Changes in the quality of post-acute care after the implementation of public reporting Rachel M. Werner R. Tamara Konetzka Elizabeth Stuart Edward Norton.

Changes in the quality of post-acute Changes in the quality of post-acute care after the implementation of care after the implementation of

public reportingpublic reporting

Rachel M. WernerRachel M. WernerR. Tamara KonetzkaR. Tamara Konetzka

Elizabeth StuartElizabeth StuartEdward NortonEdward Norton

Jeongyoung ParkJeongyoung Park

June 2008June 2008Funding: AHRQ (R01 HS016478-01)Funding: AHRQ (R01 HS016478-01)

Public Reporting and Quality Public Reporting and Quality ImprovementImprovement

History of problems with nursing home History of problems with nursing home quality, in part due to absence of typical quality, in part due to absence of typical market attributesmarket attributes– Difficult for consumers to judge qualityDifficult for consumers to judge quality– Little incentive for providers to compete on Little incentive for providers to compete on

qualityquality Public reporting of quality is intended to Public reporting of quality is intended to

improve quality by:improve quality by:– Giving consumers information needed to shop Giving consumers information needed to shop

on qualityon quality– Giving providers incentive to compete on qualityGiving providers incentive to compete on quality– Giving providers information and targets for QIGiving providers information and targets for QI

High-quality SNFs

Mid quality SNFs

Low quality SNFs

Patients

High-quality SNFs

Mid quality SNFs

Low quality SNFs

Patients

A. No Report Cards B. With Report Cards

Figure 3. Conceptual Model

ObjectiveObjective

Examine the effect of publicly Examine the effect of publicly reporting quality information on post-reporting quality information on post-acute care quality.acute care quality.

Assess the extent to which changes Assess the extent to which changes in quality may be consumer-driven in quality may be consumer-driven vs. provider-driven.vs. provider-driven.– Changes in average resident-level Changes in average resident-level

quality within market quality within market – Changes in average within-SNF quality Changes in average within-SNF quality

ContributionContribution

Others have found modest Others have found modest improvement in reported quality of improvement in reported quality of nursing home carenursing home care

We improve upon the existing We improve upon the existing literature by:literature by:– Better control for patient selectionBetter control for patient selection– Control for secular trendsControl for secular trends– Assess changes in overall qualityAssess changes in overall quality

Setting: Nursing Home Setting: Nursing Home CompareCompare

Launched November 12, 2002Launched November 12, 2002 Publicly release quality information:Publicly release quality information:

http://http://www.medicare.gov/NHComparewww.medicare.gov/NHCompare All Medicare- and Medicaid-certified All Medicare- and Medicaid-certified

NHsNHs– 17,000 nursing homes17,000 nursing homes

10 quality measures10 quality measures– 3 post-acute care3 post-acute care– 6 chronic care6 chronic care

Staffing, inspections Staffing, inspections

DataData Minimum Data Set (1999-2005)Minimum Data Set (1999-2005)

– All Medicare- and Medicaid-certified nursing homesAll Medicare- and Medicaid-certified nursing homes– Detailed clinical dataDetailed clinical data– Source to calculate quality measures for Nursing Source to calculate quality measures for Nursing

Home CompareHome Compare– Used to calculate quality measures over study Used to calculate quality measures over study

periodperiod MedPARMedPAR

– Claims data on all non-managed-care Medicare Claims data on all non-managed-care Medicare beneficiariesbeneficiaries

– Used to calculate rehospitalizations and several Used to calculate rehospitalizations and several health risk variableshealth risk variables

Empirical approach 1Empirical approach 11A. Within market: Quality1A. Within market: Qualityitit= β= β11NHCNHCt t + + βXβXjtjt

1B1B. . Within SNF: QualityWithin SNF: Qualityjtjt= β= β11NHCNHCt t + + βXβXjtjt + + jj

– QualityQualityj/jt j/jt = quality for individual i or SNF j in year t= quality for individual i or SNF j in year t

– NHCNHCt t = indicator of Nursing Home Compare= indicator of Nursing Home Compare

▪ ▪ pre-post (2000-2002 vs. 2003-2005)pre-post (2000-2002 vs. 2003-2005)

▪ ▪ set of set of year dummy variablesyear dummy variables

– XXj/itj/it = set of control variables = set of control variables

jj = SNF fixed effects = SNF fixed effects

Empirical Approach 2Empirical Approach 2

Repeat both analyses but add control Repeat both analyses but add control groupgroup

Small SNFs (roughly 30%) were excluded Small SNFs (roughly 30%) were excluded from Nursing Home Compare.from Nursing Home Compare.– They are different from large SNFsThey are different from large SNFs– But estimates not biased as long as secular But estimates not biased as long as secular

trends over time are sametrends over time are same Estimate difference-in-differences model Estimate difference-in-differences model

which uses trend in small SNFs as measure which uses trend in small SNFs as measure of secular trend in large SNFs. of secular trend in large SNFs.

Controlling for SelectionControlling for Selection

Challenge: Potential bias. Challenge: Potential bias. – Case-mix may be different before and after NHCCase-mix may be different before and after NHC– Differences may be due to provider selectionDifferences may be due to provider selection

Solution: Matched cohorts of SNF residents Solution: Matched cohorts of SNF residents pre- and post-NHC using propensity scores.pre- and post-NHC using propensity scores.– Avoids dependence on correctly specifying Avoids dependence on correctly specifying

functional form of case-mix controls.functional form of case-mix controls.– Corrects for unobserved case-mix if correlated Corrects for unobserved case-mix if correlated

with observed case-mix.with observed case-mix. Matching done within-SNF and within-Matching done within-SNF and within-

market (Dartmouth Atlas definition of market (Dartmouth Atlas definition of health care service area)health care service area)

Reported quality measuresReported quality measures

Technical definitions of measures Technical definitions of measures from CMSfrom CMS

Follow CMS conventionsFollow CMS conventions– 2 quarters2 quarters– 14-day assessment14-day assessment– Facilities with more than 20 cases Facilities with more than 20 cases

during target periodduring target period 8,137 SNFs8,137 SNFs

30-day Preventable 30-day Preventable RehospitalizationsRehospitalizations

UnreportedUnreported Good measure of Good measure of

overall SNF qualityoverall SNF quality Based on Based on allall SNF SNF

admissions, not just admissions, not just those with 14-day those with 14-day assessmentsassessments

Used AHRQ prevention Used AHRQ prevention QIs that were QIs that were applicable to 65+applicable to 65+– Bacterial pneumoniaBacterial pneumonia– COPDCOPD– DehydrationDehydration– Heart failureHeart failure– HypertensionHypertension– Short-term diabetic Short-term diabetic

complicationscomplications– Uncontrolled diabetesUncontrolled diabetes– UTIUTI

Quality MeasuresQuality Measures

Measures Description

No pain% of residents who did not have moderate or severe pain

No delirium % of residents without delirium

Improved walking % of residents whose walking improved

Rehospitalization% of residents who had a preventable rehospitalization within 30-days of admission to post-acute care*

6.9 (10.5)

7.0 (5.9)

*A preventable rehospitalization is defined as a hospitalization for chronic obstructive pulmonary disease, dehydration, diabetes complications, heart failure, hypertension, bacterial pneumonia, uncontrolled diabetes and urinary infection

Table 1. Summary Post-acute Care Performance Measures

Mean (SD)

76.3 (19.3)

96.4 (7.5)

Control variablesControl variables

All variables used in the propensity All variables used in the propensity scores to adjust for remaining small scores to adjust for remaining small differencedifference

Prior residential history for delirium Prior residential history for delirium (as specified by CMS for this (as specified by CMS for this measure)measure)

Previously developed risk adjustors Previously developed risk adjustors for preventable hospitalizationsfor preventable hospitalizations

Risk-adjusted trends in post-Risk-adjusted trends in post-acute care qualityacute care quality

No pain

0.71

0.72

0.73

0.74

0.75

0.76

0.77

0.78

1999 2000 2001 2002 2003 2004 2005

Risk-adjusted trends in post-Risk-adjusted trends in post-acute care qualityacute care quality

No delirium

0.94

0.945

0.95

0.955

0.96

0.965

0.97

1999 2000 2001 2002 2003 2004 2005

Risk-adjusted trends in post-Risk-adjusted trends in post-acute care qualityacute care quality

Improved walking

0.062

0.064

0.066

0.068

0.07

0.072

0.074

0.076

0.078

0.08

0.082

1999 2000 2001 2002 2003 2004 2005

Risk-adjusted trends in post-Risk-adjusted trends in post-acute care qualityacute care quality

Preventable rehospitalizations

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

1999 2000 2001 2002 2003 2004 2005

Results 1: Within-market Results 1: Within-market changeschanges

No pain No delirium

Improved walking

Rehosp

Change at implementation of NHC (between 2002 and 2003)

0.0192***

0.0009

0.0013*

-0.0022***

Change between pre-NHC (1999-2002) and post-NHC (2003-2005)

0.0301***

0.0062***

0.0037***

-0.0106***

Results 2Results 2

Within-SNF results similarWithin-SNF results similar– Most of the observed quality improvements Most of the observed quality improvements

attributable to provider-driven changes as attributable to provider-driven changes as opposed to consumers choosing high-quality opposed to consumers choosing high-quality facilitiesfacilities

– Some inconsistent results for rehospitalizationSome inconsistent results for rehospitalization Using small SNFs as a controlUsing small SNFs as a control

– Magnitude of improvement in pain decreasedMagnitude of improvement in pain decreased– Magnitude of improvement in walking increasedMagnitude of improvement in walking increased– Rehospitalization rates worsenedRehospitalization rates worsened

SummarySummary

Measured post-acute care quality Measured post-acute care quality improved after NHCimproved after NHC– Statistically significant but small changes (4% Statistically significant but small changes (4%

pain; 6% walking)pain; 6% walking) Results for unmeasured overall quality – Results for unmeasured overall quality –

preventable rehospitalizations – were preventable rehospitalizations – were inconsistent and less promising.inconsistent and less promising.

Most of the effect is attributable to within-Most of the effect is attributable to within-SNF changes, suggesting that changes in SNF changes, suggesting that changes in market share played a negligible rolemarket share played a negligible role

Implications and Next StepsImplications and Next Steps

Public reporting can play a positive Public reporting can play a positive but – so far – limited role in but – so far – limited role in improving quality of post-acute care.improving quality of post-acute care.

Will explicitly examine:Will explicitly examine:– Changes in market shareChanges in market share– Role of market competition and facility Role of market competition and facility

attributesattributes– Selective dischargeSelective discharge– Selective admissionsSelective admissions