Frequent Emergency Department Use Among Released Prisoners With Human Immunodeficiency Virus: Characterization Including a Novel Multimorbidity Index


  • Jaimie P. Meyer MD,

    Corresponding author
    1. Divisions of Chronic Disease Epidemiology, Yale University School of Public Health, New Haven, CT
    • Department of Medicine, Section of Infectious Diseases, AIDS Program, Yale University School of Medicine, New Haven, CT
    Search for more papers by this author
  • Jingjun Qiu MD, MS,

    1. Department of Medicine, Section of Infectious Diseases, AIDS Program, Yale University School of Medicine, New Haven, CT
    Search for more papers by this author
  • Nadine E. Chen MD, MPH,

    1. Department of Medicine, Section of Infectious Diseases, AIDS Program, Yale University School of Medicine, New Haven, CT
    2. Department of Medicine, Division of Global Public Health, University of California San Diego School of Medicine, University of California San Diego School of Medicine, San Diego, CA
    Search for more papers by this author
  • Gregory L. Larkin MD, MSPH, MS,

    1. Department of Emergency Medicine, Yale University School of Medicine, New Haven, CT
    2. Department of Surgery, Division of Emergency Medicine, University of Auckland School of Medicine, Auckland, NZ
    Search for more papers by this author
  • Frederick L. Altice MD, MA

    1. Department of Medicine, Section of Infectious Diseases, AIDS Program, Yale University School of Medicine, New Haven, CT
    2. Division of Epidemiology of Microbial Diseases, Yale University School of Public Health, New Haven, CT
    Search for more papers by this author

  • Primary funding for this research was provided by the National Institute on Drug Abuse (R01 DA017059, FLA). The authors also acknowledge career development funding from the National Institutes of Allergy and Infectious Diseases (T32 AI007517 for JPM), the National Institute of Mental Health (T32 MH020031 for JPM), the National Institute on Drug Abuse (K24 DA017072 for FLA and K23 DA033858 for JPM), and the American Foundation for Suicide Prevention (Distinguished Investigator Award for GLL). The authors have no further disclosures or conflicts of interest to report.

Address for correspondence and reprints: Jaimie P. Meyer, MD; e-mail:



The objective was to characterize the medical, social, and psychiatric correlates of frequent emergency department (ED) use among released prisoners with human immunodeficiency virus (HIV).


Data on all ED visits by 151 released prisoners with HIV on antiretroviral therapy (ART) were prospectively collected for 12 months. Correlates of frequent ED use, defined as having two or more ED visits postrelease, were described using univariate and multivariate models and generated medical, psychiatric, and social multimorbidity indices.


Forty-four (29%) of the 151 participants were defined as frequent ED users, accounting for 81% of the 227 ED visits. Frequent ED users were more likely than infrequent or nonusers to be female; have chronic medical illnesses that included seizures, asthma, and migraines; and have worse physical health-related quality of life (HRQoL). In multivariate Poisson regression models, frequent ED use was associated with lower physical HRQoL (odds ratio [OR] = 0.95, p = 0.02) and having not had prerelease discharge planning (OR = 3.16, p = 0.04). Frequent ED use was positively correlated with increasing psychiatric multimorbidity index values.


Among released prisoners with HIV, frequent ED use is driven primarily by extensive comorbid medical and psychiatric illness. Frequent ED users were also less likely to have received prerelease discharge planning, suggesting missed opportunities for seamless linkages to care.

Resumen: La Hiperfrecuentación al Servicio de Urgencias de los Presos Puestos en Libertad con VIH: Caracterización que Incluye un Índice de Multimorbilidad Novedoso


Caracterizar la correlación médica, social y psiquiátrica de la hiperfrecuentación al servicio de urgencias (SU) de los presos puestos en libertad con VIH.


Se recogieron de forma prospectiva los datos de todas las visitas al SU de 151 presos en libertad con VIH en tratamiento antirretroviral durante 12 meses. La correlación de la hiperfrecuentación al SU, definida como haber tenido dos o más consultas al SU tras su libertad, se describió usando modelos univariables y multivariables, y se generaron indices multimorbilidad social, psiquiátrica y médica.


Cuarenta y cuatro (29%) de los 151 participantes se definieron como hiperfrecuentadores del SU, lo que suponia el 81% de las 227 visitas al SU. Los hiperfrecuentadores tuvieron mayor probabilidad que los no frecuentadores o no usuarios de ser mujer, tener enfermedades médicas crónicas (que incluyeron epilepsia, asma y migraña) y peor calidad de vida relacionada con la salud. En el análisis multivariable mediante modelos de regresión de Poisson, los hiperfrecuentadores se asociaron con una peor calidad de vida relacionada con la salud (odds ratio [OR] 0,95, p = 0,02) y con no haber tenido un plan antes de la libertad (OR 3.16, p = 0.04). Los hiperfrecuentadores del SU se correlacionaron positivamente con valores elevados de multimorbilidad psiquiátrica.


Entre los presos liberados con VIH, la hiperfrecuentación del SU está condicionada principalmente por una amplia comorbilidad médica y enfermedad psiquiátrica. Los hiperfrecuentadores del SU tienen también menos probabilidad de tener un plan antes de la liberación, lo que sugiere una pérdida de oportunidades para la continuidad en la atención.

In recent years, political discourse has centered on cost-cutting to improve efficiency in U.S. health care systems. The Affordable Care Act, passed by Congress in March 2010 and deemed constitutional by the Supreme Court in 2012, has significant implications for U.S. emergency departments (EDs) and the delivery of emergency services. As the only guaranteed haven for health care in America, EDs will be increasingly challenged to balance the mandate of patient safety with ED oversubscription and the attendant financial disincentives attached to ED recidivism.[1, 2] Until health care reform is fully realized, the effect of recidivism on ED operations is an important area for exploration and discussion.

Frequent ED users, as variously defined, are recognized in both emergency medicine literature and popular media as disproportionately expensive consumers of health care.[3, 4] In Oregon, for example, 50% of Medicaid dollars spent on emergency services were attributable to just 3% of enrollees who made multiple ED visits.[5] Frequent users have been stigmatized as manipulative patients with repetitive nonurgent visits that contribute significantly to hospital costs and ED provider burnout.[6] More recent systematic evaluations, however, document frequent ED users as being a heterogeneous group who defy these stereotypes: compared to infrequent or nonusers, they are more likely to have a primary health care provider and medical insurance, be non-Hispanic white, and experience more severe medical and psychiatric comorbidities.[3, 7-9] Rather than visits simply being “inappropriate,” a label perpetuated by medical insurance companies to justify denied reimbursement, frequent ED visits may be an important way by which clinically ill patients access resources they otherwise cannot in community settings. Some authors and policy-makers have thus called for an end to the term “inappropriate” ED use.[10, 11]

Most previous research on frequent ED users has either focused on all comers to a single ED or on ED patients with a defining chronic illness, like migraines or asthma, which is associated with repeat visits.[3] Existing literature on frequent ED users has often relied on self-reported measures of ED use[12] or has minimally examined individual-level sociobehavioral characteristics that may contribute to health care use.[8] This article addresses these shortcomings with a richer database and a longitudinal design, which allows for a more robust understanding of ED recidivism. Furthermore, we describe an important subset of frequent ED users, specifically human immunodeficiency virus (HIV)-infected prisoners transitioning to the community, who have not yet been described in the ED health services literature.

Released prisoners constitute an important subgroup of frequent visitors to the ED. Indeed, many EDs remain the provider of first and last resort for released prisoners. Released prisoners overall experience disproportionately high rates of substance use disorders, mental disorders, and social instability, including homelessness, that likely contribute to their health care use.[13, 14] Within the first 2 weeks of being released from correctional settings, former prisoners experience markedly high levels of age-matched morbidity and mortality, often related to drug overdose and complications from HIV and liver disease.[15-17] Moreover, during this destabilizing period immediately following prison release, released prisoners with HIV are at high risk of discontinuous primary care and lapses in antiretroviral therapy (ART) adherence,[18-22] so that the ED becomes their only point of contact with health care systems.[23-26] Disenfranchisement with primary care may translate into reliance on ED services to meet extensive medical, psychiatric, and social needs because this out-of-care group is also more likely to be homeless, lack health insurance, and have significant underlying psychiatric comorbidity.[14, 27]

To the best of our knowledge, this is the first study of ED use by newly released prisoners, and we aimed to characterize and describe correlates of frequent ED use using both individual descriptors and multimorbidity indices. Multimorbidity indices are becoming increasingly recognized in the epidemiologic literature as a useful and convenient measure of aggregate health burden, but have not previously been applied to frequent users of ED services.[28] We hypothesized that frequent ED use would be associated with social or psychiatric instability rather than severity of medical illness. This unique cohort and methodology provides insights into frequent ED use more generally, and for released prisoners with chronic illness, in particular. Furthermore, by identifying an important patient population at risk for current and continued frequent ED use, we have addressed a major research agenda proposed by Pines et al. in Academic Emergency Medicine.[9]


Study Design

The study's analytical plan draws on the Behavioral Health Model[29, 30] that has been adapted for vulnerable populations.[31] This conceptual model proposes that health behaviors are driven by individual predisposing factors, enabling resources, and needs characteristics of populations.

In this prospective longitudinal study of ED use among released prisoners with HIV, all participants provided written informed consent to have their health care records reviewed at all designated ED sites. Ethical approval for all procedures was obtained from the Yale University School of Medicine institutional review board and the Connecticut Department of Correction Research Advisory Committee.

Study Setting and Population

The 151 study participants were HIV-infected, released prisoners on ART who were enrolled in an ongoing randomized clinical trial of directly administered antiretroviral therapy (DAART) versus self-administered antiretroviral therapy (SAT) between 2006 and 2010 ( registry NCT00786396).[32] The study design and baseline characteristics have previously been described.[33] Eligibility criteria included: age ≥ 18 years, laboratory-confirmed HIV infection, current receipt of ART, and planned return to New Haven or Hartford after release from prison. Subjects were recruited from Connecticut prisons within 90 days of release; individuals who otherwise met the inclusion criteria, but did not receive prerelease planning and were within 30 days of release, were also enrolled. After informed consent, all study participants were followed prospectively for 12 months after date of release, during which they received comprehensive case management services, adherence support with DAART (if randomized to this arm), assistance with enrollment in medical and medication entitlements, and prescribed buprenorphine if they qualified for and desired opioid substitution therapy. Subjects were also interviewed and assessed for HIV-1 RNA levels and CD4 counts quarterly. Buprenorphine procedures and preliminary outcomes of those meeting criteria for opioid dependence have previously been described.[34, 35]

Study Protocol

Enrolled participants who completed a baseline assessment and consented to release of medical information were asked to designate all health care facilities at which they had ever received or intended to use medical care. Medical chart reviews were performed according to this designation, including both EDs in New Haven and all three in Hartford, ensuring complete coverage of each urban catchment area so we could document the most extensive possible spectrum of ED utilization. The Veteran's Administration (VA) and psychiatric hospitals were not accessed for chart review (although all psychiatric units within the reviewed EDs were included and are a conduit to psychiatric inpatient care). None of the participants designated the VA hospital as a site of prior medical care. Charts were prospectively reviewed for up to 1 year from the date of prison release. ED visit data were compiled for each participant, regardless of how long he or she was retained in the parent study. No one withdrew their consent for participation in the study.

The primary outcome of interest was quantity of ED visits. By use of standardized chart review instruments, ED medical records were extracted for time and status of triage and discharge, patient-cited reason(s) for visit (i.e., “chief complaint”), provider diagnoses and associated diagnostic codes, results of urine toxicology tests, whether the ED visit resulted in a hospitalization (including psychiatric hospitalization), disposition from the ED, and discharge medications. An ED visit was considered as being for substance misuse if it was coded by an ED provider as being for acute intoxication or overdose, regardless of the type of substance used, or if the patient's chief complaint was recorded as “requesting detox.” Unless otherwise noted, all other measurements were collected as part of the baseline assessment.

Baseline demographic parameters were assessed as part of the previously validated[36] Addiction Severity Index (ASI),[37] including age, sex, education, self-reported race/ethnicity, and marital status. Anticipated housing status was categorized based on the question: “How would you describe your expected living situation on the day of release?” Subjects were considered homeless if they answered “in a shelter” or “street or other public place”; temporarily housed if they reported “with a family member or friend temporarily,” “haven't decided yet,” “short-term boarding,” or “drug treatment program”; and permanently housed if they answered “with a family member or friend permanently” or “my own place.” History of abuse was measured on the ASI by self-report of any emotional, physical, or sexual abuse in one's lifetime. These questions have been shown to be a highly sensitive screen for subjects with histories of substance use and posttraumatic stress disorders (sensitivity = 0.91).[38] Abuse was coded as a dichotomous variable indicating lifetime history of any physical or sexual abuse versus none.

Study involvement was described in terms of 1) prerelease discharge planning from the parent study (participants recruited in the 30 day “postrelease” period did not receive prerelease discharge planning); 2) randomization status in the parent study (DAART versus SAT); 3) site of residence after release (New Haven versus Hartford); and 4) mean number of months retained in the parent study, up to six.

Physical and mental health was described in terms of the following:

Surrogates of HIV Biologic Severity: Absolute CD4 count and quantitative HIV-1 RNA levels were measured at baseline and at 12-week follow-up. CD4 count (cells/μL) was described both continuously and dichotomously (<200 vs. ≥200); viral suppression was dichotomously defined as having HIV-1 viral load < 400 copies/mL or not.

Substance Use Disorders: Alcohol use disorders were measured using the Alcohol Use Disorders Identification Test (AUDIT; Cronbach's α = 0.83, range = 0.75 to 0.097), with hazardous drinking defined as a score ≥ 8 for men and ≥ 4 for women.[39, 40] Lifetime alcohol and drug use was measured by the ASI, and severity was assessed using previously validated continuous alcohol and drug composite subscores that ranged from 0 to 1 with larger values reflecting greater severity.[37] If the subject initiated and was retained on buprenorphine at 6 months, this was recorded as a dichotomous variable. Self-reported history of ever using injection drugs was also recorded as a dichotomous variable.

Current Mental Illness: Depressive symptoms at baseline were measured using the 16-item Quick Inventory of Depressive Symptomatology Self-Report (QIDS-SR; Cronbach's α = 0.86, range = 0.73 to 0.92).[41] Depression severity scores were continuous from 0 to 27 and were also classified dichotomously as major depression (score > 11) versus no or mild depression (score ≤ 11). Self-report of ever having attempted suicide was recorded as a dichotomous variable.

Other Health-related Factors: Social support was measured continuously using the validated 31-item Social Support Scale (range = 0 to 100 with higher scores reflecting greater social support).[42] Trust in physician was measured continuously using the 11-item Trust in Physician scale (Cronbach's α = 0.90).[43] Each item is scored on a five-point Likert scale, and a summary measure is equal to the unweighted mean of the responses (range 0 to 100 with higher scores reflecting higher levels of trust).[44] HIV-associated health-related quality of life (HRQoL) was measured using the 36-item Medical Outcomes Study Short Form (SF-36; Cronbach's α > 0.80)[45]; physical and mental HRQoL subscores were calculated separately using norm-based scoring (mean ± standard deviation [SD] = 50 ± 10) with higher scores representing perception of better health. HIV-related stigma was measured using the Berger HIV Stigma Scale, a 40-item assessment in which each item is rated on a four-point Likert type scale, and then summed for a total score. Higher scores represent higher levels of perceived HIV-related stigma and the scale has been shown to be reliable and valid in diverse populations with HIV (Cronbach's α = 0.96).[46]

Medical History: Previous chronic illnesses were measured as part of the baseline interview, in which participants were asked, “Have you ever been told by a health care provider that you currently have any of the following”: seizures, hypertension (high blood pressure), asthma, emphysema, tuberculosis, pneumonia, diabetes, hepatitis, cirrhosis (liver problems), abnormal Pap smear (women only), cancer, arthritis/bone pain, sickle cell, migraine headaches, depression, anxiety disorder, schizophrenia, posttraumatic stress disorder (PTSD), eating disorders, thyroid problems, or dyslipidemia (high cholesterol). Though not all-inclusive, this comprehensive list of comorbidities was derived from an extensive review of medical and psychiatric comorbidity among HIV-infected people who use drugs.[47] These comorbidities were also chosen because participants were more likely aware of medical illnesses from which they have had symptoms or are receiving medical treatment (e.g., hypertension compared to coronary artery disease). Responses were coded dichotomously for each and are presented under the Results section if the prevalence of the illness within the total cohort was >5%.

Use of other health care services was derived from the follow-up interview at 3 months. HIV clinic utilization was determined from the question: “In the last 30 days or since the last interview, did you see or talk with an HIV provider or go to an HIV clinic?” and responses were coded dichotomously. Supervised withdrawal from alcohol or drugs was coded dichotomously from the question, “In the past three months, have you been to a detox center?”

Multimorbidity Indices: In recognition that a number of factors may contribute to frequent ED use, multimorbidity indices portray a broader picture of overall health, taking into account the compound burden of multiple concurrent illnesses. To portray aggregate health burden, multimorbidity indices were generated for this population by taking into account the number and/or severity of medical, psychiatric, and social issues identified on the baseline assessment. Based on previously published literature, selection criteria for included conditions was based on: 1) prolonged duration (chronicity), 2) need for continuous treatment, 3) severe impact on those affected, and 4) high prevalence in this patient population.[28] Medical indices thus included markers of HIV biologic severity (baseline CD4 < 200, baseline HIV viral suppression), self-reported presence of comorbid conditions (seizures, hypertension, asthma or emphysema, diabetes, cirrhosis, migraine headaches, arthritis), and overall perception of physical health (lowest quartile physical HRQoL score). Psychiatric indices took into account current and prior substance use disorders (hazardous drinking, highest quartile ASI score for alcohol and for drugs, injection drug use), current and prior mental illnesses (major depression, suicide attempt, anxiety, schizophrenia), and self-perception of mental health (lowest quartile mental HRQoL score). Social indices included variables related to housing stability (homelessness), lifetime histories of physical or sexual abuse, social support (lowest quartile Social Support score), relationship status (not in a relationship), trust in physician (lowest quartile Trust in Physician score), stigma (highest quartile HIV stigma score), and ART adherence support (not randomized to DAART). The final composite score was calculated by the sum of total “yes” responses.

Data Analysis

Descriptive statistics were applied to the total cohort and subsets stratified by number of ED visits, using chi-square tests for categorical variables and Student's t-tests for continuous variables. Frequent users were defined has having two or more ED visits in the year following prison release. This definition was based on the distribution of ED use frequency in our sample and is within the range of what has been previously reported in the literature on frequent ED users.[3] To determine correlates of being a frequent user, univariate analyses were performed first using all possible covariates. Significant associations with p < 0.2 from the univariate model were then included in a multivariate model that used Poisson regression to address non-Gaussian distribution of ED count data. Exposure time, defined as up to 12 months from study enrollment, was included as an offset variable. Multiple combinations of variables were evaluated for interaction or colinearity, including depression and mental HRQoL, baseline and 12-week viral suppression, alcohol severity score, and hazardous drinking; none were found to be statistically significant. Akaike's Information Criterion and Bayesian Information Criterion were used to describe goodness of fit for final models.[48, 49] Odds ratios (ORs) with 95% confidence intervals (CIs) are reported, and p-values < 0.05 are considered to indicate a statistically significant difference. After medical, psychiatric, and social multimorbidity indices were calculated, scatter plots with fitted regression lines were used to plot number of ED visits against each index separately. All analyses were performed using statistical software STATA version 9.0 (StataCorp, College Station, TX).


Participant Characteristics

As has been described previously,[32, 33] 154 released prisoners with HIV were enrolled in an ongoing randomized clinical trial of DAART versus SAT. Three participants either declined release of medical information or did not complete a full baseline assessment, leaving 151 evaluable subjects for whom ED records were prospectively reviewed.

Table 1 describes characteristics of all 151 subjects and subsets stratified by number of ED visits. Overall, the demographic profile of the total cohort reflected the statewide correctional population in Connecticut,[50] in which the typical individual was a middle-aged, unmarried, non-Hispanic African American man facing homelessness or temporary housing upon release from prison. Nearly 80% of participants had HIV viral suppression by the time of release, despite high prevalence of other chronic illnesses like hypertension, asthma, viral hepatitis, depression, and anxiety. Drug and alcohol use disorders were also common in this sample, reflected in relatively high mean lifetime ASI severity scores for drug and alcohol. At enrollment, 94 participants met criteria for opioid dependence and 50 initiated buprenorphine.

Table 1. Characteristics of Subjects, Stratified by Number of ED Visits
Characteristics<2 ED Visits (n = 107)≥2 ED Visits (n = 44)Total Cohort (n = 151)p-value
  1. Values are reported as n (%), median (IQR 25%–75%), or mean (±SD).

  2. ASI = Addiction Severity Index; BPN = buprenorphine; CD4 = CD4 count, cells/μL; DAART = directly administered antiretroviral therapy; HRQoL = health-related quality of life; PTSD = posttraumatic stress disorder; VL = viral load.

Mean age, years45.70 (±7.13)45.23 (±6.32)45.56 (±6.88)0.975
Male92 (85.9)31 (70.5)123 (81.5)0.026
Female15 (14.0)13 (29.6)28 (18.5)
Mean years education10.71 (±2.16)10.89 (±1.92)10.76 (±2.09)0.703
Non-Hispanic white11 (10.3)9 (20.5)20 (13.3)0.212
Non-Hispanic black58 (54.2)23 (52.3)81 (53.6)
Hispanic38 (35.5)12 (27.3)50 (33.1)
Relationship status
Married8 (7.69)5 (12.2)13 (8.97)0.690
Widowed/divorced33 (31.7)12 (29.3)45 (31.0)
Never married63 (60.6)24 (58.5)87 (60.0)
Anticipated housing status
Homeless20 (19.6)9 (20.5)29 (19.9)0.136
Temporarily housed60 (58.8)19 (43.2)79 (54.1)
Permanently housed22 (21.6)16 (36.4)38 (26.0)
Lifetime abuse46 (43.0)23 (52.3)69 (45.7)0.298
Study involvement
No prerelease discharge planning11 (10.3)13 (29.5)24 (15.9)0.003
Randomized to DAART69 (64.5)31 (70.5)100 (66.2)0.481
Site of residence
Hartford 55 (51.4)21 (47.7)76 (50.3)0.682
New Haven52 (48.6)23 (52.3)75 (49.7)
Mean months retained in parent study 4.68 (±2.09)4.91 (±2.12)4.75 (±2.09)0.506
Mean baseline CD4 count401.6 (±240.2)360.7 (±234.9)389.7 (±238.6)0.851
Baseline CD4 count <20023 (21.5)12 (27.3)35 (23.2)0.445
HIV-1 RNA viral suppression
Baseline87 (82.9)31 (72.1)118 (79.7)0.139
12-weeks 70 (65.4)22 (50.0)92 (60.9)0.078
Hazardous drinking 39 (37.1)19 (43.2)58 (38.9)0.490
Mean lifetime alcohol ASI score0.23 (±0.13)0.25 (±0.15)0.24 (±0.14)0.772
Mean lifetime drug ASI score 0.17 (±0.19)0.16 (±0.22)0.17 (±0.20)0.938
Prescribed BPN35 (33.3)15 (34.1)50 (33.6)0.929
BPN retention at 6 months, n = 5026 (74.3), n = 3514 (93.3), n =1540 (80.0), n = 500.123
Injection drug use 73(75.3)32(74.4)105(75.0)0.916
Major depression 33 (32.4)18 (40.9)51 (34.9)0.320
Suicide attempt 32 (30.8)16 (38.1)48 (32.9)0.394
Median Social Support score66.0 (49.5–83.5)68.0 (48.5–89.5)66.5 (49.5–86.5)0.611
Median Trust in Physician score68.2 (63.6–70.9)67.3 (63.6–70.9)67.3 (63.6–70.9)0.500
Median physical HRQoL52.0 (39.7–56.1)43.4 (30.5–51.7)49.7 (36.0–55.3)0.693
Median mental HRQoL50.0 (34.6–57.2)45.1 (31.8–58.2)48.7 (33.4–57.4)0.693
Median HIV stigma score104.4 (92.5–115.0)103.0 (92.0–111.5)104.0 (92.0–114.0)0.097
Medical history
Seizures5 (4.85)8 (18.6)13 (8.90)0.008
Hypertension33 (32.0)9 (20.9)42 (28.8)0.177
Asthma18 (17.5)17 (39.5)35 (24.0)0.004
Tuberculosis6 (5.83)7 (16.3)13 (8.90)0.043
Pneumonia23 (22.3)21 (48.8)44 (30.1)0.001
Diabetes9 (9.09)9 (9.30)13 (8.90)0.968
Hepatitis80 (79.2)32 (74.4)112 (77.8)0.527
Cirrhosis 17 (16.5)7 (15.9)24 (16.3)0.929
Abnormal Pap smear7 (22.6)6 (33.3)13 (26.5)0.411
Arthritis21 (20.4)13 (29.6)34 (23.1)0.228
Migraine headaches15 (14.6)13 (29.6)28 (19.1)0.034
Depression51 (49.5)27 (61.4)78 (53.1)0.187
Anxiety disorder31 (30.1)18 (40.9)49 (33.3)0.203
Schizophrenia10 (9.62)5 (11.4)15 (10.1)0.747
PTSD 9 (8.65)4 (9.09)13 (8.78)0.932
Dyslipidemia11 (10.6)4 (9.09)15 (10.1)0.784

Forty-four of 151 subjects (29%) visited the ED twice or more during the year following prison release and were considered frequent ED users. Frequent ED users were more likely than infrequent or nonusers to be female and have medical histories of seizures, asthma, tuberculosis, pneumonia, and migraine headaches. Frequent ED users were also less likely to have received prerelease discharge planning, although there was no association between receipt of prerelease discharge planning and measure of physical HRQoL (OR = 0.98, 95% CI = 0.95 to 1.02), mental HRQoL (OR = 0.99, 95% CI = 0.96 to 1.02), or addiction severity (data not shown.) Participants who did not receive prerelease discharge planning were more likely than those who did to be non-Hispanic white (p = 0.02) and have a lower mean (± SD) baseline CD4 count (319 ± 22.9 vs. 403 ± 241; p = 0.01). Despite achieving parity with other participants in terms of baseline viral suppression, only half of frequent ED users maintained viral suppression by 12-week follow-up, suggesting reduced short-term persistence on ART and/or interval development of genotypic drug resistance.[51]

ED and Other Health Care Use

There were 227 total ED visits made by 151 released prisoners with HIV (1.9 ED visits per person year) who were enrolled in the parent study in the 12 months after release from prison (Table 2). Nearly one-fourth of these visits resulted in hospital admission and almost the same proportion were considered incomplete, because the patient either left the ED before being seen by a medical provider or left against medical advice. The majority of visits occurred during standard hours of typical primary care clinic operation (9 a.m. to 5 p.m.), even though 68% of subjects self-reported having established at least minimal care in a primary HIV clinic by 3 months postrelease. In contrast to this pattern of ED use, frequent ED users' visits were more likely to take place during the off hours of 5 p.m. to 12 a.m. and 12 a.m. to 9 a.m. Frequent ED users' reason for ED visit differed significantly from infrequent or nonusers: substance misuse diagnosis was more common among frequent ED users (43% vs. 8%; p < 0.001).

Table 2. Healthcare Utilization, Stratified by Number of ED Visits
Variable<2 ED Visits (n = 107)≥2 ED Visits (n = 44)Total Cohort (n = 151)p-value
  1. a

    Includes visits in which the patient left before being seen by an ED provider or left against medical advice.

Total number ED visits42185227 
Mean number ED visits per person-year, mean (±SD)0.5 (±0.8)5.2 (±5.9) 1.9 (±3.9) 
Proportion of ED visits resulting hospitalization31.022.724.20.260
Proportion of incomplete ED visitsa
Mean length of ED stay (minutes), mean (±SD)334.7 (±273.0)275.1(±146.5)297.9 (±204.3)0.276
Proportion of visits during
9 a.m.–5 p.m.53.777.365.90.022
5 p.m.–12 a.m.19.559.140.0<0.001
12 a.m.–9 a.m.24.468.247.1<0.001
Proportion having seen an HIV provider within 3 months67.968.868.10.930
Proportion admitted for detox within 3 months2.506.33.60.334
Any ED visit for substance misuse, n (%)9 (8.4)19 (43.2)28 (18.5)<0.001

Correlates of Frequent ED Use

Using multivariate Poisson regression models, correlates of being a frequent ED user included not receiving prerelease discharge planning and having poorer baseline physical HRQoL (Table 3). Specifically, subjects recruited after release from prison and who had not received prerelease discharge planning were more than three times more likely to frequently use the ED compared to those who were assessed and had adequate time to plan for postrelease activities during the period of incarceration.

Table 3. Correlates of Having Two or More ED Visits (n = 151)
VariableUnadjusted OR (95% CI)Adjusted OR (95% CI)a
  1. For categorical variables, referent is listed; for continuous variables, 0 is the referent group; for dichotomous variables, “No” is the referent group.

  2. Unadjusted models performed using univariate Poisson regression; adjusted models performed using multivariate Poisson regression.

  3. HRQoL = health-related quality of life

  4. a

    Adjusted for all covariates having statistically significant univariate associations, defined as p < 0.2, with frequent ED use.

Female2.57 (1.10–6.00)2.07 (0.65–6.57)
Non-Hispanic whiteReferentReferent
Non-Hispanic black0.48 (0.18–1.32)0.71 (0.18–2.81)
Hispanic0.39 (0.13–1.15)0.54 (0.12–2.33)
No prerelease discharge planning3.66 (1.49–8.99)3.16 (1.08–9.26)
HIV-1 RNA viral suppression
Baseline 0.53 (0.23–1.24)1.16 (0.38–3.50)
12-weeks 0.53 (0.26–1.08)0.77 (0.31–1.92)
Physical HRQoL0.95 (0.93–0.98)0.95 (0.92–0.99)
Medical comorbidities
Seizure disorder4.48 (1.37–14.61)1.86 (0.44–7.83)
Hypertension0.56 (0.24–1.30)0.51 (0.19–1.38)
Asthma3.09 (1.39–6.84)1.64 (0.57–4.70)
Tuberculosis3.14 (0.99–9.98)2.08 (0.51–8.44)
Pneumonia3.32 (1.56–7.08)1.77 (0.69–4.51)
Migraines2.46 (1.05–5.74)1.67 (0.54–5.11)
Depression1.62 (0.79–3.32)0.81 (0.32–2.06)
Akaike's Information Criterion 97.935
Bayesian Information Criterion 137.780

When we plotted number of ED visits against calculated medical and social multimorbidity indices, the relationships were not found to be significant (data not shown.) As shown in Figure 1, however, we did find a positive relationship between the number of ED visits and increasing levels of “psychiatric multimorbidity” for frequent ED users.

Figure 1.

ED utilization by psychiatric multimorbidity index, stratified by frequent ED use.


While studies by Hunt et al.[8] and Byrne et al.[52] have addressed frequent ED use as a general construct, to our knowledge this is the first study of frequent ED use by released prisoners and specifically those with HIV. Released prisoners and those living with HIV are an important subpopulation at risk of high ED utilization. In describing frequent ED use by this subpopulation, we build on our previous findings that many people with HIV visit the ED once within 30 days of prison release, often for untreated mental disorders and social access issues.[53] Similarly, previous studies of other vulnerable populations have shown that frequent ED use represents a type of repetitive health-seeking behavior that is often reflective of severe underlying medical and psychiatric illness in the context of social instability. Hunt et al. used the Community Tracking Study Household Survey to show that frequent ED users typically were poor, yet over 80% possessed both insurance and a usual source of care; like our findings, Hunt et al. found frequent ED use to be independently associated with poor physical and mental health.[8] In both studies, although far from being “inappropriate” or even unnecessary, frequent ED use by a small subset of individuals contributes significantly to ED use overall.

In a cohort with significant medical and psychiatric comorbidity, we found important sex differences, in that women were consistently more likely to be frequent ED users, compared to men. While mental health–related ED visits share a similar overall prevalence between the sexes, women are more regular users of health care services in general, with higher rates of psychosocial-related ED visits, in particular.[54] Previous data suggest that women who interact with the criminal justice system differ from their male counterparts in that women experience a higher burden of medical, psychiatric, and substance use disorders.[55] Following release from prison, women's medical and psychiatric needs may be compounded by specific psychosocial stressors such as unequal relationship power dynamics and challenges caring for children and reuniting with families.[56, 57] Women's excess burden of illness and stress following prison release is likely reflected in frequent ED use, although this remains an important area for future research.

For both men and women, the transition period from correctional settings to communities is an important window for intervening with an otherwise neglected population of people living with HIV.[14, 58, 59] To maintain the benefits of prison- or jail-based ART, continuity of HIV care is critical.[60] Unfortunately, care is often disrupted, resulting in decreased engagement in HIV primary care and reduced persistence on ART.[18, 20, 21, 60-62] In the often chaotic lives of recently released prisoners, energy is diverted away from positive health promotion activities and toward meeting basic subsistence needs and distractions from coping with the demands of drug or alcohol addiction. This element of chaos is reflected in frequent off-hour ED visits, often for issues related to substance misuse. Furthermore, the cumulative burden of psychiatric illness is represented by the psychiatric multimorbidity index, which was positively correlated with increased number of ED visits, especially for frequent ED users. This is especially critical in a population with few other positive resources and serious underlying medical and psychiatric illness.

For the population of released prisoners described here, frequent ED use is correlated with mental disorders and serious medical comorbidities and is compounded by a lack of prerelease discharge planning during incarceration. Prerelease discharge planning has been associated with increased retention in HIV care among individuals with HIV released from jail across 10 diverse U.S. sites.[63] In our study, while all participants who were recruited before release met with transitional case managers during incarceration and by our team the day of release, the group who did not receive prerelease discharge planning was released from prison directly to communities without any planning for continuity of care. For these individuals, an important opportunity for linkage to community-based care immediately after release was missed, but might have otherwise stabilized health conditions or reduced an overreliance on ED services. Findings from our study complement those from a randomized controlled trial of HIV-infected prisoners in North Carolina, who received either prerelease discharge planning or strengths-based case management; linkage to care and ED use was similar between the two groups.[64]

Our study, through its naturalistic design, confirms that prerelease discharge planning reduces ED use and is the minimum required to address basic needs and effectively transition people with HIV to the community.[14] Our findings also suggest that there may be gaps in community medical and mental health resources for newly released prisoners that translate directly into frequent ED use. Other proposed interventions aimed primarily at reducing frequent ED use have included expansion of case management services within EDs and ED diversion toward expanded primary care programs.[65] For the most part, these interventions have had mixed results,[10] perhaps because they miss critical contextual interactions between individuals and communities.[9] For released prisoners with HIV and other marginalized patient populations, more meaningful structural interventions that increase availability of or facilitate entry into evidence-based substance abuse treatment, stabilize housing, and maintain continuity of care between prison and communities will secondarily reduce the need for ED services. While some components of these interventions were undertaken in the parent DAART study described here, our observational cohort study design prohibited us from evaluating the specific effect of the DAART intervention on reducing frequent ED use.

Other studies have shown that additional beneficial services for this population are short, brief interventions and referral to treatment, evidence-based interventions that reduce the negative consequences of alcohol and drug use.[66] Support for substance abuse treatment as a means of reducing ED use is evidenced by a recent study of a similar urban population accessing mobile health care services, in which each month of retention on buprenorphine treatment was associated with a 17% reduction in expected ED use per person.[67] These findings were not replicated here, perhaps due to the small number of participants who met criteria for opioid dependence and initiated buprenorphine. Our study suggests overall that there may be gaps in community medical and mental health resources for newly released prisoners that translate directly into frequent ED use.


Despite every attempt to measure the full spectrum of ED use by subjects in the year following prison release, visits may have been inadvertently missed if they occurred outside of designated study areas. Although other data from this university community suggest that ED use in the two designated hospitals is associated with the vast majority of all ED visits, we cannot fully ensure that it is the case here.[66, 68] We were also unable to control for reincarceration during follow-up as a possible confounder to ED use. Subjects may have had somewhat restricted ED use during periods of reincarceration, which is representative of the cyclical health care use common among those who interface with the criminal justice system. This secondary analysis was limited by the use of baseline and some 3-month data, which cannot account for changes in housing stability or physical/mental health over time, reflecting instead only the period immediately before or after prison release. External validity of results may be challenged by regional variations in insurance coverage, drug use patterns, or support services available for HIV-infected released prisoners. Although the multimorbidity indices were designed to be comprehensive, items may have been missing from the baseline assessment from which the indices were derived. Of note, weighted index scores could not be calculated because limited data are available on the effect of multimorbidity on health care use in this patient population. We did approximate weight by including measures of both the presence and the severity of various medical, psychiatric, and social issues in each respective index.


In this study, we explored correlates of frequent ED use by 151 released prisoners with HIV, using individual descriptors, multivariate Poisson regression models, and generated multimorbidity indices. Frequent ED users were more likely than infrequent users or nonusers to be female, have significant medical and psychiatric comorbidities including substance misuse, and report worse physical HRQoL. They were no different, however, in terms of HIV biologic surrogates and use of other health care services either at baseline or in short-term follow-up. The cumulative burden of mental disorders experienced by frequent users was best captured in a psychiatric multimorbidity index, which was positively associated with ED utilization. Surprisingly, and in contrast to other published studies on frequent ED use, social multimorbidity indices were not correlated with number of ED visits. Perhaps most pertinent for health policy and intervention development, frequent ED users were less likely to have received prerelease discharge planning, suggesting missed opportunities for seamless linkages to care.

The authors thank the study participants and Project CONNECT research assistants and substance abuse counselors for their tireless work with and insight about program participants.