1. Hutchison, Shari L. MS, PMP
  2. Herschell, Amy D. PhD
  3. Edwards, Jamie F. LCSW
  4. Karpov, Irina MS
  5. Wasilchak, Deborah S. MA
  6. Hurford, Matthew O. MD


Purpose of Study: To examine the effectiveness of a care management intervention to decrease readmissions and to better understand clinical and social determinants associated with readmission.


Primary Practice Setting: Inpatient mental health (MH) and substance use disorder (SUD) facilities, nonhospital SUD withdrawal management and rehabilitation facilities.


Methodology and Sample: The authors identified 3,950 Medicaid-enrolled individuals who received the intervention from licensed clinical staff of a behavioral health managed care organization; 2,182 individuals were eligible but did not receive the intervention, for treatment as usual (TAU). We used logistic regression to examine factors associated with readmission. Determinants of readmission were summarized through descriptive tests.


Results: The intervention was associated with lower readmissions to SUD facilities compared with TAU (6.0% vs. 8.6%, p = .0002) and better follow-up to aftercare. Controlling for clinical differences between groups, regression results found increased odds of readmission for male gender (odds ratio [OR]: 1.33; 95% confidence interval [CI]: 1.16-1.52, p < .0001) and dual MH and SUD diagnoses (OR: 1.52; CI: 1.29-1.79, p < .0001). Prior inpatient and case management services were also associated with increased odds for readmission. In the regression model, the intervention was not associated with decreased odds for readmission. Individuals with readmission (n = 796) were more likely to report being prescribed psychotropic medication and having housing difficulties and less likely to report having a recovery plan than those without readmission.


Implications for Case Management Practice: Characteristics of Medicaid populations with hospitalization may contribute to readmission, which may be mitigated through care management intervention.


Article Content

Socioeconomic and basic needs, as well as conditions of an individual's environment, also known as social determinants of health (SDoH), may exacerbate illness symptoms and contribute to rates of unplanned emergent care and hospitalization. Within behavioral health care, SDoH have been named as potential contributors to the "revolving door" of inpatient readmissions (Habit et al., 2018). The demand for health care systems to help facilitate linkages to community-based services and resources, support the provider network, and understand the needs of individuals is critical to making an impact on whole person health as well as population health.


Likely in part of negative repercussions of coronavirus disease-2019 (COVID-19) on unemployment, rates of individuals covered under Medicaid have increased, with economic recovery expected to occur over the next 2 years or longer (Corallo & Rudowitz, 2021; Health Management Associates, 2020). Management of the increase in Medicaid enrollment is especially challenging at this time, as rates of behavioral health issues, which include mental health (MH) disorders and substance use disorders (SUDs), have increased and symptoms have intensified under isolation, fear of COVID-19, and associated mandates (Hansel et al., 2020). Although the longer-term impact of COVID-19 on behavioral health care utilization is difficult to predict, economic downturns have been shown to be associated with increased rates of suicide, substance use, depression, and anxiety; increased use of psychopharmacology for depression and anxiety; and increased rates of utilization for community-based and residential levels of care (Frasquilho et al., 2015; Hubley et al., 2016; Mulvaney-Day et al., 2019; Phillips & Nugent, 2014). Furthermore, at the time of this report, the financial impact of COVID-19-related mandates on many smaller, community-based MH centers is unknown but expected to strain limited resources and state Medicaid budgets (Garrett & Gangopadhyaya, 2020). As a result, managed care organizations look to practical strategies that address SDoH while decreasing high-cost, acute service utilization (such as inpatient MH and SUD treatment and withdrawal management) and linking individuals to ambulatory care as a secondary prevention effort (Brar et al., 2021; Gottlieb et al., 2016, 2017; Highland et al., 2020).


Community Care Behavioral Health Organization of the University of Pittsburgh Medical Center (UPMC) Insurance Services Division, a nonprofit Medicaid behavioral health managed care organization (BHMCO) in Pennsylvania established in 1999, routinely monitors rates of readmission and timely follow-up to ambulatory care. Individuals funded through Medicaid programs have been shown to have higher rates of behavioral health readmissions compared with privately insured or uninsured individuals (Fingar et al., 2017).


Efforts by the BHMCO to improve care coordination have shown success and are associated with lower rates of MH and SUD readmissions through a care management bridging strategy designed to address SDoH, facilitate referrals to community-based services, and provide education on continuity of care at time of discharge (Hutchison et al., 2019; Taylor et al., 2016). To date, the care management intervention has been targeted to individuals with 30-day readmissions, as these individuals have been shown to be at higher risk for subsequent readmission (Craig et al., 2000). Earlier evaluation found that the care management intervention is especially effective for individuals who use SUD services and those with co-occurring MH and SUD diagnoses (Hutchison et al., 2019).


The present study was designed to examine readmission and rates of follow-up after intensive service for individuals receiving the care management intervention versus treatment as usual (TAU). Determinants of health that may impact readmission will be described for individuals receiving the intervention. Expanding our knowledge of clinical and social determinants of health at time of readmission may help inform training, resources, and clinical pathways for care managers.




The sample included Medicaid-enrolled adults, ages 18 to 64 years, within the BHMCO network in calendar years 2018 and 2019 who met eligibility for the care management intervention due to readmission within 30 days of an MH or SUD inpatient or SUD residential withdrawal management or rehabilitation service. Individuals with residential SUD service were included in the intervention to ensure good continuum of care for these individuals; residential rehabilitation was not considered a readmission event. Individuals who were eligible for the care management intervention but did not receive the intervention formed a TAU comparison group. TAU individuals met eligibility criteria but did not receive the intervention for the following reasons: early discharge (27%), other reasons (e.g., not on unit, unit locked; 22%), travel/distance not possible for the care manager (14%), left against medical advice (12%), symptom related (11%), refusal (8%), medical transfer (3%), substance use withdrawal (2%), and seclusion (<1%). All activities were approved as quality improvement by the UPMC Quality Review Committee.


Study Design

A quasi-experimental design was utilized to examine effectiveness of the intervention; 3,950 individuals received the intervention, and 2,182 individuals were eligible but did not receive the intervention and formed the comparison group (TAU). Outcomes were examined comparing intervention versus TAU. Clinical and social determinants of health are described for those individuals who received the intervention and compared between those with and without readmission.



Care Management Intervention

Those who received the care management intervention were identified from a daily report of individuals with readmission events, defined as inpatient MH, inpatient SUD, or residential withdrawal management or rehabilitation service within 30 days of a previous discharge. The care management intervention consisted of a 15- to 30-minute onsite interview between the care manager and the individual prior to discharge and was focused on discharge planning, medications, SDoH that may impact hospitalization, and resources during the inpatient stay that may assist with a transition to the community. The care manager also coordinated with the individual's community-based case management providers or, if appropriate, assessed the individual for referral to acute care coordination (Taylor et al., 2016).



The comparison condition consisted of TAU, including discharge planning by hospital or residential staff with care management supports and referrals to behavioral health services and appropriate community-based resources (e.g., housing). For TAU, care managers did not meet with individuals during the readmission but did help to facilitate appropriate aftercare and referrals.



Sociodemographic Characteristics

Sociodemographic variables included gender, race, ethnicity, and age. Medicaid expansion status was determined as coverage under Pennsylvania's expanded eligibility criteria for Medicaid up to 138% of the federal poverty level (Expansion) versus historical coverage (Legacy). All sociodemographic characteristics were obtained from Medicaid administrative data provided by the Pennsylvania Department of Human Services.


Behavioral Health Service Utilization and Diagnoses

Behavioral health service utilization was defined as at least one paid claim for a service; data on utilization were obtained from the BHMCO's paid service claims. Each claim contained up to three diagnoses as listed by the clinician. Diagnoses were categorized using the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition (DSM-5; American Psychiatric Association, 2013). Individuals could have multiple diagnoses.


Clinical and Social Determinants of Health

Additional variables of interest were self-reported by the individual and recorded in the BHMCO's electronic record by the care manager. Clinical determinants of health included physical health, medications, and components of the discharge process. SDoH included issues around housing, financial, legal, and social support that contributed to readmission or were identified for assistance during the readmission stay.


Readmission and Follow-up Outcomes

Individuals were identified as readmitted if a paid claim for MH or SUD hospitalization or withdrawal management service at any network provider occurred within 30 days of discharge from the originating hospitalization or SUD residential service. Receipt of follow-up was considered any claim within 7 or 30 days of discharge from the originating hospitalization or SUD residential service for outpatient MH or SUD, or SUD residential rehabilitation, halfway house, or partial hospitalization program. The outcome variables were dichotomous. The first readmission was selected for individuals with multiple 30-day readmissions during the study period.


Statistical Analysis

The [chi]2 test and the t test of significance were utilized to examine differences in characteristics between groups. For individuals with the care management intervention, the presence of clinical and social determinants of health was compared between readmitted and nonreadmitted groups using the [chi]2 test. To determine effectiveness of the intervention, stepwise logistic regression was performed using readmission (yes/no) as the outcome and group (intervention vs. TAU) as the independent variable, controlling for variables associated with readmission in univariate analyses.


For regressions, the Hosmer-Lemeshow test statistic was used as a goodness-of-fit test; a nonsignificant result provides evidence that the model fits the data appropriately, while the significant likelihood ratio, score, and Wald tests indicate that the model with covariates is more appropriate than the null (intercept only) model (Ambrosius, 2007; Hosmer et al., 2013). The c-statistic, or concordance statistic, was determined to assist with interpretation of the results of the model. Results with p values < .05 were considered significant. Analyses were performed using the Statistical Analysis System (SAS) version 9.4.1



Characteristics of Individuals Receiving the Intervention

During the study period, 6,132 individuals were eligible to receive the care management intervention (see Table 1). A majority of the eligible population was male (60.09%) and European American (74.33%), with an average age of 37.61 +/- 12.00 years. A large portion had dual MH and SUD diagnoses (74.53%). Individuals became eligible for the intervention due to utilization of acute care including inpatient MH admission (40.28%), inpatient withdrawal management (6.02%), inpatient rehabilitation (0.90%), nonhospital withdrawal management (28.36%), nonhospital rehabilitation (21.95%), or a combination of these acute services. Of the eligible population, 3,950 (64.42%) individuals received the intervention and 2,182 (35.58%) received TAU. The intervention group, compared to TAU, had higher rates of inpatient MH treatment utilization (41.34% vs. 38.36%, p = .0226) as well as lower rates of nonhospital withdrawal management (25.82% vs. 32.95%, p < .0001) and nonhospital rehabilitation (20.86% vs. 23.92%, p = .0055) in the 6 months prior to readmission. The intervention group had fewer European Americans and more racial diversity, p = .0022, and was older (38.11 +/- 12.08 vs. 36.70 +/- 11.81 years) than TAU. Differences in behavioral health service utilization in the 6 months before the inpatient episode for intervention versus TAU included higher rates of case management (17.80% vs. 13.11%, p < .0001) and outpatient MH services (58.86% vs. 55.00%, p = .0034), and lower rates of outpatient SUD services (28.58% vs. 33.41%, p < .0001), residential withdrawal management (25.82% vs. 32.95%, p < .0001), and residential rehabilitation (20.86% vs. 23.92%, p = .0055).

Table 1 - Click to enlarge in new windowTABLE 1 Characteristics of Individuals in the Intervention and TAU Groups

Readmission and Follow-up Outcomes

The intervention was associated with significantly lower readmissions to SUD facilities compared with TAU (6.03% vs. 8.57%, p = .0002; see Table 2). The intervention was not associated with lower readmissions for MH facilities, but it was associated with increased follow-up after discharge within 7 and 30 days to community-based MH treatment (24.89% vs. 20.16%, p < .0001, 7-day; 42.94% vs. 38.13%, p = .0003, 30-day) and increased follow-up within 7 days to SUD aftercare (27.62% vs. 25.16%, p = .0372). However, no effect of the intervention was found for follow-up within 30 days to SUD aftercare.

Table 2 - Click to enlarge in new windowTABLE 2 Readmission and Follow-up Outcomes for the Intervention vs. Comparison Groups

Compared with individuals not readmitted, individuals readmitted were more likely to be male (65.69% vs. 58.64%, p < .0001), had co-occurring MH and SUD diagnoses (81.93% vs. 72.61%, p < .0001), had Medicaid legacy versus expansion status (45.25% vs. 41.19%, p = .0093), and had higher rates of prior utilization of MH services, including assertive community treatment, case management, crisis services, outpatient MH, inpatient MH, and inpatient withdrawal management (see Table 3). These variables were added to a logistic regression model to control for differences associated with readmission.

Table 3 - Click to enlarge in new windowTABLE 3 Comparison Between Readmitted and Not Readmitted Individuals

Results from logistic regression (see Table 4) did not show an association for the intervention with readmission. Thus, after controlling for possible confounds, individuals who received the intervention or TAU had equivalent odds for readmission. Individuals with male gender (1.33 [1.16, 1.52], p < .0001) and dual diagnoses (1.52 [1.29, 1.79], p < .0001) had higher odds for readmission. Individuals with prior utilization of case management (1.33 [1.13, 1.56], p = .0007), inpatient MH (2.23 [1.95, 2.54], p < .0001), and inpatient withdrawal management had higher odds for readmission (2.04 [1.60, 2.61], p < .0001). No other variable was found to be associated with odds of readmission.

Table 4 - Click to enlarge in new windowTABLE 4 Logistic Regression Results for Odds of Readmission

Clinical and Social Determinants of Health

Table 5 presents the care management intervention results for 3,950 individuals who received the intervention, comparing those with (n = 796) and without (n = 3,154) readmission. More individuals with readmission than without reported being prescribed psychotropic medication (78.77% vs. 72.61%, p = .0004). Rates of activities indicating good clinical care were high, and often higher for those with readmission versus those without, including receipt of medication instructions (74.00% vs. 63.97%, p < .0001), receiving a discharge plan (66.96% vs. 59.92%, p = .0003), understanding the discharge plan (64.95% vs. 58.18%, p = .0005), and having a scheduled aftercare appointment following the originating admission (61.81% vs. 53.96%, p < .0001). Fewer individuals with readmission than without reported having a recovery plan (59.92% vs. 65.66%, p = .0025).

Table 5 - Click to enlarge in new windowTABLE 5 Clinical and Social Determinants of Health for Individuals with and without a Readmission

Significantly more individuals with readmission versus without readmission reported housing difficulties as their reason for admission (19.72% vs. 15.73%, p = .0067) and requested help with housing resources (24.25% vs. 20.26%, p = .0137). There were no differences between groups for self-reported financial, legal, or social difficulties leading to hospitalization. Although most individuals reported receiving medication instructions and having an aftercare appointment, significantly more individuals with readmission than those without requested help with access to medications or understanding medications (15.08% vs. 12.37%, p = .0416). Fewer individuals with readmission than those without requested help understanding options for aftercare (46.98% vs. 51.68%, p = .0179). Fewer individuals with readmission than those without reported caring for family members (5.53% vs. 11.38%, p < .0001) or having employment (7.41% vs. 12.33%, p < .0001).



In the current study, mixed results were found for a care management intervention designed to decrease readmission through assessment of clinical and social determinants of health. The intervention was associated with lower rates of costly readmission to inpatient SUD service and better connection to ambulatory care after discharge for MH and SUD service (within 7 days). The intervention was not associated with lower readmission to MH facilities. Determining the immediate health care and social needs for individuals with frequent readmission may help to ensure a good continuum of care provided through ambulatory, community-based services. Differences in service availability may explain observed patterns for readmission and aftercare in the current study. Individuals with inpatient SUD treatment or withdrawal management can connect to an intermediate level of care via residential rehabilitation that is not available for individuals with MH diagnoses who utilize only the MH system of care. More support, other than ambulatory service, may be needed for some individuals leaving inpatient MH facilities. For example, the finding of increased housing issues reported for individuals with readmission points to needs beyond health care services. Additionally, the current study found co-occurring MH and SUD diagnoses and some prior service utilization were associated with increased odds for readmission. These factors may indicate illness severity. Other clinical determinants were discovered through the care management intervention and also found to be associated with readmission; prescribed psychotropic medication was more likely for individuals with readmission, which may be another indicator of illness severity.


Several differences were found between individuals with and without readmission who received the care management intervention. Counterintuitively, individuals with readmission reported higher rates of receipt of components indicating quality care: medication instructions, discharge planning, and aftercare. However, these findings provide support for the importance of addressing SDoH as factors that contribute to readmission. This information, as well as recovery planning, housing, and other SDoH, was not available for those receiving TAU and therefore was not included in regression models examining factors statistically associated with readmission. Consistent with other research, these results highlight the importance of understanding more than access and service utilization when considering health care needs. For example, a study of more than 1.6 million people found no association of improved access to health care services with self-reported well-being, decreased stress, or improved life satisfaction (Kobayashi et al., 2019). Care management interventions can address these factors in addition to other clinical determinants and SDoH.


Current emphasis on SDoH has highlighted the need to address the causal, upstream factors in addition to monitoring immediate, proximal factors associated with readmission and poor health (Bravemen et al., 2011). Horwitz et al. (2020) report that, between January 2017 and November 2019, $2.5 billion was spent across U.S. health systems to address SDoH, including housing, employment, education, food security, community well-being, incarceration, and transportation. One way the BHMCO is addressing SDoH is through increased efforts with county-level housing programs and employment opportunities for publicly insured individuals with behavioral health disorders. Individuals under Medicaid expansion with behavioral health conditions are less likely to report being employed than are enrollees without these conditions (Tipirneni et al., 2020). The BHMCO is currently working with the U.S. Social Security Administration to increase the use of work incentives, which make it possible for Medicaid-enrolled individuals to have paid employment and move toward financial wellness.



Several limitations are worth mentioning. First, the care management intervention occurred as part of operations of the BHMCO and not as a research study. Thus, the design did not include randomization, and the number of observed variables was restricted. Of note, information on cognitive impairment. Reasons for not receiving the intervention (early discharge, discharge against medical advice, etc.) point to individual characteristics that may have differed between the groups in ways not measured, potentially contributing to differences in odds for readmission. Finally, the use of the 30-day readmission time frame to investigate readmission was limiting, and these results may not be reflective of longer-term patterns of hospitalization.


Implications for Care Management

The current findings have implications for appropriately targeting intensive practices to individuals at highest risk for readmission. Care management approaches that address the needs of individuals can be successful in facilitating connections to ambulatory care and decreasing readmission. Rates of engagement and retention in SUD services are commonly low, especially among Medicaid-enrolled individuals with psychiatric diagnoses (Lind et al., 2019). It may be that the care management intervention has greater impact where connections to appropriate care are lacking and where help is needed most in navigating the behavioral health system of care. SDoH and clinical determinants of health (discharge planning, aftercare, and medication access) that contribute to readmission are complex, and care managers may fill gaps in hospital care to work with treatment teams to facilitate connection to services and other resources. Study findings highlight the need to identify unmet health care needs and other determinants that may impact an individual's quality of health care.



The authors wish to thank the High-Risk Care Management Team, and James M. Schuster, MD, MBA, Community Care Behavioral Health Organization, Pittsburgh, Pennsylvania; Advocates for Human Potential, Sudbury, Massachusetts; and The UPMC Center for High-Value Health Care, Pittsburgh, Pennsylvania, for contributions and review of this article.




Ambrosius W. T. (2007). Topics in biostatistics. Humana Press. [Context Link]


American Psychiatric Association. (2013). Diagnostic and statistical manual of mental disorders (5th ed.). Author.[Context Link]


Brar J. S., Maise A. A., Schake P., Bills L. J., Washington L., Nikolajski C., Herschell A. D. (2021). Implementing a learning collaborative for population-based physical and behavioral health integration. Community Mental Health Journal. Advance online publication.[Context Link]


Bravemen P., Egerter S., Williams D. R. (2011). The social determinants of health: Coming of age. Annual Review of Public Health, 32, 381-398.[Context Link]


Corallo B., Rudowitz R. (2021, April 8). Analysis of recent national trends in Medicaid and CHIP enrollment. Kaiser Family Foundation.[Context Link]


Craig T. J., Fennig S., Tanenberg-Karant M., Bromet E. J. (2000). Rapid versus delayed readmission in first-admission psychosis: Quality indicators for managed care? Annals of Clinical Psychiatry, 12(4), 233-238.[Context Link]


Fingar K. R., Barrett M. L., Jiang H. J. (2017). A comparison of all-cause 7-day and 30-day readmissions, 2014. HCUP Statistical Brief #230. Agency for Healthcare Research and Quality Healthcare Cost and Utilization Project.[Context Link]


Frasquilho D., Matos M. G., Salonna F., Guerreiro D., Storti C. C., Gaspar T., Caldas-de-Almeida J. (2015). Mental health outcomes in times of economic recession: A systematic literature review. BMC Public Health, 16, 115.[Context Link]


Garrett B., Gangopadhyaya A. (2020). How the COVID-19 recession could affect health insurance coverage. Urban Institute.[Context Link]


Gottlieb L. M., Garcia K., Wing H., Manchanda R. (2016). Clinical interventions addressing nonmedical health determinants in Medicaid managed care. The American Journal of Managed Care, 22(5), 370-376.[Context Link]


Gottlieb L. M., Quinones-Rivera A., Manchanda R., Wing H., Ackerman S. (2017). States' influences on Medicaid investments to address patients' social needs. American Journal of Preventive Medicine, 52(1), 31-37.[Context Link]


Habit N. F., Johnson E., Edlund B. J. (2018). Appointment reminders to decrease 30-day readmission rates to inpatient psychiatric hospitals. Professional Case Management, 23(2), 70-74.[Context Link]


Hansel T. C., Saltzman L. Y., Bordnick P. S. (2020). Behavioral health and response for COVID-19. Disaster Medicine and Public Health Preparedness, 14(5), 670-676.[Context Link]


Health Management Associates. (2020). COVID-19 impact on Medicaid, Marketplace, and the uninsured.[Context Link]


Highland J., Nikolajski C., Kogan J., Ji Y., Kukla M., Schuster J. (2020). Impact of behavioral health homes on cost and utilization outcomes. Psychiatric Services, 71(8), 796-802.[Context Link]


Horwitz L. I., Chang C., Arcilla H. N., Knickman J. R. (2020). Quantifying health systems' investment in social determinants of health, by sector, 2017-19. Health Affairs, 39(2), 192-198.[Context Link]


Hosmer D. W., Lemeshow S., Sturdivant R. X. (2013). Applied logistic regression (3rd ed.). Wiley. [Context Link]


Hubley S., Lynch S. B., Schneck C., Thomas M., Shore J. (2016). Review of key telepsychiatry outcomes. World Journal of Psychiatry, 6(2), 269-282.[Context Link]


Hutchison S. L., Flanagan J. V., Karpov I, Elliott L., Holsinger B., Edwards J., Loveland D. (2019). Care management intervention to decrease psychiatric and substance use disorder readmissions in Medicaid-enrolled adults. Journal of Behavioral Health Services & Research, 46(3), 533-543.[Context Link]


Kobayashi L. C., Altindag O., Truskinovsky Y., Berkman L. F. (2019). Effects of the Affordable Care Act Medicaid expansion on subjective well-being in the US adult population, 2010-2016. American Journal of Public Health, 109(9), 1236-1242.[Context Link]


Lind B. K., McCarty D., Gu Y., Baker R., McConnell K. J. (2019). Predictors of substance use treatment initiation and engagement among adult and adolescent Medicaid recipients. Substance Abuse, 40(3), 285-291.[Context Link]


Mulvaney-Day N., Gibbons B. J., Alikhan S., Karakus M. (2019). Mental Health Parity and Addiction Equity Act and the use of outpatient behavioral health services in the United States, 2005-2016. American Journal of Public Health, 109(S3), S190-S196.[Context Link]


Phillips J. A., Nugent C. N. (2014). Suicide and the Great Recession of 2007-2009: The role of economic factors in the 50 U.S. states. Social Science & Medicine, 116, 22-31.[Context Link]


Taylor C., Holsinger B., Flanagan J. V., Ayers A. M., Hutchison S. L., Terhorst L. (2016). Effectiveness of a brief care management intervention for reducing psychiatric hospitalization readmissions. The Journal of Behavioral Health Services & Research, 43(2), 262-271.[Context Link]


Tipirneni R., Patel M. R., Goold S. D., Kieffer E. C., Ayanian J. Z., Clark S. J., Lee S., Bryant C., Kirch M. A., Solway E., Luster J., Lewallen M., Zivin K. (2020). Association of expanded Medicaid coverage with health and job-related outcomes among enrollees with behavioral health disorders. Psychiatric Services, 71(1), 4-11.[Context Link]


1 SAS. SAS and all other SAS Institute Inc. product or service names are registered trademarks or trademarks of SAS Institute Inc. in the USA and other countries. (R) indicates USA registration. 2013. [Context Link]


For more than 46 additional continuing education articles related to Case Management topics, go to


behavioral health managed care; Medicaid; psychiatric readmission; social determinants of health; substance use disorder