Trends in Emergency Department Use by Medicaid Expansion Status - - PowerPoint PPT Presentation

Trends in Emergency Department Use by Medicaid Expansion Status

Katherine Hempstead, RWJF Joel Cantor, Rutgers University Ewan Rankin, Princeton University

Background

• Some studies have found increase in ED use after Medicaid expansion

(OR)

• Concern about utilization increases can undermine support for

program

How does Medicaid expansion affect ED utilization?

• Payer Mix factor
• Ignore whether change in coverage status affects utilization
• Size of payer mix factor is a function of:
• Size of expansion
• “Enrollment dynamics”
• Compositional factors
• Payer mix change related to enrollment change and these other factors
• Local and state variation
• No implications for Total ED volume

How does Medicaid expansion affect ED utilization?

• Utilization factor
• Change in coverage status DOES affect individual utilization
• Possible pathways:

+ Coverage effect – increase utilization + “Pent up demand” – increase utilization in SR

• Care management – substitute ambulatory care

+ Network/Access/Delivery System issues

• Program features such as co-pays

Net effect could be + or - Implications for total ED volume

Measuring the impact of expansion on ED utilization

• Payer mix effect certain
• Utilization effect uncertain

Measures

Payer mix: Change in Medicaid share of ED visits Change in Medicaid ED visits Utilization effect: Change in total ED volume

Prior studies

• Hempstead and Cantor (2016) Payor mix, no total increase
• Pines (2016) Payer mix, no total increase
• Nikpay et al (2017) ** Both!!

HAMP data

• Quarterly data from @ 25 state hospital associations
• Volume by payer – ED and IP
• Several conditions sensitive to coverage
• AHRQ PQIs, non-emergency ED visits, TKR
• Adding behavioral health

Participating states

Participating states*

Expansion Non expansion CO FL CT GA IA KS IL MD IN MO KY NE MN SC MT TN NJ VA NV WI NY WY OH *TX, LA, MI excluded

Measurement of expansion

• Coverage variable - accounts for timing and size of expansion
• Used KFF data on Medicaid eligibility relative to FPL
• Created difference and lagged difference in coverage
• Distinguishes between large and small expansions
• Distinguishes between early and late expanders

Payer Mix effect: Panel regression results

Dependent var: Change in Medicaid Share of Emerg Dept Coefficient Standard Error* Coverage level change: 𝛾_current year change 0.047 0.010 𝛾_previous year change 0.021 0.009 Time fixed effects: 𝛾_2015

• 1.35

0.95 𝛾_2016

• 1.27

0.76 * Clustered at state level Dependent var: Change in Medicaid Share of Inpatient Coefficient Standard Error* Coverage level change: 𝛾_current year change 0.023 0.002 𝛾_previous year change 0.013 0.006 Time fixed effects: 𝛾_2015

• 0.52

0.65 𝛾_2016

• 0.31

0.44 * Clustered at state level

Alternative specification: Percent change in Medicaid volume

Dependent variable: Annual percentage point change in Medicaid Emerg Dept Volumes Coefficient Standard Error* Coverage level change: 𝛾_current year change 0.25 0.07 𝛾_previous year change 0.07 0.05 Time fixed effects: 𝛾_2015

• 2.6

7.4 𝛾_2016

• 6.5

5.1 * Clustered at state level Dependent variable: Annual percentage point change in Medicaid In-Patient Volumes Coefficient Standard Error* Coverage level change: 𝛾_current year change 0.11 0.03 𝛾_previous year change 0.05 0.04 Time fixed effects: 𝛾_2015

• 2.34

5.5 𝛾_2016

• 3.49

3.5 * Clustered at state level

Interpretation

• Significant payer mix effect from expansion
• Lagged measure of expansion is smaller
• Suggests that effect of expansion declines over time

Test for utilization effect DV: Percent change in total visits

Dependent var: Percent change in total ED visits Coefficient Standard Error* Coverage level change: 𝛾_current year change 0.007 0.017 𝛾_previous year change

• .005

0.03 Time fixed effects: 𝛾_2015 1.63 4.89 𝛾_2016

• 1.24

2.31 * Clustered at state level Dependent var: Percent change in IP visits Coefficient Standard Error* Coverage level change: 𝛾_current year change

• .00016

0.0001 𝛾_previous year change 0.0000143 0.00015 Time fixed effects: 𝛾_2015

• .0005

0.025 𝛾_2016

• .009

0.014 * Clustered at state level

Interpretation

No significant effect of expansion on percent change in total ED visits or inpatient admissions No evidence of a “utilization effect” from expansion *Also: robust to a few other specifications

Findings thus far

• Finds significant “payer mix” effect, declines over time
• No support for overall “utilization” effect
• Advantages: timely, more sensitive measure of expansion
• Limitations:
• Subset of states, may not be representative
• Lacks controls, empirical work not complete
• PRELIMINARY