At the Intersection of Health, Health Care and Policy Cite this article as: Theresa Y. Kim, Karoline Mortensen and Barbara Eldridge Linking Uninsured Patients Treated In The Emergency Department To Primary Care Shows Some Promise In Maryland Health Affairs, 34, no.5 (2015):796-804 doi: 10.1377/hlthaff.2014.1102
The online version of this article, along with updated information and services, is available at: http://content.healthaffairs.org/content/34/5/796.full.html
For Reprints, Links & Permissions: http://healthaffairs.org/1340_reprints.php E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml
Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600, Bethesda, MD 20814-6133. Copyright © 2015 by Project HOPE - The People-to-People Health Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical, including photocopying or by information storage or retrieval systems, without prior written permission from the Publisher. All rights reserved.
Not for commercial use or unauthorized distribution Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
Improving Care Systems By Theresa Y. Kim, Karoline Mortensen, and Barbara Eldridge 10.1377/hlthaff.2014.1102 HEALTH AFFAIRS 34, NO. 5 (2015): 796–804 ©2015 Project HOPE— The People-to-People Health Foundation, Inc.
doi:
Theresa Y. Kim is a doctoral student in the Department of Health Services Administration at the University of Maryland, in College Park. Karoline Mortensen ([email protected]) is an assistant professor in the Department of Health Services Administration at the University of Maryland. Barbara Eldridge is the manager of quality improvement at the Primary Care Coalition of Montgomery County, in Silver Spring, Maryland.
Linking Uninsured Patients Treated In The Emergency Department To Primary Care Shows Some Promise In Maryland ABSTRACT Use of the emergency department (ED) has increased significantly over the past twenty years, especially among people who lack access to regular care, such as from a primary care provider. Not only are many ED visits avoidable, but receiving care through the ED also may disrupt continuity of care and result in increased overall health care costs. This article analyzes one of the twenty-nine local projects funded by the Centers for Medicare and Medicaid Services: the Emergency Department–Primary Care Connect initiative of the Primary Care Coalition of Montgomery County, Maryland. The initiative linked low-income or uninsured patients with local safety-net primary care providers. In the period 2009–11, five participating hospital EDs referred 10,761 low-income uninsured ED patients to four local primary care clinics. The intervention did not significantly reduce overall subsequent ED visits, but there was a significant reduction in subsequent ED visits among the subpopulation with chronic physical or behavioral conditions if they had more than two visits to the same primary care clinic. Our findings suggest that expansion of safety-net clinics, combined with strategies to link high-need patients in the ED with these primary care providers, can reduce subsequent ED use.
U
se of the emergency department (ED) has increased significantly over the past twenty years for a variety of reasons.1–3 In particular, people who face barriers to accessing a primary care provider rely more heavily on the ED for care than do patients who do not face such barriers.4–6 Almost half of all ED visits are considered avoidable and could be prevented if patients had access to care in more timely and appropriate settings, such as care from a primary care provider.7,8 Reducing the barriers to primary care in the United States could enable more patients to manage illnesses and conditions without visiting the ED. A growing body of research confirms that 79 6
Health Affa irs
M ay 20 1 5
3 4 :5
greater continuity of care with a primary care provider is associated with significant reductions in ED visits among Medicaid enrollees, children, and the general population.9–11 One strategy for reducing ED visits associated with access barriers to primary care involves connecting patients in the ED with a primary care provider. For example, an intervention in Michigan that linked underserved Detroit residents who visited the ED to a primary care provider resulted in transitioning 55 percent of active patients out of the ED and into primary care settings, and it reduced subsequent ED use.12 The intervention was particularly effective for the subpopulation of participants with chronic diseases and frequent ED use.13
Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
The Centers for Medicare and Medicaid Services (CMS) awarded funds from the Emergency Room Diversion Grant Program to twenty state Medicaid agencies in April 2008 to address the issue of improving access to primary care for ED patients in medically underserved areas through collaborations with local community hospitals.14 Twenty-nine projects were funded, including one administered by the Primary Care Coalition of Montgomery County, Maryland, just north of Washington, D.C. An independent, nonprofit organization charged with providing high-quality, evidence-based primary care services to lowincome and uninsured residents in the county, the coalition received a diversion grant in 2009 through the Maryland Department of Health and Mental Hygiene. To our knowledge, very few projects produced strong enough data to support an evaluation of their results, and few evaluations were made publicly available beyond the twelve one-page summaries collected by CMS.14 A review of the peer-reviewed literature did not find any published evaluations of the projects, and CMS has not released a final report on the grants. Thus, little is known about the success of these diversion programs. The main objective of the Primary Care Coalition’s diversion intervention, the Emergency Department–Primary Care Connect initiative, was to connect low-income, uninsured, and Medicaid patients who visited the ED to a primary care provider at one of four safety-net clinics, and thus establish a “medical home” for each patient. All of the partnering clinics added capacity for walk-in appointments; increased open access appointment scheduling; and expanded appointment hours—specifically, additional weekend days and evening hours. Informational brochures on the intervention were available in the EDs and were also mailed to patients. Patients in this intervention reported visiting the ED because clinic hours were inconvenient, they lacked access to clinics, they preferred the speed of service at the ED, or they lacked an identifiable provider, according to qualitative interviews with the patients conducted by the Primary Care Coalition—a result that is consistent with the literature.15 In this study we analyzed the efficacy of the intervention in addressing these limitations and thereby reducing subsequent ED visits. One goal of the intervention was to encourage patients to establish an ongoing relationship with a primary care provider and reduce reliance upon the ED.9,16 We defined patients who completed two or more visits to the same clinic after their “index” ED visit—the one that led to their participation in the intervention—as “sticky” pa-
tients. Two-thirds of patients who completed an initial clinic visit returned to the clinic for a second visit.17
Study Data And Methods Participants And Data The Primary Care Coalition provided a merged data set that included data from three other sources: the discharge records of all five hospitals that participated in the intervention, electronic medical records with data from the four participating safety-net clinics, and referral data entered by patient navigators (discussed below). We matched 10,761 patients across the three individual data sets with an index ED visit between March 1, 2009, and December 31, 2011. Patients were included for analysis only if their names, dates of birth, and ED dates of service matched across the hospital, clinic, and patient navigator data. The matching rate was 88 percent: 711 of the 12,222 total referred patients had different ED dates in the hospital data than in the clinic and navigator data, and 750 were missing a part of the patient record (patient first name, last name, or date of birth) in one of the three individual data sets. Montgomery County, Maryland, is an affluent area north of Washington, D.C.,18 but it has a large and poor immigrant population. During the intervention period, the county had almost one million adult residents, 120,000 of whom were uninsured and 80,000 of whom were Medicaid recipients.18 Patients were eligible for the intervention if they visited the ED at any of the five hospitals in the county: Holy Cross Hospital, Montgomery General Hospital, Shady Grove Adventist Hospital, Suburban Hospital, and Washington Adventist Hospital. All five hospitals relied on ED providers (physicians, physician assistants, and nurse practitioners) to identify eligible patients in the ED and provide referrals. To participate, patients needed to be uninsured, have incomes below 250 percent of the federal poverty level, be age eighteen or older, reside in Montgomery County, and lack a primary care provider. To determine the lack of a primary care provider, ED providers asked patients if they had a doctor they saw for medical problems. Hospitals And Partner Clinics Each hospital adopted a unique clinic arrangement and approach to the intervention. Holy Cross Hospital referred eligible patients to its partner clinic if they were identified as requiring follow-up care within two weeks. Navigators were based in clinics and made contact with referred patients by telephone. Montgomery General Hospital ED providers M ay 2 0 1 5
3 4: 5
Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
Health A ffairs
797
Improving Care Systems referred eligible patients to their partner clinic. Here the navigator was based in the hospital and spoke face-to-face with patients referred by the ED providers. The navigator also attempted to make telephone contact with self-pay patients who had not been contacted by the ED provider while in the ED. Shady Grove Adventist Hospital referred eligible patients to its partner clinic, and the clinicbased navigator contacted all patients by telephone. Suburban Hospital did not have its own partner clinic and referred eligible patients to one of three (of the four total) participating clinics based on the patient’s home address and the specific services required. Similar to Montgomery General Hospital, Suburban Hospital identified self-pay patients from a list of those who did not receive a referral from an ED provider. Suburban’s hospital-based navigator contacted all patients referred by an ED provider either face-to-face or by telephone and attempted to contact all self-pay patients by telephone. Washington Adventist Hospital referred eligible patients to its partner clinic. The clinic-based navigator contacted all ED-referred patients face-to-face or by phone. Washington Adventist was the only hospital whose clinic partner was a federally qualified health center. This allowed the hospital to also refer Medicaid enrollees without a primary care provider to the clinic (the other clinics treat uninsured patients only). Patient Navigators The patient navigators were a key component of the intervention. Navigators came from various backgrounds, but most were not licensed to be a patient navigator. Hospital-based navigators included a registered nurse, AmeriCorps volunteers, and a human resources representative. Most navigators were bilingual in Spanish and English, had completed at least some college, and had been selected for their strong interpersonal and communication skills. All navigators received training from the Primary Care Coalition and were given authority to schedule clinic appointments. Navigators met quarterly to review results, share best practices, and inform project implementation. Analysis We performed two analyses using a negative binomial hurdle model. The first analysis modeled the number of subsequent ED visits for all patients in the intervention. The second estimated a similar regression for a subset of the study population: patients who had a chronic physical condition, a behavioral condition, or both. Our tabulation of descriptive statistics showed that the patients in this subset had been more reliant on the ED before the intervention and were more likely to develop a relationship 79 8
H e a lt h A f fai r s
May 2 015
34:5
Almost half of all ED visits could be prevented if patients had access to care in more timely and appropriate settings.
with a primary care clinic, compared to patients without these conditions. A hurdle model was used to account for the excess number of zeros in our data, because our descriptive statistics showed that most patients in our data did not have a subsequent ED visit.19 The first part of the hurdle model was a logistic regression that modeled the decision to have a subsequent ED visit. The second part, a negative binomial, described the factors that made patients more likely to return for repeated ED visits, given their decision to have a subsequent visit. The model was analyzed with robust clustering on the five different hospitals using a hospital indicator variable. The covariates of our analysis were age group (younger than age 20, 20–39, 40–59, and 60 or older), sex, race, and ethnicity, when available. An indicator was included for use of the ED within one year before the index ED visit.We included a measure for status as a “sticky” patient. The chronic physical conditions and behavioral conditions documented in the hospital or the clinic records represented the most common conditions seen in the ED or clinic. We categorized patients as having a chronic physical condition if they received either a clinic or a hospital diagnosis for hypertension, hyperlipidemia, diabetes, asthma, chronic obstructive pulmonary disease, or congestive heart failure. Patients with behavioral conditions were those with clinic or hospital International Classification of Diseases, Ninth Revision (ICD-9), diagnosis codes of depression, anxiety, or any other behavioral health diagnosis (ICD-9 codes 290–319).20 One strength of our data is that all were derived from inpatient, clinic, or patient navigator records. Thus, none of these measures was selfreported. We performed all analyses in Stata/SE, version 13.1. Limitations There are some limitations to this analysis. There were some inconsistencies in
Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
how hospitals recorded race and ethnicity: Some hospitals collected data on race but not ethnicity; some classified ethnicity as a proxy for race. The sample size varied across sites, and ED and clinic capacity varied by partnership. The project was not a case-control study, and each hospital-clinic partnership was unique. Matching patients across hospital, clinic, and patient navigator data was not perfect. There were also important differences in how hospitals and clinics approached their partnerships and used patient navigators, as discussed above. To address some of the differences, we reestimated our models without clustering on the hospitals. We also excluded one hospital at a time. In neither case did our results change significantly.We found the variability across hospitals and clinics to be a strength of the study because the effects that we observed were not driven by one particularly successful approach to implementing the intervention. Left and right censoring may be a concern for our analysis. That is, we did not have data on ED use more than one year before the index ED visit or—because of resource limitations once the grant funding ended—on ED use after the intervention ended in 2011. To address left censoring—the fact that we had data on ED use for only one year before the index ED visit—we considered conducting a sensitivity analysis on only patients who had no ED use in the year before the index visit. However, limiting our sample to those without previous ED use caused the patient population with behavioral health or chronic physical health problems to decrease by 44 percent, from 2,768 to 1,549. “Sticky” patient status did not vary between those with and without previous ED visits. Thus, our key explanatory variable appears to capture the short-term effects of the intervention. The Primary Care Coalition did not have access to patient records at other safety-net clinics within and outside of Montgomery County, so not all utilization was captured. This is not likely to affect our measure of a sticky patient, since the concept is based on the establishment of a medical home instead of on visiting a variety of clinics. Montgomery County has a large population of poor immigrants who reside in an affluent suburban area. Thus, the data are not necessarily generalizable to other study populations.
Study Results Subsequent Clinic Visits And Sticky Patients Descriptive statistics for the study sample are available in the online Appendix.21 Demographic characteristics of patients, tabulated by their subsequent clinic visits, show variation
in who became a “sticky” patient (Exhibit 1). Almost 21 percent of the sample (2,257) had a subsequent clinic visit. The majority of patients (65.9 percent) who completed a subsequent clinic visit returned to the same clinic for another visit (1,487 patients), becoming sticky patients. The proclivity to become a sticky patient increased with age: 23.9 percent of patients ages sixty and older became sticky, compared to 4.6 percent of those younger than twenty. Females had a higher tendency to become sticky (17.6 percent) than males did (10.2 percent). Of patients with a chronic physical condition or behavioral health condition, 26.8 percent became sticky, compared to 9.3 percent of patients without these conditions. Sticky patient status did not vary between those with and those without ED visits before the index ED visit. Subsequent ED Visits Descriptive statistics on the relationship between subsequent ED visits and the key explanatory variables (patients with and without previous ED visits, who were and were not sticky, and who had and did not have chronic physical conditions and behavioral conditions) are shown in Exhibit 2. Among the 10,761 patients in the sample, 71.3 percent (7,676 patients) did not have a subsequent ED visit. More than 16 percent reported one subsequent ED visit, and 12.5 percent had two or more subsequent ED visits. Patients were more likely to have ED visits after the intervention if they were previous ED users: 21.8 percent of those with previous ED visits had one subsequent visit, and 28.8 percent had two or more subsequent visits. Sticky patients were more likely than nonsticky patients to have one subsequent ED visit (19.8 percent versus 15.6 percent) and two or more subsequent visits (16.0 percent versus 12.0 percent). More than 40 percent of patients with chronic conditions and more than 50 percent with behavioral conditions had subsequent ED visits. Hurdle Model Results We found that patients ages 40–59 were 42 percent more likely to have a subsequent ED visit, compared to patients who were younger than age 20 (the exponential of the coefficient in Exhibit 3). There was no significant difference in subsequent ED use between patients ages 20–39 and those ages 60 and older. Hispanic or Latino patients were 39 percent less likely to have a subsequent ED visit, compared to patients who were not Hispanic or Latino. Furthermore, those who were Hispanic or Latino were significantly less likely than other patients to have repeated visits to the ED. Use of the ED before the intervention was significantly associated with having a subsequent ED visit (Exhibit 3). Previous use was also significantly associated with repeated subsequent M ay 20 1 5
34:5
Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
H ea lt h A f fai r s
799
Improving Care Systems Exhibit 1 Demographic Characteristics Of Patients, By Number Of Visits To The Same Clinic After An Index Emergency Department (ED) Visit Number of subsequent clinic visits 0 (n = 8,504)
1 (n = 770)
Number
Percent
Number
Less than 20 20–39 40–59 60 or more
242 5,336 2,515 411
92.4 83.8 71.4 67.6
8 410 300 52
3.1 6.4 8.5 8.6
12 621 709 145
4.6 9.8 20.1 23.9
Sex Female Male Race
3,930 4,574
74.3 83.6
433 337
8.2 6.2
929 558
17.6 10.2
American or Alaskan Native Asian Black or African American White or Caucasian Other Unknown
150 271 2,737 2,592 2,167 587
59.5 75.4 79.8 80.2 80.0 75.5
21 29 238 237 183 62
8.3 8.1 6.9 7.3 6.8 8.0
81 59 455 404 359 129
32.1 16.4 13.3 12.5 13.3 16.6
Ethnicity Not Hispanic or Latino Hispanic or Latino Unknown
4,296 2,820 1,388
81.9 75.3 78.4
344 291 135
6.6 7.8 7.6
605 634 248
11.5 16.9 14.0
Chronic physical condition or behavioral condition No 6,729 84.2 Yes 1,775 64.1 Chronic physical condition
520 250
6.5 9.0
744 743
9.3 26.8
No Yes
Characteristic
2 or more (n = 1,487) Percent
Number
Percent
Age (years)
7,556 948
83.3 55.9
594 176
6.6 10.4
916 571
10.1 33.7
7,498 1,006
79.9 73.3
677 93
7.2 6.8
1,214 273
12.9 19.9
Behavioral condition No Yes
ED use before the index ED visit No Yes
6,081 2,423
79.4 78.0
539 231
7.0 7.4
1,036 451
13.5 14.5
Hospital ID number 1 2 3 4 5
1,784 479 1,243 4,175 823
67.7 79.3 79.8 83.9 83.4
250 39 108 306 67
9.5 6.5 6.9 6.2 6.8
602 86 206 496 97
22.8 14.2 13.2 10.0 9.8
SOURCE Authors’ analysis of data from the Primary Care Coalition of Montgomery County, Maryland. NOTES The index ED visit is defined in the text. Differences for all demographic characteristics were significant (p
The online version of this article, along with updated information and services, is available at: http://content.healthaffairs.org/content/34/5/796.full.html
For Reprints, Links & Permissions: http://healthaffairs.org/1340_reprints.php E-mail Alerts : http://content.healthaffairs.org/subscriptions/etoc.dtl To Subscribe: http://content.healthaffairs.org/subscriptions/online.shtml
Health Affairs is published monthly by Project HOPE at 7500 Old Georgetown Road, Suite 600, Bethesda, MD 20814-6133. Copyright © 2015 by Project HOPE - The People-to-People Health Foundation. As provided by United States copyright law (Title 17, U.S. Code), no part of Health Affairs may be reproduced, displayed, or transmitted in any form or by any means, electronic or mechanical, including photocopying or by information storage or retrieval systems, without prior written permission from the Publisher. All rights reserved.
Not for commercial use or unauthorized distribution Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
Improving Care Systems By Theresa Y. Kim, Karoline Mortensen, and Barbara Eldridge 10.1377/hlthaff.2014.1102 HEALTH AFFAIRS 34, NO. 5 (2015): 796–804 ©2015 Project HOPE— The People-to-People Health Foundation, Inc.
doi:
Theresa Y. Kim is a doctoral student in the Department of Health Services Administration at the University of Maryland, in College Park. Karoline Mortensen ([email protected]) is an assistant professor in the Department of Health Services Administration at the University of Maryland. Barbara Eldridge is the manager of quality improvement at the Primary Care Coalition of Montgomery County, in Silver Spring, Maryland.
Linking Uninsured Patients Treated In The Emergency Department To Primary Care Shows Some Promise In Maryland ABSTRACT Use of the emergency department (ED) has increased significantly over the past twenty years, especially among people who lack access to regular care, such as from a primary care provider. Not only are many ED visits avoidable, but receiving care through the ED also may disrupt continuity of care and result in increased overall health care costs. This article analyzes one of the twenty-nine local projects funded by the Centers for Medicare and Medicaid Services: the Emergency Department–Primary Care Connect initiative of the Primary Care Coalition of Montgomery County, Maryland. The initiative linked low-income or uninsured patients with local safety-net primary care providers. In the period 2009–11, five participating hospital EDs referred 10,761 low-income uninsured ED patients to four local primary care clinics. The intervention did not significantly reduce overall subsequent ED visits, but there was a significant reduction in subsequent ED visits among the subpopulation with chronic physical or behavioral conditions if they had more than two visits to the same primary care clinic. Our findings suggest that expansion of safety-net clinics, combined with strategies to link high-need patients in the ED with these primary care providers, can reduce subsequent ED use.
U
se of the emergency department (ED) has increased significantly over the past twenty years for a variety of reasons.1–3 In particular, people who face barriers to accessing a primary care provider rely more heavily on the ED for care than do patients who do not face such barriers.4–6 Almost half of all ED visits are considered avoidable and could be prevented if patients had access to care in more timely and appropriate settings, such as care from a primary care provider.7,8 Reducing the barriers to primary care in the United States could enable more patients to manage illnesses and conditions without visiting the ED. A growing body of research confirms that 79 6
Health Affa irs
M ay 20 1 5
3 4 :5
greater continuity of care with a primary care provider is associated with significant reductions in ED visits among Medicaid enrollees, children, and the general population.9–11 One strategy for reducing ED visits associated with access barriers to primary care involves connecting patients in the ED with a primary care provider. For example, an intervention in Michigan that linked underserved Detroit residents who visited the ED to a primary care provider resulted in transitioning 55 percent of active patients out of the ED and into primary care settings, and it reduced subsequent ED use.12 The intervention was particularly effective for the subpopulation of participants with chronic diseases and frequent ED use.13
Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
The Centers for Medicare and Medicaid Services (CMS) awarded funds from the Emergency Room Diversion Grant Program to twenty state Medicaid agencies in April 2008 to address the issue of improving access to primary care for ED patients in medically underserved areas through collaborations with local community hospitals.14 Twenty-nine projects were funded, including one administered by the Primary Care Coalition of Montgomery County, Maryland, just north of Washington, D.C. An independent, nonprofit organization charged with providing high-quality, evidence-based primary care services to lowincome and uninsured residents in the county, the coalition received a diversion grant in 2009 through the Maryland Department of Health and Mental Hygiene. To our knowledge, very few projects produced strong enough data to support an evaluation of their results, and few evaluations were made publicly available beyond the twelve one-page summaries collected by CMS.14 A review of the peer-reviewed literature did not find any published evaluations of the projects, and CMS has not released a final report on the grants. Thus, little is known about the success of these diversion programs. The main objective of the Primary Care Coalition’s diversion intervention, the Emergency Department–Primary Care Connect initiative, was to connect low-income, uninsured, and Medicaid patients who visited the ED to a primary care provider at one of four safety-net clinics, and thus establish a “medical home” for each patient. All of the partnering clinics added capacity for walk-in appointments; increased open access appointment scheduling; and expanded appointment hours—specifically, additional weekend days and evening hours. Informational brochures on the intervention were available in the EDs and were also mailed to patients. Patients in this intervention reported visiting the ED because clinic hours were inconvenient, they lacked access to clinics, they preferred the speed of service at the ED, or they lacked an identifiable provider, according to qualitative interviews with the patients conducted by the Primary Care Coalition—a result that is consistent with the literature.15 In this study we analyzed the efficacy of the intervention in addressing these limitations and thereby reducing subsequent ED visits. One goal of the intervention was to encourage patients to establish an ongoing relationship with a primary care provider and reduce reliance upon the ED.9,16 We defined patients who completed two or more visits to the same clinic after their “index” ED visit—the one that led to their participation in the intervention—as “sticky” pa-
tients. Two-thirds of patients who completed an initial clinic visit returned to the clinic for a second visit.17
Study Data And Methods Participants And Data The Primary Care Coalition provided a merged data set that included data from three other sources: the discharge records of all five hospitals that participated in the intervention, electronic medical records with data from the four participating safety-net clinics, and referral data entered by patient navigators (discussed below). We matched 10,761 patients across the three individual data sets with an index ED visit between March 1, 2009, and December 31, 2011. Patients were included for analysis only if their names, dates of birth, and ED dates of service matched across the hospital, clinic, and patient navigator data. The matching rate was 88 percent: 711 of the 12,222 total referred patients had different ED dates in the hospital data than in the clinic and navigator data, and 750 were missing a part of the patient record (patient first name, last name, or date of birth) in one of the three individual data sets. Montgomery County, Maryland, is an affluent area north of Washington, D.C.,18 but it has a large and poor immigrant population. During the intervention period, the county had almost one million adult residents, 120,000 of whom were uninsured and 80,000 of whom were Medicaid recipients.18 Patients were eligible for the intervention if they visited the ED at any of the five hospitals in the county: Holy Cross Hospital, Montgomery General Hospital, Shady Grove Adventist Hospital, Suburban Hospital, and Washington Adventist Hospital. All five hospitals relied on ED providers (physicians, physician assistants, and nurse practitioners) to identify eligible patients in the ED and provide referrals. To participate, patients needed to be uninsured, have incomes below 250 percent of the federal poverty level, be age eighteen or older, reside in Montgomery County, and lack a primary care provider. To determine the lack of a primary care provider, ED providers asked patients if they had a doctor they saw for medical problems. Hospitals And Partner Clinics Each hospital adopted a unique clinic arrangement and approach to the intervention. Holy Cross Hospital referred eligible patients to its partner clinic if they were identified as requiring follow-up care within two weeks. Navigators were based in clinics and made contact with referred patients by telephone. Montgomery General Hospital ED providers M ay 2 0 1 5
3 4: 5
Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
Health A ffairs
797
Improving Care Systems referred eligible patients to their partner clinic. Here the navigator was based in the hospital and spoke face-to-face with patients referred by the ED providers. The navigator also attempted to make telephone contact with self-pay patients who had not been contacted by the ED provider while in the ED. Shady Grove Adventist Hospital referred eligible patients to its partner clinic, and the clinicbased navigator contacted all patients by telephone. Suburban Hospital did not have its own partner clinic and referred eligible patients to one of three (of the four total) participating clinics based on the patient’s home address and the specific services required. Similar to Montgomery General Hospital, Suburban Hospital identified self-pay patients from a list of those who did not receive a referral from an ED provider. Suburban’s hospital-based navigator contacted all patients referred by an ED provider either face-to-face or by telephone and attempted to contact all self-pay patients by telephone. Washington Adventist Hospital referred eligible patients to its partner clinic. The clinic-based navigator contacted all ED-referred patients face-to-face or by phone. Washington Adventist was the only hospital whose clinic partner was a federally qualified health center. This allowed the hospital to also refer Medicaid enrollees without a primary care provider to the clinic (the other clinics treat uninsured patients only). Patient Navigators The patient navigators were a key component of the intervention. Navigators came from various backgrounds, but most were not licensed to be a patient navigator. Hospital-based navigators included a registered nurse, AmeriCorps volunteers, and a human resources representative. Most navigators were bilingual in Spanish and English, had completed at least some college, and had been selected for their strong interpersonal and communication skills. All navigators received training from the Primary Care Coalition and were given authority to schedule clinic appointments. Navigators met quarterly to review results, share best practices, and inform project implementation. Analysis We performed two analyses using a negative binomial hurdle model. The first analysis modeled the number of subsequent ED visits for all patients in the intervention. The second estimated a similar regression for a subset of the study population: patients who had a chronic physical condition, a behavioral condition, or both. Our tabulation of descriptive statistics showed that the patients in this subset had been more reliant on the ED before the intervention and were more likely to develop a relationship 79 8
H e a lt h A f fai r s
May 2 015
34:5
Almost half of all ED visits could be prevented if patients had access to care in more timely and appropriate settings.
with a primary care clinic, compared to patients without these conditions. A hurdle model was used to account for the excess number of zeros in our data, because our descriptive statistics showed that most patients in our data did not have a subsequent ED visit.19 The first part of the hurdle model was a logistic regression that modeled the decision to have a subsequent ED visit. The second part, a negative binomial, described the factors that made patients more likely to return for repeated ED visits, given their decision to have a subsequent visit. The model was analyzed with robust clustering on the five different hospitals using a hospital indicator variable. The covariates of our analysis were age group (younger than age 20, 20–39, 40–59, and 60 or older), sex, race, and ethnicity, when available. An indicator was included for use of the ED within one year before the index ED visit.We included a measure for status as a “sticky” patient. The chronic physical conditions and behavioral conditions documented in the hospital or the clinic records represented the most common conditions seen in the ED or clinic. We categorized patients as having a chronic physical condition if they received either a clinic or a hospital diagnosis for hypertension, hyperlipidemia, diabetes, asthma, chronic obstructive pulmonary disease, or congestive heart failure. Patients with behavioral conditions were those with clinic or hospital International Classification of Diseases, Ninth Revision (ICD-9), diagnosis codes of depression, anxiety, or any other behavioral health diagnosis (ICD-9 codes 290–319).20 One strength of our data is that all were derived from inpatient, clinic, or patient navigator records. Thus, none of these measures was selfreported. We performed all analyses in Stata/SE, version 13.1. Limitations There are some limitations to this analysis. There were some inconsistencies in
Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
how hospitals recorded race and ethnicity: Some hospitals collected data on race but not ethnicity; some classified ethnicity as a proxy for race. The sample size varied across sites, and ED and clinic capacity varied by partnership. The project was not a case-control study, and each hospital-clinic partnership was unique. Matching patients across hospital, clinic, and patient navigator data was not perfect. There were also important differences in how hospitals and clinics approached their partnerships and used patient navigators, as discussed above. To address some of the differences, we reestimated our models without clustering on the hospitals. We also excluded one hospital at a time. In neither case did our results change significantly.We found the variability across hospitals and clinics to be a strength of the study because the effects that we observed were not driven by one particularly successful approach to implementing the intervention. Left and right censoring may be a concern for our analysis. That is, we did not have data on ED use more than one year before the index ED visit or—because of resource limitations once the grant funding ended—on ED use after the intervention ended in 2011. To address left censoring—the fact that we had data on ED use for only one year before the index ED visit—we considered conducting a sensitivity analysis on only patients who had no ED use in the year before the index visit. However, limiting our sample to those without previous ED use caused the patient population with behavioral health or chronic physical health problems to decrease by 44 percent, from 2,768 to 1,549. “Sticky” patient status did not vary between those with and without previous ED visits. Thus, our key explanatory variable appears to capture the short-term effects of the intervention. The Primary Care Coalition did not have access to patient records at other safety-net clinics within and outside of Montgomery County, so not all utilization was captured. This is not likely to affect our measure of a sticky patient, since the concept is based on the establishment of a medical home instead of on visiting a variety of clinics. Montgomery County has a large population of poor immigrants who reside in an affluent suburban area. Thus, the data are not necessarily generalizable to other study populations.
Study Results Subsequent Clinic Visits And Sticky Patients Descriptive statistics for the study sample are available in the online Appendix.21 Demographic characteristics of patients, tabulated by their subsequent clinic visits, show variation
in who became a “sticky” patient (Exhibit 1). Almost 21 percent of the sample (2,257) had a subsequent clinic visit. The majority of patients (65.9 percent) who completed a subsequent clinic visit returned to the same clinic for another visit (1,487 patients), becoming sticky patients. The proclivity to become a sticky patient increased with age: 23.9 percent of patients ages sixty and older became sticky, compared to 4.6 percent of those younger than twenty. Females had a higher tendency to become sticky (17.6 percent) than males did (10.2 percent). Of patients with a chronic physical condition or behavioral health condition, 26.8 percent became sticky, compared to 9.3 percent of patients without these conditions. Sticky patient status did not vary between those with and those without ED visits before the index ED visit. Subsequent ED Visits Descriptive statistics on the relationship between subsequent ED visits and the key explanatory variables (patients with and without previous ED visits, who were and were not sticky, and who had and did not have chronic physical conditions and behavioral conditions) are shown in Exhibit 2. Among the 10,761 patients in the sample, 71.3 percent (7,676 patients) did not have a subsequent ED visit. More than 16 percent reported one subsequent ED visit, and 12.5 percent had two or more subsequent ED visits. Patients were more likely to have ED visits after the intervention if they were previous ED users: 21.8 percent of those with previous ED visits had one subsequent visit, and 28.8 percent had two or more subsequent visits. Sticky patients were more likely than nonsticky patients to have one subsequent ED visit (19.8 percent versus 15.6 percent) and two or more subsequent visits (16.0 percent versus 12.0 percent). More than 40 percent of patients with chronic conditions and more than 50 percent with behavioral conditions had subsequent ED visits. Hurdle Model Results We found that patients ages 40–59 were 42 percent more likely to have a subsequent ED visit, compared to patients who were younger than age 20 (the exponential of the coefficient in Exhibit 3). There was no significant difference in subsequent ED use between patients ages 20–39 and those ages 60 and older. Hispanic or Latino patients were 39 percent less likely to have a subsequent ED visit, compared to patients who were not Hispanic or Latino. Furthermore, those who were Hispanic or Latino were significantly less likely than other patients to have repeated visits to the ED. Use of the ED before the intervention was significantly associated with having a subsequent ED visit (Exhibit 3). Previous use was also significantly associated with repeated subsequent M ay 20 1 5
34:5
Downloaded from content.healthaffairs.org by Health Affairs on May 27, 2015 at Karolinska institutet
H ea lt h A f fai r s
799
Improving Care Systems Exhibit 1 Demographic Characteristics Of Patients, By Number Of Visits To The Same Clinic After An Index Emergency Department (ED) Visit Number of subsequent clinic visits 0 (n = 8,504)
1 (n = 770)
Number
Percent
Number
Less than 20 20–39 40–59 60 or more
242 5,336 2,515 411
92.4 83.8 71.4 67.6
8 410 300 52
3.1 6.4 8.5 8.6
12 621 709 145
4.6 9.8 20.1 23.9
Sex Female Male Race
3,930 4,574
74.3 83.6
433 337
8.2 6.2
929 558
17.6 10.2
American or Alaskan Native Asian Black or African American White or Caucasian Other Unknown
150 271 2,737 2,592 2,167 587
59.5 75.4 79.8 80.2 80.0 75.5
21 29 238 237 183 62
8.3 8.1 6.9 7.3 6.8 8.0
81 59 455 404 359 129
32.1 16.4 13.3 12.5 13.3 16.6
Ethnicity Not Hispanic or Latino Hispanic or Latino Unknown
4,296 2,820 1,388
81.9 75.3 78.4
344 291 135
6.6 7.8 7.6
605 634 248
11.5 16.9 14.0
Chronic physical condition or behavioral condition No 6,729 84.2 Yes 1,775 64.1 Chronic physical condition
520 250
6.5 9.0
744 743
9.3 26.8
No Yes
Characteristic
2 or more (n = 1,487) Percent
Number
Percent
Age (years)
7,556 948
83.3 55.9
594 176
6.6 10.4
916 571
10.1 33.7
7,498 1,006
79.9 73.3
677 93
7.2 6.8
1,214 273
12.9 19.9
Behavioral condition No Yes
ED use before the index ED visit No Yes
6,081 2,423
79.4 78.0
539 231
7.0 7.4
1,036 451
13.5 14.5
Hospital ID number 1 2 3 4 5
1,784 479 1,243 4,175 823
67.7 79.3 79.8 83.9 83.4
250 39 108 306 67
9.5 6.5 6.9 6.2 6.8
602 86 206 496 97
22.8 14.2 13.2 10.0 9.8
SOURCE Authors’ analysis of data from the Primary Care Coalition of Montgomery County, Maryland. NOTES The index ED visit is defined in the text. Differences for all demographic characteristics were significant (p
