Alzheimer’s & Dementia 9 (2013) 30–38

Potentially avoidable hospitalizations among Medicare beneficiarieswith Alzheimer’s disease and related disorders

Pei-Jung Lina,*, Howard M. Fillitb, Joshua T. Cohena, Peter J. Neumanna

aCenter for the Evaluation of Value and Risk in Health, Institute for Clinical Research and Health Policy Studies, Tufts Medical Center, Boston, MA, USAbAlzheimer’s Drug Discovery Foundation, New York, NY, USA

Abstract Background: Individuals with Alzheimer’s disease and related disorders (ADRD) have more fre-

This work was sup

therapy Research & D

Abstracts of this

meeting of AcademyH

Alzheimer’s Associati

*Corresponding au

E-mail address: pl

1552-5260/$ - see fro

http://dx.doi.org/10.10

quent hospitalizations than individuals without ADRD, and some of these admissions may be pre-ventable with proactive outpatient care.Methods: This study was a cross-sectional analysis of Medicare claims data from 195,024 fee-for-service ADRD beneficiaries aged �65 years and an equal number of matched non-ADRD controlsdrawn from the 5% random sample of Medicare beneficiaries in 2007–2008. We analyzed the pro-portion of patients with potentially avoidable hospitalizations (PAHs, as defined by theMedicare Am-bulatory Care Indicators for the Elderly) and used logistic regression to examine patientcharacteristics associated with PAHs. We used paired t tests to compare Medicare expenditures byADRD status, stratified by whether there were PAHs related to a particular condition.Results: Compared with matched non-ADRD subjects, Medicare beneficiaries with ADRD weresignificantly more likely to have PAHs for diabetes short-term complications (OR 5 1.43; 95% CI1.31–1.57), diabetes long-term complications (OR 5 1.08; 95% CI 5 1.02–1.14), and hypertension(OR5 1.22; 95%CI1.08–1.38), but less likely to havePAHs for chronic obstructive pulmonarydisease(COPD)/asthma (OR 5 0.85; 95% CI 0.82–0.87) and heart failure (OR 5 0.89; 95% CI 0.86–0.92).Risks of PAHs increased significantlywith comorbidity burden.Among beneficiarieswith a PAH, totalMedicare expenditures were significantly higher for those subjects who also had ADRD.Conclusion: Medicare beneficiaries with ADRD were at a higher risk of PAHs for certain uncon-trolled comorbidities and incurred higher Medicare expenditures compared with matched controlswithout dementia. ADRD appears to make the management of some comorbidities more difficultand expensive. Ideally, ADRD programs should involve care management targeting high-risk patientswith multiple chronic conditions.� 2013 The Alzheimer’s Association. All rights reserved.

Keywords: Alzheimer’s disease; Hospitalization; Comorbidity; Medicare expenditures; Quality of care; High-risk patients;

Diabetes; Hypertension; CDPD/asthma; Heart failure

1. Introduction

Managing comorbidities has been a long-standing chal-lenge in the care of patients with Alzheimer’s disease and re-lated disorders (ADRD). Compared with individuals withoutdementia, Medicare beneficiaries with ADRD are more

ported by a grant from Janssen Alzheimer Immuno-

evelopment and Pfizer to the Tufts Medical Center.

study were presented at the 2012 annual research

ealth and the 2012 international conference of the

on.

thor. Tel.: 617-636-4616; Fax: 617-636-5560.

in@tuftsmedicalcenter.org

nt matter � 2013 The Alzheimer’s Association. All rights r

16/j.jalz.2012.11.002

likely to be hospitalized, have longer hospital stays, and incurhigher expenditures for their comorbidities [1–4]. Further,hospital admissions represent the largest component ofhealthcare expenditures for individuals with ADRD [5,6]and constitute more than half of total expenditures amongthe most expensive patients with prominent comorbidities[7]. Additional time spent in the hospital also increasessuffering and the risk of adverse events. Therefore, reducingunnecessary admissions is clearly an important and under-served need in ADRD care.

Some hospital admissions are expected as part of thenatural course of treatment, whereas others are consideredpotentially avoidable and “indicators of poor care or missed

eserved.

P.-J. Lin et al. / Alzheimer’s & Dementia 9 (2013) 30–38 31

opportunities to better coordinate care” [8]. Potentiallyavoidable hospitalization (PAH) is an important quality indi-cator because, unlike process-of-care indicators, PAH di-rectly measures the end result of care as experienced bythe patient. Such outcome measures are a critical componentof initiatives to provide quality-of-care information that res-onates with patients, providers, and payers [9]. A recentstudy showed that the incidence of dementia was signifi-cantly associated with increased risk of all-cause hospitali-zations and PAHs for conditions such as pneumonia andulcer, based on a longitudinal cohort of 3019 enrollees ina private, integrated health system [10]. Although priorstudies have reported more frequent hospitalizations amongADRD patients (and some of these admissions may be pre-ventable with proactive outpatient care) [1,10,11], theresulting expenditures are less clear.

Our analysis extends the existing literature by comparingPAHs as well as Medicare expenditures for selected comor-bidities in a matched cohort of beneficiaries with and with-out ADRD, using more recent data from a large, nationalsample of Medicare enrollees. We also examined patientcharacteristics associated with PAHs to shed light on theidentification of high-risk patients for which policies anddisease management may be particularly valuable.

2. Methods

2.1. Data and sample

We used the national 5% sample Medicare BeneficiaryAnnual Summary File (BASF) and corresponding Medicareclaims records for 2007–2008. The BASF contains demo-graphic, enrollment, and date-of-death information for eachbeneficiary, as well as flags for 19medical conditions definedby the CMS Chronic Condition Data Warehouse (CCW)condition categories. The claims files contain date of service,diagnosis codes, and payment information for all healthcareservices (inpatient, physician office, outpatient, postacuteskilled nursing facility [SNF], home health, hospice, durablemedical equipment) provided to aMedicare beneficiary. Thisstudy was approved by the institutional review board of theTufts Medical Center.

Figure 1 describes our sample selection process. From the2007Medicare sample (n5 2,585,349), we excluded benefi-ciaries who were ,65 years of age (n 5 394,775). We alsoexcluded beneficiaries who died during 2007–2008 or whoenrolled in 2008 (n 5 208,488) because they lacked 2 yearsof claims data, which are needed to construct the PAH indica-tors. Managed care enrollees (n 5 133,778) were excludedbecause CMS did not require managed care plans to submitclaims with diagnoses. Thus, we retained fee-for-service(FFS) beneficiaries who had 11 or 12 months of both PartA and Part B FFS coverage during a year. Finally, beneficia-ries who used hospice care (n 5 14,842) were excludedbecause, by definition, they opt not to extend medical care

(e.g., hospital admission) and thus are not an appropriatepopulation for studying avoidable hospitalizations.

Among these 1,833,466 eligible Medicare beneficiaries,we identified individuals as having ADRD if they had at leastone claim as of December 31, 2008 with an ICD-9-CM (ninthedition of the InternationalClassification ofDiseases,ClinicalModification) diagnosis code defined by the CCW: 331.0,331.1, 331.11, 331.19, 331.2, 331.7, 290.0, 290.1, 290.10,290.11, 290.12, 290.13, 290.20, 290.21, 290.3, 290.40,290.41, 290.42, 290.43, 294.0, 294.1, 294.10, 294.11, 294.8,and 797. The final unmatched sample included 198,693beneficiaries with ADRD (10.8%) and 1,634,773 beneficia-ries without ADRD (89.2%). The ADRD prevalence rates inour sample were comparable to other estimates [12–14].

2.2. Indicators of potentially avoidable hospitalizations

PAHs have been defined as hospital admissions prevent-able by good ambulatory care, or by early intervention toavoid severe disease [15,16]. We classified a hospitalizationas a PAH if it was directly related to one of five avoidableadverse events designated by the Medicare AmbulatoryCare Indicators for the Elderly (MACIEs) [16,17]: (1)serious short-term complications of diabetes; (2) seriouslong-term complications of diabetes; (3) chronic obstructivepulmonary disease (COPD) or asthma; (4) hypertension; and(5) heart failure (see Appendix 1).

The MACIEs were developed by the Medicare PaymentAdvisory Commission (MedPAC) to analyze ambulatorycare quality through Medicare claims files. To conduct thisanalysis, we needed 2 years of claims data for each benefi-ciary so that follow-up clinical information could be checkedfor the PAH events. For example, among patients with out-patient or inpatient visits for heart failure in 2007, 2008data were used to see if they had hospital admissions relatedto heart failure during that year [16].

The MedPAC has used MACIEs to examine quality ofcare associated with various geographic and socioeconomicfactors [17]. Detailed methodology and rationale for devel-oping the MACIEs has been described elsewhere [16,18].The five PAH conditions designated by MACIEs are alsoexamined in the Prevention Quality Indicators endorsed bythe Agency for Healthcare Research and Quality (AHRQ)and validated from the clinician perspective [19].

2.3. Analytic approach

2.3.1. Unmatched analysisWe calculated the proportion of beneficiaries having PAH

for each condition of interest, stratified bywhether the patienthad ADRD. To examine whether ADRD beneficiaries weremore likely than their peers without ADRD to have PAHs,we used logistic regression to calculate odds ratios (ORs),controlling for age, gender, race, Medicare and Medicaiddual eligibility, whether an individual lived in a metropolitan

2007 5% Medicare sampleN=2,585,349

Age 65n=2,190,574

Had both 2007 & 2008 datan=1,982,086

Age < 65n=394,775

Died during 2007-2008 or newly enrolled in 2008

n=208,488

Full or nearly full FFS*

n=1,848,308Non-FFS

n=133,778

Non-hospice beneficiariesn=1,833,466

Hospice beneficiariesn=14,842

Beneficiaries with ADRD†

n=198,693 (10.8%)Beneficiaries without ADRD

n=1,634,773 (89.2%)

Beneficiaries with ADRDn=195,024

Beneficiaries without ADRDn=195,024

One-to-one propensity score matching with logistic regression adjusted for age, sex, race, Medicaid eligibility, MSA, and

number of CCW conditions

Fig. 1. Sample selection flowchart. ADRD, Alzheimer’s disease and related disorders; CCW, Chronic Condition Data Warehouse; FFS, fee-for-service; MSA,

metropolitan statistical area. Asterisk (*): full or nearly full fee-for-service (FFS) indicates the beneficiary had 11 or 12 months of both Part A and Part B fee-for

service coverage. Dagger (y): beneficiaries were identified as having ADRD or senile dementia (ADRD) if they had at least one inpatient, skilled nursing facility,

home health agency, hospice, or physician office claims with the following ICD-9-CM diagnosis codes as of December 31, 2007: 331.0, 331.1, 331.11, 331.19,

331.2, 331.7, 290.0, 290.1, 290.10, 290.11, 290.12, 290.13, 290.20, 290.21, 290.3, 290.40, 290.41, 290.42, 290.43, 294.0, 294.1, 294.10, 294.11, 294.8, 797, as

defined by the CMS CCW Condition Categories.

P.-J. Lin et al. / Alzheimer’s & Dementia 9 (2013) 30–3832

statistical area (MSA), and the number of chronic conditionsdefined by the CMS CCW (excluding ADRD).

2.3.2. Propensity score matchingTo further account for differences between subjects with

and without ADRD, we used logistic regression to calculatepropensity scores and then used a “greedy algorithm” [20]to match subjects from the two groups. A propensity scorerepresents the probability that a beneficiary with a given setof characteristics will be classified as having ADRD. Select-ing beneficiaries matched on propensity score mitigates thepossibility that the two groups being compared differ in termsof a confounding factor that could influence the PAH risk in-dependently of ADRD. The propensity score model includedage, gender, race,Medicare andMedicaid dual eligibility, res-idence in an MSA, and the number of CCW conditions (ex-

cluding ADRD) (concordance index 5 0.79). Despite thelarge sample of non-ADRD beneficiaries (n 5 1,634,773),we used one-to-one matching to avoid the possible bias ofmany-to-one matching [21]. This process identified 195,024ADRD patients (of 198,693 possible; 98.2%) and an equalnumber of non-ADRD controls.

2.3.3. Patient characteristics associated with PAHsIn subgroup analysis of beneficiaries with ADRD, we

used logistic regression to identify patient characteristics as-sociated with PAHs, controlling for a same set of predictorsin our propensity scoremodel. Because the ICD-9-CM codesfor ADRD do not denote disease severity, we created an in-dicator for late-stage disease based on the presence ofmedical complications, including pressure ulcers, eatingdisorders and malnutrition, aspiration pneumonia, and

P.-J. Lin et al. / Alzheimer’s & Dementia 9 (2013) 30–38 33

incontinence. The rationale for selecting these complicationsas indicative of advanced ADRD has been reported previ-ously [22–26]. All other ADRD beneficiaries without thesecomplications were classified as having earlier stage disease.

2.3.4. Medicare expenditures by ADRD and by PAHcondition

To examine the excess expenditures attributable toADRD, we used paired t tests to compare total Medicare ex-penditures for propensity-matched ADRD beneficiaries andnon-ADRD subjects, stratified by whether they had PAHs re-lated to a particular condition. Total Medicare expendituresincluded payments for all medical services in 2008.

3. Results

3.1. Sample characteristics

In the unmatched sample (left three numerical columns inTable 1), ADRDbeneficiarieswere significantly older, dispro-portionately female, non-white, and dually eligible for Medi-care and Medicaid services compared with non-ADRDbeneficiaries (all P, .001). Comorbidity burden was signifi-cantly higher among ADRD beneficiaries: 80.4% had one ormore conditions, compared with 54.8% among non-ADRDbeneficiaries (P,.001). The proportion of beneficiaries livingin an MSA differed significantly between the two groups, butthemagnitude of the differencewas,1%. The right three col-umns in Table 1, which compare ADRD and non-ADRD ben-eficiaries matched by propensity score, shows that these twogroups had well-balanced characteristics.

Table 1

Characteristics of unmatched and propensity-matched ADRD beneficiaries and no

Unmatched sample

ADRD

(n 5 198,693)

Non-ADRD

(n 5 1,634,773

Age in years (mean 6 SD) 82.6 (7.4) 75.8 (7.2)

65–69 4.6% 23.2%

70–74 11.1% 25.9%

75–79 17.5% 21.4%

80–84 24.7% 16.1%

�85 42.2% 13.3%

Female 70.7% 61.2%

Race

White 84.5% 86.7%

Black 9.8% 7.9%

Hispanic 2.8% 1.9%

Other/Unknown 2.9% 3.6%

Medicare and Medicaid dual-eligible 26.7% 10.6%

Metropolitan statistical area 77.8% 78.4%

Number of CCW conditions*

0 19.6% 45.2%

1–2 32.3% 33.6%

3–4 29.0% 15.8%

�5 19.1% 5.4%

Abbreviations: ADRD, Alzheimer’s disease and related disorders; CCW, Chron

*CCW conditions include acute myocardial infarction, atrial fibrillation, cata

failure, diabetes, glaucoma, hip/pelvic fracture, ischemic heart disease, depression

attack, female breast cancer, colorectal cancer, prostate cancer, lung cancer, and e

3.2. PAHs among ADRD beneficiaries and controls

Asdetailed in the portionofTable 2, a substantial proportionof ADRD beneficiaries had PAHs for uncontrolled chronic ob-structivepulmonarydisease (COPD)/asthma (20.2%)andheartfailure (11.7%). Compared with propensity-score-matchedsubjects, Medicare beneficiaries with ADRD were morelikely to have PAHs for diabetes short-term complications(OR5 1.43; 95%confidence interval [CI] 1.31–1.57), diabeteslong-term complications (OR5 1.08; 95%CI 1.02–1.14), andhypertension (OR5 1.22; 95%CI 1.08–1.38), but less likely tohave PAHs for COPD/asthma (OR 5 0.85; 95% CI0.82–0.87) and heart failure (OR5 0.89; 95% CI 0.86–0.92).

3.3. Association between patient characteristics and PAHs

Table 3 details the association between various patientcharacteristics and PAHs among ADRD beneficiaries. Therisks of PAHs increased significantly with the number of co-morbidities (ORs ranging from 1.25 to 8.72). Compared tothose with earlier stage disease, beneficiaries with advancedADRD (17.4% of the patient cohort) had significantly higherrisks of PAHs for diabetes long-term complications (OR 51.23; 95% CI 1.13–1.34), hypertension (OR 5 1.23; 95%CI 1.02–1.47), COPD/asthma (OR 5 1.79; 95% CI 1.71–1.87), and heart failure (OR 5 1.25; 95% CI 1.19–1.32).

3.4. Medicare expenditures by ADRD and by PAH

Total Medicare expenditures were 51% higher amongADRD beneficiaries compared with propensity-matched

n-ADRD controls—2008 Medicare claims data

Matched sample

) P-value

ADRD

(n 5 195,024)

Non-ADRD

(n 5 195,024) P-value

,.001 82.5 (7.4) 82.0 (7.1) ,.001

,.001 4.7% 4.7% .992

11.3% 11.3%

17.8% 17.8%

25.1% 25.2%

41.1% 41.1%

,.001 70.4% 70.3% .801

,0.001 .175

84.6% 84.4%

9.7% 9.8%

2.8% 2.9%

2.9% 2.9%

,.001 25.3% 25.3% .988

,.001 77.7% 77.5% .109

,.001 .861

20.0% 20.0%

32.9% 32.8%

28.7% 28.8%

18.4% 18.4%

ic Condition Data Warehouse.

ract, chronic kidney disease, chronic obstructive pulmonary disease, heart

, osteoporosis, rheumatoid arthritis/osteoarthritis, stroke/transient ischemic

ndometrial cancer.

Table 2

Potentially avoidable hospitalizations in unmatched and propensity-score matched cohorts of ADRD beneficiaries and non-ADRD controls—2007–2008

Medicare claims data

Unmatched analyses

ADRD (n 5 198,693) Non-ADRD (n 5 1,634,773) OR (95% CI)*

Diabetes (n 5 98,741) (n 5 445,047)

Hospitalization for short-term

complications

1.31% 0.86% 1.44 (1.34–1.54)

Hospitalization for long-term

complications

2.78% 2.13% 1.16 (1.11–1.22)

Hypertension (n 5 134,910) (n 5 696,787)

Hospitalization for hypertension 0.46% 0.24% 1.34 (1.21–1.48)

COPD/Asthma (n 5 55,681) (n 5 177,124)

Hospitalization for COPD/asthma 20.21% 20.75% 0.89 (0.86–0.91)

Heart failure (n 5 72,764) (n 5 224,642)

Hospitalization for heart failure 11.98% 8.14% 1.02 (0.99–1.05)

Matched analyses

ADRD (n 5 195,024) Non-ADRD (n 5 195,024) OR (95% CI)

Diabetes (n 5 96,117) (n 5 81,979)

Hospitalization for short-term

complications

1.31% 0.92% 1.43 (1.31–1.57)

Hospitalization for long-term

complications

2.78% 2.58% 1.08 (1.02–1.14)

Hypertension (n 5 131,797) (n 5 120,283)

Hospitalization for hypertension 0.46% 0.37% 1.22 (1.08–1.38)

COPD/Asthma (n 5 53,955) (n 5 38,369)

Hospitalization for COPD/asthma 20.18% 23.03% 0.85 (0.82–0.87)

Heart failure (n 5 70,529) (n 5 50,393)

Hospitalization for heart failure 11.71% 12.97% 0.89 (0.86–0.92)

Abbreviations: ADRD, Alzheimer’s disease and related disorders; CCW, Chronic Condition DataWarehouse; COPD, chronic obstructive pulmonary disease.

*Unmatched analyses: Odds ratios were calculated based on logistic regressions adjusted for age, gender, race, Medicaid eligibility, metropolitan statistical

area, and number of CMS CCW categories.

P.-J. Lin et al. / Alzheimer’s & Dementia 9 (2013) 30–3834

subjects without ADRD ($13,949 vs $9237 in 2008, P ,.001). The elevated expenditures among the ADRD cohortwere also observed when attention was restricted to benefi-ciaries who had PAHs directly related to their comorbiditydiagnoses (Figure 2). For example, expenditures averaged$42,931 for ADRD beneficiaries with PAHs related to diabe-tes long-term complications. For propensity-matched non-ADRD subjects who also had PAHs for diabetes long-termcomplications, expenditures averaged $35,030. The differ-ence of $7901 (P , .001) represents the excess expenditureassociated with ADRD status. For beneficiaries withoutPAHs related to diabetes long-term complications, but diag-nosed with diabetes, excess expenditures attributable toADRD were $6574 ($24,593 vs. $18,019, P , .001).

Figure 2 can also be used to infer the impact of PAHs ontotal expenditures. For example, total Medicare expenditureswere $21,961 (121%) higher (P , .001) for ADRD benefi-ciaries with a PAH related to heart failure ($40,061) thanfor other ADRD beneficiaries without a heart failure-related PAH ($18,100).

4. Discussion

ADRD appears to make the management of other comor-bidities more difficult and expensive. When compared with

propensity-score-matched subjects without ADRD, Medi-care beneficiaries with ADRD were more likely to havePAHs related to diabetes complications and hypertension.Many ADRD beneficiaries had PAHs for COPD/asthmaand heart failure, although the PAH risk for these beneficia-ries was lower compared with matched subjects withoutADRD. Among beneficiaries with a PAH related to a partic-ular condition, total Medicare expenditures were signifi-cantly higher for those subjects who also had ADRD.

TheNational Alzheimer’s Project Act (NAPA)was signedinto law in 2011 [27]. Central to the NAPA is the goal of pro-viding patients with high-quality and efficient care. For con-ditions such as diabetes, hypertension, COPD/asthma, andheart failure, high-quality outpatient management reducesthe risk of almost all types of serious hospitalizations [28].Our data show that many ADRD beneficiaries were nonethe-less admitted to the hospital for these conditions (e.g., 20%for COPD/asthma and 12% for heart failure). The PAH find-ings in our study may signal suboptimal use of appropriateambulatory care [28].Moreover, current treatment guidelinesfor the PAH conditions generally emphasize patient self-management, which is especially problematic for ADRD pa-tients. As disease progresses, ADRD patients may lose theirself-care skills and have difficulty complying with care man-agement instructions (e.g., adherence to medications) [4],

Table 3

Odds ratios of patient characteristics associated with potentially avoidable hospitalizations among ADRD beneficiaries

Diabetes short-term

complications

Diabetes long-term

complications Hypertension COPD/asthma Heart failure

Age in years

65–69 (reference) — — — — —

70–74 0.64 (0.51–0.80) 0.91 (0.77–1.07) 0.92 (0.61–1.40) 1.00 (0.90–1.12) 0.92 (0.80–1.06)

75–79 0.63 (0.51–0.78) 0.83 (0.71–0.97) 0.78 (0.52–1.17) 0.89 (0.81–0.99) 1.05 (0.92–1.20)

80–84 0.51 (0.41–0.63) 0.64 (0.55–0.75) 0.84 (0.57–1.24) 0.78 (0.71–0.86) 1.20 (1.05–1.36)

�85 0.36 (0.29–0.45) 0.41 (0.35–0.48) 0.86 (0.59–1.25) 0.62 (0.56–0.68) 1.48 (1.30–1.68)

Female 0.99 (0.88–1.12) 1.21 (1.12–1.31) 0.87 (0.72–1.05) 1.17 (1.11–1.22) 1.08 (1.02–1.14)

Race

White (reference) — — — — —

Black 1.76 (1.51–2.05) 1.79 (1.61–1.99) 2.11 (1.71–2.61) 0.75 (0.69–0.81) 1.22 (1.13–1.32)

Hispanic 2.16 (1.69–2.76) 1.51 (1.24–1.84) 0.79 (0.44–1.41) 0.84 (0.73–0.96) 0.86 (0.73–1.01)

Other/unknown 1.18 (0.86–1.64) 1.25 (1.01–1.56) 0.59 (0.30–1.15) 0.79 (0.68–0.91) 0.89 (0.75–1.05)

Dual eligible 1.18 (1.04–1.33) 1.10 (1.01–1.20) 0.96 (0.79–1.16) 1.11 (1.06–1.17) 1.05 (0.99–1.11)

MSA 0.94 (0.82–1.08) 1.06 (0.96–1.16) 1.08 (0.89–1.33) 0.89 (0.84–0.94) 0.86 (0.82–0.92)

Number of CCW conditions

�2 (reference) — — — — —

3–4 1.59 (1.36–1.86) 1.29 (1.15–1.45) 1.43 (1.17–1.74) 1.25 (1.18–1.33) 2.58 (2.36–2.81)

�5 2.38 (2.05–2.77) 2.70 (2.43–2.99) 1.93 (1.58–2.36) 2.01 (1.90–2.12) 8.72 (8.06–9.44)

Late-stage ADRD 0.83 (0.72–0.96) 1.23 (1.13–1.34) 1.23 (1.02–1.47) 1.79 (1.71–1.87) 1.25 (1.19–1.32)

Abbreviations: ADRD, Alzheimer’s disease and related disorders; CCW, Chronic Condition DataWarehouse; COPD, chronic obstructive pulmonary disease;

MSA, metropolitan statistical area.

P.-J. Lin et al. / Alzheimer’s & Dementia 9 (2013) 30–38 35

a situation that is particularly challenging for those who lacksupport networks (family, informal caregiver).

Although we used different methodology and criteria toidentify PAHs, our results are consistentwith previous studiesthat found ADRD patients were more likely than otherMedi-care beneficiaries to have PAHs related to diabetes (OR 53.61; 95% CI 2.32–5.63) and hypertension (OR 5 2.60;95%CI 2.16–3.12) [1], and less likely to be admitted to a hos-pital for heart failure (OR 5 0.80; 95% CI 0.50–1.26) [3].However, our findings contrast with a recent longitudinal co-hort study by Phelan and colleagues, suggesting an increasedrisk of hospitalization for congestive heart failure exacerba-tion among dementia patients (OR 5 1.73; 95% CI

$35,088

$42,931

$25,889

$29,911

$35,030

$19,476

$24,970 $24,593

$16,3$18,352 $18,019

Diabetes short-term complications

Diabetes long-term complications

Hypertens

ADRD with PAH Non-ADRD with PAH

Fig. 2. Overall Medicare expenditures by type of potentially avoidable hospitaliz

(2008$). ADRD, Alzheimer’s disease and related disorders; COPD, chronic obstr

1.15–2.60) [10]. The difference might be due to cohort char-acteristics (e.g., older, fee-for-service Medicare beneficiariesin our study vs. younger, healthier enrollees in Group HealthCooperative in the Phelan et al. study), case and control groupidentification (ADRD diagnosis recorded in claims files inour study vs. incident dementia in Phelan et al.), differentlevels of ADRD severity, and the PAHmeasures used. More-over, some comorbidities (e.g., heart failure, COPD) aremoresevere than others (e.g., diabetes, hypertension) and com-monly reflect “end-of-life conditions” [16,29]. Possibly,less aggressive medical services for end-of-life conditionsare given to patients with ADRD. To build on the study pre-sented here, future researchers could assess other end-of-

$26,374

$40,061

$21,911

$34,995

91

$21,044$18,100

$11,811

$16,081$13,769

ion COPD/Asthma Heart failure

ADRD, no PAH Non-ADRD, no PAH

ation in propensity-matched ADRD beneficiaries and non-ADRD controls

uctive pulmonary disease; PAH, potentially avoidable hospitalization.

P.-J. Lin et al. / Alzheimer’s & Dementia 9 (2013) 30–3836

life conditions and explore patient and caregiver preferencesfor care, as well as physicians’ practice patterns [26].

Our analysis shows thatADRDpatientswith higher comor-bidity burden were at significantly increased risk of PAHs.These findings suggest that complex care management(CCM) isneeded to bettermanagemultiplemedical conditionsamonghigh-risk patients.Examples ofCCMstrategies includecare managers to coordinate care [30,31] and clinicalinterventions to improve discharge planning and to ensurecontinuing care [32]. These interventionsmay improve overallcare quality and patients’ quality of life and caregiver support,as well as reduce unnecessary hospitalizations [30,31,33].

Aggregate payments for ADRD are projected to increasefrom $183 billion in 2011 to $1.1 trillion in 2050 in the U.S.[14]. Curbing these rising costs has become a priority amongpolicymakers and payers. Past findings that hospitalizationscontribute substantially to healthcare expenditures amongADRD beneficiaries [1,3,6,7] are consistent with ourfindings that average Medicare expenditures weresignificantly higher among those who had PAHs. Forexample, total expenditures were 121% higher for ADRDbeneficiaries who had PAHs related to heart failure than forother ADRD beneficiaries diagnosed with heart failure butno directly related PAH. For the other conditions, thecorresponding increases associated with having a relatedPAH were 25% for COPD/asthma, 41% for diabetes short-term complications, 58% for hypertension, and 75% fordiabetes long-term complications. A sensitivity analysis re-stricting attention toMedicare inpatient expenditures (insteadof overallMedicare expenditures) yielded similar findings (re-sults not shown). Future cost containment efforts should in-volve effective care coordination and tailored ADRDmanagement strategies to decrease unnecessary inpatient ex-penditures. However, without changing Medicare’s FFS pay-ment structure, coordinated care programs have had limitedsuccess at reducing cost while improving quality [34,35].

Prior CMS demonstrations for disease management alsohave had limited impact on improving beneficiary knowl-edge, increasing preventive care, or reducing PAHs, emer-gency room visits, or mortality [34–36]. During 2011, twoMedicare Advantage (MA) Special Needs Plans (SNPs)have been implemented to enroll exclusively beneficiarieswith dementia, providing full Medicare coverage(prescription drug coverage and Parts A and B healthbenefits) as well as extra services to ADRD beneficiariesand their caregivers (such as a paid consultative visit forselecting hospice care) [37]. Currently, these dementiaSNPs have enrolled fewer than 200 ADRD patients and theperformance evidence for SNPs in general is scant [38]. Stud-ies evaluating whether the enrollees in dementia SNPs obtainhigher quality care and have better outcomes than individualsin traditional FFS and regular MA plans will help determinewhether these plans should be implemented on a large scale.

TheCMSQuality Initiatives and newAffordable CareActinitiatives, such as Accountable Care Organizations and ret-rospective bundling payments, all aim to shift the system

away from paying providers based on volume and insteadpaying them based on quality and value of care [9,39]. ForADRD care, provision of good ambulatory care that canprevent hospitalizations may reduce expenditures. Byallowing providers to design a more integrated healthcaredelivery structure and to determine how payments will beallocated among participating providers, these initiativesmay lower expenditures and promote better coordinationbetween acute inpatient care and postdischarge recovery.Linking the care quality to payments would furtherencourage providers to improve provider performance byreducing unnecessary hospitalizations.

Our study’s strengths include a large Medicare sample,evaluation of PAH conditions particularly important in olderpopulations, examination of patient characteristics associ-ated with PAHs, and use of a statistical matching approachto assess PAH and expenditure differences in beneficiarieswith and without ADRD. Nonetheless, there are limitationsthat should be considered when interpreting our findings.First, our sample included Medicare fee-for-service benefi-ciaries, so the results might not generalize to managed careenrollees or nursing home residents. A study analyzing1991–1993 data demonstrated that nursing home residentswith ADRD were more likely to be hospitalized for certainconditions such as gastroenteritis and kidney/urinary trackinfections [11]. Further, other conditions, such as COPDexacerbation and heart failure, could be treated in nursinghomes without prompting transfer to hospital. Second, theuse of Medicare claims data has inherent limitations in iden-tifying ADRD patients [40]. Mild cases might be underdiag-nosed and uncoded in claims files because of reimbursementissues, failure to recognize symptoms, and resistance toan ADRD diagnosis [41,42]. Nonetheless, resultingmisclassifications would likely underestimate the PAH andexpenditure differences between ADRD and non-ADRDbeneficiaries. Third, Medicare expenditures may be underes-timated as a result of the absence in our analysis of Part Ddrug expenditure data and our exclusion of managed care en-rollees. According to MedPAC, Managed Advantage planswere, on average, paid higher per enrollee than traditionalFFS plans (e.g., 114% of FFS costs in 2009) [17]. We alsolacked data on Medicaid payments for dually eligible bene-ficiaries. Fourth, our indicator for advanced ADRD wasbased on the presence of complications commonly seen inlate-stage patients. In the future it would be useful to validatethis indicator by using other data sources that contain diseasestaging information, such as electronic health records. Fifth,we could not determine the relationship between provisionof ambulatory care and PAHs. Future research should iden-tify ambulatory care elements that may reduce PAHs and ex-penditures. Finally, becausewe needed 2 years of claims datato check for PAH events, beneficiaries with partial-year data(e.g., subjects who died during 2007–2008) were excludedfrom our analysis. The impact is likely to be minimal, how-ever, as a sensitivity analysis including beneficiaries withoutcomplete data did not produce notably different findings.

P.-J. Lin et al. / Alzheimer’s & Dementia 9 (2013) 30–38 37

Our analysis suggests that Medicare beneficiaries withADRD have substantial PAHs for certain uncontrolledcomorbidities. For some conditions, such as diabetes and hy-pertension, ADRD beneficiaries were at a higher risk ofcondition-related PAHs compared with matched controls.The resulting Medicare expenditures highlight the opportu-nity to reduce cost while improving quality. Future ADRDmanagement programs should involve CCM services target-ing high-risk patients with multiple chronic conditions.Continued monitoring and reporting of PAHs is essentialto help CMS and providers to establish performance bench-marks and identify areas of care in need for quality improve-ment. By combining PAH outcomes with other qualitymeasures, such as resource use, future research can assessthe efficiency of care and guide incentive designs to encour-age value-driven care [9].

Acknowledgments

This work was supported by a grant from Janssen AlzheimerImmunotherapy R&D and Pfizer to the TuftsMedical Center.Publication was not contingent on Janssen AlzheimerImmunotherapy R&D and Pfizer’s approval. We are gratefulto Chi-Hui Fang for assistance with data analysis.

RESEARCH IN CONTEXT

1. Systematic review: We searched PubMed forEnglish-language original research that examinedpotentially avoidable hospitalizations (PAHs) amongindividuals with Alzheimer’s disease and relateddisorders (ADRD). Three studies were identified[1,10,11], all measuring PAHs as defined usingambulatory-care sensitive conditions.

2. Interpretation: We analyzed five PAH conditions, asdefined by the Medicare Ambulatory Care Indicatorsfor the Elderly. Compared with matched non-ADRDsubjects, Medicare beneficiaries with ADRD weresignificantly more likely to have PAHs for diabetesshort-term complications, diabetes long-term com-plications, and hypertension, but less likely to havePAHs for COPD/asthma and heart failure. Risks ofPAHs significantly increased with comorbidity bur-den. Among beneficiaries with a PAH, total Medicareexpenditures were significantly higher for those sub-jects who also had ADRD.

3. Future directions: Ideally, ADRD programs shouldinvolve care management targeting high-risk patientswith multiple chronic conditions. Continued moni-toring and reporting of PAHs is essential to helpCMS and providers to establish performance bench-marks and identify areas of care in need for qualityimprovement.

References

[1] Bynum JP, Rabins PV, Weller W, Niefeld M, Anderson GF, Wu AW.

The relationship between a dementia diagnosis, chronic illness, Medi-

care expenditures, and hospital use. JAmGeriatr Soc 2004;52:187–94.

[2] Sloan FA, Taylor DH Jr. Effect of Alzheimer disease on the cost of

treating other diseases. Alzheimer Dis Assoc Disord 2002;16:137–43.

[3] Zhao Y, Kuo TC, Weir S, Kramer MS, Ash AS. Healthcare costs and

utilization for Medicare beneficiaries with Alzheimer’s. BMC Health

Serv Res 2008;8:108.

[4] Hill JW, Futterman R, Duttagupta S, Mastey V, Lloyd JR, Fillit H. Alz-

heimer’s disease and related dementias increase costs of comorbidities

in managed Medicare. Neurology 2002;58:62–70.

[5] Zhu CW, Scarmeas N, Torgan R, Albert M, Brandt J, Blacker D, et al.

Clinical features associated with costs in early AD: baseline data from

the Predictors Study. Neurology 2006;66:1021–8.

[6] Weiner M, Powe NR, Weller WE, Shaffer TJ, Anderson GF. Alz-

heimer’s disease under managed care: implications fromMedicare uti-

lization and expenditure patterns. J Am Geriatr Soc 1998;46:762–70.

[7] Lin PJ, Biddle AK, Ganguly R, Kaufer DI, Maciejewski ML. The con-

centration and persistence of health care expenditures and prescription

drug expenditures in Medicare beneficiaries with Alzheimer disease

and related dementias. Med Care 2009;47:1174–9.

[8] Medicare Payment Advisory Commission. Report to the Congress:

promoting greater efficiency in Medicare. Washington, DC: MedPAC;

2007.

[9] CMS, roadmap for quality measurement. https://www.cms.gov/

QualityInitiativesGenInfo/downloads/QualityMeasurementRoadmap_

OEA1-16_508.pdf/. Accessed October 1.

[10] Phelan EA, Borson S, Grothaus L, Balch S, Larson EB. Association of

incident dementia with hospitalizations. JAMA 2012;307:165–72.

[11] Carter MW, Porell FW. Vulnerable populations at risk of potentially

avoidable hospitalizations: the case of nursing home residents with

Alzheimer’s disease. Am J Alzheimers Dis Other Dement 2005;

20:349–58.

[12] Hebert LE, Scherr PA, Bienias JL, Bennett DA, Evans DA. Alzheimer

disease in the US population: prevalence estimates using the 2000 cen-

sus. Arch Neurol 2003;60:1119–22.

[13] Plassman BL, Langa KM, Fisher GG, Heeringa SG, Weir DR,

Ofstedal MB, et al. Prevalence of dementia in the United States: the

aging, demographics, and memory study. Neuroepidemiology 2007;

29:125–32.

[14] Thies W, Bleiler L. 2011 Alzheimer’s disease facts and figures. Alz-

heimers Dement 2011;7:208–44.

[15] Jiang HJ, Wier LM, Potter DEB, Burgess J. Potentially preventable

hospitalizations among Medicare-Medicaid dual eligibles, 2008.

Rockville, Maryland: AHRQ; 2010.

[16] Westrick E, Kogut S. Medicare ambulatory care indicators for the el-

derly: Refinement of the access to care for the elderly project indica-

tors: MagnaCare Health Services Improvement, Inc. for MedPAC;

2006.

[17] Medicare Payment Advisory Commission. Report to the Congress:

Medicare payment policy. Washington, DC: MedPAC; 2011.

[18] Medicare Payment Advisory Commission. Appendixes: Physician and

other health professional services (online-only appendixes for Chapter

4, March 2011 report). http://www.medpac.gov/chapters/Mar11_

Ch04_APPENDIX.pdf/. Accessed October 1.

[19] Davies S,McDonaldKM, Schmidt E, Schultz E,Geppert J, RomanoPS.

Expanding the uses of AHRQ’s prevention quality indicators: Validity

from the clinician perspective. Med Care 2011;49:679–85.

[20] Parsons LS. Performing a 1:N Case-control match on propensity score.

Paper presented at: Proceedings of the 29th Annual SAS Users Group

International Conference; May 2004; Montreal, Canada.

[21] Austin PC. Balance diagnostics for comparing the distribution of base-

line covariates between treatment groups in propensity-score matched

samples. Stat Med 2009;28:3083–107.

P.-J. Lin et al. / Alzheimer’s & Dementia 9 (2013) 30–3838

[22] Fillit H, Hill JW, Futterman R. Health care utilization and costs of Alz-

heimer’s disease: the role of co-morbid conditions, disease stage, and

pharmacotherapy. Fam Med 2002;34:528–35.

[23] Berlowitz DR, Brandeis GH, Morris JN, Ash AS, Anderson JJ,

Kader B, et al. Deriving a risk-adjustment model for pressure ulcer de-

velopment using the Minimum Data Set. J Am Geriatr Soc 2001;

49:866–71.

[24] Andrieu S, Reynish W, Nourhashemi F, Ousset PJ, Grandjean H,

Grand A, et al. Nutritional risk factors for institutional placement in

Alzheimer’s disease after one year follow-up. J Nutr Health Aging

2001;5:113–7.

[25] Horner J, Alberts MJ, Dawson DV, Cook GM. Swallowing in Alz-

heimer’s disease. Alzheimer Dis Assoc Disord 1994;8:177–89.

[26] Mitchell SL, Teno JM, Kiely DK, Shaffer ML, Jones RN,

Prigerson HG, et al. The clinical course of advanced dementia. N

Engl J Med 2009;361:1529–38.

[27] Obama PB. Draft. National plan to address Alzheimer’s disease, http://

aspe.hhs.gov/daltcp/napa/NatlPlan.shtml/. Accessed March 25.

[28] Agency for Healthcare Research and Quality. Guide to prevention

quality indicators: Hospital admission for ambulatory care sensitive

conditions (Revision 3). AHRQ Publication No. 02–R0203. Rockville,

MD: AHRQ; January 9, 2004.

[29] Lorenz K, Lynn J, Morton SC, Dy S, Mularski R, Shugarman L, et al.

End-of-Life care and outcomes: Summary, evidence report/technology

assessment no. 110. (prepared by the Southern California Evidence-

based Practice Center, under Contract No. 290-02-0003). AHRQ Pub-

lication No. 05-E004-1. Rockville, MD: AHRQ; December 2004.

[30] Vickrey BG, Mittman BS, Connor KI, Pearson ML, Della Penna RD,

Ganiats TG, et al. The effect of a disease management intervention on

quality and outcomes of dementia care: a randomized, controlled trial.

Ann Intern Med 2006;145:713–26.

[31] Callahan CM, Boustani MA, Unverzagt FW, Austrom MG,

Damush TM, Perkins AJ, et al. Effectiveness of collaborative care

for older adults with Alzheimer disease in primary care: a randomized

controlled trial. JAMA 2006;295:2148–57.

Appendix

Descriptionof potentially avoidable hospitalization indicators defined by the Med

Indicator Description

1. Hospitalizations for serious short term complications

of diabetes

Admissions for di

diabetes mellitu

mellitus within

2. Hospitalizations for serious long term complications

of diabetes

Hospitalizations fo

diabetes mellitu

out-patient visi

a calendar year

3. Hospitalizations for hypertension Hospitalizations w

older, with 2 or

measurement y

4. Hospitalizations for COPD/asthma Admissions for res

2 visits (out- or

5. Hospitalizations for heart failure Admissions for he

1 inpatient visi

Source: Westrick E, Kogut S. Medicare Ambulatory Care Indicators for the E

(final report to MedPAC, January 2006).

[32] Epstein AM, Jha AK, Orav EJ. The relationship between hospital

admission rates and rehospitalizations. N Engl J Med 2011;

365:2287–95.

[33] Boustani MA, Sachs GA, Alder CA, Munger S, Schubert CC, Guer-

riero Austrom M, et al. Implementing innovative models of dementia

care: the Healthy Aging Brain Center. Aging Ment Health 2011;

15:13–22.

[34] McCall N, Cromwell J. Results of the Medicare Health Support

disease-management pilot program. N Engl J Med 2011;365:1704–12.

[35] Peikes D, Chen A, Schore J, Brown R. Effects of care coordination on

hospitalization, quality of care, and health care expenditures among

Medicare beneficiaries: 15 randomized trials. JAMA 2009;

301:603–18.

[36] Bott DM, Kapp MC, Johnson LB, Magno LM. Disease management

for chronically ill beneficiaries in traditional Medicare. Health Aff

(Millwood) 2009;28:86–98.

[37] CMS,SNPcomprehensive report.www.cms.gov/mcradvpartdenroldata/

snp/List.asp/. Accessed October 1.

[38] Maciejewski ML, Wang V, Grabowski DC, Lin PJ. Early evidence on

the quality of care provided by special needs plans. Med Care 2011;

49:891–6.

[39] U.S. Department of Health and Human Services. Accountable

Care Organizations: Improving care coordination for people

with Medicare. http://www.healthcare.gov/news/factsheets/2011/03/

accountablecare03312011a.html/. Accessed November 1.

[40] Lin PJ, Kaufer DI, Maciejewski ML, Ganguly R, Paul JE, Biddle AK.

An examination of Alzheimer’s disease case definitions using Medi-

care claims and survey data. Alzheimers Dement 2010;6:334–41.

[41] Fillit H, Geldmacher DS, Welter RT, Maslow K, Fraser M. Optimiz-

ing coding and reimbursement to improve management of Alz-

heimer’s disease and related dementias. J Am Geriatr Soc 2002;

50:1871–8.

[42] Newcomer R, Clay T, Luxenberg JS, Miller RH. Misclassification and

selection bias when identifying Alzheimer’s disease solely fromMedi-

care claims records. J Am Geriatr Soc 1999;47:215–9.

icare Ambulatory Care Indicators for the Elderly (MACIE)

abetic, hyperosmolar, and ketotic coma and admissions for uncontrolled

s among patients with 2 outpatient visits or 1 inpatient visit with diabetes

a calendar year.

r renal, ophthalmologic, neurologic and circulatory complications of

s and non-traumatic lower extremity amputation, in patients with 2 or more

ts or 1 in-patient visit with a diagnosis code for diabetes mellitus within

.

ith hypertension as the primary diagnosis, in patients aged 65 years and

more out- or in-patient visits with a diagnosis code for hypertension in the

ear.

piratory diagnoses among patients with COPD (including asthma) defined as

inpatient) with coded for COPD or asthma in the measurement year.

art failure in the measurement year among patients with 2 outpatient visits or

t with heart failure in the year prior to the measurement year.

lderly: Refinement of the Access to Care for the Elderly Project Indicators