Impact of the Affordable Care Act Medicaid ...

YAJEM-56413; No of Pages 6 American Journal of Emergency Medicine xxx (2017) xxx–xxx

Contents lists available at ScienceDirect

American Journal of Emergency Medicine journal homepage: www.elsevier.com/locate/ajem

Impact of the Affordable Care Act Medicaid expansion on emergency department high utilizers with ambulatory care sensitive conditions: A cross-sectional study Daniel B. Gingold, MD, MPH a,⁎, Rachelle Pierre-Mathieu, MD, MPP a, Brandon Cole, MD, MPH a, Andrew C. Miller, MD b, Joneigh S. Khaldun, MD, MPH a,c,d a

Department of Emergency Medicine, University of Maryland, Baltimore, MD, USA Department of Emergency Medicine, West Virginia University, Morgantown, WV, USA Baltimore City Health Department, Baltimore, MD, USA d Department of Emergency Medicine, Henry Ford Hospital, Detroit, MI, USA b c

a r t i c l e

i n f o

Article history: Received 28 October 2016 Received in revised form 31 December 2016 Accepted 11 January 2017 Available online xxxx Keywords: Affordable Care Act Medicaid expansion ED utilization Ambulatory care sensitive conditions

a b s t r a c t Objectives: The effect of the Affordable Care Act on emergency department (ED) high utilizers has not yet been thoroughly studied. We sought to determine the impact of changes in insurance eligibility following the 2014 Medicaid expansion on ED utilization for ambulatory care sensitive conditions (ACSC) by high ED utilizers in an urban safety net hospital. Methods: High utilizers were defined as patients with ≥4 visits in the 6 months before their most recent visit in the study period (July–December before and after Maryland's Medicaid expansion in January 2014). A differences-in-differences approach using logistic regression was used to investigate if differences between high and low utilizer cohorts changed from before and after the expansion. Results: During the study period, 726 (4.1%) out of 17,795 unique patients in 2013 and 380 (2.4%) of 16,458 during the same period in 2014 were high utilizers (p-value b0.001). ACSC-associated visit predicted being a high utilizer in 2013 (OR 1.66 (95% CI [1.37, 2.01])) and 2014 (OR 1.65 (95% CI [1.27, 2.15])) but this was not different between years (OR ratio 0.99, 95% CI [0.72, 1.38], p-value 0.97). Conclusion: Although the proportion of high utilizers decreased significantly after Maryland's Medicaid expansion, ACSC-associated ED visits by high ED utilizers were unaffected. © 2017 Published by Elsevier Inc.

1. Introduction A number of recent health policy changes have aimed to decrease inappropriate emergency department (ED) use [1,2]. One area of concern has been high utilizers of the ED, the so-called “Super Users”, who allegedly contribute to higher hospital costs and insurance premiums owing to uncompensated care, resource utilization, and ED crowding [2,3]. Anecdotally this population is often described as uninsured, minority, and using (or misusing) the ED for minor complaints better served in a primary care or non-acute care setting [2,3]. At first glance, increased primary care access via health insurance expansion may be an attractive policy choice to reduce ED over-utilization [4]. It has been proposed that insurance expansion may decrease ED visits, specifically those by frequent ED utilizers with ambulatory care sensitive conditions (ACSCs). ACSCs are conditions whose appropriate management in a primary care setting would abrogate need for management in an acute care setting such as the ED [5]. Some policy makers ⁎ Corresponding author at: Department of Emergency Medicine, University of Maryland School of Medicine, 614 Wyeth St., Baltimore, MD 21230, USA. E-mail address: [email protected] (D.B. Gingold).

have targeted ED over-utilization for ACSCs as one means to reduce costs [1]. It has been hypothesized that ACSC-associated ED visits from frequent ED utilizers may be more sensitive to the effects of insurance expansion than emergent acute care visits [1]. Despite optimism that insurance expansion could reduce ED visits and therefore cost from both high utilizers and ACSC-related visits in general, emerging research has cast doubt on this notion [4,6-9]. It has been reported that high ED utilizers may actually be adequately insured, and more likely to have significant chronic health conditions requiring more frequent acute ED and chronic primary care management than the general population [2,3,8,10]. Some have suggested that in medically underserved areas, primary care access could become more difficult immediately after Medicaid expansion as an already overextended primary care system receives an influx of new patients [7]. It might also increase overall ED use as recently insured patients with health needs but no established primary care access may turn to the ED for evaluation and treatment [7]. The Affordable Care Act (ACA) was signed into law on March 23, 2010, and its associated Medicaid expansion was enacted in Maryland on January 1, 2014, in addition to subsidized privatize insurance exchanges [11,12]. The effects of the ACA on ED visits by high ED utilizers

http://dx.doi.org/10.1016/j.ajem.2017.01.014 0735-6757/© 2017 Published by Elsevier Inc.

Please cite this article as: Gingold DB, et al, Impact of the Affordable Care Act Medicaid expansion on emergency department high utilizers with ambulatory care sensitive conditi..., American Journal of Emergency Medicine (2017), http://dx.doi.org/10.1016/j.ajem.2017.01.014

2

D.B. Gingold et al. / American Journal of Emergency Medicine xxx (2017) xxx–xxx

with ACSCs remain unknown. This study aims to characterize ED visits by high ED utilizers and those for ACSCs both before and after the 2014 Medicaid expansion at a public safety net hospital in Maryland. We hypothesize that Medicaid and health exchange expansion will not decrease overall ED utilization, visits by high ED utilizers, or visits for by high utilizers associated with ACSCs. 2. Methods 2.1. Study design We conducted a retrospective cross-sectional study in adult patients presenting to the emergency department at Prince George's Hospital Center (PGHC), a public safety-net hospital in Cheverly, Maryland, a suburb of Washington D.C. PGHC is a 300-bed hospital and level 2 trauma center with an annual ED volume of approximately 88,000 visits. All parts of the study were reviewed according to the Strengthening the Reporting of Observational Studies in Epidemiology: ‘STROBE’ Statement [13,14]. Cases occurring between January 1, 2013 and December 31, 2014 were included. The study was approved by the investigative review board at PGHC and the need for patient consent was waived. All patients age ≥ 18 years were included. Visits that were entered in error were excluded, and duplicate medical record numbers (MRNs) were consolidated. Patients were identified as high utilizers if they had ≥4 ED visits in the 6 months immediately preceding the most recent visit for each year [15]. 2.2. Data collection The study institution began using the ED information system Picis EDPulseCheck® (OptumInsight Inc., Eden Prairie, MN), in February 2013. Cases occurring between February 1, 2013 and December 31, 2014 were identified by querying Picis EDPulseCheck®. Cases occurring in January 2013 were identified by query of billing software RTI (Reimbursement Technologies, Inc., Conshohocken, PA). The rollout for ACA Medicaid expansion in Maryland occurred on January 1, 2014. The pre- and post-expansion study periods were July 1–December 31, 2013 and 2014 respectively. This is the 6 months immediately prior to Maryland Medicaid expansion, and then 6–12 months after expansion. The same portion of the calendar year was used to avoid seasonal bias in visits and allow for the initial program enrollment period. Cases were assigned a unique visit number as cross-reference, and then cases were filtered to identify patients with multiple MRNs. Data was then de-identified for the remainder of the analysis. Information on patient demographics, insurance status, clinical information, final ED diagnosis and ED disposition was collected. Race was collected to assess for race-based disparity in health provision, and was self-identified and then consolidated by the investigators. Missing insurance information was categorized as “Other”. Mean ED length-of-stay was determined by the averaging the length of each ED visits during the 6 months preceding and including the patient's most recent visit in the calendar year. Likelihood of admission was determined by the number of visits resulting in admission for each patient in the six months preceding and including their most recent visit in the calendar year divided by total number of visits in the same period. Likelihoods of other dispositions (discharge, observation, left prior to treatment complete (LPTC), expired) were calculated similarly. Diagnosis information was free text and diagnoses indicating an ACSC-related visit were determined by the primary researcher (ICD10 had not yet been implemented), using ACSCs definition similar to Johnson et al. [1] We grouped diabetic complications together, and also identified psychiatric/behavioral/substance related diagnoses. Angina visits included those with diagnosis for chest pain as well as ACS. The likelihood for each patient of having an ACSC-related visit was calculated as the number of ACSC-related visits in the study period divided by the total visits in the same period.

2.3. Data analysis All data analysis was performed using SAS University Edition Studio version 3.4 (SAS Institute Inc., Cary, NC). The two tailed Student's t-test was used to compare continuous variables both between years and between the high and low utilizer groups. A pvalue ≤ 0.05 was considered significant. No power calculation was performed as all eligible patients in study period were included in the convenience sample. p-Value determination incorporated the Satterthwaite approximation to account for unequal variance. Categorical variables were compared using the Pearson χ 2 test. For selected demographic variables with different measures between high and low utilizers as well as ACSC-visit likelihoods, logistic regression was used to identify odds ratios for high vs. low utilizers within each year using the model high utilizer = variable + year + variable ∗ year. The coefficient for the interaction term was used to find the ratio of odds ratios comparing 2013 to 2014. This ratio of odds ratios identifies the effect of each variable on the likelihood the patient is a high utilizer changes between before and after Medicaid expansion. 3. Results There were 17,795 unique adult patients with visits during the study period in 2013, and 16,458 in 2014. Distribution of visits in six months from most recent visit for patients in each year is shown in Fig. 1. Descriptive statistics for the cohort of each year are shown in Table 1. There were 726 (4.1%) high utilizer patients in 2013 and 380 (2.3%) in 2014 (p-value b 0.001). In 2014 a higher proportion of all patients in 2014 identified as Hispanic (8.9% vs. 10%, p-value b 0.001) and lower proportion identified as black (78% vs. 75.9%, p-value b 0.001) compared to 2013. Mean ED length-of-stay increased slightly from 5.5 h to 5.8 h (p-value b0.001). Uninsured and Medicare rates were similar between years, although in 2014 there was a slight increase in the percentage of patients with Medicaid (25.3% vs. 26.4%, p-value 0.02) and decrease in percentage with private insurance (30.6% vs. 26.4%, p-value b 0.001) compared to 2013. A slightly lower percentage of patients were discharged in 2014 (61.6% vs. 59.4%, p-value b 0.001). Following expansion in 2014, the likelihood of patients having an ACSC diagnosis decreased from 15.7% to 14% (p-value b0.001). Tables 2 and 3 compare the descriptive statistics of high and low utilizers in 2013 and 2014. In 2013 before Medicaid expansion, frequent ED utilizers were more likely than others to be black (86.6% vs. 77.6%, p-value b 0.001), insured with Medicaid (51.4% vs. 24.2%, p-value b0.001) or Medicare (12.8% vs. 6.8%, p-value b 0.001), have a longer ED length-of-stay (6.0 vs. 5.5 h, p-value b0.001), be admitted to the hospital (37.1% vs. 31.6%, p-value b 0.001) or psychiatric unit (6.1% vs. 3.5%, p-value b0.001), and have an ACSC diagnosis (22.2% vs. 15.4%, p-value b0.001). They were less likely to be male (40.8% vs. 45.5%, p-value 0.013), Hispanic (4.1% vs. 9.1%, p-value b0.001), privately insured (11% vs. 31.4%, p-value b0.001), uninsured (17.2% vs. 30%, p-value b0.001) or discharged (53.5% vs. 62%, p-value b0.001) (Table 2). There were similar notable differences between cohorts in race, Medicaid, Medicare, private insurance, uninsured rates, and likelihood of admission and ACSC diagnosis in 2014 after Medicaid expansion (Table 3). Table 4 shows odds ratios (OR) associated with being a high utilizer from logistic regression analysis. It compares high utilizers to low utilizers within years, but also compares these ORs as a ratio of odds ratios to identify “differences in differences” between high utilizers and low utilizers before and after Medicaid expansion. The OR of high utilizers having an ACSC diagnosis was 1.66 (95% CI [1.37, 2.01]) in 2013 and 1.65 (95% CI [1.27, 2.15]) in 2014. These ORs do not differ significantly (OR ratio 0.99, 95% CI [0.72, 1.38], p-value 0.97). In 2013 high utilizers were more likely to have diabetes and CHF, and in 2014 high utilizers were more likely to have COPD or asthma. In both years they were more likely to have psychiatric disorders compared to low utilizers,



3

Fig. 1. Frequency of patents with N1 ED visit within 6 months, 2013 and 2014.

but these odds ratios were not significantly different across years (OR ratios not significantly different from 1.0). Again, high utilizers were slightly less likely to be male. There were no statistically significant differences between ORs for predictors of high utilization between 2013 and 2014, except that having private insurance more strongly predicted high utilization in 2014 compared to 2013 (OR 0.4 vs. 0.27, OR ratio 1.47, 95% CI [1.003, 2.16], p-value 0.05). Tables 2 and 3 also show that high utilizers with insurance increased from 11% in 2013 to 12.9% in 2014, while private insurance among low utilizers fell (31.4% to 27.2%).

4. Discussion 4.1. Summary In the immediate aftermath of ACA-associated health insurance expansion, we observed a 7.5% decrease in the number of unique patients visiting the ED in the 6-month study period (Table 1). Total visits between years were essentially the same (51,895 in 2013 and 51,705 in 2014). Other reports have reported an increase in ED visits after

Table 1 Demographics of patient population, before and after Medicaid expansion. Variable Male (%) Median age (95% CI) Race Black (%) White (%) Hispanic (%) Asian (%) Other (%) Mean length of stay (95% CI) Insurance (most recent visit) Medicaid (%) Medicare (%) Private (%) Uninsured (%) Other (%) High utilizers (%) % likelihood of disposition (95% CI) Discharge (95% CI) Obs (95% CI) Admit (95% CI) Psych (95% CI) Floor (95% CI) ICU (95% CI) Transfer (95% CI) OR (95% CI) LPTC (95% CI) Expired (95% CI) % likelihood of ACSC diagnosis (95% CI)

Before expansion 2013 (N = 17,795)

After expansion 2014 (N = 16,458)

p value

8057 (45.3) 40.7 (40.5, 41)

7480 (45.5) 41.1 (40.8, 41.3)

0.75 0.1

13,871 (78) 1482 (8.3) 1584 (8.9) 99 (0.6) 416 (2.3)

12,492 (75.9) 1369 (8.3) 1651 (10.0) 96 (0.6) 405 (2.5)

b0.001 0.97 b0.001 0.74 0.46

5.5 (5.5, 5.6)

5.8 (5.7, 5.9)

b0.001

4500 (25.3) 1259 (7.1) 5447 (30.6) 5240 (29.5) 1349 (7.6) 726 (4.1)

4337 (26.4) 1147 (7.0) 4415 (26.8) 4875 (29.6) 1684 (10.2) 380 (2.3)

0.02 0.7 b0.001 0.72 b0.001 b0.001

61.6 (60.9, 62.3) 0.17 (0.11, 0.23) 31.8 (31.2, 32.5) 3.7 (3.5, 4) 21.4 (20.9, 22) 4.5 (4.2, 4.7) 1.5 (1.3, 1.7) 0.7 (8.1, 0.1) 5.8 (5.5, 6.1) 0.5 (0.4, 0.6)

59.4 (58.7, 60.1) 0.12 (0.07, 0.17) 32.8 (32.1, 33.5) 4.2 (3.9, 4.5) 22.6 (22, 23.2) 4.2 (3.9, 4.5) 1.3 (1.1, 1.4) 0.5 (0.4, 0.6) 7.1 (24.1, 0.2) 0.6 (0.5, 0.7)

b0.001 0.16 0.05 0.01 0.01 0.21 0.06 0.03 b0.001 0.31

15.7 (15.2, 16.2)

14 (13.5, 14.6)

b0.001

Table 2 Demographics of high vs. low utilizers, 2013 before Medicaid expansion.

Variable Male (%) Median age (95% CI) Race Black (%) White (%) Hispanic (%) Asian (%) Other (%) Mean length of stay (95% CI) Insurance (most recent visit) Medicaid (%) Medicare (%) Private (%) Uninsured (%) Other (%) % likelihood of disposition (95% CI) Discharge (95% CI) Obs (95% CI) Admit (95% CI) Psych (95% CI) Floor (95% CI) ICU (95% CI) Transfer (95% CI) OR (95% CI) LPTC (95% CI) Expired (95% CI) Likelihood of ACSC diagnosis (95% CI)

Low utilizers (N = 17,069)

High utilizers (N = 726)

p-Value

7761 (45.5) 40.7 (40.5, 41)

296 (40.8) 41.2 (39.9, 42.4)

0.013 0.49

13,242 (77.6) 1431 (8.4) 1554 (9.1) 97 (0.6) 403 (2.4) 5.5 (5.5, 5.6)

629 (86.6) 51 (7) 30 (4.1) 2 (0.3) 13 (1.8) 6 (5.7, 6.2)

b0.001 0.19 b0.001 0.3 0.32 b0.001

4127 (24.2) 1166 (6.8) 5367 (31.4) 5115 (30) 1294 (7.6)

373 (51.4) 93 (12.8) 80 (11) 125 (17.2) 55 (7.6)

b0.001 b0.001 b0.001 b0.001 0.996

62 (61.3, 62.7) 0.17 (0.11, 0.23) 31.6 (30.9, 32.3) 3.6 (3.4, 3.9) 21.2 (20.6, 21.8) 4.5 (4.2, 4.8) 1.5 (1.3, 1.7) 0.7 (8.2, 0.1) 5.7 (5.4, 6) 0.5 (0.4, 0.6) 15.4 (14.9, 15.9)

53.6 (51.1, 56.1) 0.26 (0.09, 0.43) 37.2 (34.6, 39.7) 6.1 (4.9, 7.4) 26.6 (24.3, 28.9) 3.2 (2.4, 4.1) 0.9 (0.4, 1.5) 0.2 (0, 0.3) 8.4 (14.5, 0.5) 0.1 (0, 0.2) 22.2 (20.2, 24.2)

b0.001 0.33 b0.001 b0.001 b0.001 0.004 0.051 b0.001 b0.001 b0.001 b0.001


4


Table 3 Demographics of high vs. low utilizers, 2014 after Medicaid expansion.

Variable Male (%) Median age (95% CI) Race Black (%) White (%) Hispanic (%) Asian (%) Other (%) Mean length of stay (95% CI) Insurance (most recent visit) Medicaid (%) Medicare (%) Private (%) Uninsured (%) Other (%) % likelihood of disposition (95% CI) Discharge (95% CI) Obs (95% CI) Admit (95% CI) Psych (95% CI) Floor (95% CI) ICU (95% CI) Transfer (95% CI) OR (95% CI) LPTC (95% CI) Expired (95% CI) Likelihood of ACSC diagnosis (95% CI)

Low utilizers (N = 16,078)

High utilizers (N = 380)

p-Value

7325 (45.6) 41.1 (40.8, 41.3)

155 (40.8) 40.3 (38.5, 42.1)

0.065 0.39

12,155 (75.6) 1348 (8.4) 1638 (10.2) 93 (0.6) 15,678 (97.5) 5.8 (5.7, 5.8)

337 (88.7) 21 (5.5) 13 (3.4) 3 (0.8) 375 (98.7) 5.8 (5.5, 6.2)

b0.001 0.05 b0.001 0.59 0.14 0.72

4143 (25.8) 1109 (6.9) 4366 (27.2) 4821 (30) 1639 (10.2)

194 (51.1) 38 (10) 49 (12.9) 54 (14.2) 45 (11.8)

b0.001 0.02 b0.001 b0.001 0.29

59.6 (58.8, 60.3) 0.12 (0.06, 0.17) 32.7 (32, 33.4) 4.2 (3.9, 4.5) 22.4 (21.8, 23.1) 4.2 (3.9, 4.5) 1.3 (1.1, 1.4) 0.5 (7, 0.1) 7.1 (6.7, 7.4) 0.6 (0.5, 0.7) 13.9 (13.4, 14.4)

52.1 (48.6, 55.5) 0.15 (−0.03, 0.32) 37.6 (34, 41.1) 5.8 (4.1, 7.5) 28.1 (24.9, 31.4) 2.8 (1.8, 3.9) 0.7 (0.1, 1.4) 0.1 (0, 0.2) 10.2 (15.8, 0.8) 0.1 (−0.1, 0.2) 20.4 (17.6, 23.1)

b0.001 0.73 0.01 0.07 b0.001 0.01 0.1 b0.001 b0.001 b0.001 b0.001

Medicaid expansion across entire states [4,6,7]. We observed a slight increase in the percentage of patients insured by Medicaid, a decline in private insurance rates, and a no significant change in uninsured rates. National level data comparing 2013 to 2014 showed modest increases in Medicaid and private coverage, and a decrease in uninsured rates [16,17]. Our study was performed in the year immediately after expansion; it is possible future data looking further after Medicaid expansion may show similarly improved trends in uninsured rates at our site.

The percentage of high utilizers within our population was consistent with prior reports [2,10,15,18]. Interestingly, both overall number of unique patient visits and the percentage of high utilizer patients decreased in 2014 following Medicaid expansion. This could be attributable to the ACA or due to more local shifts in visits between other area hospitals. Similar to prior descriptions of high ED utilizers, we found that when compared to low ED utilizers, high utilizers are more likely to be a racial minority, insured by Medicare or Medicaid, higher acuity of complaint, higher severity of diagnosis, and more often require admission [2,3,10,18]. In our population high utilizers were less likely to be male and more likely to have a psychiatric diagnosis than low utilizers, although these differences did not reach statistical significance in 2014 due to a smaller cohort. Psychiatric visits in our population were approximately 30% substance related, 30% for depression/suicidal ideation, and 20% for psychosis symptoms, with the remainder being non-specific behavioral issues (data not shown). Additionally, high utilizers were more likely to visit the ED for an ACSC-related diagnosis compared to low utilizers, although these findings did not change in the first year following Medicaid expansion. Although the overall share of privately insured patients decreased moderately in 2014, the percentage of high utilizer patients with private insurance actually increased (Tables 2, 3 and 4). It is possible that high utilizers in our population had a lot to gain by accessing private insurance on subsidized exchanges, also started in Maryland January 2014, and were more likely to interface with healthcare access through frequent healthcare visits. Multiple visits during insurance exchange enrollment periods may have made it more likely that high utilizers of health care were identified as candidates for private insurance through the new subsidized exchanges. Also, high utilizers with multiple co-morbidities may have been more likely to benefit from the ACA's elimination of pre-existing condition restrictions from private insurers, which was also rolled out in January 2014 [11]. ACSC-related visits overall decreased only slightly in the year following Medicaid expansion. It is possible that six months after Medicaid expansion is not sufficient for outpatient capacity and availability to change appreciably to alter acute care seeking behaviors. High utilizers were more likely than low utilizers to have ACSC-associated visits both before and after Medicaid expansion. They were also more likely to have

Table 4 Single variable logistic regression: Adjusted Odds Ratios of predictors of being a high utilizer. 2013

2014

Year interaction

Variable ACSC visit likelihood Dehydration PNA UTI Diabetes Hypertension CHF Angina COPD Asthma Psych/substance abuse Other variables Male Black Hispanic Medicaid Medicare Private Admit Discharge LPTC

OR

95% CI

1.66 0.78 0.91 1.11 2.47 1.08 2.88 1.18 1.81 1.67 1.8

1.37 0.22 0.24 0.59 1.21 0.47 1.52 0.79 0.7 0.96 1.37

0.83 1.69 0.45 3.31 2 0.27 1.31 0.68 1.62

0.71 1.34 0.3 2.85 1.6 0.21 1.12 0.58 1.22

OR

95% CI

OR ratio

95% CI

2.01 2.79 3.43 2.08 5.04 2.49 5.45 1.77 4.71 2.9 2.37

1.65 0.65 1.9 0.9 1.44 0.66 2.16 0.99 4.88 3.3 1.68

1.27 0.12 0.46 0.33 0.46 0.14 0.81 0.51 1.86 1.69 1.16

2.15 3.49 7.78 2.48 4.51 3.14 5.79 1.93 12.83 6.42 2.43

0.99 0.83 2.1 0.81 0.58 0.61 0.75 0.84 2.69 1.97 0.932

0.72 0.1 0.3 0.25 0.15 0.1 0.23 0.39 0.69 0.83 0.59

1.38 6.81 14.58 2.69 2.24 3.57 2.43 1.84 10.49 4.69 1.48

p value 0.97 0.86 0.45 0.73 0.43 0.59 0.63 0.67 0.67 0.15 0.12

0.96 2.13 0.68 3.85 2.51 0.34 1.54 0.8 2.17

0.82 2.42 0.27 3.01 1.5 0.4 1.26 0.72 1.57

0.67 1.77 0.15 2.45 1.07 0.29 1.01 0.58 1.1

1.01 3.31 0.49 3.69 2.11 0.54 1.57 0.89 2.24

0.997 1.43 0.59 0.91 0.75 1.47 0.96 1.05 0.96

0.77 0.97 0.28 0.7 0.5 1.003 0.73 0.81 0.61

1.29 2.12 1.21 1.17 1.13 2.16 1.26 1.37 1.53

0.98 0.07 0.15 0.45 0.16 0.05 0.77 0.71 0.88



a psychiatric-related primary diagnosis, highlighting a potentially underserved need for effective outpatient behavioral health resources within the community (Table 4). We did not stratify by super-utilizers (N20 ED visits per year) and high utilizers as in Doupe et al. [15] and Doran et al. [18], as absolute numbers of super-utilizers was relatively low (Fig. 1). Previous literature showed that super-utilizer patients are very likely to have behavioral health or substance abuse issues, and have a low likelihood of admission [15,18]. 4.2. Public health policy implications In this study, ACSC-related visits among high utilizers were not decreased in the period immediately following Maryland's Medicaid expansion. It is likely that additional time is needed for usage patterns to adjust to new insurance coverage. In order to decrease ED visits for ACSCs, increased availability of well-coordinated and accessible outpatient care as well as increased numbers of primary care providers is necessary. Simply improved insurance access is likely necessary but not sufficient solution to reduce non-emergent ED utilization and overcrowding, as prior studies have also stated [8,10,18]. This highlights the need for policy-makers to recognize that insurance is a small part of healthcare access, and the importance of developing comprehensive delivery systems, increased primary care access, improved care coordination, and alternative payment models. Interestingly, in January 2014 the Centers for Medicaid and Medicare Services (CMS) granted Maryland a waiver for hospital Medicare payments, creating global budgets for hospitals and financial penalties for spending over these budgets [19]. While the success of this model remains to be seen, it is intended to create significant financial incentives for hospitals and health systems to avoid unnecessary ED visits and hospitalizations by investing in care coordination and robust outpatient medical and social services [19]. 4.3. Future research Further investigations are needed to characterize the long-term effects of insurance expansion on ED utilization patterns. Given fewer uninsured patients, outpatient services may expand to reduce the burden of non-emergent ACSC visits in EDs. Increased access to preventative care services may reduce the number of emergent visits as well, but this time frame is years if not decades in the future. Increased testing of innovative healthcare models that take an integrative approach including social services, welfare, housing, behavioral health, public health, criminal justice, and other systems that impact vulnerable populations will be important to further elimination of healthcare barriers. This study is meant to be descriptive and hypothesis generating with respect to the barriers of reducing high ED utilization in this population. Previous literature defining the direction of utilization research calls for qualitative research to identify barriers to care in the high utilizer cohort, especially while stratifying by disease [3]. Qualitative studies will inform future research as well as community-based interventions to improve access to primary care, reduce avoidable ED utilization, and improve health outcomes. 4.4. Limitations This was a single center study, and changes in ED utilization outside the single center are unable to be accounted for. Large, regional multicenter studies would be ideal to answer this question [3]. Sample size may have limited the power to detect small differences, especially using logistic regression technique, although such small differences may not be significant from a policy perspective. Diagnosis data was from a drop-down menu as well as free-text, not ICD codes, which may result in categorization of diagnosis not consistent with other literature. The time frame from this study was only shortly after the insurance expansion, and it may take longer than six

5

months for the insurance expansion to have an effect on utilization patterns. We were also not able to follow individual patients before and after the utilization, which could perhaps elucidate bigger changes when focusing on patients whose insurance status changed as a result of the policy. This is an appealing line of questioning for future research, but this technique was not possible with this dataset. 5. Conclusion In the year after Medicaid expansion, there was an overall increase in ED utilization at our center, while there was a small but statistically significant reduction in the proportion of ED patients that were high utilizers or visited for ACSC related diagnoses. Insurance expansion did not correlate with a change in Medicaid coverage of high utilizers, although it did correlate with a reduction in the disparity in private insurance coverage between high and low ED utilizers. ACSC-associated ED visits by high ED utilizers remained stable following the state's Medicaid expansion. Contributions Substantial contributions to conception and design, or acquisition or analysis of data, are credited to DBG, RPM, CB, ACM, JK. Manuscript drafting and revision was performed by DBG, RPM, CB, ACM, JK. Disclosure Dr. Khaldun also worked for the Baltimore City Health Department, and currently is Medical Director for City of Detroit Health Department. Any opinions expressed herein are those of the authors and not official policy of Baltimore City or City of Detroit Governments. No other authors have sources of funding or conflicts of interest to disclose. Acknowledgements Marcus Mitchell, for research assistance. Stephanie Gaboda for help with data abstraction from billing software. References [1] Johnson P, Ghildayal N, Ward A, et al. Disparities in potentially avoidable emergency department (ED) care: ED visits for ambulatory care sensitive conditions. Med Care 2012;50(12):1020–8. [2] LaCalle E, Rabin E. Frequent users of emergency departments: the myths, the data, and the policy implications. Ann Emerg Med 2010;56(1):42–8. [3] Pines JM, Asplin BR, Kaji AH, et al. Frequent users of emergency department services: gaps in knowledge and a proposed research agenda. Acad Emerg Med 2011;18(6): e64–9. [4] Taubman SL, Allen HL, Wright BJ, et al. Oregon's health insurance experiment. Science 2014;343(6168):263–8. [5] Purdy S, Griffin T, Salisbury C, et al. Ambulatory care sensitive conditions: terminology and disease coding need to be more specific to aid policy makers and clinicians. Public Health 2009;123(2):169–73. [6] Smulowitz PB, Lipton R, Wharam JF, et al. Emergency department utilization after the implementation of Massachusetts health reform. Ann Emerg Med 2011;58(3): 225–234.e1. [7] Smulowitz PB, O'Malley J, Yang X, et al. Increased use of the emergency department after health care reform in Massachusetts. Ann Emerg Med 2014;64(2): 107–115.e3. [8] Gindi RM, Black LI, Cohen RA. Reasons for emergency room use among U.S. adults aged 18–64. National Health Interview Survey, 2013 and 2014; 2016. [9] Joynt KE, Gawande Aa, Orav EJ, et al. Contribution of preventable acute care spending to total spending for high-cost Medicare patients. JAMA 2013;309(24):2572–8. [10] Hunt KA, Weber EJ, Showstack JA, et al. Characteristics of frequent users of emergency departments. Ann Emerg Med 2006;48(1):1–8. [11] US house of representatives. Patient Protection and Affordable Care Act; 2010. p. 1–955. [12] Maryland General Assembly. Maryland Health Progress Act of 2013; 2013 1–20. [13] von Elm E, Von EE, Altman DG, et al. The strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. Ann Intern Med 2007;147(8):573–8.


6


[14] Vandenbroucke JP, von Elm E, Altman DG, et al. Strengthening the reporting of observational studies in epidemiology (STROBE): explanation and elaboration. PLoS Med 2007;4(10):1628–54. [15] Doupe MB, Palatnick W, Day S, et al. Frequent users of emergency departments: developing standard definitions and defining prominent risk factors. Ann Emerg Med 2012;60(1):24–32. [16] Smith JC, Medalia C, US Census Bureau. Health insurance coverage in the United States: 2014. Washington, DC. https://www.census.gov/content/dam/Census/library/publications/2015/demo/p60-253.pdf; 2015. [Accessed April 11, 2016].

[17] Sommers B, Gunja M, Finegold K, et al. Changes in self-reported insurance coverage, access to care, and health under the affordable care act. JAMA 2015;314(4):366–74. http://dx.doi.org/10.1001/jama.2015.8421. [18] Doran KM, Raven MC, Rosenheck R. a. What drives frequent emergency department use in an integrated health system? National data from the Veterans Health Administration. Ann Emerg Med 2013;62(2):151–9. [19] Rajkumar R, Patel A, Murphy K, et al. Maryland's all-payer approach to deliverysystem reform. N Engl J Med 2014;370(6):493–5.
