: To examine the association of Alzheimer’s disease (AD) with common chronic conditions, acute care events, and risk of hospitalization.

: Community-dwelling subjects with a diagnosis of and/or medication for AD were matched to subjects without AD based on age, sex, and geographic region. Administrative claims from commercially insured health plans for medical and pharmacy services provided from January 1, 2000, to March 31, 2006 (inclusive) were analyzed. The Deyo Charlson Index (DCI) was used to assess the number of chronic conditions. The outcomes of interest were risk of fractures and hospitalization.

: Among 5396 persons with AD and a matched cohort of 5396 persons without the condition, subjects with AD were more likely to have a diagnosis for any of the DCI components, had a higher rate of fractures (17.7% vs 7.9%, 

<.00) and other urgent medical events (eg, pneumonia 14.0% vs 6.3%, 

<.00), and were more likely to be hospitalized (odds ratio = 1.7; 95% confidence interval = 1.5, 1.9). There were significant differences in the medication use between the 2 groups, with the use of psychotics/tranquilizers 9-fold higher among persons with AD.

: Persons with AD have higher odds of experiencing a fracture, being hospitalized, and requiring other acute care medical services than those without AD. The disease also is associated with a higher prevalence of common chronic conditions.

Alzheimer’s disease is a debilitating condition that affects both community and institutionalized elderly persons.

Persons with newly diagnosed Alzheimer’s disease have a higher prevalence of cardiovascular disease, diabetes, and various other conditions than age- and sex-matched controls.

The risk of bone fracture is significantly higher among persons with Alzheimer’s disease.

Persons living in the community with Alzheimer’s disease are hospitalized more frequently than age- and sex-matched controls.

With an aging society, the burden of Alzheimer’s disease (AD) is becoming a greater concern for society and especially healthcare payers. Alzheimer’s disease is typically characterized by progressive memory loss, confusion, and a variety of other cognitive and functional deficits. Although the exact mechanism of AD is not completely understood, recent studies suggest that the presence of beta-amyloid plaques and deposits combined with tangles, twists, or strands of the protein tau may be responsible for neuronal and cognitive loss.

 In 2007, the number of persons in the United States with AD was estimated to be approximately 5 million, with more than 24 million worldwide.

 The overall prevalence of AD among persons over age 65 years is 5.7%, increasing from approximately 3% for persons age 65 to 70 years to more than 30% for persons age 85 years or older.

The estimated lifetime cost of persons with AD has been estimated to be $174,000.

 In 2005, Medicare spent $91 billion for healthcare for persons with AD, with $21 billion associated with long-term care services.

 The excess cost burden to managed care organizations of persons with AD has been estimated to be more than $4000 per person per year.

To date, the bulk of the literature on the economic burden of AD has focused on 3 aspects: costs associated with long-term care, informal care costs (ie, the costs associated with loved ones providing around-the-clock care to AD patients living at home), and the incremental impact of AD on the cost of treating other chronic conditions.

 This body of literature suggests that direct medical costs are primarily focused on the later stages of the disease (long-term care costs) or are associated with more subtle effects such as poor treatment adherence leading to worse outcomes in other illnesses such as diabetes or hypertension. However, AD can be associated with significant events that directly increase medical expenditures. Previous research has shown AD to be a risk factor for falls in the elderly, and hospitalization rates have been shown to increase with increasing AD severity.

The purpose of this analysis was to expand on these reports by providing a more detailed look at resource use in AD patients, focusing on those who still are living in the community and therefore are theoretically at an earlier and less severe stage of disease. Specifically, AD patients and age- and sex-matched peers from a number of managed care plans across the United States were compared with respect to rates of comorbid chronic illness, presence of acute events, pharmacologic treatment, and associated direct medical costs in hopes of providing a broader perspective of the economic impact of AD.

This study used a historical cohort design to examine comorbid conditions and costs among persons with AD and persons without the condition. Administrative claims from commercially insured health plans for medical and pharmacy services provided from January 1, 2000, to March 31, 2006 (inclusive) were analyzed. HealthCore, as a subsidiary of Well-Point Inc, maintains the integrated research database used for the analysis. The data consisted of a deidentified, limited-use data set and were compliant with applicable state and federal laws, including the Health Insurance Portability and Accountability Act, and regulations concerning privacy and security of patient identifiable health information.

During the study period, approximately 20.4 million lives were included in the data set, with 2% coming from the mid- Atlantic, 26% from the southeastern, 8% from the central, and 64% from the western United States; 8% were over age 65 years; and approximately 8% of the total population were individuals participating in Medicare Advantage programs.

Two cohorts were identified: persons with AD and persons without AD. Persons with AD were required to have at least 1 medical claim with a diagnosis indicating AD—based on 

International Classification of Diseases, Ninth Revision, Clinical Modification

) code 331.0x—and/or a pharmacy claim for a cholinesterase inhibitor or memantine. (For a list of the specific agents, see 

.) Subjects with a diagnosis of alcohol- or drug-induced dementia were excluded, as were subjects with fewer than 24 months of continuous data centered on the first diagnosis or medication claim (index date). The comparison cohort was matched to the AD cohort based on age (± 1 year), sex, and geographic region. Non-AD subjects were assigned an index date that corresponded to that of the matched person with AD. Subjects younger than age 50 years on the index date were excluded from the study. Subjects with AD and no matched non-AD subject were excluded (n = 3027), as were AD patients with prescription claims for acetylcholinesterase inhibitors and/or memantine before their index date (n = 1143).

Comorbid conditions were identified based on diagnoses associated with medical encounters. For this study we used the Deyo modification of the Charlson index (DCI).

 Because dementia is one of the conditions contained in the DCI, we excluded diagnosis codes for dementia from the DCI calculation. We also examined the prevalence of other conditions believed to be commonly associated with AD and/or common medical conditions using 

 codes we evaluated the frequency of infections, injuries, and other acute events occurring during the study period. Fractures were of particular interest because persons with AD are known to be at higher risk of falls. We examined the types of fractures experienced by subjects in both the AD and non-AD groups.

We examined the use of medications that are broadly classified into the therapeutic class of psychotropic medications. For this study psychotropic medications included antiepilepsy therapies, antidepressants, antipsychotics/tranquilizers, anxiolytics/sedatives, hypnotic agents/sleep aids, and narcotic analgesics. These medications were identified via National Drug Code numbers on the pharmacy claims.

We compared the summed DCI score for both the 12 months before and the 12 months after the index date. A nonparametric Savage scores test was used to compare DCI comorbidity scores. The prevalence of each of the specific DCI conditions also was evaluated. The McNemar 

 test for matched pairs was performed to compare the test and control groups for the presence of individual DCI components and other conditions/comorbidities. Where expected cell counts were <5, Fisher exact 

 values were obtained in the corresponding comparisons. In addition, the presence of conditions considered by the investigators to be clinically relevant to AD was compared between the 2 groups. Because it was hypothesized that fracture diagnoses would be common among the AD cohort, the types of fractures that occurred during the course of the study period were more closely examined to determine whether the incidence differed between skeletal sites. We modeled the risk of any fracture using logistic regression, controlling for comorbid conditions, age, sex, and use of narcotics, antidepressants, anxiolytics/sedatives, antipsychotics, and hypnotics. The prevalence of psychotropic medication use was compared between the groups using the McNemar 

 test. Finally, we also evaluated the risk of hospitalization controlling for age, sex, and comorbid conditions.

A total of 5396 subjects with AD met all inclusion criteria. A cohort of 5396 matched non-AD subjects also were included in the study. 

 displays the demographic characteristics of the 2 cohorts. Overall, 38.3% were male and the mean (SD) age was 76.9 (10.2) years; median age for the sample was 78.7 years. For persons in the AD cohort, more than half (n = 3141, 58.2%) were identified as having AD based on medication use, with few subjects (n = 113, 2.1%) having both an AD diagnosis and a prescription claim for an AD medication on the same date. Also, most AD patients (n = 4242, 78.6%) had 2 or more AD claims during the study period.

, the mean (SD) DCI score was 2.1 (2.2) for the AD cohort versus 1.2 (1.8) for the non-AD cohort (

 <.0001). The prevalence of 12 of the 14 conditions that comprise the DCI was significantly higher in the AD cohort than in the non-AD cohort (

<.003). Table 2 also lists other chronic comorbidities occurring in the AD and non-AD cohorts. All of these comorbidities (anxiety, cardiac arrhythmias, cardiomyopathy, depression, dyslipidemia, epilepsy, hypertension, osteoporosis, other nonorganic psychoses, Parkinson’s disease) were more frequently identified among those persons in the AD cohort (

<.0001). The percentage of patients with at least 1 claim for Parkinson disease was 10-fold higher in the AD cohort than in the non-AD cohort. Epilepsy was 6 times more prevalent, depression was 5 times more prevalent, and anxiety was 3 times more prevalent in the AD cohort than in the non-AD cohort.

Persons with AD were significantly more likely to have diagnoses associated with adverse health outcomes (anemia, respiratory infections, burns, complications of surgical or medical care, complications of diabetes, fractures, malnutrition, open wounds, pneumonia, poisonings, sprains and strains, superficial injuries, and venous thrombosis) compared with persons without AD (

<.001) (data not shown). There were no significant differences with respect to diagnoses associated with dislocations or diabetes with ophthalmic manifestations.

Subjects with AD were significantly more likely to have a diagnosis for a fracture than subjects without AD. Specifically, 955 (17.7%) subjects in the AD group had at least 1 fracture, compared with 428 (7.9%) in the non-AD cohort (

<.001). The most common fracture type in both groups was fracture of the neck of the femur (5.3% of the AD cohort and 1.4% of the non-AD cohort, 

<.01). Vertebral fractures also were relatively common (3.3% of AD cohort vs 1.2% of non-AD cohort, 

<.01), as were a number of other fracture types in the AD cohort (

A logistic regression analysis was conducted to evaluate predictors of a fracture (

). The adjusted odds ratio for the presence of a fracture was 1.9 (95% confidence interval = 1.6, 2.1) for persons with AD compared with persons without the condition, after controlling for comorbid conditions and patient characteristics. Other signification characteristics that were associated with higher odds ratios for a fracture included being female; older age; diagnosis of osteoporosis, congestive heart failure, cerebrovascular disease, or liver disease; and use of narcotic or antidepressant medications.

Of the 5396 subjects in the AD cohort, 1338 (24.8%) had at least 1 acute care inpatient stay. Among the non-AD group, 638 (11.8%) subjects had at least 1 hospitalization. We examined factors that were significantly associated with hospitalization (

). All independent predictors, except for presence of mild liver disease and AIDS, were associated with higher odds ratios for being hospitalized. After adjusting for these other conditions, having AD was associated with an odds ratio of 1.68 (95% CI = 1.51, 1.86) for being hospitalized during the study period.

The primary findings of this study demonstrate the substantial burden that AD places on those unfortunate enough to be afflicted with the condition. This study found that persons with AD have an increased number of chronic comorbidities, experience more negative health events including falls and infections, receive more psychotropic medications, and are more likely to be hospitalized than an age-, sex-, and geographic region—matched cohort of subjects without AD. Taken separately, these findings are not new. However, this study is the first known analysis of these factors performed simultaneously, in a population with what might be considered relatively early AD (ie, patients who remain in the community), wherein AD is an independent contributor to morbidity.

The results of our study are supported by others who also have found a higher prevalence of chronic conditions in persons with AD. Recently, Joyce et al found a higher prevalence of diabetes, hypertension, acute myocardial infarction, coronary artery disease, cerebrovascular diseases, and peripheral vascular diseases in persons with AD compared with a matched cohort of persons without AD.

 More specifically, the mean (SD) DCI score among persons with AD was 1.7 (1.9), compared with 1.2 (1.7) for persons without AD. Those findings are similar to ours; in our study the DCI was 2.1 (2.2) and 1.2 (1.8) for persons with and without AD, respectively. Similar results have been found in other studies.

Other research also documented the excess burden of comorbidities in persons with AD. Hill et al found that persons with AD have a higher prevalence of cerebrovascular disease (39% vs 10%), congestive heart failure (30% vs 14%), chronic pulmonary disease (25% vs 20%), diabetes (22% vs 16%), peripheral vascular disease (16% vs 9%), myocardial infarction (13% vs 5%), and renal disease (8% vs 3%) compared with age- and sex-matched controls.

The high rate of fractures and other acute events among persons with AD is striking. Our study found that persons with AD had an increased risk of 86% for a fracture, with most occurrences being hip fractures (ie, neck of femur, pelvis). Previous studies have noted a 3-fold higher risk of falls among persons with AD, though the majority of these studies focused on nursing home populations.

 Weller and Schatzker reported an increased risk of 78% (95% CI = 1.01, 3.14) for hip fractures for persons with AD.

 Other studies have reported similar odds ratios for risk of hip factures.

 In a 3-year cohort study of 157 persons with Alzheimer-type dementia, 50% of the subjects fell or became unable to walk.

 Fifteen percent of the subjects experienced a fracture, most commonly of the hip. These findings parallel the results from our study, which reports a fracture rate of 17.7%. As these events were identified via medical claims (ie, the patients were treated at a physician’s office, clinic, or hospital), it is also possible that the rates reported here are an underestimation of the true rate, as some patients may not have sought treatment.

We also found higher rates of contusions, sprains/strains, superficial injury, and open wounds, which may be indicative of falls. Recent research has highlighted that persons with AD have disturbances in cognitive function that affects gait function and movement.

 Persons with dementia have a shorter step, lower gait speed, greater step-to-step variability, and larger sway than persons without AD.

More importantly, falls in the elderly have been linked to higher mortality rates.

It has been established that use of certain psychoactive medications increases the risk of falls and hip fractures among the general elderly population.

 Use of benzodiazepines is a known risk factor for falls among the elderly.

 Use of 3 or more psychoactive medications also has been linked to an increased risk of falling.

Our study found that persons with AD were significantly more likely to be hospitalized—even after controlling for patient demographics and other chronic conditions. Hospitalization and long-term care contribute to higher direct medical care costs among persons with AD. Joyce and colleagues found that total healthcare costs for AD patients were 5 times greater than those for a matched cohort.

 The major cost driver in that analysis was inpatient care costs, accounting for 76% of average annual expenditures. Similar results have been reported by Fillit and colleagues.

This study has several limitations. Administrative claims do not capture disease events and episodes in which treatment is not sought or in which treatment is delivered by a provider not within the health plan. It is possible that differences between AD and non-AD groups are due to factors not considered or available within claims databases. However, a thorough and broad inclusion of relevant demographic and clinical characteristics was done in our multivariate analyses in the attempt to control for differences between groups. Miscoding and misdiagnosis does occur in administrative data sets like the one used in this analysis. In addition, administrative data provide no information on the severity of disease. Finally, these patients are not representative of Medicare beneficiaries because the study sample included only Medicare Advantage enrollees.

This study documents the high burden of AD and also highlights the risk among AD patients for acute events such as falls, injuries, infections, and poisonings. Persons with AD are at significant risk of hospitalization and fractures, with approximately 1 in 6 having a fracture claim and 1 in 4 being hospitalized for some period of time during the 12-month study period. Although further research will be required to better understand the relationship between AD severity and risk of these high-cost medical events, the high rate of psychotropic medication use in this population suggests a potential target for intervention.

: From the College of Pharmacy (DCM), University of Arizona, Tucson, AZ; Elan Pharmaceuticals (TPM, CL, HMA), San Francisco, CA; Healthcore, Inc (PMW, MJC), Wilmington, DE; and Wyeth Research (LMM), Collegeville, PA.

: This study was funded by Elan Pharmaceuticals and Wyeth Research.

Articles

CLINICAL

Community-based persons with Alzheimer’s disease have a higher risk of fractures, hospitalization, and various comorbidities than persons without the disease.

Burden of Alzheimer's Disease and Association With Negative Health Outcomes

: Guided Care (GC) is a model of proactive, evidence-based comprehensive healthcare provided by physician-nurse teams for people with several chronic health conditions. Our objective was to evaluate the preliminary effects of GC on health service utilization and costs.

: Cluster-randomized controlled trial of GC involving 14 primary care teams (49 physicians) and 904 of their chronically ill patients age 65 years or older.

: Using insurance claims, we compared the health services used by patients who received GC with the health services used by patients who received usual care during the first 8 months of the study.

: After adjustment for baseline characteristics, GC patients experienced, on average, 24% fewer hospital days (95% confidence interval [CI]: 49% fewer, 13% more), 37% fewer skilled nursing facility days (95% CI: 65% fewer, 5% more), 15% fewer emergency department visits (95% CI: 38% fewer, 18% more), and 29% fewer home healthcare episodes (95% CI: 53% fewer, 8% more), as well as 9% more specialist visits (95% CI: 8% fewer, 29% more). Based on current Medicare payment rates and GC costs, these differences in utilization represent an annual net savings of $75,000 (95% CI: −$244,000, $150,900) per nurse, or $1364 per patient. 

: Initial introduction of GC into primary care practices may be associated with less use of expensive health services and a net savings in healthcare costs among older patients with several chronic health conditions. Final results from the remaining 2 years of this ongoing study will be published in 2011.

Guided Care is a model of proactive, comprehensive healthcare by physician-nurse teams for people with several chronic health conditions. If its promising early effects persist and reach statistical significance during the remainder of an ongoing cluster-randomized controlled trial and are replicated during upcoming demonstration programs, Guided Care may:

Attract public and private health insurance coverage in the future.

Become adopted widely by US primary care practices.

Healthcare for older persons with chronic conditions is very expensive.

 Growing recognition of these costs has led to the development and promulgation of numerous new disease and care management strategies designed to provide high-quality, costeffective chronic care. One such approach is Guided Care (GC).

 Guided Care is provided by a practice-based team that includes a registered nurse, 2 to 5 physicians, and the other members of the office staff. This team provides 8 clinical services to a panel of 50 to 60 of the practice’s highest-risk older patients. For each patient, the Guided Care nurse (GCN) (1) performs a comprehensive assessment at home, (2) creates an evidence-based care guide, (3) monitors and coaches the patient monthly, (4) coordinates the efforts of all of the patient’s providers of healthcare, (5) smooths the patient’s transitions between sites of care, (6) promotes patient self-management, (7) educates and supports family caregivers, and (8) facilitates access to appropriate community resources.

In a pilot study during 2003-2004, GC appeared to improve the quality of chronic care

 In the initial months of the present cluster-randomized controlled trial (cRCT), patients and physicians rated GC more highly than usual care (UC).

 The aim of the present analysis was to evaluate the effects of GC on the use and costs of health services during the first 8 months of this cRCT. We report these preliminary results now to help inform business and policy decisions related to improving chronic care that must be made before the publication of the study’s final results in 2011.

In 2006, we launched a cRCT of GC in 8 community-based primary care practices located in urban and suburban neighborhoods in the Baltimore, MD—Washington, DC, metropolitan area. Three practices were operated by Kaiser Permanente, a group-model managed care organization; 4 were operated by Johns Hopkins Community Physicians, a statewide network of community-based practices; and 1 was operated by Medstar Physician Partners, a multisite group practice. The study population comprised older persons insured by traditional fee-for-service Medicare (34%) or managed care plans (45% Kaiser Permanente, 21% Tricare/USFHP, a federal health insurance program for retired military personnel and their dependents). Additional details about the study, which was approved by 3 relevant institutional review boards, have been published previously.

We screened the insurance claims of all patients of the 8 practices to identify those who were age 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.

 “High risk” was equated with HCC scores of 1.2 or higher, which identifies persons in the highest quartile of risk. Eligible high-risk patients who provided informed consent completed in-home baseline interviews.

Within the 8 practices, we identified 14 “pods,” each of which consisted of 2 to 5 primary care physicians and their consenting high-risk patients. The study’s statistician, blinded to the identities of the pods, randomly allocated each pod (physicians and their patients) to either GC (7 pods) or UC (7 pods).

Registered nurses who had completed a course in Guided Care Nursing joined their assigned pods in May 2006. During the following 6 to 8 months, each GCN established a caseload of 50 to 60 GC patients. The date on which each patient’s care guide was developed was his/her start date for receiving GC.

Patients in the UC group continued to receive the same care they had received before enrolling in the study. We randomly assigned their start dates for receiving UC to reflect the GC patients’ distribution of start dates.

Information about baseline characteristics was obtained from the prerandomization in-home interviews. Data about health service utilization were obtained from paid insurance claims for services used during the period from 1 year before the person’s start date through June 30, 2007. To maximize the capture of most health services, we waited until late November 2007 to begin the process of querying, cleaning, and merging the 3 insurers’ claims databases. We calculated participants’ baseline HCC scores from claims for services used during the 12 months before their start dates. We grouped these health services into 9 categories: hospital, skilled nursing facility, primary care physician, specialty physician, home healthcare, emergency department, durable medical equipment, tests (radiologic and laboratory), and nonmedication treatments (eg, physical therapy, outpatient procedures).

 we imputed values for missing baseline interview questions. We compared the GC and UC patients’ use of health services during the intervals between their start dates and June 30, 2007. For each category of healthcare service, we constructed a marginal regression model to estimate the effect of GC (vs UC) on the mean units of service used per person per year (

). For each category, we modeled the logarithm of the mean as a linear function of treatment group, age, race, sex, education, 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. The adjusted treatment effect for each outcome was interpreted as the ratio of the mean units of service used per GC recipient (vs UC recipient) over a common exposure period.

To facilitate computation of the difference in costs associated with GC, we first quantified the average units of service used annually by the recipients of UC and multiplied these units by 55 (the average number of patients in a GCN’s caseload). Then we computed the analogous average units of service used by 55 recipients of GC by multiplying the UC units of service by the treatment effects estimated by the marginal regression models. Finally, we calculated the difference between the 2 groups’ aggregate healthcare costs by multiplying the differences between the groups’ annualized use of services by Medicare’s average payment per unit of service.

 Excluded from these estimates are the costs of durable medical equipment, emergency department visits, tests, and treatments, which are extremely heterogeneous and for which average Medicare payment rates are not available.

Confidence intervals for the annualized use of services and cost differences were derived by using a 2-stage bootstrap procedure. In the first stage, individuals were resampled within practices. In the second stage, treatment effect estimates (on a log scale) were sampled from a multivariate normal distribution in which the mean was equal to the estimated treatment effect (on a log scale) and the variance-covariance matrix was equal to that obtained from the generalized estimating equation procedure. A total of 100,000 bootstrap samples were generated, and confidence intervals were formed using the percentile method.

 13,534 older patients were screened for eligibility, 1763 were offered the option of participating, and 904 were randomized by pod to receive either GC (n = 485) or UC (n = 419). We excluded from these analyses utilization by participants whose insurance claims were unavailable (1.4%) and those who rescinded consent (1.8%), died before their start dates (2.2%), or did not have start dates before June 30, 2007 (2.2%). We analyzed data on the remaining 835 participants (n = 433 in the GC group, n = 402 in the UC group).

At baseline, the demographic features, comorbidity, and HCC risk scores of the 2 groups were similar, but the GC participants were less likely to experience economic challenge (10.4% vs 15.5% lacked “enough money at the end of the month”; 

= .05) or functional impairment (21.7% vs 28.3% reported difficulty with 2+ instrumental activities of daily living; 

= .03), and they were more likely to have better health (23.1% vs 17.3% reported “excellent” or “very good” health; 

<.001) and to be insured by Tricare/USFHP (26.8% vs 16.5%, 

= .001). Other details of these baseline characteristics have been reported previously.

 The mean period of observation—from participants’ start dates through June 30, 2007&mdash;was longer for the GC group (8.3 months vs 7.8 months; 

<.05); 83% of all participants were observed for at least 6 months.

Table 1 reports the mean annual per capita use of health services by GC and UC patients, as well as the crude and adjusted ratios of GC to UC service use. On average, GC patients tended to use fewer units of hospital care, skilled nursing facility care, emergency care, and home healthcare, but more specialist care, durable medical equipment, tests, and treatments. Of the GC and UC patients, respectively, a relatively small percentage used hospital (26% vs 30%), home health (20% vs 22%), and skilled nursing facility (8% vs 9%) services.

 displays the financial savings and costs associated with providing GC to 55 high-risk patients for a year. Guided Care produced a net savings of $75,000 (95% confidence interval: −$244,000, $150,900), two-thirds of which accrued from reductions in hospital utilization.

Guided Care is a model of comprehensive care designed for older adults with chronic conditions that addresses all of the elements of the Chronic Care Model

 in the primary care setting. From the preliminary analyses reported here, GC appears to reduce multimorbid older patients’ use of hospital care, skilled nursing facility rehabilitation, home healthcare, and emergency care. The avoided costs of these services appear more than sufficient to offset the costs of providing GC, as well as the cost of a slight increase in care from specialist physicians. We observed an average net savings of $75,000 per GCN per year. The wide confidence intervals around these estimated averages indicate that these 8-month findings are not precise; these findings should be interpreted with appropriate caution. The remaining 24 months of this cRCT will provide additional utilization data, from which financial conclusions may be drawn with greater certainty.

There are several limitations to these preliminary findings. Two problems are related to the use of pods for randomization. First, the provision of GC to patients in 1 pod within a practice could have “contaminated” the care provided to patients in the UC pod within the practice. Although we saw no evidence that this occurred, it had the potential to reduce the measured differences between the GC and UC groups throughout the study. Second, our analyses assumed a common treatment effect across pairs of pods. Although some pods may have implemented GC more effectively than others, this study was not powered to evaluate such heterogeneity.

Third, the use of the Centers for Medicare & Medicaid Services national average payment amounts to calculate the costs of the health services used by the GC and UC groups (and the use of mid-Atlantic nurse salaries) limits the applicability of the reported net savings in geographic areas where the actual costs of these services are higher or lower. Local estimates could be derived by substituting local costs of services and nurses. Fourth, the heterogeneity and lack of estimates for the average costs of emergency care, durable medical equipment, tests, and treatments prohibited including these items in our aggregate cost analyses.

Finally, the observed differences between the GC and UC groups’ use of health services during the initial 8-month rollout of GC did not meet the traditional criterion for statistical significance (ie, 

<.05). We hypothesize that as the GCNs, physicians, patients, and family caregivers become increasingly familiar with GC, the reductions in the use and cost of health services observed during the first 8 months will become more pronounced. This hypothesis is supported by the trajectory of utilization observed in a recent RCT of a similar model of care; the difference between the 2 groups was not statistically significant until the second year of the study.

We present these preliminary findings now to help inform the many decisions that will be made in the next few years regarding the transformation of chronic care in the United States.

The authors acknowledge the invaluable contributions to this analysis made by data analysts Wade Kramer, MHSA, and Susan Kim, BS, CPA, of Kaiser Permanente, and Paula Norman, BS, of Johns Hopkins HealthCare.

: From the School of Medicine (BL, KDF, DOS, CMB, CB) and the School of Public Health (BL, LR, KDF, DOS, CMB, KF, CB), Johns Hopkins University, Baltimore, MD; and Kaiser Permanente Mid- Atlantic States (LK), Rockville, MD.

: This study was supported by the Agency for Healthcare Research and Quality (grant R0I HS14580) The John A. Hartford Foundation (grant 2004-0335), The National Institute on Aging (grant R01 HS14580), Kaiser Permanente Mid-Atlantic States, Johns Hopkins HealthCare, the Jacob and Valeria Langeloth Foundation (grant IR0I AG022010), and the Roger C. Lipitz Center for Integrated Healthcare.

: All authors (BL, LR, KDF, DOS, CMB, KF, LK, and CB) report no relationship or financial interest with any entity that would pose a conflict of interest with the subject matter of this article. Authorship Information: Concept and design (BL, DOS, CMB, KF, CB); acquisition of data (LR, DOS, CMB, KF, CB); analysis and interpretation of

data (BL, LR, KDF, DOS, CMB, LK, CB); drafting of the manuscript (BL, LR, KDF, DOS, CB); critical revision of the manuscript for important intellectual content (BL, LR, KDF, DOS, CMB, LK, CB); statistical analysis (LR, DOS, CB); provision of study materials or patients (KF, LK); obtaining funding (DOS, CB); administrative, technical, or logistic support (KF, LK, CB); and supervision (KF, LK, CB).

: Bruce Leff, MD, School of Medicine, Johns Hopkins University, 5505 Hopkins Bayview Cir, Baltimore, MD 21224. Email: bleff@jhmi.edu.

 Prevalence, expenditures, and complications of multiple chronic conditions in the elderly. 

 Guided care for multi-morbid older adults. 

 A pilot test of the effect of guided care on the quality of primary care experiences for multimorbid older adults. 

4. Sylvia ML, Griswold M, Dunbar L, Boyd C, Park M, Boult C.

 Guided care: cost and utilization outcomes in a pilot study. 

 The early effects of “guided care” on the quality of health care for multi-morbid older persons: a clusterrandomized controlled trial. 

 Risk adjustment of Medicare capitation payments using the CMS-HCC model. 

7. Centers for Medicare & Medicaid Services.

Discharge, Total Days of Care, Total Charges, and Program Payments for Medicare Beneficiaries Discharged From Short-Stay Hospitals, by Type of Entitlement: Calendar Years 1972-2006.

 http://www.cms.hhs.gov/MedicareMedicaidStatSupp/downloads/2007Table5.1b.pdf. Accessed October 17, 2008.

 Improving primary care for patients with chronic illness: the chronic care model, part 2. 

9. Counsell SR, Callahan CM, Clark DO, et al.

 Geriatric care management for low-income seniors: a randomized controlled trial. 

TRENDS FROM THE FIELD

The early months of a cluster-randomized trial suggest that Guided Care is associated with a trend toward a net reduction in total healthcare expenditures.

Guided Care and the Cost of Complex Healthcare: A Preliminary Report

: To calculate intervention costs and the potential cost offset of a care management intervention that substantially improved the quality of dementia care.

: From both a payer perspective and a social planner perspective, we analyzed data from a cluster randomized controlled trial (RCT) evaluating this intervention versus usual care. The RCT included 408 pairs of older adults with dementia and their caregivers. Caregivers were surveyed at baseline, at 12 months, and at 18 months to assess patient healthcare utilization and out-of-pocket costs.

: We calculated fixed and per-patient intervention costs from RCT records. We combined the monthly per-patient costs of healthcare services, informal caregiving, and out-of-pocket costs, and we conducted multivariate analyses comparing this sum (potential cost offset) for intervention versus usual care patients. Covariates included patient age, sex, and baseline costs. We limited the main analysis to patients who survived until the 12-month survey or the 18-month survey.

: The intervention required a start-up cost of $70,256 and mean intervention per-patient permonth costs of $118. There were no significant differences in the mean monthly cost of healthcare and caregiving services for intervention versus usual care patients using the societal perspective (difference of −$555 per month, 

= .28) or the payer perspective (difference of −$219 per month [including nursing home costs], 

= .55; difference of −$256 per month [excluding nursing home costs], 

: Although this analysis of a dementia care management intervention did not demonstrate a significant cost offset, the intervention may represent a worthwhile approach to improving the quality of care and health outcomes for patients with dementia and their caregivers.

This cost analysis of a dementia care management program proven to improve quality and outcomes did not show a definitive cost offset over 18 months but may represent a cost-effective approach to providing high-quality dementia care in managed care settings.

Few programs that demonstrably improved quality of dementia care have been reported.

This program successfully connected patients and caregivers to existing community services, as recommended in guidelines.

This detailed cost analysis from the payer and social planner perspectives provides managed care decision makers with data on fixed and ongoing program costs, as well as on potential reductions in other healthcare and caregiving costs.

Dementia, defined as an acquired persistent impairment in 2 or more cognitive areas (eg, language and memory), is a leading cause of disability in older adults.

 Dementia of the Alzheimer type, the most common variant, affects more than 4 million Americans.

Dementia care represents a substantial expense for healthcare organizations and society, with the combined costs for formal services (inpatient and outpatient medical care) and informal caregiving exceeding $25,000 per patient per year in most studies.

 The overall cost burden will escalate significantly over the coming decades as the American population ages and the prevalence of dementia increases.

Existing practice guidelines for dementia care reflect evidence from randomized controlled trials (RCTs) showing that assistance to caregivers can reduce the severity of patients’ symptoms, decrease the use of acute care services, and delay institutionalization, potentially lowering costs over time.

 Unfortunately, adherence to these  guidelines is variable across health systems.

 Collaborative care management programs emphasize coordinated and comprehensive approaches to improving quality of care. These programs are most effective for diseases such as dementia that are high-volume conditions primarily managed in the outpatient setting, have substantial variability in treatment, and rely on coordination with community agencies for recommended social services.

 Large healthcare organizations,  particularly managed care systems, have an opportunity to implement care management programs for patients with dementia that improve quality of care, reduce the burden on caregivers, and potentially have long-term cost offsets.

In a recent evaluation of an RCT of a care management intervention for dementia, it was found that patient health-related quality of life, caregiving quality, social support, level of unmet caregiving  assistance needs, and adherence to published dementia care guidelines were better for intervention patients compared with patients receiving usual  care.

 Although this intervention substantially and broadly improved the quality of dementia care, health plan administrators need data on start-up costs and cost offsets to weigh the value of such programs. Some or all of the initial financial expenses may be recovered over time if the intervention provides an ongoing net savings (eg, by  reducing the need for costly services such as hospitalizations).

To understand the economics of this intervention, we calculated the associated start-up and annual fixed costs and projected the per-patient variable costs of the intervention operating at full capacity. We hypothesized that, among patients who survived for the entire study period, the costs of healthcare and caregiving services would be lower for patients in the intervention arm compared with patients receiving usual care from both payer and societal perspectives. We also hypothesized that savings associated with the intervention would represent a complete offset of the associated start-up and annual fixed costs, such that budget neutrality would be achieved within several years.

The components of the care management intervention, as well as specific study-related outcomes that included adherence to each of 23 dementia guideline recommendations, have been described previously.

 In brief, patients with dementia receiving services in 18 primary care clinics within 3 healthcare organizations in the San Diego, California, area, together with their informal caregivers, were enrolled in a cluster RCT with an 18-month follow-up. Patient-caregiver dyads in the intervention arm were assigned a care manager, who was trained in the use of Internet-based care management software. The care managers performed a structured home assessment, identified problems, initiated care plan actions, and sent a summary to the primary care physician and other designated providers. Care managers provided ongoing follow-up as needed, with in-home reassessments every 6 months.

Three community agencies collaborated with the participating healthcare organizations to provide specific care services (eg, access to respite care). A physician champion was established within each healthcare organization. At each of the intervention clinics, providers received standardized interactive educational seminars on relevant care issues such as the evaluation of acute behavior changes. Selected intervention tools and documents with more detailed descriptions can be accessed at the Alzheimer’s Disease Coordinated Care for San Diego Seniors Web site (http://www.adc.ucla.edu/access/access.swf). Patients, caregivers, and providers in the usual care group did not receive any of the intervention protocol, and patients received care as usual.

Institutional review boards at the University of California, Los Angeles, and all participating sites approved the study, and informed consent was obtained from all study subjects (or from their proxies). Patients were eligible for the intervention if they were at least 65 years of age, enrolled in Medicare (either fee-for-service or managed care plans), had a dementia diagnosis, and had an informal caregiver at least 18 years of age. Potentially eligible patients were identified by a dementia diagnosis code or a cholinesterase inhibitor prescription recorded in administrative claims data. Clinic providers were contacted to confirm each diagnosis and to suggest additional patients as potential participants. Patient-caregiver dyads were enrolled between August 2001 and November 2002. After enrollment, caregivers were surveyed at baseline, at 12 months, and at 18 months. The surveys collected detailed information about patient healthcare utilization, paid and unpaid caregiving hours, costs of paid nonprofessional caregivers, and out-of-pocket expenses in the 6 months before enrollment through 18 months after enrollment.

Within each healthcare organization, clinics were matched by patient volume and subsequently were randomized to either the intervention (9 clinics) or usual care (9 clinics). By the end of the enrollment period, each full-time care manager was managing approximately 50 patient-caregiver dyads. We used this figure to calculate the per-patient variable costs of the intervention, which depended on the scale of the intervention. We used an intent-to-treat framework based on the initial randomization and limited the main analyses to patients with complete utilization data at baseline who survived for the entire study period and had follow-up data at 12 months, 18 months, or both. Patients who were institutionalized during the study period were included in the analyses. We excluded from the main analyses patients who died, as the substantial costs associated with end-of-life care

 might overshadow any cost differences related to the intervention. Although mortality rates did not differ significantly between the intervention and usual care groups at any point during the study,

 we performed sensitivity analyses that included patients who died before the study ended.

Intervention Cost and Potential Cost Offset

In brief, we estimated the cost of the intervention by summing fixed 1-time start-up costs, annual fixed costs, and per-patient variable costs (

). We did not have true cost data for healthcare and caregiving services not directly attributable to the intervention (ie, potential cost offset), but we estimated this information using expenditure data (

). We determined the units of each service utilized by each patient and applied standard unit costs from sources such as the Medical Expenditure Panel Survey and the Centers for Medicare & Medicaid Services. Although we did not include pharmacy costs because sufficiently detailed information was unavailable to us, we performed an ancillary analysis examining the effect of the intervention on the rates of cholinesterase inhibitor use.

We considered 2 different perspectives in calculating the potential cost offset, a “payer” perspective and a “social planner” perspective (

). The payer perspective is limited to the formal costs of care borne by the health plan. We calculated estimates for this perspective under 2 different assumptions with and without the cost of nursing home stays included. The social planner perspective implies that costs and benefits accruing to all members of society are measured.

 This perspective includes not only the formal costs of care but also the opportunity costs associated with informal caregiving, which represent a substantial proportion of the costs of dementia care.

In unadjusted analyses, we compared baseline characteristics and costs of healthcare for the intervention and usual care arms. We used the 

 test for categorical variables and the Wilcoxon 2-sample rank sum test or 2-sample 

 test for continuous variables, depending on their distributions. In ordinary least squares regression analyses, we compared the mean monthly costs of healthcare and caregiving services for intervention and usual care patients who survived the entire 18-month study period and had follow-up data available at 12 months, 18 months, or both. The estimated intervention effect was similar when using square root and log transformations of the dependent variable, so for ease of interpretation we report the results when using the original (nontransformed) cost variable.

To account for the complex sample design, including the clustering of patient-caregiver dyads within clinics, we used generalized estimating equations and adjusted for clinic as the clustering variable using statistical software (STATA 8.2; StataCorp LP, College Station, TX). In addition to the indicator variable for the intervention group versus the control group, we included as covariates patient age and sex, indicators for the healthcare organizations, an indicator of managed care versus fee-for-service insurance, and the patient’s mean monthly costs of healthcare and caregiving services during the 6 months before baseline. The analyses were weighted based on the number of months of follow-up data (maximum, 18 months) available for each subject to increase the influence of data from patients with longer study enrollment and to produce more reliable monthly cost means.27 We conducted 2 sensitivity analyses that included (1) patients with at least 1 follow-up survey regardless of whether they died during the study and (2) patients who survived for the entire study period and had follow-up data at both 12 months and 18 months.

Of 1043 patients initially identified from claims data and contacted about enrollment through their provider, 91 were ineligible, 308 declined to participate, and 236 did not respond, comparable to other quality improvement trials that used registries or claims data to identify potentially eligible subjects for recruitment.

 Four hundred eight patientcaregiver dyads were enrolled (238 in the intervention group and 170 in the usual care group), and all but 1 had complete utilization data at baseline. Survey response rates were 88% at 12 months and 82% at 18 months, excluding 32 deaths in the intervention group and 26 deaths in the usual care group. There were no significant differences in patient or caregiver characteristics (including age, sex, living status, education, marital status, race/ethnicity, dementia severity and duration, and comorbidity) between follow-up survey respondents and nonrespondents. For the main analyses, data were available for 296 of 408 enrolled patients who survived for the entire study period and completed at least 1 follow-up survey. For the sensitivity analyses, data were available for 354 patients who completed at least 1 follow-up survey and for 260 patients who survived for the entire study period and completed surveys at both 12 months and 18 months.

At baseline, the intervention and usual care groups did not differ with regard to demographic and clinical characteristics (

). The mean age was approximately 80 years, and more than half of the patients were female. Approximately three-quarters had dementia of the Alzheimer type, and patients had received a dementia diagnosis on average between 2 and 3 years before study onset. Approximately 77% were in Medicare managed care settings, with the remainder in Medicare fee-for-service arrangements.

There were no significant differences in inpatient or outpatient utilization between the 2 study groups at baseline or at follow-up  (

). At baseline, less than 20% of patients in either group had been hospitalized, and less than 5% of patients overall had been admitted to a nursing home. During the course of follow-up, approximately 30% of patients in each group were hospitalized, and less than 15% were admitted to a nursing home. Although (as previously noted) we could not include medication costs in the cost analyses, there was an increase of approximately 10 percentage points in the use of cholinesterase inhibitors among patients in the intervention group at follow-up versus no change among patients in the usual care group.

The intervention had a fixed start-up cost of $70,256 and annual fixed costs of $24,162 to each of the healthcare organizations regardless of the number of patient-caregiver dyads enrolled (eAppendix A). Assuming that the intervention enrolled 200 patient-caregiver dyads, the variable cost to the healthcare organizations would be $98 per patient per month. Examining the payer perspective with the cost of nursing home stays included, the mean monthly adjusted costs of healthcare and caregiving services during the follow-up period were $219 less for the intervention group compared with the usual care group ($1402 vs $1621 per patient per month), although the difference was not statistically significant (

). Results were similar if the cost of nursing home stays was excluded, with a nonsignificant reduction of $256 in costs for the intervention group ($1119 vs $1375 per patient per month, 

To examine the social planner perspective, we combined the costs accrued by the community agencies with those of the healthcare organizations and included out-of-pocket costs. The fixed start-up cost for each of the 3 community agencies was $1240, with annual fixed costs varying from $4042 to $10,812 across the agencies (eAppendix A). The ongoing mean cost of the intervention was approximately $118 per patient per month for the healthcare organizations and community organizations together. The mean monthly adjusted costs of healthcare and caregiving services during the follow-up period, representing the  potential cost offset, were $555 less for the intervention group compared with the usual care group ($5332 vs $5887 per patient per month), although the difference was again not statistically significant (

). Sensitivity analyses using different inclusion criteria also found no difference in costs between the 2 study arms regardless of whether the social planner perspective or the payer perspective was used. In a 1-sided hypothesis test, we found no evidence that the intervention costs would be less than the cost offset to healthcare organizations as calculated over an 18-month period (

 = .72) (data not shown). Finally, total costs did not differ for patients enrolled in managed Medicare versus fee-for-service Medicare using either analytic perspective.

In this cost evaluation of a dementia care management intervention, we found no evidence of lower costs compared with the costs of usual care at 18 months’ follow-up. However, this intervention has been shown to significantly improve multiple measures of dementia care quality, as well as patient and caregiver outcomes, and may represent a cost-effective approach to improving quality of care even if it is not cost saving.

 and the American Association for Geriatric Psychiatry

 have published guidelines advocating the increased use of comprehensive dementia care strategies. The intervention may be attractive to healthcare organizations that prioritize improving adherence to these guidelines.

We are unaware of any prior cost analyses of dementia care management interventions that were designed to simultaneously target both healthcare and community settings. The Medicare Alzheimer’s Disease Demonstration and Evaluation project examined the effect of case management and collaboration with community organizations, but not health plans, on  formal costs of care.

 The total cost of the project reached budget neutrality at only 1 of 8 sites. The Resources for Enhancing Alzheimer’s Caregivers Health II home-based intervention reported a cost of $5 for each hour of informal caregiving time saved by the intervention.

 The present study adds to the literature by providing cost estimates on a more comprehensive care management intervention that, on average, doubled rates of adherence to 23 indicators of dementia care quality and improved patient and caregiver outcomes.

The likelihood of similar cost estimates if our intervention was replicated in other settings depends on several assumptions, particularly the enrollment of 200 patient-caregiver dyads within a system to achieve full operating capacity. In our study, this enrollment level was reached after 15 months, but delayed or limited enrollment in other settings would increase the per-patient variable costs. Replicating the recruitment approach used in this randomized trial, in which potential participants were identified based on case-finding strategies such as a registry or an administrative database and were then approached and automatically offered the opportunity to enroll, would help to ensure adequate enrollment rather than relying solely on referrals triggered by “crisis” situations.

Given the intervention start-up cost of approximately $75,000, whether managed care organizations ultimately adopt this approach will depend on the existing cost structure of dementia care overall, as well as on projected patient turnover and mortality rates. There are several financial obstacles to adopting this type of intervention in managed care, including the  absence of risk sharing for long-term care. Although long-term care is the most costly aspect of dementia care, Medicaid and patient out-of-pocket payments cover two-thirds of the cost.

 Managed care organizations have little financial incentive to invest in quality improvement strategies that may reduce institutionalization, as they are generally not responsible for expenses related to long-term dementia care.

Another financial barrier to the adoption of care management for dementia is a problematic reimbursement system. Although dementia is a contributing factor to many hospital admissions and increases the cost of inpatient and outpatient care,

 it is rarely recorded as a primary or secondary diagnosis.

A dementia care management program proven to improve quality and outcomes showed a statistically nonsignificant cost offset (both payer and societal perspectives) over 18 months.

August 2009

Burden of Alzheimer's Disease and Association With Negative Health Outcomes

Guided Care and the Cost of Complex Healthcare: A Preliminary Report

Cost Evaluation of a Coordinated Care Management Intervention for Dementia

A Telephone-Based Intervention for Increasing the Use of Osteoporosis Medication: A Randomized Controlled Trial

Distal Upper and Lower Limb Fractures Associated With Thiazolidinedione Use

Outcomes of Patients Discharged From Pharmacy-Managed Cardiovascular Disease Management

Can a Nationwide Media Campaign Affect Antibiotic Use?

Cost Minimization of Medicare Part D Prescription Drug Plan Expenditures

Medicare Part D After 2 Years

Cost Comparison of Peritoneal Dialysis Versus Hemodialysis in End-Stage Renal Disease