Abstract
The annual number of emergency department (ED) visits in the United States increased 23% between 1997 and 2007. 1 The uninsured are more likely to lack a usual source of care, require avoidable hospitalizations, and use the ED for ambulatory sensitive conditions.2-4 Though many uninsured patients are able to access primary care through the federal safety net, these patients often lack access to care coordination and specialty care, including diagnostics and specialty providers. 5
Studies have failed to show that access to primary care alone decreases utilization of the ED and in fact, patients with a usual source of care may be driving the recent increase in ED visits.6-8 Project Access, using a model including provision of free care to uninsured individuals lacking a primary care provider, was able to demonstrate a reduction in ED utilization at several program sites, 9 leading to the hypothesis that access to primary plus specialty care may contribute to reduction in ED utilization among uninsured patients. We sought to evaluate whether a program that targets uninsured patients with a primary care provider and facilitates primary care engagement as well as access to specialty care services and care coordination would decrease ED utilization among program enrollees relative to nonenrollees.
The Access Partnership (TAP) program is a novel collaboration between primary and specialty care clinicians at an academic medical center and East Baltimore Medical Center (EBMC), a primary care center that serves a low-income population with a high rate of uninsured individuals. 10 TAP facilitates access to specialty services and care coordination for uninsured patients with a primary care provider at EBMC. TAP also promotes engagement in primary care by promoting communication between primary care providers and specialists and timely follow-up visits.
We hypothesized that enrollment in TAP would improve access to outpatient care and might therefore result in decreased ED utilization among this population, compared with those referred to the program who chose not to enroll. We further hypothesized that although rates of inpatient admission might not change significantly, more comprehensive outpatient primary and specialty care might result in fewer ED visits that did not result in admission.
Methods
Study Design
We conducted a quasi-experimental multiple time series analysis using difference-in-difference estimation to determine the impact of TAP on monthly ED and inpatient hospitalization rates. Approval for this study and HIPAA (Health Insurance Portability and Accountability Act) waiver were granted by the School of Medicine Institutional Review Board. We retrospectively examined administrative data on ED utilization and inpatient hospitalization at the academic medical center and affiliated community hospital during 48 months from May 2007 through April 2011.
Intervention
Beginning in May 2009, internal medicine providers at EBMC referred uninsured and underinsured patients in their practice from 5 neighboring zip codes who needed specialty care to the program. Uninsured patients were those who were seen in the practice on a sliding scale fee. The underinsured were patients seen in the practice who were covered through the state Primary Adult Care program. No patients referred to the program had insurance to cover specialty care, diagnostics, or hospitalizations.
Referrals to TAP were evaluated for medical necessity. If approved, a TAP navigator invited the patient into the program and explained the program to the patient, the importance of each referral. The navigator, a certified medical assistant at the clinic, then collected a small fee of $10 to $20, scheduled the patient’s appointment, helped the patient negotiate transportation, and followed up to remind the patient of the upcoming appointment. The fee was waived in cases of reported significant financial hardship. TAP-covered services included all specialty care available at the academic medical center. Subsequent visits, diagnostics, and hospitalizations stemming from the referral were covered by the program as well. 10 The hospital faculty practice funded physician services for the TAP program, and physicians were not aware of which patients were covered via TAP.
The Access Partnership navigators also facilitated engagement in primary care. The navigator communicated with the primary care physician about the status of the referral and arranged timely follow-up with the primary care physician after the specialty visit.
Study Population
The study population included patients eligible to join TAP during the first 21 months of the program, May 2009 through January 2011. Patients were categorized as TAP patients (the intervention group) if they joined the program by paying the fee or requesting a fee waiver. Patients who paid the fee continued to be classified as TAP patients, even if they did not follow through with the referral. Eligible patients who chose not to join the program were classified as non-TAP patients and served as the comparison group for this study. Patients whose referrals were not approved for medical reasons or who obtained specialty care coverage prior to scheduling their specialty visit were not included in the comparison group.
Outcome Measures
Key outcome measures included rates of ED and inpatient admission per month prior to TAP referral, compared with after referral. Patient information was obtained from program administrative records. ED and inpatient visits were collected from the hospital’s inpatient records.
We calculated ED visit rates for each patient depending on their time in the program. This was done to adjust for the fact that patients were referred to TAP at different time points during the study and to prevent misclassification bias. Time prior to TAP (pre-TAP) was calculated for each patient as the number of months from the start of the study on May 1, 2007 until date of first referral to TAP. Time subsequent to TAP (post-TAP) was calculated as the number of months from date of first referral through the end of the study on April 30, 2011. Monthly rates of emergency department utilization and inpatient admission were calculated for the pre-TAP and post-TAP interval for each patient. These rates were compared for TAP patients and non-TAP patients.
Data Analysis
Multiple time series analysiswas performed using 2-tailed t tests, difference-in-differences estimation, 11 and multivariate linear regression using STATA IC 11 (College Station, TX). Difference-in-differences estimation was performed using a simple t test to test the premise that in the absence of other factors, ED utilization would increase equally among the comparison and intervention groups. This strategy was used to determine whether enrollment in TAP was associated with lower monthly ED utilization and inpatient admission rates, after adjusting for baseline characteristics, including age, gender, and zip code of residence.
Results
During the first 21 months of TAP, 623 patients were referred to TAP. 374 patients paid the fee or requested a waiver and joined the program. In all, 249 patients chose not to enter the program, referred to as non-TAP patients in the present article (i.e., comparison group). Demographic characteristics of TAP and non-TAP patients are shown in Table 1. TAP patients were not significantly different from non-TAP patients with respect to age, gender, or zip code, with the exception of zip code 21224, which was disproportionately represented among TAP patients (30% of TAP patients compared with 22% of non-TAP patients, P = .01). Race and ethnicity data are not kept in the program’s administrative data, but more than 95% of the patients at EBMC are African American.
Demographic Characteristics of TAP and Non-TAP Patients.
Abbreviations: TAP, The Access Partnership patients; Non-TAP, comparison group of eligible patients who did not join the program.
P < .05 comparing TAP and non-TAP patients.
As shown in Figure 1, rates of ED visits per month increased in both groups, from 3.9 to 5.1 visits per 100 person-months among TAP patients, a 31% increase from baseline, and 5.8 to 8.5 visits per 100 person-months among non-TAP patients, a 47% increase from baseline. Compared with non-TAP patients, TAP patients had 1.5 fewer total ED visits per 100 patient-months post-TAP, but this result was not statistically significant (P = .16, 95% confidence interval [CI] = −0.6 to 3.6). In multivariable linear regression, adjusting for gender, age, and zip code of residence, there was no significant association between TAP status and ED utilization (P = .17, 95% CI = −0.6 to 3.0; Table 2).
Emergency department (ED) visits per 100 patient-months among The Access Partnership (TAP) and comparison groups, pre- and post-TAP.
Multiple Linear Regression Results for TAP Versus Non-TAP Patients. a
Abbreviations: TAP, The Access Partnership patients; Non-TAP, comparison group of eligible patients who did not join the program; ED, emergency department; 95% CI, 95% confidence interval.
Comparing ED utilization per 100 person-months, pre-TAP enrollment versus post-TAP enrollment. Adjusted for age, gender, and zip code of residence.
Examining the subgroup of ED visits that did not result in inpatient admission, the number of visits per month was relatively stable among TAP patients, whereas this rate increased among non-TAP patients. As shown in Figure 1, TAP patients had 3.0 non-inpatient visits per 100 patient-months pre-TAP and 3.6 visits per 100 patient-months post-TAP (difference post − pre = 0.6), a 20% increase from baseline, whereas non-TAP patients had 4.5 visits per 100 patient-months pre-TAP and 7.1 visits per 100 patient-months post-TAP (difference post − pre = 2.6), a 58% increase from baseline. Compared with non-TAP patients, TAP patients had 2.0 fewer ED visits not leading to admission per 100 patient-months post-TAP (P = .03, 95% CI = 0.2-3.9). As shown in Table 2, in multiple linear regression, after controlling for age, gender, and zip code, TAP status was a significant predictor of ED visits not leading to admission (P = .04, 95% CI = 0.1-3.9).
Discussion
Prior studies have not shown that having a usual source of care alone reduces ED use among uninsured patients, whereas access to adequate primary and specialty care may reduce ED utilization. 9 By using a multiple time series analysis and difference-in-differences estimation, we were able to determine the change in ED utilization relative to enrollment in TAP while minimizing bias due to unmeasured confounders, and controlling for secular trends that might affect ED utilization.11,12 We found a significantly lower rate of ED visits that did not result in inpatient admission subsequent to joining the program among TAP patients, relative to the comparison group.
Similar to national trends seen among low-income patients over the past decade, 13 rates of ED utilization increased in this sample over the 4-year study period. Difficulties getting to clinic visits due to work, childcare, or transportation conflicts as well as low health literacy may contribute to the use of emergency services. By promoting engagement in primary care, communication between providers, and access to specialty care as well as care coordination services, the TAP program attempts to reduce some of these barriers.
Limitations
First, we only were able to assess ED utilization at the 2 hospitals affiliated with the academic medical center. Previous survey data with TAP patients indicated that 82% of patients reported using 1 of these 2 hospitals primarily for ED use, leading us to believe that we were capturing the majority of ED utilization through administrative data from these 2 hospitals. 10
Second, though both groups were eligible for TAP due to lack of insurance, income eligibility, and need for specialty care, our comparison group may differ from our intervention group in terms of motivation toward obtaining needed health care, burden of illness, or barriers to use of outpatient care services. Indeed, as shown in Figure 1, non-TAP patients had a significantly higher baseline level of ED utilization than TAP patients (5.8 vs 3.9 visits per 100 patient-months, P = .01). Although we adjusted for some demographic variables, we were unable to adjust for race, ethnicity, or for medical comorbidities. For example, the 21224 zip code, which was disproportionately represented among TAP patients, is home to a higher proportion of Hispanic patients than other zip codes studied. 14 Residual confounding is therefore a limitation of this study. There may have been crossover if study subjects elected to pay the fee but did not attend any TAP visits. In trying to capture an “intention-to-treat” population, we analyzed this group as TAP patients.
Third, the time frame for this study was short. Stronger results might result from more time in the program in which to have medical needs addressed and to change behaviors.
Finally, judging “appropriateness” of ED visits is difficult and wrought with assumptions. 15 Our decision to stratify ED visits into those resulting in admission and those not resulting in admission is based on our hypothesis that the TAP program would result in improved access to outpatient care, thereby avoiding some nonurgent ED visits.
In conclusion, increasing ED utilization is in part driving the rising cost of health care in the United States. Improved access to outpatient services may promote continuity of care while decreasing reliance on the ED. Our evaluation found that although overall rates of ED utilization did not change, TAP patients had a somewhat decreased rate of ED visits resulting in inpatient admission relative to a comparison group of patients who did not receive the intervention.
Footnotes
Acknowledgements
We would like to acknowledge Dr. Barbara G. Cook for her leadership and The Access Partnership and East Baltimore Medical Center staff for their hard work to make the program a success.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received following financial support for the research, authorship, and/or publication of this article: The Urban Health Institute at Johns Hopkins University partially funded the research of this paper.