facilities (and possibly other health services) among patients who were insured by and who received their primary care from Kaiser-Permanente. The reason for this finding is unknown, but it could include several factors. We checked to ensure that randomization was successful within the Kaiser-Permanente subgroup, and there were no significant differences between the Kaiser-Permanente patients or physicians who were randomly assigned to the guided care and usual care groups. The nurses assigned to the Kaiser-Permanente practices were similar in background, age, and skill level to the nurses assigned to the other study sites.
In contrast to the fee-for-service environment, however, Kaiser-Permanente has a sophisticated interoperable electronic medical record system and a medical center model that comprises a variety of health services, both of which are manifestations of a well-established culture that promotes and rewards team care, prevention, and avoidance of unnecessary care. Thus, guided care may have been assimilated into this culture more readily than into the practice cultures where the study participants insured by other carriers received their care. Guided care may simply have extended the existing approach at Kaiser-Permanente, whereas it may have required challenging changes in other organizations. Unfortunately, we did not design this study to measure environmental influences on the effectiveness of guided care, so we did not collect the information that would be necessary to analyze such influences. Thus, we can only speculate as to reasons why the patient outcomes were better in 1 environment than in the others. Additional studies will be necessary to determine with certainty which environmental factors facilitate or impede the effectiveness of this organizational intervention.
Several methodologic limitations to this study should be considered when interpreting its results. First, there are 2 threats to the study's external validity. The consent rate (37.8%), although better than average among randomized trials of health care interventions for chronically ill older persons,24 may have led to some selection bias, probably toward greater participation by healthier persons and less participation by sicker persons. In addition, the sampling universe was restricted to urban and suburban areas in the mid-Atlantic region of the United States, so its results may not apply to the populations of other geographic areas.
Second, the sample for this study was calculated to provide 80.0% statistical power to detect changes in clinical outcomes; thus, it provided only limited ability to detect differences in health care use, a secondary outcome, with statistical significance, especially in the subgroup analyses. Third, the 20-month period of observation may have been too short to detect medium- to long-term effects of the intervention, such as reductions in the vascular complications of diabetes mellitus. Fourth, the effect of guided care on patients' use of medications, although important and costly, was not evaluated because medications were not included consistently in the health insurance claims made available for this study. Medicare Part D was implemented during the study. Fifth, unmeasured baseline differences between the experimental and control groups could have biased the observed results of the study.
Finally, we are unable to report the actual aggregate costs of the 2 groups' health care because the prices of the services varied considerably among the 3 insurers and were considered proprietary by 1 insurer. Readers wishing to estimate the guided care model's costs and economic benefits in their local markets can (1) project the total costs for employing a guided care nurse in their markets (eg, $95 900 per year in the study reported here)16 and (2) convert the differences in use reported here to differences in health care costs by inserting their actual local prices for units of these services.
The search will continue for innovative models that improve the quality of care and the quality of life for older persons with chronic conditions. Because the resources needed to operate most such models would increase patients' total cost of care at least temporarily, it is important to evaluate the models' downstream effects on the use and cost of care. Models that reduce subsequent use and costs of care enough to offset their own costs are financially more feasible to adopt and sustain than models that increase the aggregate costs of health care, assuming that some of the reductions in insurers' expenditures are allocated to subsidizing the models' operating costs.
As heterogeneous forms of patient-centered medical homes, accountable care organizations, and similar entities are tested during the next several years, health care leaders will need to address all the challenges discussed herein: which models to use, whom to target, what incentives to create, and how to manage clinical teams. Research regarding innovative models will need to be designed rigorously to assess multiple clinical and financial outcomes and the role of organizational factors. Society will need to decide how much it is willing to pay for high-quality, complex chronic care.
Correspondence: Chad Boult, MD, MPH, MBA, Department of Health Policy and Management, Johns Hopkins Bloomberg School of Public Health, 624 N Broadway, Room 693, Baltimore, MD 21205 (email@example.com).
Accepted for Publication: July 31, 2010.
Author Contributions: Drs Boult, Reider, Leff, Frick, Boyd, Wegener, and Scharfstein had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Boult, Reider, Leff, Frick, Boyd, Wolff, Frey, Karm, Wegener, Mroz, and Scharfstein. Acquisition of data: Boult, Reider, and Frey. Analysis of data: Boult, Reider, Leff, Frick, Boyd, Mroz, and Scharfstein. Drafting of the manuscript: Boult and Reider. Criticalrevision of the manuscript for important intellectual content: Boult, Reider, Leff, Frick, Boyd, Wolff, Frey, Karm, Wegener, Mroz, and Scharfstein. Statistical analysis: Boult, Reider, Leff, Frick, Boyd, Wolff, and Scharfstein. Obtaining funding: Boult. Administrative,technical, and material support: Boult, Reider, Frey, Wegener, and Mroz. Studysupervision: Boult, Reider, Frey, and Karm.
Financial Disclosure: None reported.
Funding/Support: This study was supported by grant R-01HS14580 from the Agency for Healthcare Research and Quality, grant R01-HS14580 from the National Institute on Aging, and grant 2004-0335 from The John A. Hartford Foundation, as well as invaluable in-kind contributions from Kaiser-Permanente of the Mid-Atlantic States, Johns Hopkins HealthCare, and the Roger C. Lipitz Center for Integrated Health Care.
Role of the Sponsors: The sponsors had no role in the study design, data collection, data management, data analysis, or the preparation or approval of the manuscript.
Additional Contributions: The study would not have been possible without the expert technical assistance of Wade Kramer, MHSA, and Susan Kim, BS, CPA, of Kaiser-Permanente, Paula Norman, BS, of Johns Hopkins HealthCare, and Taneka Lee, BS, of the Lipitz Center—all of whose contributions were made within the scopes of their primary jobs.
Institute of Medicine, Crossing the Quality Chasm: A New Health System for the 21st Century. Washington, DC National Academy Press2001;
et al. The quality of health care delivered to adults in the United States. N Engl J Med
2635- 2645PubMedGoogle ScholarCrossref
NS Quality of care for older persons at the dawn of the third millennium. J Am Geriatr Soc
S346- S350PubMedGoogle ScholarCrossref
et al. Primary care experiences of Medicare beneficiaries, 1998 to 2000. J Gen Intern Med
991- 998PubMedGoogle ScholarCrossref
et al. Department of Veterans Affairs Cooperative Study Group on Home-Based Primary Care, Effectiveness of team-managed home-based primary care. JAMA
2877- 2885PubMedGoogle ScholarCrossref
et al. IMPACT Investigators. Improving Mood-Promoting Access to Collaborative Treatment, Collaborative care management of late-life depression in the primary care setting: a randomized controlled trial. JAMA
2836- 2845PubMedGoogle ScholarCrossref
et al. Effectiveness of collaborative care for older adults with Alzheimer disease in primary care: a randomized controlled trial. JAMA
2148- 2157PubMedGoogle ScholarCrossref
et al. Geriatric care management for low-income seniors: a randomized controlled trial. JAMA
2623- 2633PubMedGoogle ScholarCrossref
B Successful models of comprehensive care for older adults with chronic conditions. J Am Geriatr Soc
2328- 2337PubMedGoogle ScholarCrossref
et al. Early effects of “Guided Care” on the quality of health care for multimorbid older persons: a cluster-randomized controlled trial. J Gerontol A Biol Sci Med Sci.
321- 2327Google ScholarCrossref
et al. The effects of guided care on the perceived quality of health care for multi-morbid older persons. J Gen Intern Med
235- 242PubMedGoogle ScholarCrossref
et al. Guided care and the cost of complex healthcare: a preliminary report. Am J Manag Care
555- 559PubMedGoogle Scholar
et al. Risk adjustment of Medicare capitation payments using the CMS-HCC model. Health Care Financ Rev
119- 141PubMedGoogle Scholar
C Expanding the gerontological nursing role in Guided Care. Geriatr Nurs
358- 364PubMedGoogle ScholarCrossref
DB Multiple Imputation for Nonresponse in Surveys. New York, NY John Wiley & Sons1987;
S Analysis of Longitudinal Data. Oxford, England Oxford University Press2002;
PJ The behavior of maximum likelihood estimates under nonstandard conditions. Proceedings of the Fifth Berkeley Symposium on Mathematical Statistics and Probability.
Vol 1. Berkeley University of California Press1967;221- 233Google Scholar
H A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity. Econometrica
817- 830Google ScholarCrossref
P Soliciting defined populations to recruit samples of high-risk older adults. J Gerontol A Biol Sci Med Sci
M379- M384PubMedGoogle ScholarCrossref