The Effects of Global Budgeting on Emergency Department Admission - - PowerPoint PPT Presentation

1

The Effects of Global Budgeting on Emergency Department Admission Rates

Jessica E. Galarraga, MD, MPH

Department of Health Services Research, MedStar Health Research Institute

@JGalarragaMD

2

Acknowledgements

• Key Collaborators:
• Bernard Black, PhD
• Laura Pimentel, MD
• Arvind Venkat, MD
• William J. Frohna, MD
• John P. Sverha, MD
• Daniel L. Lemkin, MD
• Jesse M. Pines, MD, MBA, MSCE

3

Financial Disclosures

• No financial conflicts of interest
• Grant Funding
• Emergency Medicine Foundation Health Policy Research Scholar Award
• MedStar Health Research Institute New Investigator Award

4

Maryland’s All-Payer Global Budget Revenue Model

• In 2014, Maryland launched a statewide reform that replaced fee-for-

service hospital payments with a population-based payment model – a “global budget revenue” (GBR) model.1-4

• Hospitals have a fixed revenue target, independent of patient volume
• r services provided, tied to all-payer quality metrics.1,5
• Readmission Reduction Inventive Program
• Maryland Hospital Acquired Conditions Program
• Potentially Avoidable Utilization Savings Policy
• We examined the effects of GBR on emergency department (ED)

admission rates.

5

Methods – Study Population

• Retrospective study of all-payer medical record and billing data
• January 1, 2012 to December 31, 2015
• Hospital-based adult ED encounters from Maryland and other Medicaid Expansion states

– 10 Maryland GBR sites – 5 Maryland Total Patient Revenue (TPR) sites

• Exclusions:

– Elopement, leaving AMA, left without being seen, expiration, and psychiatric encounters

• Final Study Sample:
• 3,175,210 ED encounters:

– 1,397,560 GBR, 1,471,331 Non-Maryland matches, and 306,319 TPR

• Mean Unadjusted ED Hospitalization Rate: 20.9% (95% CI: 17.8, 24.0)

– 10 Non-Maryland matched sites

6

Methods – Study Variables

• ED hospitalization rate: # hospital stays (inpatient and observation) / # ED visits
• GBR adoption: start date of hospital’s GBR contract to account for 6-month

transition period from January-July 2014

• Subgroup Analysis by:
• Ambulatory Care Sensitive Conditions (ACSCs)

– Defined by 11 out of 13 AHRQ Prevention Quality Indicators6

• Excludes low birth weight (not relevant for adults) and lower extremity

amputations among diabetics (relates to hospital inpatient course)

• Primary Clinical Diagnosis

– Based on the ICD-9/10-CM diagnosis categorized using multi-level clinical classification software (MCCS) codes

7

Methods - Analysis

• Difference-in-differences analysis using hospital-fixed effect regression
• Leads and Lags analysis to examine temporal trends
• Regressions included hospital and year fixed effects, clustering of standard errors
• n hospital-level, and adjustment for:
• Patient Factors:

Age, Sex, Insurance Status

• Encounter Factors:

Primary Diagnosis (Clinical Classification Software Category)7,8

• Hospital Factors:

Annual ED volume, Trauma Level, Number of Hospital Beds, Teaching Status

• Community Factors: Per Capita Income, Primary Care Provider to Population Ratio,

Metropolitan Residency

8

Figure 1. Monthly levels of risk-adjusted ED hospitalization rates relative to the global budget revenue (GBR) model implementation date in 2014 (month 0). Vertical lines show 95% CIs, with standard errors clustered on the hospital level.

9

Table 1. Difference-in-differences (DiD) analysis of ED admission and transfer rates in GBR versus Non- Maryland and TPR hospitals.

ED Hospitalization Rate PreGBR (2012-2013) PostGBR (2014-2015) Absolute Rate Difference [95% CI] Regression-based DiD [95% CI] GBR vs. Non-MD Matches † Admission from ED GBR 20.1% 19.3%

• 0.8% (-1.03, -0.58)**
• 0.5% (-0.72, -0.39)**

Non-MD Matches 21.3% 21.0%

• 0.3% (-0.47,-0.03)*

Transfer from ED to Another Hospital GBR 1.6% 1.6%

• 0.0% ( -0.06, 0.06)

+0.1% (0.08, 0.17)** Non-MD Matches 0.4% 0.3%

• 0.1% (-0.19,-0.07)**

GBR vs. TPR‡ Admission from ED GBR 20.1% 19.3%

• 0.8% (-1.03, -0.58)**
• 1.9% (-2.27, -1.80)**

TPR 16.1% 17.2% +1.1% (0.83, 1.46)**

Linear probability model. Dependent variable is ED admission (0-1), including inpatient and observation stays. Predictor variable for regressions is GBR dummy * post dummy (=1 for 2014-2015). Regressions include hospital and year fixed effects, primary clinical condition, and patient, hospital, and community factors. Standard errors are clustered on the hospital level.

† GBR hospitals are hospitals subject to Maryland global budget revenue (GBR) program starting in 2014. Non-MD matches are

matched hospitals in Medicaid expansion states other than Maryland.

‡ TPR hospitals are five rural Maryland hospitals included in total patient revenue (TPR) pilot program starting 2010. * Hospitalization rate difference with p-value < .05, ** Hospitalization rate difference with p-value < .0001

10

Title Arial Bold – 34pt font

Subtitle Arial – 25pt font

Presenter’s Name Office or Department Name

Table 2. Difference-in-differences (DiD) analysis of ED admissions for ambulatory-case-sensitive conditions (ACSC) versus other conditions, in GBR versus Non-Maryland and TPR hospitals.

ED Hospitalization Rate PreGBR (2012-2013) PostGBR (2014-2015) Absolute Rate Difference [95% CI] Regression-based DiD [95% CI] GBR vs. Non-MD Matches† ACSC GBR 36.4% 35.1%

• 1.3% (-2.45, -0.06)**
• 1.1% (-2.13, -0.57)**

Non-MD Matches 49.1% 48.9%

• 0.2% (-1.36, 1.05)

Non-ACSC GBR 19.4% 18.6%

• 0.8% (-1.06,-0.61)*
• 0.3% (-0.39,-0.05)*
• 0.5% (-0.8, -0.4)*

Non-MD Matches 20.3% 20.0% GBR vs. TPR‡ ACSC GBR 36.4% 35.1%

• 1.2% (-2.45, -0.06)**

+1.0% (-0.95,1.71) TPR 38.6% 36.4%

• 2.2% ( -3.69,-0.76)**

Non-ACSC GBR 19.4% 18.6%

• 0.8% (-1.06,-0.61)*
• 2.1% (-2.57, -1.99)*

TPR 15.1% 16.4% +1.3% (1.02, 1.65 )*

Linear probability model. Dependent variable is ED admission (0-1), including inpatient and observation stays. Predictor variable for regressions is GBR dummy * post dummy (=1 for 2014-2015). Regressions include hospital and year fixed effects, primary clinical condition, and patient, hospital, and community factors. Standard errors are clustered on the hospital level.

† GBR hospitals are hospitals subject to Maryland global budget revenue (GBR) program starting in 2014. Non-MD matches are

matched hospitals in Medicaid expansion states other than Maryland.

‡ TPR hospitals are five rural Maryland hospitals included in total patient revenue (TPR) pilot program starting 2010. * Hospitalization rate difference with p-value < .05, ** Hospitalization rate difference with p-value < .0001

11

Title Arial Bold – 34pt font

Subtitle Arial – 25pt font

Presenter’s Name Office or Department Name

Table 3. Difference-in-differences (DiD) analysis of ED hospitalizations in GBR versus Non-Maryland control hospitals by primary diagnosis, 2012-2015.

ED Hospitalization rate Regression-based DiD [95% CI] MCCS Category Name (MCCS Code)‡ PreGBR (2012-2013) PostGBR (2014-2015) Absolute Rate Difference [95% CI] Diseases of the heart (7.2) GBR 41.0% 39.8%

• 1.2% (-2.3,0.0)**

+0.7% (-1.5, 0.2) p = .1106 Non-MD Matches 44.7% 42.8%

• 1.9% (-3.0,0.7)**

Symptoms, signs, and ill-defined conditions (17.1) GBR 15.1% 15.4% +0.3% (-2.1,0.8 ) +0.5% (-0.07, 9.7) p =.0894 Non-MD Matches 11.7% 11.5%

• 0.2% (-1.8,0.4)

Lower respiratory disease (8.8)*** GBR 28.7% 28.1%

• 0.6% (-2.4,-1.2) **
• 1.45% (-2.8 -0.1)

p =.0309 Non-MD Matches 21.9% 22.8% +0.8% (-1.0,2.7) COPD and bronchiectasis (8.2) GBR 40.2% 40.6% +0.4% (-2.3,3.1) +0.9% (-1.2, 3.1) p = .3877 Non-MD Matches 72.3% 71.8%

• 0.5% (-3.3,2.2)

Cerebrovascular disease (7.3) GBR 78.7% 79.0% +0.3% (-2.4,3.1)

• 1.1% (-3.4, 1.1)

p =.3205 Non-MD Matches 83.0% 84.4% +1.4% (-1.4,4.3)

We show results for the top 10 most common clinical conditions among ED admissions. GBR includes encounters exposed to global budget revenue (GBR) implementation in Maryland hospitals. Non-MD matches includes encounters in matched hospitals located in Medicaid expansion states external to Maryland, not exposed to GBR.

*Hospitalization rate difference with p < .0001, **Hospitalization rate difference with p < .05, ***DiD result statistically significant, p < .05. ‡MCCS = Multi-level clinical classification software category.

12

Title Arial Bold – 34pt font

Subtitle Arial – 25pt font

Presenter’s Name Office or Department Name

Table 3 (continued). Difference-in-differences (DiD) analysis of ED hospitalizations in GBR versus Non-Maryland control hospitals by primary diagnosis, 2012-2015.

ED Hospitalization Rate Regression-based DiD [95% CI] MCCS Category Name (MCCS Code)‡ PreGBR (2012-2013) PostGBR (2014-2015) Absolute Rate Difference [95% CI] Fractures (16.2) GBR 27.3% 26.9%

• 0.4% (-2.5,1.5)
• 0.3% (-1.9, 1.3)

p =.7262 Non-MD Matches 36.9% 36.7%

• 0.2% (-1.3,2.3)

Fluid and electrolyte disorders (3.8) *** GBR 51.9% 53.3% +1.4% (-1.5,4.4) +4.5% (2.1, 6.9) p =.003 Non-MD Matches 39.5% 36.4%

• 3.1% (-6.0,-0.1)**

Lower gastrointestinal disorders (9.6) GBR 57.1% 56.4%

• 0.6% (-4.4,3.1)
• 0.9% (-2.0, 3.8)

p = .0.5501 Non-MD Matches 32.8% 31.3%

• 1.5% (-5.3,2.3)

Diseases of the urinary system (10.1) GBR 16.7% 15.5%

• 1.2% (-3.4,1.0)

+0.8% (-0.8, 2.4) p = .3337 Non-MD Matches 24.6% 22.6%

• 2.0% (-4.3,-0.3)**

Diabetes mellitus with complications (3.3) *** GBR 48.3% 43.1%

• 5.2% (-9.5,1.0) **
• 6.3% (-10.1, -2.6)

p =.0009 Non-MD Matches 66.9% 68.1% +1.1% (-3.4,5.7)

We show results for the top 10 most common clinical conditions among ED admissions. GBR includes encounters exposed to global budget revenue (GBR) implementation in Maryland hospitals. Non-MD matches includes encounters in matched hospitals located in Medicaid expansion states external to Maryland, not exposed to GBR.

*Hospitalization rate difference with p < .0001, **Hospitalization rate difference with p < .05, ***DiD result statistically significant, p < .05. ‡MCCS = Multi-level clinical classification software category.

13

Conclusions

• Implementation of the GBR payment structure is

associated with significant, although moderate, declines in ED admission rates.

• Admission rate decline more pronounced among

encounters for:

• Ambulatory Care Sensitive Conditions
• Lower Respiratory Disease
• Diabetes

14

Implications for Policy

• Helps meet health care spending goals of GBR:
• Medicare hospital spending $538 million in its first 3 years.9
• Future research needed to examine impact on quality

beyond its CMS waiver metrics and longer-term effects of GBR on health care delivery.

• If GBR continues to curtail health care costs, without

unintended compromises in quality, then Maryland’s reform may become a model for improving health care delivery in other states.

15

References

• 1. The Maryland Health Services Cost Review Commission. Maryland All-Payer Model Agreement.

http://www.hscrc.state.md.us/documents/md-maphs/stkh/MD-All-Payer-Model-Agreement-(executed).pdf Accessed May 18, 2018.

• 2. Rajkumar R, Patel A, Murphy K, et al. Maryland's all-payer approach to delivery-system reform. N Engl J Med 2014

Feb 6;370(6):493-5.

• 3. Coyle C. Maryland’s Progress On The Path to The Triple Aim. http://healthaffairs.org/blog/2015/11/12/marylands-

progress-on-the-path-to-the-triple-aim/. Accessed May 27, 2018.

• 4. Robeznieks A. Global budgets pushing Maryland hospitals to target population health. Modern Healthcare.

http://www.modernhealthcare.com/article/20141206/MAGAZINE/312069983. Accessed May 27, 2018.

• 5. Cragun E, Martucci K, Su PN. Maryland’s All-Payer Global Budget Cap Model and Its Implications for Providers: The

Advisory Board Company; 2016. https://www.advisory.com/health-policy/resources/2016/maryland-all-payer-model- and-implications-for-providers. Accessed May 18, 2018.

• 6. Agency for Healthcare Research and Quality. Prevention Quality Indicators. Rockville, MD. June 2018. Available at:

http://qualityindicators.ahrq.gov/Modules/pqi_resources.aspx. Accessed May 25, 2018.

• 7. Healthcare Cost and Utilization Project. Clinical Classifications Software (CCS) for ICD-9-CM. https://www.hcup-

us.ahrq.gov/toolssoftware/ccs/ccs.jsp. Accessed May 27, 2017.

• 8. Healthcare Cost and Utilization Project. Clinical Classifications Software (CCS) for ICD-10-CM/PCS. https://www.hcup-

us.ahrq.gov/toolssoftware/ccs10/ccs10.jsp. Accessed May 27, 2018.

• 9. Maryland Hospital Association. Tracking Our All-Payer Experiment: Waiver Dashboard.

http://www.mhaonline.org/transforming-health-care/tracking-our-all-payer-experiment/waiver-dashboard. Accessed May 27, 2018.