ADD5107: Week 8 Discussion: Evidence-Based Approaches and Co-Occurring Disorders

ORIGINAL ARTICLE

High inpatient utilization among Veterans Health Administration patients with substance-use disorders and co-occurring mental health conditions Janelle M. Painter, PhDa, Carol A. Malte, MSWb,c, Anna D. Rubinsky, PhDb,c,d, Timothy R. Campellone, PhDa,e, Amanda K. Gilmore, PhDf, John S. Baer, PhDb,c,g, and Eric J. Hawkins, PhDb,c,h

aVA Puget Sound Health Care System, Seattle Division, Seattle, WA, USA; bHealth Services Research & Development (HSR&D) Seattle Center of Innovation for Veteran Centered and Value-Driven Care, VA Puget Sound Health Care System, Seattle, WA, USA; cCenter of Excellence for Substance Abuse Treatment and Education (CESATE), VA Puget Sound Health Care System, Seattle, WA, USA; dKidney Health Research Collaborative (KHRC), Department of Medicine, University of California, San Francisco, and VA San Francisco Health Care System, San Francisco, CA, USA; eDepartment of Psychology, University of California, Berkeley, Berkeley, CA, USA; fDepartment of Psychiatry & Behavioral Science, Medical University of South Carolina, National Crime Victims Research and Treatment Center, Charleston, SC, USA; gDepartment of Psychology, University of Washington, Seattle, WA, USA; hDepartment of Psychiatry & Behavioral Sciences, University of Washington, Seattle, WA, USA

ABSTRACT Background. Substance-use disorders (SUDs) are common and costly conditions. Understanding high inpatient utilization (HIU) among patients with SUD can inform the development of treatment approaches designed to reduce healthcare expenditures and improve service quality. Objectives. To examine the prevalence, type, and predictors of HIU among patients with SUD and co-occurringmental health conditions. Methods. Service utilization and demographic and clinical variables were extracted from a national sample of Veterans Health Administration (VA) patients with SUD-only [n = 148,960 (98.3% male)], SUD plus serious mental illness ([i.e. schizophrenia- and/or bipolar-spectrum disorders; SUD/SMI;n=75,913 (91.6%male)], and SUDplus othermental illness [SUD/MI;n=245,675 (94.6%male)]. Regressionmodels were used to examine HIU during a follow-up year. Results. Prevalence of HIU among the SUD-only group was 6.2% (95% confidence interval (CI): 6.1%–6.3%) compared with 22.7% (95% CI: 22.4%–23.0%) and 9.7% (95% CI: 9.6%–9.8%) among the SUD/SMI and SUD/MI groups, respectively. Patients with SUD/MI represented nearly half of the HIU sample. Primary type of inpatient service use varied by comorbidity: SUD-only = medicine; SUD/SMI = psychiatric; SUD/MI similar use of psychiatric, SUD-related, andmedicine. Predictors of HIUwere generally similar across groups: older age, unmarried, homelessness, suicide risk, pain diagnosis, alcohol/opioid/sedative-use disorders, and prior-year emer- gency department/inpatient utilization. Conclusions. Substantial reductions in HIU among an SUD population will likely require treatment approaches that target patients with less-severe mental health conditions in addition to SMI. Cross-service collaborations (e.g., integration of medical providers in SUD care) and interventions designed to target issues and/or conditions that lead to HIU (e.g., homeless care services) may be critical to reducing HIU in this population.

ARTICLE HISTORY Received 4 May 2017 Revised 6 September 2017 Accepted 9 September 2017

KEYWORDS Veterans; high inpatient utilization; substance-use disorders; mental illness

Introduction

Substance-use disorders (SUDs) are common and costly conditions accounting for a large percentage of the overall global disease burden (1,2). For a subset of patients, SUDs are chronic, severe, and recurrent conditions that are com- plicated to treat (3). Such patients are characterized by high rates of acute care admissions and readmissions (4), recur- ring relapse (5,6), mental health and medical comorbidity (7,8), and complicating social conditions such as home- lessness (9). Recent studies suggest that this subset of patients returns to treatment multiple times and uses dis- proportionate substance use, mental health, and medical services (10).

Evaluations of healthcare utilization consistently indi- cate that 5% of the United States (US) population accounts for 50% of the healthcare costs (11). With such large expenditures devoted to relatively few individuals,managed care systems have increased their focus on methods for identifying individuals at risk for high utilization and pro- viding access to appropriate levels of less-costly care (12,13). Inpatient services and conditions associatedwith high inpa- tient utilization (HIU) are often the focus of such efforts, as hospital stays account for nearly one-third of all healthcare expenditures in the US (14). However, little is known about HIU patterns, and factors associated with those patterns, among patients with SUDs. It is imperative to better

CONTACT Janelle M. Painter [email protected] VA Puget Sound Health Care System, 1660 S. Columbian Way (S-116-PRRC), Seattle, WA 98108, USA.

Supplemental data for this article can be accessed on the publisher’s website.

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 2018, VOL. 44, NO. 3, 386–394 https://doi.org/10.1080/00952990.2017.1381701

© 2018 Taylor & Francis

identify those at risk ofHIU to inform the development and delivery of services that more effectively address the unique needs of these patients and reduce hospitalizations.

Recent studies among Medicaid patients with HIU suggest that psychotic-, mood-, and alcohol-related dis- orders are among the most common conditions related to hospital stays (15) and readmissions (16). Furthermore, serious mental illness (SMI), chronic mental health conditions that result in substantial func- tional impairment, and less-chronic and impairing mental health conditions (e.g., depression, anxiety dis- orders) are prevalent among the SUD populations (17– 19). Although it is likely that a large proportion of patients with SUD and HIU have co-occurring mental health disorders, little information exists on whether the rates or predictors of HIU vary among patients with SUD depending on mental health comorbidity.

As the country’s largest integrated healthcare delivery system and one of the principal providers of substance- use treatment (20), the Veterans Health Administration (VA) provides a unique opportunity to examine the HIU patterns of patients with SUDs, depending on the pre- sence of co-occurring mental health conditions. The pre- sent study extends our understanding of inpatient service utilization among patients with SUD and co-occurring mental health conditions by specifically examining whether the prevalence rates and predictors of HIU differ among those with SUD-only, SUD and co-occurring ser- ious mental illness (SUD/SMI), and SUD and other co- occurring mental health disorders (SUD/MI). The aims of this study include: (1) estimating the unadjusted annual prevalence and 95% confidence intervals (CI) of HIU among VA patients with SUD-only, SUD/SMI, and SUD/MI; (2) determining whether VA patients with SUD/MI and SUD/SMI are at increased odds of HIU compared with those with SUD-only; (3) describing the patterns of utilization across inpatient service type among VA patients with SUD and HIU and determine whether utilization differs by mental health comorbidity; and (4) given potential differences in symptom severity and func- tioning across these groups, identifying demographic and clinical characteristics associated with HIU among VA patients with SUD-only, SUD/SMI, and SUD/MI.

Methods

Source of data and study sample

This study used administrative medical records data from the VA Informatics and Computing Infrastructure (VINCI), a national data repository for all VA facilities that includes patient-level data on VA service utilization, as well as demographics and clinical diagnoses. VA patients

aged 18 or older were eligible for study inclusion if they had a documented primary or secondary diagnosis of SUD from at least one outpatient visit or inpatient discharge at a VA facility between 1 October 2011 and 30 September 2012. Given that SUDs are often associated with recurring relapse, patients with current or remitted SUD diagnoses were identified from Internal Classification of Diseases, Ninth Revision Clinical Modification (ICD-9-CM) codes for alcohol-use disorders and drug-use disorders (see Table 1 for a list of diagnostic codes).

To assess HIU, patients were followed for 12 months (hereafter referred to as the SUD follow-up year) following their initial SUD diagnosis (SUD index diagnosis). They were excluded if they died during the SUD follow-up year to ensure that all patients included in this sample had equal opportunity for HIU. Study approval was obtained from the VA Puget Sound Institutional Review Board.

High inpatient utilization

Total number of days inpatient during the SUD follow-up year was classified into three mutually exclusive categories (medical, psychiatric, and SUD-related (e.g., detoxification) services) and combined to create a total inpatient utilization variable. Days inpatient was determined from inpatient location codes and diagnosis-related group (DRG) codes and did not include residential treatment stays. HIU was defined as the upper 10% of the study sample in terms of total inpatient utilization distribution and was coded into a binary measure. While various thresholds for identifying “high utilization” have been used, the top 10% is common (21,22).

SUD and co-occurring mental health condition classifications

Patients were categorized into one of three groups: SUD-only, SUD/SMI, and SUD/MI. Co-occurring mental health conditions were determined from pri-

Table 1. ICD-9 codes by diagnostic group. Group ICD-9 Code

Substance-Use Disorders

Alcohol: 303.0x, 303.9x, 305.0x Stimulant: 304.2x, 304.4x, 305.6x, 305.7x Cannabis: 304.3x, 305.2x Opioid: 304.0x, 304.7x, 305.5x Sedative: 304.1x, 305.4x Other: 304.5x, 304.6x, 304.9x, 305.3x, 305.8x, 305.9x

Severe Mental Illness

Schizophrenia-spectrum disorders: 295.0X-295.4X, 295.6X-295.9X, 297.X-298.X Bipolar-spectrum disorders: 296.0X-296.1X, 296.4X- 296.6X, 296.7, 296.8, 296.81, 296.89

Other Mental Illness

Depressive disorders: 296.2X-296.3X, 296.82, 300.4, 311 Anxiety disorders: 300.0X, 300.2X, 300.3 PTSD: 309.81 Eating disorders: 307.1, 307.5, 307.51, 307.59

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 387

mary or secondary ICD-9-CM codes (see Table 1) documented from at least one outpatient visit or inpatient discharge one year before to 90 days after the SUD index diagnosis. This timeframe was created to identify chronic conditions that may not have been documented in the prior year (e.g., patient was new to VA care) and likely influenced inpatient utilization during the SUD follow-up year. As there is general consensus that psychotic (e.g., schizophrenia, schi- zoaffective disorders) and bipolar disorders are among the most impairing mental illnesses (23,24), patients with schizophrenia-spectrum or bipolar-spec- trum disorders were assigned to the SUD/SMI group. Patients with posttraumatic stress disorder (PTSD), depressive, anxiety, or eating disorders were assigned to the SUD/MI group, and those without documenta- tion of these mental health disorders were assigned to the SUD-only group.

Potential predictors of HIU

We used the Gelberg–Andersen Behavioral Model for Vulnerable Populations (25) as a conceptual framework and prior studies (4,15,16,26–29) to identify the poten- tial predictors of HIU among patients within the three groups. Predictors were selected from the following three domains: predisposing characteristics (i.e., non- modifiable variables), enabling resources (i.e., modifi- able variables), and perceived/evaluated need (i.e., patient experiences that might influence inpatient uti- lization). Predictors were identified from administrative data one year prior and 90 days after the SUD index diagnosis unless otherwise specified for the same reason as described above.

Predisposing factors included age, gender, race, eth- nicity, engagement in Operation Enduring Freedom/ Operation Iraqi Freedom (OEF/OIF) conflicts, and VA service-connected disability rating >50% (i.e., an injury or illness incurred or aggravated during active military service and determines VA healthcare eligibil- ity and benefits) at the time of the SUD index diagnosis.

Enabling resources included marital status at the time of the SUD index diagnosis and housing status. Housing status was a binary measure based on the absence or presence of homelessness as determined by at least one of the following indicators: 1) ICD-9-CM code v60.0 indicating lack of housing, and 2) VA out- patient clinic codes and/or residential treatment speci- alty codes indicating use of homeless services (30).

Perceived and evaluated need factors included sui- cide risk, medical comorbidity, pain diagnosis, specific SUD diagnoses, combined alcohol and drug use, specific

mental health diagnoses, and prior-year service use. Suicide risk was determined by at least one of the follow- ing indicators to calculate a binary measure: 1) patient record flags documenting high suicide risk and/or 2) ICD-9-CM codes documenting a suicide attempt (31). Medical comorbidity was measured via a modified ver- sion (32) of the Charlson Comorbidity Index (CCI, 33) by creating a weighted score from 17 health conditions determined by ICD-9-CM codes. For the purposes of this study, CCI scores were assigned to three groups: 0, 1, 2 or greater, with higher scores reflecting greater comorbidity severity. A binary indicator of a pain diag- nosis was determined by at least one ICD-9-CM diag- nostic code indicating a chronic pain diagnosis (34). Specific substance use and mental health disorder diag- noses were identified using ICD-9-CM diagnostic codes (see Table 1). Combined alcohol and drug-use disorder was coded for any patient with at least one alcohol-use disorder and one drug-use disorder diagnosis to create a binary measure. Three binary measures of prior-year service use were created based on any primary care, outpatient mental health (including SUD specialty ser- vices), or emergency department use obtained from administrative data in the 12 months prior to the SUD index diagnosis. A measure of prior-year inpatient service use was based on the number of days inpatient in the 12 months prior to the SUD index diagnosis and cate- gorized into no days, 1–7 days, and 8+ days.

Data analysis

The annual prevalence of HIU was estimated across SUD groups with simple proportions and 95% CI. The odds (and 95% CIs) of HIU among the SUD/ SMI and SUD/MI groups compared with the SUD- only group and among the SUD/SMI group compared with the SUD/MI group were estimated using unad- justed logistic regression. The number of inpatient days by specific service type and overall (medicine, psychiatric, SUD-related, and total) was compared in the HIU sample using unadjusted negative binomial regression models and estimated with incidence rate ratios and 95% CIs to describe the patterns in inpa- tient utilization (35). Among all the SUD patients, separate multivariable logistic regression models were used to identify the characteristics associated with HIU in the SUD-only, SUD/SMI, and SUD/MI groups and estimate the adjusted odds ratios and 95% CIs.

All multivariable logistic regression models included age, gender, race, ethnicity, OEF/OIF service, VA ser- vice-connected disability rating >50%, marital status, homelessness, suicide risk, CCI score, pain diagnosis,

388 J. M. PAINTER ET AL.

specific SUD diagnoses, specific mental health diag- noses, combined alcohol and drug use, prior-year pri- mary care, mental health, emergency department visits, and prior-year days inpatient. Logistic and negative binomial regressions were modeled with cluster robust estimates of variance to account for clustering/correla- tion between observations at the VA facility level. Because small differences can be statistically significant in large data sets, a p-value of <0.001 was the threshold for statistical significance in the analyses.

Results

Patient characteristics

We identified 470,548 VA patients eligible for this study: 31.7% (n = 148,960) met the criteria for the SUD-only group, 16.1% (n = 75,913) met the criteria for the SUD/SMI group, and 52.2% (n = 245,675) met the criteria for the SUD/MI group. Table 2 summarizes the measured sociodemographic and clinical character- istics and prior-year service utilization for the three patient groups. Most patients were male, 50 years or older, non-Hispanic, white, unmarried, had a docu- mented alcohol-use disorder, and had a pain diagnosis. Of note, over one-third of the SUD/SMI group was homeless, nearly one-half used both alcohol and drugs, and had at least one ED visit in the year prior, and more than one-quarter was hospitalized in the year prior to the SUD index diagnosis.

Prevalence and odds of high inpatient utilization

Overall, the mean number of inpatient days during the SUD follow-up year was 2.9 (SD = 10.6), with 1.2 (5.8), 1.1 (7.3), and 0.7 (4.1) mean days in medicine, psychia- tric, and SUD-related services, respectively. The top 10% of VA patients with respect to inpatient utilization had 8 or more inpatient days, which was used as the cut-point to identify patients with HIU in the SUD follow-up year.

The prevalence of HIU among the SUD-only group was 6.2% (95% CI: 6.1%–6.3%) compared with 22.7% (95% CI: 22.4%–23.0%) among the SUD/SMI group and 9.7% (95% CI: 9.6%–9.8%) among the SUD/MI group. Compared with VA patients with SUD-only, VA patients with SUD/SMI (OR = 4.44; 95% CI: 4.18–4.73) and SUD/MI (OR = 1.63; 95% CI: 1.54– 1.72) had increased odds of HIU. Compared with the SUD/MI group, the SUD/SMI group had increased odds of HIU (OR = 2.73; 95% CI: 2.63–2.84).

Patterns of high inpatient utilization by service type

A total 50,212 VA patients were classified as high inpatient utilizers [SUD-only = 9,211 (18.34%); SUD/ SMI = 17,200 (34.25%); SUD/MI = 23,801 (47.40%)]. Table 3 summarizes the number of days of specific inpatient service use during the SUD follow-up year among the three HIU groups. Overall, the primary type of inpatient service use varied by comorbidity with the SUD-only group primarily using medicine, the SUD/ SMI group primarily using psychiatric, and the SUD/ MI group using psychiatric, SUD-related, and medicine services to a similar degree.

Predictors of high inpatient utilization by group

The results of the multivariable logistic regression mod- els for each patient group are shown in Table 4. For all three groups, the following characteristics were asso- ciated with greater odds of HIU: older age, divorced/ separated/never married (compared with married), homelessness, suicide risk, one or more medical comor- bidities, pain diagnosis, alcohol use, opioid use- and sedative-use disorders, at least one prior-year emer- gency department visit, and prior-year inpatient service use. Having a primary-care or mental healthcare visit in the year prior to the index SUD diagnosis appears protective for all groups, as evidenced by the lower odds of HIU.

VA service connection disability rating >50% was associated with greater odds of HIU for the SUD-only and SUD/SMI groups. Stimulant-use disorders, other SUDs, and depression were associated with greater odds of HIU for the SUD/SMI and SUD/MI groups. Racial differences also emerged for the SUD/SMI and SUD/MI groups, with black patients having lower odds of HIU than white patients. For the SUD/SMI group, OEF/OIF service and cannabis-use disorders were asso- ciated with greater odds of HIU. Although associated with greater odds of HIU in the SUD/MI group, PTSD and anxiety were not associated with greater odds of HIU in the SUD/SMI group. For the SUD/MI group, patients with combined alcohol and drug-use disorders had greater odds of HIU.

Discussion

This study takes an important first step in describing HIU among patients with SUD and mental health comorbidity. We identified differences in where care is received and examined the predictors of HIU. Although the prevalence of HIU was markedly higher

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 389

among the SUD/SMI group, with nearly one in four patients meeting the HIU criteria in a one-year period, the SUD/MI group, due to its larger size, represented the majority of the high inpatient utilizers with SUD. Furthermore, the type of inpatient service use varied by

group, with the SUD-only group primarily utilizing medicine services, the SUD/SMI group primarily utiliz- ing psychiatric services, and the SUD/MI group utiliz- ing medicine, psychiatric, and SUD-related inpatient services equally. Factors associated with HIU were similar across the SUD comorbidity groups, with few characteristics uniquely predicting HIU in specific groups. Together, these findings suggest that different approaches to service development and implementation may be necessary to reduce HIU among those with SUD and mental health comorbidity.

Consistent with prior literature on the utilization and costs associated with patients with SUD and mental health concerns (4,27–29,36), those with SUD and co- occurring SMI or other mental health conditions repre- sented nearly 82% of HIU in our sample. Of note, the SUD/MI group, which is typically considered less func- tionally impaired than the SUD/SMI population, accounted for the largest proportion of HIU (47%). Thus, efforts to substantially reduce HIU among VA patients with SUD will require approaches to care that identify and treat the needs of patients with less-severe mental health conditions. Furthermore, findings sug- gest that outreach efforts, coordination of services, and collaborations with VA specialty mental health and primary-care clinics, where many patients with co- occurring SUD and other mental health conditions are treated, are necessary to ensure that patients with SUD at risk for HIU are linked to appropriate services.

Comprehensive care management programs are designed to provide proactive, inclusive, and collabora- tive care across multiple services to improve service quality and lower costs associated with unnecessary acute care. However, findings regarding the efficacy of care management models for patients with SUD are mixed (37–39), which may be related to the diverse needs of this chronic and complex population. Our finding that the type of inpatient service use varied by SUD group provides greater clarity into where VA patients with SUD and co-occurring mental health con- ditions access acute care services and has implications for developing and locating care management approaches that are responsive to patient needs. Patients in the SUD/SMI group averaged over 16 times more psychiatric days hospitalized than patients in the SUD-only group and over two times more than those in the SUD/MI group, underscoring a need for care management interventions that foster collabora- tions across acute emergency, outpatient psychiatric and SUD services, and at least equal emphasis on the treatment of both mental health and substance-use symptoms in traditional outpatient SUD specialty care for this population in particular.

Table 2. National sample patient characteristics by diagnostic group*.

SUD only SUD/SMI SUD/MI

(n = 148,960) (n = 75,913) (n = 245,675)

n (%) n (%) n (%) Age, M(SD) 57.6 (11.9) 51.5 (11.5) 52.2 (13.3) Male 146,389 (98.3) 69,548 (91.6) 232,362 (94.6) Race White 95,020 (63.8) 48,634 (64.1) 165,747 (67.5) Black 39,130 (26.3) 20,659 (27.2) 55,621 (22.6) Other 5,698 (3.8) 4,652 (6.1) 11,506 (4.7) Missing/Not reported

9,112 (6.1) 1,968 (2.6) 12,801 (5.2)

Ethnicity Not Hispanic/Latino 127,552 (85.6) 67,399 (88.8) 214,337 (87.2) Hispanic/Latino 8,445 (5.7) 5,364 (7.1) 18,053 (7.4) Missing/Not reported

12,963 (8.7) 3,150 (4.2) 13,285 (5.4)

OEF/OIF Service 8,331 (5.6) 6,286 (8.3) 43,241 (17.6) >50% VA Service Connection

13,782 (9.3) 24,439 (32.2) 78,594 (32.0)

Marital Status Married 53,375 (35.8) 18,872 (24.9) 94,009 (38.3) Divorced/Separated 55,460 (37.2) 29,554 (38.9) 84,384 (34.4) Never married/Single 39,063 (26.2) 27,139 (35.8) 65,645 (26.7) Missing/Not reported

1,062 (0.7) 348 (0.5) 1,637 (0.7)

Homeless 19,476 (13.1) 25,691 (33.8) 48,830 (19.9) Suicide Risk 233 (0.2) 7,614 (10.0) 10,102 (4.1) Charlson Score 0 76,855 (51.6) 38,069 (50.2) 133,998 (54.5) 1 55,628 (37.3) 29,776 (39.2) 87,380 (35.6) 2+ 16,477 (11.1) 8,068 (10.6) 24,297 (9.9) Pain Diagnosis 88,305 (59.3) 54,026 (71.2) 177,799 (72.4) Substance-Use Disorders

Alcohol 120,802 (81.1) 57,379 (75.6) 196,057 (79.8) Cannabis 15,762 (10.6) 23,278 (30.7) 50,314 (20.5) Stimulant 19,932 (13.4) 26,633 (35.1) 46,264 (18.8) Opioid 10,325 (6.9) 11,046 (14.6) 29,458 (12.0) Sedative 742 (0.5) 3,567 (4.7) 6,818 (2.8) Other 22,025 (14.8) 28,488 (37.5) 58,200 (23.7) Alcohol + Drug Disorder

22,485 (15.1) 34,364 (45.3) 72,509 (29.5)

MI Diagnoses Depressive DO 0 (0.0) 15,245 (20.1) 185,651 (75.6) PTSD 0 (0.0) 26,671 (35.1) 120,531 (49.1) Anxiety DO 0 (0.0) 21,185 (27.9) 82,933 (33.8) Bipolar DO 0 (0.0) 45,171 (59.5) 0 (0.0) Psychotic DO 0 (0.0) 38,890 (51.2) 0 (0.0) Eating DO 0 (0.0) 357 (0.5) 608 (0.3) ≥1 YP PC Visit 105,926 (71.1) 64,194 (84.6) 205,492 (83.6) ≥1 YP MH Visit 31,931 (21.4) 66,433 (87.5) 177,996 (72.5) ≥1 YP ED Visit 35,780 (24.0) 35,006 (46.1) 81,217 (33.1) YP Days Inpatient None 137,998 (92.6) 54,488 (71.8) 211,493 (86.1) 1–7 Days 7,279 (4.9) 8,628 (11.4) 19,582 (8.0) 8+ Days 3,683 (2.5) 12,797 (16.9) 14,600 (5.9)

Note. SUD = substance-use disorder; SMI = serious mental illness; MI = mental illness; M = mean; SD = standard deviation; OEF/ OIF = Operation Enduring Freedom/Operation Iraqi Freedom; DO = disorder; PTSD = posttraumatic stress disorder; YP = year prior to SUD index diagnosis; PC = primary care; MH = mental health; ED = emergency department. *Group differences were significant at the p < 0.001 level for all variables.

390 J. M. PAINTER ET AL.

SUD-only patients’ use of inpatient medicine ser- vices points to a need for enhanced coordination between outpatient medical and SUD services. This approach to care is consistent with a recent VA direc- tive prompting the implementation of access to mental health services within primary-care settings (40). Although integrated primary care and mental health services appear helpful for connecting patients present- ing to primary care with mental health resources, this approach likely does not meet the clinical care needs of patients with chronic and severe conditions. Indeed, Trivedi and colleagues (41) found relatively little change in inpatient service use for veterans with SUD and SMI who received care in integrated primary care/ mental health settings compared with veterans with other mental health conditions that are more conducive to time-limited interventions (e.g., depression, anxiety). These findings suggest that integrating medical services within SUD specialty-care clinics and strengthening coordination between primary care and SUD speci- alty-care may be necessary to adequately serve the needs of this complex and chronic patient group.

This study expanded upon prior work by specifically identifying the risks factors for HIU among patients with SUD and SMI or other mental health comorbid- ities. Several of the factors associated with increased odds of HIU in this study are similar to those pre- viously reported in the mental health and substance- use literatures, including older age, being unmarried or separated, homelessness, suicide risk, chronic medical illnesses, alcohol- and opioid-use disorders, emergency department visits, and prior inpatient use (27– 29,36,42,43). Although not commonly reported in the substance-use literature, our findings that indicate pri- mary care and mental healthcare visits as protective factors for all groups are consistent with findings in the general medical literature (44), and support an underlying goal of care management approaches to promote the coordination of services that address the spectrum of patients’ needs (45).

Identified risk factors also point to key services such as acute crisis management, chronic pain services,

chronic disease management, and homeless care ser- vices as well as evidenced-based behavioral and phar- macotherapy treatments for SUD that should be considered when designing a program to decrease HIU among those with SUD. As stimulant-use disor- ders were associated with increased odds of HIU for the SUD/SMI and SUD/MI groups, evidence-based stimu- lant misuse interventions, such as contingency manage- ment, deserve specific consideration. In addition to being effective for reductions in cocaine, opiate, and poly-drug use (46), contingency management has also demonstrated efficacy in enhancing treatment atten- dance (47).

This study had several limitations. First, we were unable to assess the association between the timing or progression of SUD and risk factors and the likelihood of HIU (i.e., which came first or which might cause or exacerbate the other). Second, although we attempted to capture all SUD-related inpatient care by using a combination of bed section and DRG codes, it is pos- sible that we underestimated the SUD care received on medical and psychiatric inpatient units if multiple types of care were delivered. Third, there may be important characteristics that predict HIU among patients with SUD that were not available in VA administrative data (e.g., psychosocial stressors, barriers to treatment engagement). Similarly, the study did not have access to non-VA service utilization and, thus, we may have underestimated the prevalence of HIU in our sample. Fourth, this study relied on ICD-9-CM codes docu- mented in the medical records to classify diagnostic groups. Although this method of classification is sub- ject to error and conditions may go undiagnosed, this is a widely used approach to classifying VA administra- tive data and diagnoses are generally considered valid (48). Finally, our findings may not generalize to veter- ans receiving healthcare from other systems or to non- veteran populations.

Of note, this study examined inpatient utilization for only one year. HIU observed here may be due to acute medical and/or mental health conditions (e.g., relapse after a period of sustained sobriety) or external factors

Table 3. Inpatient utilization among high inpatient utilizers by diagnostic group*. SUD-only SUD/SMI SUD/MI (n = 9,211) (n = 17,200) (n = 23,801)

Mean (SD) Mean (SD) IRR (95% CI)a Mean (SD) IRR (95% CI)b IRR (95% CI)c

Total Days Inpatient 19.8 (18.9) 27.8 (30.2) 1.41 (1.37–1.44) 21.3 (19.9) 1.08 (1.05–1.10) 1.31 (1.28–1.33) Total Medicine Days 15.8 (18.2) 4.8 (12.2) 0.31 (0.29–0.32) 8.6 (15.7) 0.55 (0.53–0.57) 0.56 (0.53–0.58) Total Psychiatric Days 1.1 (7.7) 17.5 (28.4) 16.62 (14.28–19.34) 6.4 (13.8) 6.08 (5.22–7.07) 2.74 (2.64–2.84) Total SUD-Related Days 3.0 (7.7) 5.5 (12.2) 1.87 (1.75–1.99) 6.3 (11.6) 2.13 (2.01–2.25) 0.88 (0.84–0.91)

Note. SUD = substance-use disorders; SMI = serious mental illness; MI = mental illness; SD = standard deviation; IRR = unadjusted incidence rate ratio; CI = confidence interval; a SUD/SMI compared with SUD-only, b SUD/MI compared with SUD-only, c SUD/SMI compared with SUD/MI. *All group comparisons were significant at the p < 0.001 level.

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 391

such as unemployment or homelessness that, when addressed by traditional healthcare services, result in a return to less-intense utilization patterns. It remains unknown whether the VA patients identified in this sample are persistent or transient high utilizers. If tran- sient, they may benefit from more traditional outpati- ent services whereas persistent high utilizers may benefit from more comprehensive care management programs (13). Future studies could profitably explore whether the observed patterns in HIU among VA patients with SUD and co-occurring mental illness are stable over time, providing greater specificity in the identification of patients who will most likely benefit

from comprehensive care management programs aimed at improving care and reducing healthcare expenditures.

In sum, this study identified differences in the pre- valence rates of HIU and the type of inpatient utiliza- tion among VA patients with SUD depending on the presence and severity of co-occurring mental health conditions. Taken together, these findings point to the importance of engaging patients with SUD and less- severe mental health conditions in interventions designed to decrease HIU. Furthermore, differences in utilization across inpatient service type provide insights into where to focus outreach attempts and cross-service

Table 4. Characteristics associated with high inpatient utilization by diagnostic group. SUD Only SUD/SMI SUD/MI

(n = 148,960) (n = 75,913) (n = 245,675)

AOR (95% CI) AOR (95% CI) AOR (95% CI)

Age 1.01 (1.01–1.02)* 1.01 (1.00–1.01)* 1.01 (1.01–1.01)* Male 1.37 (1.13–1.67) 1.09 (1.01–1.19) 1.13 (1.04–1.23) Race White Black 1.02 (0.94–1.12) 0.88 (0.82–0.95)* 0.83 (0.77–0.90)* Other 1.19 (1.05–1.34) 1.04 (0.95–1.14) 0.98 (0.89–1.07) Missing/Not reported 0.78 (0.67–0.91) 0.78 (0.68–0.90)* 0.67 (0.60–0.74)* Ethnicity Not Hispanic/Latino Hispanic/Latino 1.05 (0.87–1.27) 0.93 (0.84–1.04) 0.90 (0.77–1.06) Missing/Not reported 0.72 (0.62–0.82)* 0.84 (0.76–0.93)* 0.84 (0.76–0.92)* OEF/OIF Service 0.76 (0.62–0.94) 1.27 (1.16–1.39)* 1.01 (0.95–1.08) >50% VA Service Connection 1.18 (1.10–1.27)* 1.23 (1.18–1.29)* 1.05 (1.01–1.09) Marital Status Married Divorced/Separated 1.27 (1.21–1.34)* 1.14 (1.09–1.20)* 1.19 (1.15–1.23)* Never married/Single 1.33 (1.25–1.42)* 1.21 (1.14–1.27)* 1.18 (1.13–1.24)* Missing/Not reported 1.35 (1.03–1.78) 1.00 (0.70–1.44) 1.02 (0.82–1.28) Homeless 1.57 (1.41–1.74)* 1.50 (1.42–1.58)* 1.73 (1.66–1.81)* Suicide Risk 3.21 (2.09–4.93)* 2.20 (2.04–2.36)* 2.85 (2.63–3.08)* Charlson Score 0 1 3.16 (2.99–3.33)* 1.56 (1.49–1.64)* 1.90 (1.83–1.97)* 2+ 2.93 (2.71–3.18)* 1.58 (1.49–1.68)* 1.76 (1.67–1.85)* Pain Diagnosis 1.29 (1.23–1.37)* 1.16 (1.10–1.22)* 1.28 (1.23–1.32)* Substance-Use Disorders Alcohol 1.39 (1.27–1.52)* 1.16 (1.08–1.24)* 1.20 (1.14–1.27)* Cannabis 0.89 (0.81–0.98) 1.15 (1.09–1.22)* 1.02 (0.96–1.08) Stimulant 1.11 (1.02–1.21) 1.17 (1.11–1.23)* 1.18 (1.11–1.25)* Opioid 1.43 (1.30–1.57)* 1.22 (1.14–1.31)* 1.49 (1.42–1.57)* Sedative 2.02 (1.44–2.84)* 1.39 (1.27–1.53)* 1.53 (1.40–1.67)* Other 1.05 (0.98–1.14) 1.12 (1.05–1.19)* 1.27 (1.20–1.35)* Alcohol + Drug Disorder 1.07 (0.97–1.19) 1.10 (1.03–1.17) 1.19 (1.13–1.26)* Mental Health Disorders Depression 1.28 (1.18–1.38)* 1.48 (1.40–1.57)* PTSD 1.03 (0.98–1.08) 1.09 (1.04–1.14)* Anxiety 1.08 (1.03–1.14) 1.09 (1.05–1.14)* Eating 1.19 (0.89–1.59) 1.27 (0.96–1.66) Bipolar 1.88 (1.72–2.06)* Psychotic 2.40 (2.23–2.57)* ≥1 YP PC Visit 0.65 (0.61–0.70)* 0.56 (0.53–0.59)* 0.59 (0.56–0.63)* ≥1 YP MH Visit 0.74 (0.68–0.80)* 0.53 (0.50–0.56)* 0.59 (0.56–0.61)* ≥1 YP ER Visit 1.93 (1.79–2.07)* 1.44 (1.33–1.56)* 1.60 (1.49–1.72)* YP Days Inpatient 0 1–7 Days 2.03 (1.86–2.22)* 1.35 (1.25–1.44)* 1.74 (1.64–1.85)* 8+ Days 4.07 (3.71–4.47)* 2.46 (2.26–2.67)* 3.48 (3.22–3.76)*

Note. SUD = substance-use disorders; SMI = serious mental illness; MI = mental illness; AOR = adjusted odds ratio; CI = confidence interval; OEF/ OIF = Operation Enduring Freedom/Operation Iraqi Freedom; YP = year prior to SUD index diagnosis; PC = primary care; MH = mental health; ED = emergency department; *p < .001. Group sample sizes decreased in adjusted logistic regression analyses compared with the initial group sample sizes reported in Table 1 due to patients missing one or more covariates.

392 J. M. PAINTER ET AL.

collaborations (e.g., integration of medical providers in SUD care). Finally, patient characteristics among SUD patients associated with the highest odds of HIU were generally consistent across groups and provide insights into key services to include in HIU interventions designed to treat patients with SUD and co-occurring mental health concerns (e.g., acute crisis management, chronic disease management, and homeless care services).

Declaration of interest

The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the article.

Funding

The research reported here was funded by the VA Center of Excellence in Substance Abuse Treatment and Education and the National Institute on Alcohol Abuse and Alcoholism (#1R21AA02894-01A1).

References

1. Funk M, Drew N, Freeman M, Faydi E. Mental health and development: targeting people with mental health conditions as a vulnerable group. Geneva: World Health Organization; 2010.

2. Whiteford HA, Degenhardt L, Rehm J, Baxter AJ, Ferrari AJ, Erskine HE, Charlson FJ, Norman RE, Flaxman AD, Johns N, et al. Global burden of disease attributable to mental and substance use disorders: findings from the Global Burden of Disease Study 2010. Lancet. 2013 Nov 15;382(9904):1575–86. doi:10.1016/S0140-6736(13)61611-6.

3. Substance Abuse and Mental Health Services Administration. Treatment episode data set (TEDS) highlights: 2007 national admissions to substance abuse treatment services. Rockville, MD: SAMHSA; 2009.

4. Walley AY, Paasche-OrlowM, Lee EC, Forsythe S, Chetty VK, Mitchell S, Jack BW. Acute care hospital utilization among medical inpatients discharged with a substance use disorder diagnosis. J Addict Med. 2012 March;6 (1):50–56. doi:10.1097/ADM.0b013e318231de51.

5. McKay JR. Is there a case for extended interventions for alcohol and drug use disorders? Addiction. 2005 Nov 1;100(11):1594–610. doi:10.1111/add.2005.100. issue-11.

6. McLellan AT, Lewis DC, O’Brien CP, Kleber HD. Drug dependence, a chronic medical illness: implica- tions for treatment, insurance, and outcomes evalua- tion. Jama. 2000 Oct 4;284(13):1689–95. doi:10.1001/ jama.284.13.1689.

7. Compton WM, Thomas YF, Stinson FS, Grant BF. Prevalence, correlates, disability, and comorbidity of DSM-IV drug abuse and dependence in the United States: results from the national epidemiologic survey on alcohol and related conditions. Archives Gen Psy.

2007 May 1;64(5):566–76. doi:10.1001/archpsyc.6 4.5.566.

8. Dickey B, Normand SL, Weiss RD, Drake RE, Azeni H. Medical morbidity, mental illness, and substance use disorders. Psychiatr Serv. 2002 Jul;53(7):861–67. doi:10.1176/appi.ps.53.7.861.

9. Shelton KH, Taylor PJ, Bonner A, Van Den Bree M. Risk factors for homelessness: evidence from a popula- tion-based study. Psychiatr Serv. 2015 April;60(4):465– 72. doi:10.1176/ps.2009.60.4.465.

10. Neighbors CJ, Sun Y, Yerneni R, Tesiny E, Burke C, Bardsley L, McDonald R, Morgenstern J. Medicaid care management: description of high-cost addictions treat- ment clients. J Subst Abuse Treat. 2013 Sep 30;45 (3):280–86. doi:10.1016/j.jsat.2013.02.009.

11. Stanton MW, Rutherford MK. The high concentration of US health care expenditures. Rockville, MD: Agency for Healthcare Research and Quality; 2006 AHRQ Pub. No. 06-0060.

12. Orszag PR, Emanuel EJ. Health care reform and cost control. New England J Med. 2010 Aug 12;363(7):601– 03. doi:10.1056/NEJMp1006571.

13. Williams BC. Limited effects of care management for high utilizers on total healthcare costs. Am J Manage Care. 2015 May;7(21):e244–6.

14. Gonzalez J. National Health Care Expenses in the U.S. Civilian Noninstitutionalized Population, 2011. Rockville, MD: Agency for Healthcare Research and Quality; 2013 Statistical Brief #425.

15. Jiang HJ, Weiss AJ, Barrett ML, Sheng M. Characteristics of hospital stays for super-utilizers by payer, 2012. Rockville, MD: Agency for Healthcare Research and Quality; 2015 Statistical Brief# 190.

16. Hines AL, Barrett ML, Jiang HJ, Steiner CA. Conditions with the largest number of adult hospital readmissions by payer, 2011. Rockville, MD: Agency for Healthcare Research and Quality; 2014 Statistical Brief# 172.

17. Hedden SL. Behavioral health trends in the United States: results from the 2014 National survey on drug use and health. 2015 HHS Publication No. SMA 15-4927, NSDUH Series H-50, HHS Publication; 2015

18. Lai HM, Cleary M, Sitharthan T, Hunt GE. Prevalence of comorbid substance use, anxiety and mood disorders in epidemiological surveys, 1990–2014: a systematic review and meta-analysis. Drug Alcohol Depend. 2015 Sep 1;154:1–3. doi:10.1016/j.drugalcdep.2015.05.031.

19. Yoon G, Petrakis IL, Rosenheck RA. Correlates of major depressive disorder with and without comorbid alcohol use disorder nationally in the veterans health administration. Am J Addict. 2015 Aug 1;24(5):419–26. doi:10.1111/ajad.v24.5.

20. Humphreys K, McLellan AT. A policy-oriented review of strategies for improving the outcomes of services for substance use disorder patients. Addiction. 2011 Dec 1;106(12):2058–66. doi:10.1111/add.2011.106.issue-12.

21. Farley JF, Harley CR, Devine JW. A comparison of comorbidity measurements to predict healthcare expenditures. Am J Manag Care. 2006 Feb 1;12 (2):110–18.

22. Fleishman JA, Cohen JW. Using Information on Clinical Conditions to Predict High-Cost Patients.

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 393

Health Serv Res. 2010 Apr 1;45(2):532–52. doi:10.1111/ j.1475-6773.2009.01080.x.

23. McAlpine DD, Mechanic D. Utilization of specialty mental health care among persons with severe mental illness: the roles of demographics, need, insurance, and risk. Health Serv Res. 2000 Apr;35(1 Pt 2):277.

24. Mechanic D. Mental health and social policy: the emer- gence of managed care. Boston, MA: Allyn and Bacon; 1999.

25. Gelberg L, Andersen RM, Leake BD. The behavioral model for vulnerable populations: application to med- ical care use and outcomes for homeless people. Health Serv Res. 2000 Feb;34(6):1273.

26. Hawkins EJ, Malte CA, Baer JS, Kivlahan DR. Prevalence, predictors, and service utilization of patients with recurrent use of Veterans Affairs sub- stance use disorder specialty care. J Subst Abuse Treat. 2012 Sep 30;43(2):221–30. doi:10.1016/j. jsat.2011.11.002.

27. Lorine K, Goenjian H, Kim S, Steinberg AM, Schmidt K, Goenjian AK. Risk factors associated with psychia- tric readmission. J Nerv Ment Dis. 2015 Jun 1;203 (6):425–30. doi:10.1097/NMD.0000000000000305.

28. Mark T, Tomic KS, Kowlessar N, Chu BC, Vandivort- Warren R, Smith S. Hospital readmission among med- icaid patients with an index hospitalization for mental and/or substance use disorder. J Behav Health Serv Res. 2013 Apr 1;40(2):207–21. doi:10.1007/s11414- 013-9323-5.

29. Smith MW, Stocks C, Santora PB. Hospital readmis- sion rates and emergency department visits for mental health and substance abuse conditions. Community Ment Health J. 2015 Feb 1;51(2):190–97. doi:10.1007/ s10597-014-9784-x.

30. Himelhoch S, McCarthy JF, Ganoczy D, Medoff D, Dixon LB, Blow FC. Understanding associations between serious mental illness and HIV among patients in the VA health system. Psychiatr Serv. 2007 Sep;58(9):1165–72. doi:10.1176/ps.2007.58.9.1165.

31. Department of Veterans Affairs. Use of patient record flags to identify patients at high risk for suicide. Washington, DC: Veterans Health Administration; 2008 VHA directive No.: 2008–036.

32. Quan H, Sundararajan V, Halfon P, Fong A, Burnand B, Luthi JC, Saunders LD, Beck CA, Feasby TE, Ghali WA. Coding algorithms for defining comorbidities in ICD-9- CM and ICD-10 administrative data. Med Care. 2005 Nov 1;43:1130–39. doi:10.1097/01.mlr.0000182534.19832.83.

33. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987 Jan 1;40(5):373–83. doi:10.1016/ 0021-9681(87)90171-8.

34. Dobscha SK, Morasco BJ, Kovas AE, Peters DM, Hart K, McFarland BH. Short-term variability in outpatient pain intensity scores in a national sample of older veterans with chronic pain. Pain Med. 2015 May 1;16 (5):855–65. doi:10.1111/pme.12643.

35. Gardner W, Mulvey EP, Shaw EC. Regression analyses of counts and rates: poisson, overdispersed Poisson, and negative binomial models. Psychol Bull. 1995 Nov;118(3):392. doi:10.1037/0033-2909.118.3.392.

36. Hunter G, Yoon J, Blonigen DM, Asch SM, Zulman DM. Health care utilization patterns among high-cost VA patients with mental health conditions. Psychiatr Serv. 2015 May 1;66(9):952–58. doi:10.1176/appi. ps.201400286.

37. Saitz R, Cheng DM, Winter M, Kim TW, Meli SM, Allensworth-Davies D, Lloyd-Travaglini CA, Samet JH. Chronic care management for dependence on alcohol and other drugs: the AHEAD randomized trial. Jama. 2013 Sep 18;310(11):1156–67. doi:10.1001/jama.20 13.277609.

38. Weisner C, Mertens J, Parthasarathy S, Moore C, Lu Y. Integrating primary medical care with addiction treat- ment: A randomized controlled trial. Jama. 2001 Oct 10;286(14):1715. doi:10.1001/jama.286.14.1715.

39. Willenbring ML, Olson DH. A randomized trial of integrated outpatient treatment for medically ill alco- holic men. Archives of Intern Med. 1999 Sept 13;159 (16):1946–52. doi:10.1001/archinte.159.16.1946.

40. Department of Veterans Affairs. Patient aligned care team (PACT) handbook. Washington, DC: Veterans Health Administration; 2014 VHA Handbook No.: 1101.10.

41. Trivedi RB, Post EP, Sun H, Pomerantz A, Saxon AJ, Piette JD, Maynard C, Arnow B, Curtis I, Fihn SD, et al. Prevalence, comorbidity, and prognosis of mental health among US veterans. Am J Pub Health. 2015 Dec;105(12):2564–69. doi:10.2105/AJPH.2015.302836.

42. Beard C, Hearon BA, Lee J, Kopeski LM, Busch AB, Björgvinsson T. When Partial Hospitalization Fails: risk Factors for Inpatient Hospitalization. J Nerv Ment Dis. 2016 Jun 1;204(6):431–36. doi:10.1097/ NMD.0000000000000515.

43. Chang G, Raffi E, Tang M, Fernando GI, Zucker J, Schein AZ. High utilization of inpatient detoxification: predictors among US veterans. Am J Drug Alcohol Abuse. 2016 May 3;42(3):279–86. doi:10.3109/ 00952990.2015.1089256.

44. Wang L, Porter B, Maynard C, Evans G, Bryson C, Sun H, Gupta I, Lowy E, McDonell M, Frisbee K, et al. Predicting risk of hospitalization or death among patients receiving primary care in the Veterans Health Administration. Med Care. 2013 Apr 1;51(4):368–73. doi:10.1097/MLR.0b013e31 827da95a.

45. Bodenheimer T, Wagner EH, Grumbach K. Improving primary care for patients with chronic illness. Jama. 2002 Oct 9;288(14):1775–79. doi:10.1001/jama.288.14.1775.

46. Prendergast M, Podus D, Finney J, Greenwell L, Roll J. Contingency management for treatment of substance use disorders: A meta-analysis. Addiction. 2006 Nov 1;101(11):1546–60. doi:10.1111/add.2006.101.issue-11.

47. Jones HE, Haug N, Silverman K, Stitzer M, Svikis D. The effectiveness of incentives in enhancing treatment attendance and drug abstinence in methadone-main- tained pregnant women. Drug Alcohol Depend. 2001 Feb 1;61(3):297–306. doi:10.1016/S0376-8716(00) 00152-6.

48. O’Malley KJ, Cook KF, Price MD, Wildes KR, Hurdle JF, Ashton CM. Measuring diagnoses: ICD code accu- racy. Health Serv Res. 2005 Oct 1;40(5p2):1620–39. doi:10.1111/hesr.2005.40.issue-5p2.

394 J. M. PAINTER ET AL.

Copyright of American Journal of Drug & Alcohol Abuse is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.