Nursing Home Costs and Quality of Care Outcomes

NURSING ECONOMIC$/July-August 2004/Vol. 22/No. 4178

SUBSTANTIAL UNCERTAINTY sur-rounds the issue of the rela-tionship between nursinghome costs and quality ofcare. When individuals need long-term services for themselves, orfamily members, because they canno long manage at home, they aredemanding that the necessary nurs-ing home care be available, andthat the care be good quality.Consumers are also demandingmechanisms that will allow them

Nursing Home Costs and QualityOf Care Outcomes

LANIS L. HICKS, PhD, is Professor, HealthManagement and Informatics, School ofMedicine, University of Missouri-Columbia, Columbia, MO.

MARILYN J. RANTZ, PhD, RN, FAAN, isProfessor, Sinclair School of Nursing andDepartment of Family and CommunityMedicine, School of Medicine, andUniversity Hospital Professor of Nursing,University of Missouri-Columbia, SinclairSchool of Nursing, Columbia, MO.

GREGORY F. PETROSKI, MS, isStatistician, Office of Medical Research,School of Medicine, University ofMissouri-Columbia, Columbia, MO.

DANA B. MUKAMEL, PhD, is AssociateProfessor, Department of Medicine, andSenior Fellow, Center for Health PolicyResearch, University of California, Irvine.

Executive Summary Two major issues facing most

skilled nursing facilitiesinclude concerns over qualityof care and declining Medicaidreimbursement.

The authors examined therelationship between variablecosts and four specific qualitymeasures: decline in ADLs,development of pressureulcers, weight loss, and psy-chotropic drug use.

The results revealed that vari-able costs can be influencednegatively by quality of care,particularly when all dimen-sions of quality are examinedtogether.

As expected, declining ADLsand worsening pressure ulcersaccelerate care costs.

Risk-adjusted patient daysexplained the largest variancein cost suggesting that bulk ofpatient care consuming thelargest amount of staffing dol-lars must be provided regard-less of the quality of caredelivered.

to evaluate the quality of care pro-vided in nursing homes. In addi-tion to consumers, policymakers,who are responsible for oversightof the public funding of more than70% of patient days in nursinghome care and approximately two-thirds of expenditures on nursinghome care, want assurance thatmonies are contributing to betterquality care (Grabowski & Hirth,2003; Levit et al., 2003).

The inadequacy of the quality

Lanis L. HicksMarilyn J. RantzGregory F. PetroskiDana B. Mukamel

ACKNOWLEDGMENT: The authors wish toacknowledge the contributions of otherUniversity of Missouri-Columbia MDS andNursing Home Quality Research Teammembers and the Quality ImprovementProgram for Missouri nurses. We gratefullyacknowledge the support of the MissouriDepartment of Health and Senior Servicesstaff; they are truly committed to helpingnursing homes embrace quality improve-ment. Research activities were partiallysupported by a cooperative agreement withthe Missouri Department of Health andSenior Services and the Sinclair School ofNursing and Biostatistics Group of theSchool of Medicine, University of Missouri-Columbia,Contract #AOC03380064. Fundsfrom the Missouri Department of Healthand Senior Services included partial sup-port from the Centers for Medicare andMedicaid Services. Opinions are those ofthe authors and do not represent theMissouri Department of Health and SeniorServices or the Centers for Medicare andMedicaid Services.

179NURSING ECONOMIC$/July-August 2004/Vol. 22/No. 4

of care delivered in nursing homeshas long been an issue. Followinga 1986 Institute of Medicine reporton prevalent problems with thequality of care in nursing homes(IOM, 1986), a number of laws,regulations, and efforts have beenintroduced to address the prob-lems. Despite improvements in anumber of areas, problems withquality of care continue to persist,as evidenced by an increase in 13of 25 quality of care deficienciesand an increase in ombudsmancomplaints in recent years (Officeof Inspector General, 1999). Also,a recent study found that over25% of nursing homes hadpotentially life threatening prob-lems in delivering care and wereharming residents (Harrington,2001, p. 507).

A common perception is thathigher quality is associated withhigher costs. If this is found to betrue, then improving quality innursing homes will face substan-tial obstacles, especially in thesetimes of state budget deficits andthe federal proposals to convertMedicaid funding to block grants(U.S. Department of Health andHuman Services, 2003). SinceMedicaid is a major fundingsource for nursing homes, payingover 50% in 2002 (Levit et al.,2003), if there is a direct relation-ship between costs and quality,then efforts to improve quality innursing homes may be increasing-ly difficult.

A number of studies (Arling,Nordquist, & Capitman, 1987;Birnbaum, Bishop, Lee, & Jenson,1981; Hicks et al., 1997; Holmes,1996) attempted to assess the fac-tors contributing to the variationsin the costs of care in nursinghomes. An underlying assumptionin these studies was the existenceof a direct relationship betweencosts and the efficient provision ofservices appropriate to patientneeds. However, two prior studies(Fleming, 1991; Mukamel &Spector, 2000) in nursing homesand hospitals, respectively, havefound that higher quality may be

ities. Two General AccountingOffice (GAO) reports (Ochinko,2002a; Ochinko, 2002b) have alsoexamined the issues of accuracyand public reporting and, whilerecognizing the limitations ofsome of the measures and theopportunity for confusion oninterpretation, conclude the datacan be used to measure good andpoor quality of care practices. Themultidimensional resident-specif-ic aspects of the MDS data providemore direct measures of quality ofcare than do more indirect proxymeasures, such as facility surveycitations, which measure qualityas compliance with minimumstandards.

MethodsThe relationship between

nursing home variable costs andfour quality of care outcome mea-sures decline in ADLs, devel-opment of pressure ulcers, weightloss, and psychotropic drug use was examined. These quality indi-cators were selected because of theperceived direct relationshipbetween clinical actions and qual-ity outcomes and the ability to beaffected by care management.Final 1999 Medicaid cost reportand MDS assessment data wereused in this analysis for all certi-fied nursing facilities in Missouri.The MDS was selected because ithas established reliability (Hawes,Phillips, Mor, Fries, & Morris,1992; Hawes et al., 1995; Morris etal., 1990; Morris et al., 1994;Phillips, Chu, Morris, & Hawes,1993; Phillips, Hawes, Mor, Fries,& Morris, 1996; Snowden et al.,1999); has focused attention oncare needs and resident outcomesin nursing homes (Hawes et al.,1992; Hawes et al., 1995; Rantz etal., 1996; Rantz et al., 2001); andhas been previously used inresearch (Brandeis, Baumann,Hossain, Morris, & Resnick, 1997;Mor et al., 1997; Rantz, 1995;Williams & Betley, 1995), includ-ing research conducted by thisteam (Rantz et al., 1996; Rantz etal., 1997a; Rantz et al., 1997b;

associated with lower costs. Thisfinding is also supported by thetenets of the quality improvementmovement, which maintains thatby re-engineering the processes ofcare, improvements in quality andreductions in cost can be achievedsimultaneously. If it can be shownthat it is possible to improve qual-ity without increasing costs, thennursing homes may be able toenhance quality, even in this eraof dwindling resources.

The objective of this studywas to test the hypothesis thathigher quality may be associatedwith lower costs in an environ-ment and for quality dimensionsdifferent from those studied earli-er by Mukamel and Spector(2000). Specifically, data wereexamined from a state that is notas highly regulated and does nothave as rich a Medicaid paymentsystem as New York, and incorpo-rates different quality measuresthan previously analyzed. In thisstudy, indicators of resident con-ditions contained in theMinimum Data Set (MDS) wereused to reflect quality, rather thandata from the New York PatientReview Instrument on residenthealth status, case mix, medicalconditions, and treatments. TheMDS is a multidimensional resi-dent-specific instrument routinelyused on all residents at time ofadmission, at times of significantchange in condition, annually,and quarterly for selected items;the MDS is federally mandated forall facilities participating inMedicaid and Medicare. The MDSprovides assessment data on abroad range of resident care needsand treatments, such as skin con-dition, incontinence, decline inactivities of daily living (ADL),medication use, and weight loss.

While many question theaccuracy of the MDS and the qual-ity indicators (QIs) derived fromthese data, a national validationstudy (Morris et al., 2002) con-cluded that many of the MDS QIscapture important aspects of theperformance of nursing care facil-


Rantz et al., 2000; Rantz et al.,2001).

Sample. Cost data from 474nonhospital-based nursing facili-ties in Missouri were matched tofacility-specific MDS data. Twenty-eight facilities were excluded dueto missing data on key variables orto implausible values on the costvariables (for example, negativesalary expenses), resulting in asample of 446. The characteristicsof the omitted facilities were var-ied in terms of size, ownership,and location.

Cost and facility characteris-tics. Costs were based on 1999expenditures, excluding costs lessresponsive to changes in servicedelivery and less likely to affectquality of care. Facility character-istics included in the initial costmodel were patient days, RN andnursing assistant (aides and order-lies) wages, competition, facilitysize, location, and ownership.Patient days were case-mix adjust-ed using resident level RUGS IIIscores. RN and nursing assistantwages were based on total compen-sation (wages plus benefits).Competition was measured by theHerfindhal-Hirschman Index, whichranges between 1 (monopoly) and 0(perfect competition), and is definedas the sum of squared market shares(Feldstein, 1999; Grabowski & Hirth,2003). Facility size was categorizedas small ( 60 beds), medium (61-120beds), and large (121 or morebeds), and was used to account foreconomy of scale in the model.Facility size is conceptualized as acharacteristic of the facilities, notas a measure of capital cost.Location was defined as metropol-itan, urban, or rural according to aU.S. Department of Agriculturecounty typology (Cook & Mizer,1989). Ownership was defined asinvestor-owned or tax-exempt(including governmental).

The dependent variable in theanalysis was the natural logarithmof variable costs reported by eachfacility. Variable costs weredefined as expenditures related topatient care, ancillary services,

reflecting total dependence onstaff help, giving the summatedscale a range of 0 to 16. Incidenceof ADL decline was defined as atleast a two-point increase in thisscale from admission to first quar-terly assessment. A minimumchange of two points was selectedbecause such a change over a rela-tively short time period (90 days)reflects either a dramatic declinein ability to perform one late-lossADL, or a more general decline inmultiple late-loss ADLs. Residentswith a score of 16 on admissionwere excluded from the analysis,because their status could notdecline.

The other quality measuresused in this study were weightloss, as measured by the MDS itembeing checked that a resident hadexperienced a 10% loss in 6months. Pressure ulcers were mea-sured by the MDS item indicatingthat the resident had a stage 1 to 4pressure ulcer. The psychotropicdrug use measure was defined asthe resident having a new useindicated for antipsychotics orantidepressants.

Most of the independent vari-ables used in the risk models aresingle MDS items. The exceptionsare the ADL scale discussedabove, an indicator for the signsand symptoms of depression onadmission (Burrows, Morris,Simon, Hirdes, & Phillips, 2000),and the Cognitive PerformanceScale, a seven-point scale withlarger values reflecting greatercognitive impairment (Morris etal., 1990).

Table 1 provides an overviewof the quality measures by facilityownership type and facility size inthe study. Table entries are themedian values of the expectedminus observed incidence ratesused as measures of quality. Thesummary statistics in Table 1 arefor the quality measures expressedon the percent scale. Ninety-fivepercent confidence intervals forthe median values are also includ-ed to appreciate variability inthese measures better.

and administration as reported onSchedule B of the state Title XIX(Medicaid) Cost Reports. The cate-gories of capital, depreciation,taxes, and other were excludedfrom the costs considered, asthese reflect more fixed costs, andare less responsive to changes inservices provided to residents thatare likely to influence quality ofcare.

Quality measures. In thisstudy, four resident-level mea-sures of quality from the MDSdata were used: ADL (activities ofdaily living) decline, pressureulcers, weight loss, and use ofpsychotropic drugs (antidepres-sant and antipsychotics use).These four measures were select-ed because they are sensitive tochanges in resident functionaldecline. In this analysis, qualitymeasures are based on admissionMDS data and the changes thatoccurred in a later followupassessment. Each of the outcomesis an incidence measure, becausesubjects positive for the conditionon admission were excluded fromthe analysis. The followup periodwas the quarterly review 90 dayspost admission, defined as theassessment closest to 90 dayswithin a window of 45 to 135 daysfrom admission. For the qualitymeasures, residents were includ-ed only when their MDS assess-ments, both admission and fol-lowup, occurred within theirfacilitys fiscal reporting periodplus or minus 180 days from thestart or close of the reporting peri-od. To examine relevant associa-tions between cost and quality,MDS assessments must be nearthe fiscal period, yet the windowmust be wide enough to yield thelargest sample size possible.

ADL functioning was mea-sured as a summated scale formedfrom four MDS items assessingself-performance of bed mobility,transfers, eating, and toileting.Each of the MDS ADL self-perfor-mance items was scored on a five-point scale, with 0 indicatingindependent performance and 4


The risk adjustments for eachquality of care measure werebased on resident-level character-istics identified by the clinicalexperts in the research team.Random intercept logistic regres-sion methods were used to identi-fy a combination of resident-levelvariables that predict each qualitymeasure. The inclusion of a ran-dom intercept allows the baselinelog-odds of an event to vary fromfacility to facility, while the effectsof the risk factors are assumed tobe the same across facilities. Foreach quality measure, an initiallogistic model was fit that includedall candidate risk factors. Stati-stically nonsignificant (p>0.05)

models (Hastie & Tibshirani, 1990)were used to explore the need fornonlinear terms for quantitativepredictors and to identify possibleways to collapse ordinal levelindependent variables.

In the resident-level risk mod-els, the importance of each predic-tor variable was quantified by theodds ratio, along with 95% confi-dence intervals for the odds ratio.The ability of the risk models todiscriminate between high-riskand low-risk residents for a givenoutcome was quantified by the C-statistic. The c-statistic is the areaunder the receiver operating char-acteristic curve (Hanley & McNeil,1982). For any reasonable model

terms were sequentially removed,starting with the least significant, toderive simplified resident-levelrisk models. The reduced modelswere then used to form risk-adjusted quality measures.

In a few cases, predictors thatwere not significant at the 0.05level were retained because theywere judged to be clinicallyimportant, or because they wereincluded in statistically signifi-cant interaction terms. For exam-ple, age and gender were not indi-vidually significant in the riskmodel for pressure ulcers, butwere retained because of the sta-tistically significant age-genderinteraction. Generalized additive

Table 1.Quality Measures by Facility Ownership and Facility Size

Median Values with 95% Confidence Intervals of the Median

Ownership

Investor-owned

Investor-owned

Investor-owned

Tax-exempt

Tax-exempt

Tax-exempt

BedSize

60

61-120


Table 2.Risk-Adjustment Models for Quality of Care Measures

ADL DECLINEC Statistic 0.68

Effect

Intercept

Age

Cancer

Stroke

Infection

Unstable disease

End-stage disease

Urinary incontinence

Admission ADL score

Quadratic term for admission ADL score

CPS1

CPS2

CPS3

CPS4

CPS5

CPS6

RegressionCoefficient

-3.07

0.01

0.33

0.10

0.09

0.15

0.72

0.28

0.08

-0.01

-0.14

-0.05

0.18

0.42

0.45

0.79

StandardError

0.20

0.00

0.07

0.05

0.50

0.05

0.16

0.05

0.02

0.00

0.07

0.07

0.06

0.10

0.09

0.21

SignificanceLevel


Table 2. (continued)Risk-Adjustment Models for Quality of Care Measures

WEIGHT LOSSC Statistic 0.68

Effect

Intercept

Age

Male

Depressed

Cancer

COPD

Stroke

Infection

Edema

Unstable disease

End-stage disease

Deteriorating incontinence

Admission ADL score

RegressionCoefficient

3.42

0.01

0.13

0.28

0.25

0.16

-0.11

0.11

0.23

0.25

0.46

0.18

0.05

StandardError

0.23

0.00

0.05

0.07

0.08

0.06

0.06

0.05

0.05

0.05

0.20

0.06

0.01

SignificanceLevel


(that is, the model does not doworse than chance in predictingan outcome), C has a range from0.5 to 1.0, with 0.5 indicating thata model has no discriminatorypower, and 1.0 reflecting perfectdiscrimination.

Based on the fitted logisticmodels, each resident has an esti-mated probability of having thecondition under consideration in effect, his/her risk. These prob-abilities are then aggregated to thefacility level to form the facilitysexpected proportion of residentshaving the condition. The qualitymeasures used as independentvariables in this study are the dif-ferences between expected andobserved quality scores, where theobserved score is simply the pro-portion of residents positive for

tion with constant variance. Inthis data set, the distribution ofvariable costs had substantial pos-itive skew. The log transformationproduced a better fit to normalityas reflected, and effectively stabi-lized the error variance. Graphicalmethods were used to check theregression residuals from the log-transformed model for normalityand constant variance. Bothassumptions were reasonably sat-isfied.

Quality measures were en-tered as a cubic polynomial, thusproviding flexible models thatinclude simple linear relation-ships as a special case. To deter-mine if any relationship existsbetween cost and quality as theyare defined here, preliminaryregression analyses were per-

the condition, without risk adjust-ments. Because the quality mea-sures are defined as the differencebetween expected and observedproportions, positive values reflectbetter quality (the expected inci-dence was greater than theobserved incidence). Similarly,negative values reflect poorer qual-ity because the expected incidencewas lower than the observed inci-dence. From this point forward inthis manuscript, quality measurerefers to these expected, observeddifference scores.

Cost analysis. The dependentvariable for the cost analyses isthe natural logarithm of variablecost. Several factors motivated thedecision to use the log transforma-tion. Ordinary linear regressionassumes a normal error distribu-

Table 3.Facility Characteristics

LocationMetropolitanUrbanRural

OwnershipInvestor-ownedGovernmentTax-exempt

Occupancy

Average case mix

Case mix adjusted resident days

Resident days

HHI: Market share

RN wages

Nurse assistant wage

Log of Variable Costs

Mean (SD)Investor-ownedGovernment/Tax-exempt

TotalN=446

21117263

3293582

Mean

79.36%

0.77

23,323

30,181

0.26

$19.81/hour

$8.37/hour

Total

N=446329117

60 Bedsn=102

275025

651621

(SD)

(13.85)

(0.05)

(12,992.46)

(1,627.87)

(0.24)

(3.16)

(1.49)

120 Bedsn=93

75171

696

18

>120 Beds

n=9315.07 (0.43)15.30 (0.50)


formed in which log-variable-costwas regressed on a cubic polyno-mial for each of the quality mea-sures alone. This was followed bydevelopment of a facility charac-teristic model without considera-tion of quality. Once this modelwas determined, the quality mea-

quality measures (for example,100% of residents reporting on anindicator is implausible, indicat-ing a potential data reportingproblem). These homes were alsoexcluded from the summary statis-tics for the quality measures,hence the n=443 in Table 6.

Linear, quadratic, and cubicterms are included. In the accom-panying tables, the quadratic andcubic terms are denoted by vari-ables suffixed with 2 and 3,respectively. Facility characteris-tics and quality measures were fitseparately to investor-owned andtax-exempt homes. Linear, qua-dratic, and cubic measures wereused to examine different shapesof relationships in the model.

Statistically significant rela-tionships between cost and quali-ty were further explored by trans-forming log costs back to the actu-al dollars scale and constructingplots of actual costs as a functionof changes in the quality mea-sures. This was done for eachfacility size, with wage, case mix,and market dominance variablesfixed at their mean values for thefacility size and type of owner-ship. In retransforming costs toactual dollars, the nave estimator(Duan, 1983) of expected costswas used, along with bootstrapmethods (Efron & Tibshirani,1993) to construct point-wise 90%confidence intervals for expectedcosts.

ResultsResident-level risk models.

The resident-level risk modelsused in forming the risk-adjustedquality measures are given inTable 2. As indicated by C-statis-tics in the range 0.68 to 0.70, theresident-level risk models do sub-stantially better than chance atidentifying the residents at risk foreach of the conditions.

Facility characteristics. Table3 provides a summary of facilitycharacteristics used in this study.As reflected in this table, the dom-inant home in Missouri is between61 to 120 beds, located in a metro-

sures were individually added tothe facility model. The statisticalsignificance of the quality mea-sures was assessed using both uni-variate and multivariate tests.Three facilities were excludedfrom the regression analyses dueto extreme outlying values on the

Table 4. Polynomial Regression of Log of Variable Costs on

Quality Measures

Quality Measure

ADL declineR2=0.08

Pressure ulcersR2=0.12

Weight lossR2=0.05

Psychotropic drug useR2=0.06

RegressionTerm

InterceptLinearQuadraticCubic




Parameter Estimate(SE)

14.57 (0.028)-1.42 (0.6411)-22.76 (4.872)-14.23 (34.11)

14.61 (0.030)-3.65 (0.799)-86.28 (15.37)-257.17 (98.31)

14.56 (0.029)0.05 (0.674)-22.57 (5.016)-82.02 (40.00)

14.56 (0.028)-1.01 (0.355)-7.25 (1.881)-6.12 (6.069)

p


politan area, and is investor-owned. The top part of the tableprovides descriptive statistics onthe number of homes in the statein terms of size, location, andownership. The center portion ofTable 3 provides overall summarystatistics on facility case mix,

occupancy, market share, residentdays, and wages for RNs and nurseaides. The lower portion of thetable provides summary statisticsfor log-variable costs by facilitysize and type of ownership.

Relationship between costs,quality, and facility characteris-

tics. The results of the initialregression analyses using only thecubic polynomial for each of thequality measures and costs, wherecosts reflect the variable costs ofthe institution, are presented inTable 4. Although not all terms arealways statistically significant, the

1.21

1.20

1.19

1.18

1.17

1.16

1.15

1.14

1.13

1.12

1.11

1.10

1.09

1.08

1.07

1.06

1.05

1.04

1.03

-0.10 -0.08 -0.06 -0.04 -0.02 -0.00 0.02 0.04 0.06 0.08 0.10

Risk Adjusted Expected Incidence Observed Incidence

Var

iab

le C

ost

s (x

1,0

00.0

00)

4.00

3.95

3.90

3.85

3.80

3.75

3.70

3.65

3.60

3.55

3.50

3.45

3.40

3.35-0.10 -0.08 -0.06 -0.04 -0.02 -0.00 0.02 0.04 0.06 0.08 0.10


Var

iab

le C

ost

s (x

1,0

00.0

00)

2.06

2.05

2.04

2.03

2.02

2.01

2.00

1.99

1.98

1.97

1.96

1.95

1.94

1.93

1.92

1.91

1.90

1.89

1.88

1.87

1.86

1.85

1.84

1.83

1.82

1.81

1.80

1.79

1.78-0.10 -0.08 -0.06 -0.04 -0.02 -0.00 0.02 0.04 0.06 0.08 0.10


Var

iab

le C

ost

s (x

1,0

00.0

00)

Small Facilities 120 beds

Figure 1.Variable Costs Associated with Activities of Daily Living Quality

1.55

1.50

1.45

1.40

1.35

1.30

1.25

1.20

-0.10 -0.08 -0.06 -0.04 -0.02 -0.00 0.02 0.04 0.06 0.08 0.10


Var

iab

le C

ost

s (x

1,0

00.0

00)

5.505.455.405.355.305.255.205.155.105.055.004.954.904.854.804.754.704.654.604.554.504.454.404.354.304.254.202.154.104.054.00

-0.10 -0.08 -0.06 -0.04 -0.02 -0.00 0.02 0.04 0.06 0.08 0.10


Var

iab

le C

ost

s (x

1,0

00.0

00)

2.80

2.75

2.70

2.65

2.60

2.55

2.50

2.45

2.40

2.35

2.30

2.25

2.20

-0.10 -0.08 -0.06 -0.04 -0.02 -0.00 0.02 0.04 0.06 0.08 0.10


Var

iab

le C

ost

s (x

1,0

00.0

00)

Small Facilities 120 beds

Figure 2.Variable Costs Associated with Weight Loss Quality


R2 for these regressions range fromapproximately 5% to 12%, sug-gesting that quality may influencevariable costs to a meaningfuldegree.

The preliminary regressionincluded all proposed facilitycharacteristics and the resulting R2

value was 0.94, indicating therewas not a great deal of variabilityin costs to be contributed by othervariables. The location designa-tions of metropolitan and urbanwere not statistically significant atthe 0.05 level; they were thenexcluded and the model was refit-

ted. Table 5 provides the finalregression results of the log ofvariable costs for the facilitymodel with only the statisticallysignificant (p


Table 6. (continued)Regression Results: Log of Variable Costs and Quality of Care Outcomes

WEIGHT LOSSFull Sample

R-Square 0.9422N=442

Variable

Intercept

Log of case-mix adjusted patient day

Log of hourly RN wage

Log of hourly aides/orderly wages

Herfindhal-Hirschman Index

Midsize

Large Size

Tax-exempt

WLs linear

WLs squared

WLs cubic

Joint test for weight loss terms (p=0.0317)

Weight Loss

Investor-owned

R-Square 0.9430

WLs linear

WLs squared

WLs cubic


Weight Loss

Tax-exempt/government

R-Square 0.9423

WLs linear

WLs squared

WLs cubic


Parameter Estimate

4.16

0.91

0.10

0.47

-0.11

0.05

0.11

0.10

-0.04

4.18

15.31

0.055

2.0

6.03

-0.51

9.78

50.16

Standard Error

0.22

0.02

0.04

0.04

0.03

0.02

0.03

0.01

0.17

1.43

11.15

0.20

1.70

15.44

0.39

2.80

20.34

p



PRESSURE ULCERSFull sample


Variable

Intercept





Midsize

Large Size

Tax-exempt

PUs linear

PUs squared

PUs cubic

Joint test for pressure ulcer terms (p=0.1061)

Pressure Ulcers

Investor-Owned

R-Square 0.9434

PUs linear

PUs squared

PUs cubic


Pressure Ulcers


R-Square 0.9383

PUs Linear

PUs Squared

PUs Cubic


Parameter Estimate

4.27

0.89

0.10

0.48

-0.12

0.05

0.11

0.09

-0.32

2.69

6.41

-0.22

1.87

-24.58

-0.38

12.54

60.51

Standard Error

0.22

0.02

0.04

0.04

0.03

0.02

0.03

0.01

0.22

4.35

32.19

0.28

4.78

43.88

0.45

11.35

68.28

p



PSYCHOTROPIC DRUG USEFull Sample


Variable

Intercept





Midsize

Large Size

Tax-exempt

PDs linear

PDs squared

PDs cubic

Joint test for PD terms (p=0.7082)

Psychotropic Drug Use

Investor-0wned

R-Square 0.9430

PDs linear

PDs squared

PDs cubic


Psychotropic Drug Use


R-Square 0.9386

PDs linear

PDs squared

PDs cubic


Parameter Estimate

4.26

0.89

0.11

1.48

-.011

0.05

0.11

0.10

-0.10

0.21

1.41

-0.14

0.21

2.02

0.27

-0.29

-10.71

Standard Error

0.22

0.02

0.04

0.04

0.03

0.02

0.03

0.01

0.09

0.50

1.56

0.10

0.53

1.62

0.24

1.67

7.59

p


measure, the investor-owned facil-ities contributed to the signifi-cance of the quality measure.However, in the weight-loss mea-sure, it was the tax-exempt facili-ties that contributed to the signifi-cance of the quality measure.

In the model presented in thisstudy, an overwhelming amountof the variance was explained byrisk-adjusted patient days in thefacilities. This result is plausibleand logical, given the largeamount of basic care and servicesthat must be provided to patientsin a nursing home, regardless ofquality of care delivered. As aresult, the majority of the costsincurred in a facility exist regard-less of the quality of care providedin that facility. However, theresults from these four, althoughvery specific and narrow dimen-sions of quality, indicate that qual-ity of care does have an impact ona portion of the costs. The curvesin this study suggest that variablecosts may be minimized at a cer-tain quality level, specificallywhen staffing or other facilityresources are allocated in such away as to reduce incidence levelsto slightly less than would beexpected for the case mix of thefacility. Sacrificing quality doesnot appear to be an effectivemethod of cost containment; how-ever, as illustrated by the graphs,further reduction of incidence lev-els below the expected may resultin increased costs. Therefore,nursing homes that successfullyfocus on providing quality of carethrough innovative protocols andcare management strategies canhave a positive impact on thecosts of the home.

These findings, using MDSdata to measure quality of care, aresimilar to the findings reported byMukamel and Spector (2000). Theresults of this study are also con-sistent with the results providedin the recent GAO report(Dummit, 2002) showing that theaverage share of total expendituresthat is spent on basic resident careactivities is relatively high and

Bliesmer, M.M., Smayling, M., Kane, R.L.,& Shannon, I. (1998). The relationshipbetween nursing staffing levels andnursing home outcomes. Journal ofAging and Health, 10(3), 351-371.

Brandeis G.H., Baumann, M.N., Hossain,M., Morris, J.N., & Resnick, N.M.(1997). The prevalence of potentiallyremediable urinary incontinence infrail older people: A study using theminimum data set. Journal of theAmerican Geriatrics Society, 45(2),179-184.

Burrows, A.B., Morris, J.N., Simon, S.E.,Hirdes, J.P., & Phillips, C. (2000).Development of a minimum data set-based depression rating scale for usein nursing homes. Age and Ageing,29(2), 165-172.

Cook, P.J., & Mizer, K.L. (1989). The revisedERS county topology: An overview.Rural Economy Division, EconomicResearch Service, US Department ofAgriculture, Rural DevelopmentReport 89. Washington, DC: U.S.Government Printing Office.

Duan, N. (1983). Smearing estimate: A non-parametric retransformation method.Journal of the American StatisticalAssociation, 89, 605-610.

Dummit, L.A. (2002). Nursing homes:Quality of care more related tostaffing than spending. Washington,DC: United States General AccountingOffice.

Efron, B., & Tibshirani, R.J. (1993). Anintroduction to the bootstrap. NewYork: Chapman & Hall.

Feldstein, P.J. (1999). Health care econom-ics (5th ed.). Albany, NY: DelmarPublishers.

Felton, B.B. (1993). How organization ofnursing care and resident health sta-tus affect nursing home costs. NursingOutlook, 16(4), 15-23.

Fleming, S.T. (1991). The relationshipbetween quality and cost: Pure andsimple? Inquiry, 28(1), 29-38.

Grabowski, D.C., & Hirth, R.A. (2003).Competitive spillovers across non-profit and for-profit nursing homes.Journal of Health Economics, 22(1), 1-22.

Hanley, J.A., & McNeil, B.J. (1982). Themeaning and use of the area under areceiver operating characteristic(ROC) curve. Radiology, 143(1), 29-36.

Harrington, C. (2001). Residential nursingfacilities in the United States. BritishMedical Journal, 323(7311), 507-510.

Harrington, C., Zimmerman, D., Karon,S.L., Robinson, J., & Beutel, P. (2000).Nursing home staffing and its rela-tionship to deficiencies. Journal ofGerontology: Social Science, 55B(5),S278-S287.

Hastie, R.J., & Tibshirani, T.J. (1990).Generalized additive models. NewYork: Chapman and Hall.

Hawes, C., Phillips, C.D., Mor, V., Fries, B.,& Morris, J.N. (1992). MDS data set

stable, but that other costs showsubstantial variations and thatsome of that variation is related toquality of care in the homes.

In other studies (Harrington,Zimmerman, Karon, Robinson, &Beutel, 2000; Johnson-Pawlson &Infeld, 1996; Spector & Takada,1991), increased staffing was asso-ciated with better quality of care.Bliesmer, Smayling, Kane, andShannon (1998) also found thatlicensed nursing hours were relat-ed to improved functional ability,increased probability of dischargeto home, and decreased probabili-ty of death. Felton (1993) foundthat a higher ratio of registerednurses was associated with higherresident care costs, but lower totalcosts per resident per day. Studiesusing simple linear models toshow the relationship betweencosts and quality are absent fromthe literature.

While this study providesadditional information on the asso-ciation between quality and costsin nursing homes, more research isneeded into the specific cost fac-tors that influence different qualityof care measures. While the recentGAO report examined aggregatespending on nursing (Dummit,2002), more detailed examinationneeds to be made on not only thetype and mix of nursing careincluded in the overall cost, butalso the impact of such events astenure and turnover among staffand management, training levels,and actual processes of careemployed in the facility. Theresults of this study indicate thatproviding substandard care doesnot result in cost reductions innursing homes.$

REFERENCESArling, G, Nordquist, R.H., & Capitman,

J.A. (1987). Nursing home cost andownership type: Evidence of interac-tion effects. Health Services Research,22(2), 255-269.

Birnbaum, H., Bishop, C., Lee, A.J., &Jenson, G. (1981). Why do nursinghome costs vary? The determinants ofnursing home costs. Medical Care,19(11), 1095-1107.


should be used for research. TheGerontologist, 32(4), 563-564.

Hawes, C., Morris, J.N., Phillips, C.D., Mor,V., Fries, B.E., & Nonemaker, S. (1995).Reliability estimates for the minimumdata set for nursing home assessmentand care screening (MDS). TheGerontologist, 35(2), 172-178.

Hicks, L.L., Rantz, M.J., Petroski, G.F.,Madsen, R.W., Conn, V.S., Mehr, D.R.,& Porter, R. (1997). Assessing contrib-utors to cost of care in nursing homes.Nursing Economic$, 15(4), 205-212.

Holmes, J.S. (1996). The effects of owner-ship and ownership change on nurs-ing home industry costs. HealthServices Research, 31(3), 327-346.

Institute of Medicine, Committee onNursing Home Regulation. (1986).Improving the quality of care in nurs-ing homes. Washington, DC: NationalAcademy Press.

Johnson-Pawlson, J., & Infeld, D.L. (1996).Nursing staff and quality of care innursing facilities. Journal ofGerontological Nursing, 22(8), 36-45.

Levit, K., Smith, C., Cowan, C., Lazenby,H., Sensenig, A., & Catlin, A. (2003).Trends in U.S. Health Care Spending,2001. Health Affairs, 22(1), 154-164.

Mor, V., Intrator, O., Fries, B.E., Phillips, C.,Teno, J., Hiris, J., et al. (1997). Changesin hospitalization associated withintroducing the resident assessmentinstrument. Journal of the AmericanGeriatrics Society, 45(8), 1002-1010.

Morris, J.N., Moore, T., Jones, R., Mor, V.,Angelelli, J., Berg, K., et al. (2002).Validation of long-term and post-acutecare quality indicators. Executivesummary of report to CMS, Contract #500-95-0062/TO#2.

Morris, J.N., Fries, B.E. Mehr, D.R., Hawes,C., Phillips, C., Mor, V., & Lipsitz. L.A.(1994). MDS cognitive performancescale, Journal of Gerontology, 49(4),M174-82.

Morris, J.N., Hawes, C., Fries, B., Phillips,C., Mor, V., Katz, S., et al. (1990).

Rantz, M.J., Petroski, G., Madsen, R., Mehr,D.R., Popejoy, L., Hicks, L.L., et al.(2000). Setting thresholds for qualityindicators derived from MDS data fornursing home quality improvementreports: An update. Joint CommissionJournal on Quality Improvement,26(2), 101-10.

Rantz, M.J., Petroski, G.F., Madsen, R.W.,Scott, J., Mehr, D.R., Popejoy, L., et al.(1997a). Setting thresholds for MDSquality indicators for nursing homequality improvement reports. Journalon Quality Improvement, 23(11), 602-611.

Rantz, M.J., Popejoy, L., Mehr, D.R.,Zwygart-Stauffacher, M., Hicks, L.L.,Grando, V., et al. (1997b). Verifyingnursing home care quality using min-imum data set quality indicators andother quality measures. Journal ofNursing Care Quality, 12(2), 54-62.

Rantz, M.J., Popejoy, L., Petroski, G.F.,Madsen, R.W., Mehr, D.R., Zwygart-Stauffacher, M., et al. (2001).Randomized clinical trial of a qualityimprovement intervention in nursinghomes. The Gerontologist, 41(4), 525-538.

Snowden, M., McCormick, W., Russo, J.,Srebnik, D., Comtois, K., Bowen, J., etal. (1999). Validity and responsive-ness of the minimum data set. Journalof the American Geriatrics Society,47(8), 1000-1004.

Spector, W.D., & Takada, H.A. (1991).Characteristics of nursing homes thataffect resident outcomes. Journal ofAging and Health, 3(4), 427-454.

U.S. Department of Health and HumanServices. (2003). FY 2004 budget inbrief. Washington, DC: Author.

Williams, B., & Betley, C. (1995).Inappropriate use of nonpsychotropicmedications in nursing homes.Journal of the American GeriatricsSociety, 43(5), 513-519.)

Designing the national residentassessment instrument for nursinghomes. The Gerontologist, 30(3), 293-307.

Mukamel, D.B., & Spector. W.D. (2000).Nursing home costs and risk-adjustedoutcome measures of quality. MedicalCare, 38(1), 78-89.

Ochinko, W. (2002a). Nursing homes:Federal efforts to monitor residentassessment data should complementstate activities. GAO-02-279.Washington, DC: U.S. GeneralAccounting Office.

Ochinko, W. (2002b). Nursing homes:Public reporting of quality indicatorshas merit, but national implementa-tion is premature. GAO-03-187.Washington, DC: U.S. GeneralAccounting Office.

Office of Inspector General. (1999). Qualityof care in nursing homes: An overview.OEI-02-99-00060. Washington, DC: U.S.Department of Health and HumanServices.

Phillips, C.D., Chu, C.W., Morris, J.N., &Hawes, C. (1993). Effects of cognitiveimpairment on the reliability of geri-atric assessments in nursing homes.Journal of the American GeriatricsSociety, 41(2), 136-142.

Phillips, C.D., Hawes, C., Mor, V., Fries,B.E., & Morris, J.N. (1996). Evaluationof the nursing home resident assess-ment instrument: Executive summary.Report to the Health Care FinancingAdministration by the ResearchTriangle Institute, 1-12

Rantz, M.J. (1995). Examining MDS resi-dent assessments for the impact ofpublic policy. The Journal of Long-Term Care Administration, 23(3), 18-21.

Rantz, M.J., Mehr, D.R., Conn, V., Hicks,L.L. Porter, R., Madsen, R.W., et al.(1996). Assessing the quality of nurs-ing home care: The foundation forimproving resident outcomes. Journalof Nursing Care Quality, 10(4), 1-9.

Honor Society Releases 3-year Study ResultsOn Global Nursing Practices

The Honor Society ofNursing, Sigma Theta TauInternational, has pub-lished A Report of Arista3,the results of a series of fiveglobal think-tank meetingsheld to develop strategiesfor changing nursing practice.

One hundred nine health care experts represent-ing nursing, medicine, health policy, government,economics and finance, administration, and non-governmental organizations explored: The environments in which nurses practice. The way nurses lead.

The knowledge nurses need. The settings in which nurses work. The conditions under which the profession oper-

ates. The influence of nurses. The way nurses learn.

The Arista3 report documents common currentand emerging nursing and health care themes, as wellas region-specific challenges, in Africa and the NearEast, Europe, the Pacific Rim, the Americas and theCaribbean. A print version of the complete 68-pagereport may be ordered online for a fee of $25 (mem-bers) or $30 (nonmembers), including shipping andhandling; or by calling (888) 634-7575.

Nursing Home Costs and Quality of Care Outcomes

Documents

Transcript of Nursing Home Costs and Quality of Care Outcomes