Abstract
Objective:
Irregular discharge is a concern among mental health populations and associated with poor outcomes. Little is known about the relationship between irregular discharge and treatment setting. Because care processes differ between acute inpatient and residential settings, it is important to evaluate irregular discharge in these settings.
Method:
A retrospective study was conducted in patients with mental health conditions admitted to acute inpatient or residential mental health settings in the Department of Veterans Affairs, 2003–2019. Logistic regression and multivariate Cox proportional hazards were used to evaluate factors associated with irregular discharge risk in the first 90- days of admission.
Results:
Among 1.8 million discharges, 7.4% had an irregular discharge within 90- days of admission. Younger age was a central predictor of risk. Irregular discharge rates were four-fold higher in residential versus acute settings. When accounting for length of stay (LOS) across settings, there was a modest higher risk of irregular discharge from acute versus residential settings (HR = 1.06, 95% Confidence Interval 1.04–1.07).
Conclusions:
Patients are at high risk for irregular discharge from acute and residential settings when they are young. LOS is an important determinant of irregular discharge risk.. Interventions are needed to address drivers of irregular discharge.
Keywords: Against medical advice discharge, Acute mental health treatment, Residential treatment, Irregular discharge
1. Introduction
Irregular discharges, including discharge against medical advice (AMA) as well as other unplanned hospital discharges such as self-initiated discharge, are a particular concern for the mental healthcare system [1–8]. In an analysis using the National (or Nationwide) Inpatient Sample (NIS), Onukwugha et al. found that in 2015, the rate of irregular discharge among all patients discharged with substance use disorders (SUD) was 10%, while the rate of irregular discharges among all patients discharged with physical health conditions including heart failure, acute myocardial infarction, and pneumonia was only 1.2% [4]. Irregular discharge is associated with higher rates of readmission [9–11], morbidity [9], and mortality including suicide [2,3,9]. There is a need to identify predictors of irregular discharge in mental health populations in order to improve patient care and mitigate harm.
There has been robust study of predictors of irregular discharge in general hospital settings [5,6,8,9,12–14]. This work has identified factors that are associated with irregular discharge such as male gender [5,6,8,12–14], younger age [5,6,8,12–14], having a SUD [5,6,9] and urban location [5,12–14]. Only a few of these studies, however, have examined whether these results apply to patients with mental health conditions [5,6,8]. In an analysis of hospital discharges reported in the NIS between 2002 and 2011, Spooner et al. (2017) found among patients with mental health conditions that male sex and younger age predicted irregular discharge, but urban location did not [5]. Studies of irregular discharge in mental health populations also have notable limitations. First, studies have relied upon discharge data from the United States (US) Healthcare Cost and Utilization Project (e.g. NIS) [5,8] or the National Hospital Discharge Survey [6] to conduct their analysis. Thus, their findings are generalizable to US community hospitals but not necessarily to the Department of Veterans Affairs (VA), the largest integrated healthcare system in the US [15]. Second, studies have not stratified their results by unit type. The NIS excludes psychiatric hospitals and alcoholism or chemical dependency treatment facilities [15]. Therefore, studies provide no information on predictors of irregular discharge in psychiatric settings.
The lack of information on predictors of irregular discharge in psychiatric settings is an important finding. While patients with mental health conditions may be admitted to a general medical floor, they are more commonly admitted to acute inpatient or residential mental health treatment settings [16,17].An acute inpatient stay is short and focused on crisis stabilization [16]. Conversely, residential stays are longer in duration and provide a greater range of services focused on long-term symptom management [17]. Therefore, predictors of irregular discharge may be unique within these settings and length of stay (LOS) may contribute to the relationship. In a literature review of AMA discharge in psychiatric settings Brook et al. (2006) also observed that there is conflicting evidence about whether factors such as male gender predict irregular discharge in psychiatric settings [7]. Brook et al. noted that there has been a downward trend in publications since 1980 and studies have relied heavily on univariate analysis to draw conclusions [7]. Future research should examine temporal trends in irregular discharge in psychiatric settings and use more robust methods to evaluate explanatory factors [7].
In order to address existing gaps in the literature, we examined patterns of irregular discharges among patients discharged from acute inpatient or residential mental health treatment settings within the VA healthcare system. We had three objectives. First, we aimed to understand how a broad range of individual factors predict irregular discharge in patients discharged from mental health treatment settings. Second, we aimed to understand how an acute inpatient versus residential mental health stay influences irregular discharge. Third, we aimed to evaluate how risk changes with LOS by using survival analysis. Based on the literature, we hypothesized that patients would be at greater risk for irregular discharge if they were male, of younger age and had a SUD. We hypothesized that the risk for irregular discharge would be greater among those leaving an acute inpatient mental health setting compared to a residential mental health treatment setting. By understanding factors associated with irregular discharge, our findings could aid in identifying those at highest risk and developing strategies to improve the process of discharge from these settings.
2. Methods
2.1. Study design
We conducted a retrospective study using the VA Corporate Data Warehouse (CDW). Our cohort consisted of VA users with an admission to VA acute inpatient or residential mental health treatment settings (herein after referred to as acute inpatient and residential) between 2003 and 2019. We assigned acute inpatient versus residential setting using bed section codes. Patients could be admitted multiple times during the study period. We included discharge types (irregular and regular) that align with definitions used by the Center for Medicare and Medicaid Services (CMS), excluding deaths and transfers [18]. Irregular discharges included both irregular and other unplanned discharges. Because our study focuses on mental health conditions, we excluded patients who were discharged with a primary diagnosis of delirium, dementia or medical rehabilitation, representing approximately 6% of admissions.
This study was approved by the Veteran’s institutional review board of Northern New England.
2.2. Covariates
Individual-level covariates included age, gender, race/ethnicity (Black, Hispanic, White, Other), marital status, pre-admission diagnoses, risk for homelessness, rurality, primary discharge diagnosis, setting, LOS, and year of admission. We created the following age categories: 18–35, 36–49, 50–59, 60–69, and 70+ years. We assessed marital status as of the time of admission. We assessed pre-admission diagnoses and homelessness in the two years prior to admission. We measured homelessness using a combination of ICD codes and clinic codes for use of homelessness-related services [19]. We summarized preadmission diagnoses as mental and physical health indices using a published list of related International Classification of Diseases Version 9 and 10 (ICD-9 and ICD-10) codes [20]. We coded the number of Diagnostic and Statistical Manual of Mental Disorders Version 5 categories (0–1, 2–3, and 4+ ) [21]. We coded physical health diagnoses based on the number of non-mental health Elixhauser conditions (0, 1, and 2+ ) [22]. We excluded hypertension without complications, collapsed diabetes with and without complications into a single condition, and collapsed cancer diagnoses into a single condition. We created primary discharge diagnosis categories: SUD, Alcohol Use Disorders (AUD), Bipolar Disorders, Depressive Disorders, Psychotic Disorders, Trauma-Related Disorders, and Other Mental Health Disorders. For all diagnostic groupings, we required a single instance from an inpatient facility or two or more outpatient encounters 7 to 365 days apart. We identified zip code of residence annually and used the Rural-Urban Commuting Area (RUCA) classification scheme to define RUCA codes 1–3 as urban and all others as rural [20,23]..
Of note, in our adjusted models, we divided year of admission into four time periods (2003–07, 2008–11, 2012–15, and 2016–19) for ease of interpretation. In our trend analysis, we looked at yearly rates.
2.3. Calculation of irregular discharge rate
We calculated irregular discharges rates as follows: 1) the numerator included the total number of discharges that were coded as irregular within the first 90 days of admission, and 2) the denominator included the sum total of discharges that were coded as irregular plus those that were coded as regular. We coded any discharges that were associated with stays longer than 90 days as ‘regular.’ While a LOS of up to 90 days is common in some VA residential treatment programs [24], we were concerned that LOS greater than 90 days were unusual and might indicate a housing challenge, rather than ongoing mental health treatment.
2.4. Descriptive analysis
We performed descriptive analyses in order to characterize the study population. We stratified the population by discharge type, comparing strata using chi-squared tests for dichotomous outcomes and independent samples t-test for continuous outcomes. We repeated this approach in order to examine differences in characteristics based on treatment setting (acute inpatient versus residential).
To evaluate trends in irregular discharge rates over time, we calculated annual crude and adjusted irregular discharge rates (stratified by treatment setting) and plotted these results by year (2003–2019).
2.5. Logistic regression analysis
We used logistic regression to calculate odds ratio (OR) and 95% confidence intervals (CI) for irregular discharge across the study population, using the covariates described above as predictors. We performed the analysis stratified on treatment setting (acute inpatient, residential) because of differences in LOS across treatment settings. While we censored patients at 90 days LOS in our primary analysis, we conducted sensitivity analysis that censored patients at 30, 60 and 182 days. Because patients could contribute discharge data multiple times during the study period, we used a Sandwich estimator to produce appropriate standard errors [25].
2.6. Survival analysis
We used a multivariate Cox proportional hazards regression model to calculate hazard ratios (HR) and 95% CI for irregular discharge across the study population, using the covariates described earlier as predictors. Because this approach can account for the differences in LOS among acute inpatient and residential settings, we did not further stratify by setting. However, for descriptive purposes, we generated two Kaplan-Meier estimates and survival curves in order to compare the probability of no irregular discharge in the first 90 days of admission among acute inpatient and residential settings. Similar to our logistic regression analysis, we performed a sensitivity analysis whereby we created additional models that censored data at 30, 60 and 182 days.
We performed data management and analysis using SAS Version 9.4 (SAS Institute, Cary NC).
3. Results
There were 1,816,817 total admissions from VA acute inpatient or residential settings between 2003 and 2019 (see Supplemental Appendix). Patient characteristics were generally similar across strata with a few notable exceptions. First, residential patients were far more likely to be of Black race (35.0%), have a SUD (27.8%) or AUD (43.7%) and be at risk for homelessness (44.4%). Second, LOS was over six-fold higher in residential (median 37.0 days (Interquartile range (IQR): 57.0)) versus acute inpatient settings (median 6.0 days (IQR 8.0)). A large proportion of residential stays had LOS greater than 90 days as compared to acute stays (22% versus 0.7%). Third, the irregular discharge rates were four-fold higher in residential versus acute inpatient settings. Related to these latter findings, over the past 16 years, the rates of irregular discharge from residential settings have well outpaced those of acute settings. While rates in acute inpatient settings modestly fell in the past 10 years, the rates in residential settings continued to rise.
There were 594,432 unique patients admitted to VA acute inpatient or residential settings between 2003 and 2019. Table 1 describes these patients as of their last admission; 6.3% (37,627) of which ended in an irregular discharge within 90 days of admission. While almost all differences between patients with irregular versus regular discharge were significant in this large cohort, patients with an irregular discharge most notably were more often admitted to a residential versus acute inpatient setting. In addition, patients with irregular discharge had shorter LOS, had more prior mental health admissions, were more likely to be at risk for homelessness, more likely to be divorced, and more often had a primary discharge diagnosis of SUD or AUD. Conversely, patients with irregular discharges were much less likely to have a primary discharge diagnosis of mental health disorders such as psychotic or depressive disorders. Finally, while irregular discharge was slightly more common in younger age groups, those over 70 were far less likely to have an irregular versus regular discharge.
Table 1.
Characteristics of patients based on last admission in acute inpatient or residential settings, stratified by discharge type, Department of Veterans Affairs 2003–2019.*
| Discharge type | |||
|---|---|---|---|
|
| |||
| Overall % (N) | Regular % (N) | Irregular % (N) | |
| Unique Patients | 100.0 (594,432) | 93.7 (556,805) | 6.3 (37,627) ¥ |
| Gender, female | 8.6 (51,405) | 8.7 (48,707) | 7.2 (2698) ¥ |
| Gender, male | 91.4 (543,027) | 91.3 (508,098) | 92.8 (34,929) ¥ |
| Age at admission | |||
| Mean years (SD) | 50.6 (13.8) | 50.8 (13.9) | 47.8 (12.7) ¥ |
| 18–35 years | 18.1 (107,392) | 17.8 (99,113) | 22.0 (8279) ¥ |
| 36–49 years | 23.1 (137,439) | 22.8 (127,042) | 27.6 (10,397) ¥ |
| 50–59 years | 31.7 (188,270) | 31.7 (176,339) | 31.7 (11,931) |
| 60–69 years | 20.9 (124,312) | 21.2 (118,261) | 16.1 (6051) ¥ |
| 70+ years | 6.2 (37,019) | 6.5(36,050) | 2.6 (969) ¥ |
| Marital status | |||
| Never | 26.6 (158,051) | 26.5 (145,565) | 27.9 (10.486) ¥ |
| Divorced/Widowed/Separated | 43.1 (256,233) | 42.8 (238,564) | 47.0 (17,669) ¥ |
| Married | 29.0 (172,433) | 29.3 (163,274) | 24.3 (9159) ¥ |
| Unknown | 1.3 (7715) | 1.3 (7402) | 0.8 (313) ¥ |
| Race/Ethnicity | |||
| Black | 26.8 (159,011) | 26.7 (148,898) | 26.9 (10,113) |
| Hispanic | 5.1 (30,318) | 5.1 (28,473) | 4.9 (1845) |
| White | 64.0 (380,248) | 64.0 (356,131) | 64.1 (24,117) |
| Other | 2.8 (16,754) | 2.8 (15,604) | 3.1 (1150) § |
| Unknown | 1.4 (8101) | 1.4 (7699) | 1.1 (402) ¥ |
| Primary discharge diagnosis | |||
| Substance use disorders | 14.6 (86,802) | 13.7 (76,113) | 28.4 (10,689) ¥ |
| Alcohol use disorders | 24.5 (145,784) | 23.8 (132,684) | 34.8 (13,104) ¥ |
| Bipolar disorders | 7.5 (44,483) | 7.7 (42,767) | 4.6 (1717) ¥ |
| Depressive disorders | 21.6 (128,171) | 22.3 (124,126) | 10.8 (4045) ¥ |
| Psychotics disorders | 10.4 (61,606) | 10.8 (59,877) | 4.6 (1729) ¥ |
| Trauma-related disorders | 19.0 (112,840) | 19.2 (107,117) | 15.2 (5723) ¥ |
| Other MH disorders | 2.5 (14,741) | 2.5 (14,121) | 1.6 (620) ¥ |
| Pre-existing conditions | |||
| At risk for homelessness | 26.8 (159,415) | 26.0 (144,995) | 38.3 (14,420) ¥ |
| 0–1 MH conditions | 26.6 (158,276) | 27.1 (150,702) | 20.1 (7574) ¥ |
| 2–3 MH conditions | 34.1 (202,824) | 34.2 (190,185) | 33.6 (12,639) § |
| 4+ MH conditions | 39.3 (233,332) | 38.8 (215,918) | 46.3 (17,414) ¥ |
| 0 PH condition | 48.9 (290,473) | 48.7 (271,264) | 51.1 (19,209) ¥ |
| 1 PH condition | 22.9 (136,229) | 22.9 (127,442) | 23.4 (8787) § |
| 2+ PH conditions | 28.2 (167,730) | 28.4 (158,099) | 25.6 (9631) ¥ |
| Urban-rural classification | |||
| Urban | 81.1 (481,795) | 81.0 (451,245) | 81.2 (30,550) |
| Rural | 18.9 (112,575) | 19.0 (105,499) | 18.8 (7076) |
| Unknown | 0.0 (62) | 0.0 (61) | 0.0 (1) |
| LOS, mean days (SD) | 29.6 (51.1) | 0.3 (52.4) | 20.2 (22.8) ¥ |
| Treatment Setting | |||
| Acute Inpatient | 65.1 (386,762) | 67.2 (374,336) | 33.0 (12,426) ¥ |
| Residential | 34.9 (207,670) | 32.8 (182,469) | 67.0 (25,201) ¥ |
| Number of mental health admissions | |||
| Mean | 3.1 (4.8) | 3.0 (4.7) | 3.6 (5.3) ¥ |
| One | 50.4 (299,536) | 51.1 (283,827) | 41.7 (15,709) ¥ |
| Two | 18.4 (109,359) | 18.3 (101,905) | 19.8 (7454) ¥ |
| Three + | 31.2 (185,537) | 30.7 (171,073) | 38.4 (14,464) ¥ |
LOS = Length of Stay; MH = Mental Health; N = Number; pH = Physical Health; SD = Standard Deviation; % = Percentage;
Events are censored at 90 days such that any admission >90 days is characterized as a ‘regular discharge’.
Comparison between irregular and regular discharges, p < 0.001.
Comparison between irregular and regular discharges, p ≤ 0.04.
Of note, race information was missing for some patients but overall was rare (<1.5%).
3.1. Logistic regression analysis
As shown in Table 2, for acute inpatient stays as well as residential stays younger age had a large effect on irregular discharge (OR > 2.0). In addition, there was a large effect of time period on irregular discharge when comparing years 2016 and 2019 versus 2003–2007 (OR < 0.50), although this finding only held true for acute inpatient stays. While several other variables were significantly associated with irregular discharge, all had relatively small effects. For example, in the acute inpatient setting, covariates such as female gender and Black race had ORs of 0.9 and 0.8, respectively.
Table 2.
Adjusted odds of irregular discharge within 90 days of admission stratified by treatment setting, Department of Veterans Affairs, 2003–2019.*
| Variable | Level | Odds Ratio | Lower 95% CI | Upper 95% CI |
|---|---|---|---|---|
| Acute inpatient setting | ||||
| Age in years | ||||
| 18–35 | 311 | 2.88 | 3.37 | |
| 36–49 | 2.41 | 2.23 | 2.60 | |
| 50–59 | 1.96 | 1.81 | 2.12 | |
| 60–69 | 1.52 | 1.41 | 1.65 | |
| 70+ | Reference | Reference | Reference | |
| Primary discharge diagnosis | ||||
| Substance Use Disorders | 1.60 | 1.49 | 1.71 | |
| Alcohol Use Disorders | 1.56 | 1.46 | 1.67 | |
| Bipolar Disorders | 1.02 | 0.95 | 1.09 | |
| Depressive Disorders | 0.73 | 0.68 | 0.78 | |
| Psychotics Disorders | 0.97 | 0.91 | 1.05 | |
| Trauma-Related Disorders | 0.85 | 0.79 | 0.91 | |
| Other MH Disorders | Reference | Reference | Reference | |
| Pre-existing health conditions | ||||
| At Risk for Homelessness | 1.26 | 1.22 | 1.29 | |
| 2–3 MH conditions | 1.05 | 1.02 | 1.09 | |
| 4+ MH conditions | 1.12 | 1.09 | 1.16 | |
| 0–1 MH conditions | Reference | Reference | Reference | |
| 1 PH condition | 0.98 | 0.96 | 1.01 | |
| 2+ PH conditions | 0.99 | 0.96 | 1.02 | |
| 0 PH condition | Reference | Reference | Reference | |
| Gender | Female | 0.93 | 0.89 | 0.97 |
| Race/Ethnicity | ||||
| Black | 0.77 | 0.75 | 0.80 | |
| Hispanic | 0.82 | 0.77 | 0.87 | |
| Other | 0.90 | 0.83 | 0.97 | |
| White | Reference | Reference | Reference | |
| Marital status | ||||
| Never | 0.95 | 0.92 | 0.98 | |
| Divorced/Widowed/Separated | 1.01 | 0.98 | 1.04 | |
| Unknown | 0.70 | 0.62 | 0.80 | |
| Married | Reference | Reference | Reference | |
| Urban-rural classification | ||||
| Urban | 1.12 | 1.09 | 1.16 | |
| Year | ||||
| 2008–2011 | 0.78 | 0.76 | 0.80 | |
| 2012–2015 | 0.59 | 0.58 | 0.61 | |
| 2016–2019 | 0.47 | 0.45 | 0.49 | |
| 2003–2007 | Reference | Reference | Reference | |
| Residential Setting | ||||
| Age in years | ||||
| 18–35 | 2.0 | 1.86 | 2.16 | |
| 36–49 | 1.48 | 1.37 | 1.59 | |
| 50–59 | 1.21 | 1.12 | 1.30 | |
| 60–69 | 1.08 | 1.00 | 1.17 | |
| 70+ | Reference | Reference | Reference | |
| Primary Discharge Diagnosis | ||||
| Substance Use Disorders | 1.48 | 1.36 | 1.61 | |
| Alcohol Use Disorders | 1.11 | 1.02 | 1.21 | |
| Bipolar Disorders | 1.16 | 1.06 | 1.28 | |
| Depressive Disorders | 1.01 | 0.92 | 1.11 | |
| Psychotics Disorders | 1.29 | 1.17 | 1.42 | |
| Trauma-Related Disorders | 0.83 | 0.76 | 0.90 | |
| Other MH Disorders | Reference | Reference | Reference | |
| Pre-Existing Health Conditions | ||||
| At Risk for Homelessness | 1.42 | 1.40 | 1.45 | |
| 2–3 MH conditions | 1.16 | 1.13 | 1.19 | |
| 4+ MH conditions | 1.46 | 1.42 | 1.50 | |
| 0–1 MH conditions | Reference | Reference | Reference | |
| 1 PH condition | 1.02 | 1.00 | 1.04 | |
| 2+ PH conditions | 1.01 | 0.99 | 1.03 | |
| 0 PH condition | Reference | Reference | Reference | |
| Gender | Female | 090 | 0.87 | 0.94 |
| Race/Ethnicity | ||||
| Black | 0.96 | 0.94 | 0.98 | |
| Hispanic | 1.09 | 1.04 | 1.14 | |
| Other | 1.16 | 1.10 | 1.23 | |
| White | Reference | Reference | Reference | |
| Marital Status | ||||
| Never | 1.07 | 1.05 | 1.10 | |
| Divorced/Widowed/Separated | 1.11 | 1.08 | 1.13 | |
| Unknown | 1.01 | 0.90 | 1.14 | |
| Married | Reference | Reference | Reference | |
| Urban-Rural Classification | ||||
| Urban | 1.04 | 1.02 | 1.07 | |
| Year | ||||
| 2008–2011 | 0.83 | 0.81 | 0.85 | |
| 2012–2015 | 0.95 | 0.93 | 0.97 | |
| 2016–2019 | 1.05 | 1.03 | 1.08 | |
| 2003–2007 | Reference | Reference | Reference | |
CI = Confidence Intervals; MH = Mental Health; PH = Physical Health.
Events are censored at 90 days such that any admission >90 days is characterized as a ‘regular discharge’.
In our sensitivity analysis with censoring at 30, 60 and 182 days, we found that our results did not notably change.
3.2. Survival analysis
In the time to event analysis, younger age remained a central predictor of irregular discharge for individuals between the ages of 18–35 as well as 36–49, compared to those aged 70 or greater (HR > 2) (see Table 3). While a few other covariates significantly predicted irregular discharge (HRs: 1.06–1.6), the only covariate that stood out was SUD (HR = 1.57). There were also some covariates with HR <1 such as Black race.
Table 3.
Hazard ratios for irregular discharge within 90 days of admission among patients discharged from acute inpatient or residential settings, Department of Veterans Affairs, 2003–2019.*
| Variable | Level | Hazard Ratio | Lower 95% CI | Upper 95% CI |
|---|---|---|---|---|
| Age in years | ||||
| 18–35 | 2.97 | 2.92 | 3.03 | |
| 36–49 | 2.15 | 2.10 | 2.21 | |
| 50–59 | 1.78 | 1.72 | 1.83 | |
| 60–69 | 1.49 | 1.43 | 1.54 | |
| 70+ | Reference | Reference | Reference | |
| Primary discharge diagnosis | ||||
| Substance Use Disorders | 1.57 | 1.51 | 1.62 | |
| Alcohol Use Disorders | 1.30 | 1.24 | 1.35 | |
| Bipolar Disorders | 0.85 | 0.80 | 0.91 | |
| Depressive Disorders | 0.72 | 0.66 | 0.77 | |
| Psychotics Disorders | 0.72 | 0.66 | 0.78 | |
| Trauma-Related Disorders | 0.78 | 0.73 | 0.83 | |
| Other MH Disorders | Reference | Reference | Reference | |
| Pre-existing health conditions | ||||
| At Risk for Homelessness | 1.28 | 1.27 | 1.30 | |
| 2–3 MH conditions | 1.06 | 1.04 | 1.08 | |
| 4+ MH conditions | 1.29 | 1.27 | 1.31 | |
| 0–1 MH conditions | Reference | Reference | Reference | |
| 1 PH condition | 1.02 | 1.00 | 1.03 | |
| 2+ PH conditions | 1.02 | 1.00 | 1.04 | |
| 0 PH condition | Reference | Reference | Reference | |
| Gender | Female | 0.94 | 0.91 | 0.96 |
| Race/Ethnicity | ||||
| Black | 0.86 | 0.85 | 0.88 | |
| Hispanic | 0.93 | 0.90 | 0.97 | |
| Other | 1.00 | 0.95 | 1.04 | |
| White | Reference | Reference | Reference | |
| Marital status | ||||
| Never | 0.92 | 0.90 | 0.94 | |
| Divorced/Widowed/Separated | 0.98 | 0.96 | 1.00 | |
| Unknown | 0.79 | 0.71 | 0.87 | |
| Married | Reference | Reference | Reference | |
| Urban-rural classification | ||||
| Urban | 1.06 | |||
| Year | ||||
| 2008–2011 | 0.82 | 0.80 | 0.84 | |
| 2012–2015 | 0.79 | 0.77 | 0.81 | |
| 2016–2019 | 0.80 | 0.78 | 0.82 | |
| 2003–2007 | Reference | Reference | Reference | |
| Treatment setting | ||||
| Acute Inpatient | 1.06 | 1.04 | 1.07 | |
| Residential | Reference | Reference | Reference | |
CI = Confidence Intervals; MH = Mental Health; PH = Physical Health.
Events are censored at 90 days such that any admission >90 days is characterized as a ‘regular discharge’.
Fig. 1 highlights that irregular discharge rates in acute inpatient and residential settings were very similar for the first week or so of admission. In assessing the hazard function for the first month (results now shown), we noted higher risk for irregular discharge for the first four days in the acute setting. After that time, the hazard was then lower for acute stays.
Fig. 1.
Probability of no irregular discharge in the first 90 days of admission among acute inpatient and residential mental health treatment settings within the Department of Veterans Affairs healthcare system, 2003–2019.
4. Discussion
Among more than 1.8 million discharges from acute inpatient or residential settings within the VA healthcare system, the most striking finding was a four-fold elevated rate of irregular discharge from the residential (15.9%) versus acute inpatient (3.9%) treatment setting. However, this finding was driven by longer LOS, a defining characteristic of the residential setting [17,24]. When accounting for time using a multivariate Cox proportional hazards model, there was actually a slightly higher risk of irregular discharge from the acute inpatient setting (HR = 1.06). Evaluating the shape of the survival curves, we found that the risk for irregular discharge in acute inpatient settings was primarily concentrated in the first week. Conversely, the risk for irregular discharge in the residential setting appeared constant over time such that doubling the LOS approximately doubled the risk for irregular discharge. Younger age was a central predictor of irregular discharge. Clinicians working in residential settings should be aware that almost 1 in 5 admissions will end in an irregular discharge and address potential barriers to treatment engagement early in course of admission, especially for younger patients.
In our stratified model examining the odds of irregular discharge in the acute inpatient setting, our findings were consistent with prior research, although studies have only reported on general hospital settings [5,6,8,12–14]. There is conflicting evidence about whether Black race predicts irregular discharge [5,6,13,14]. While Spooner et al. (2017) [5] and Ibrahim et al. (2007) [13] both found elevated overall adjusted odds of irregular discharge for Black patients (OR = 1.25 and 1.35, respectively), Spooner et al. (2017) found lower odds of irregular discharge among the stratum of black patients admitted with mental health or substance abuse conditions (OR = 0.87) [5]. Thus, our finding (OR = 0.77) was consistent and indicates the importance of accounting for individual- and system-level factors when evaluating factors that may predict irregular discharge. Both Spooner et al. (2017) [5] and Ibrahim et al. (2007) [13] found a higher risk of irregular discharge among urban patients, which was consistent with our finding (OR = 1.12). Unlike our results, however, Spooner et al. (2017)found among the stratum of patients admitted with mental health or substance abuse conditions that urban setting was not associated with irregular discharge [5]. Spooner et al. (2017) also found by far the highest prevalence of irregular discharge among patients admitted to a general hospital with a primary diagnosis of SUD [5]. This was consistent with our finding for acute inpatient stays, where patients with a primary discharge diagnosis of SUD had the highest adjusted odds of irregular discharge (1.60 and 1.56, respectively). Paralleling our findings that younger patients (18–35 years) had increased odds of irregular discharge when compared to older patients (70 years or greater) (OR = 3.11), Spooner et al. (2017) found that the oldest patients (80 years or greater) had the lowest odds of irregular discharge when compared to the youngest patients (age 18–29) (OR = 0.10) [5].
Our finding of a lower adjusted odds of irregular discharge (OR = 0.47) in the most recent period (2016–2019), compared to the earliest period (2003–2007) aligns with a declining trend in irregular discharge for patients with a primary mental health discharge diagnosis as noted by Spooner et al.5 The reasons for these declining trends, however, remain unclear. Over the past five years, the VA has implemented policies to improve discharge planning in mental health settings and mitigate harms [26]. These policies include a requirement that clinicians offer patients with an irregular discharge a follow-up visit within 24 h and within a week of discharge [27]. It is also important to note that unlike our results, Onukwugha et al. (2019) found that among patients with SUD diagnoses, the rates of irregular discharge have risen between 2012 and 2015 [4]. While we observed a rise in irregular discharge rates between 2012 and 2015, this was only in residential settings. The discrepancy may reflect differences in our analysis approach or included populations.
As with our results, studies highlight that drop-out (i.e. failure to complete treatment) [28] is a concern in residential settings [29–33]. In an analysis of data from the 2011 Substance Abuse and Mental Health Service Administration (SAMHSA) Treatment Episode Data Set (TEDS-D), Stahler et al (2016) found that a moderate proportion of patients (65.0%) completed the full extent of residential treatment [30]. Using a moderated logistic regression model, Stahler et al (2016) concluded that patients who were older were significantly more likely to complete the full extent of residential treatment than those who were younger (35–44 years versus 18–24 years, OR = 1.07). In a separate study of 3965 patients who were enrolled in a VA residential PTSD program, Smith et al (2019) found that younger age predicted early termination of treatment and more than one in four patients ended treatment early [31].
It is notable that the magnitude of our adjusted odds for irregular discharge in residential settings is smaller across clinical and demographic predictors. Thus, it appears to be more difficult to determine which patients will have an irregular discharge from a residential setting, as compared to an acute inpatient setting. It is possible that there are unmeasured factors that may contribute to irregular discharge in residential populations and our current models of risk adjustment do not sufficiently account for these covariates. Prior research has pointed to several predictors such as lack of access to on-site psychological services [34], poor program climate [35], drug of choice (cannabis, cocaine, heroin) [33], more drug users in social network [36], cognitive problems [37], higher symptom severity [35,37], hostility [34] and lower readiness to change [29]. In a qualitative study, Simon et al (2019) highlighted that reasons for irregular discharge among SUD populations included undertreatment of symptoms, poorly controlled pain, concerns about stigma and highly restrictive unit policies [35]. One major difference we noted is that the adjusted odds of irregular discharge in the residential setting have increased in the most recent period (OR = 1.05), rather than improved as they have in the acute inpatient setting. The reason for this is unclear. It could be because irregular discharge from residential settings has been under studied, and thus, may not have been the focus of improvement efforts.
Because patients leaving acute inpatient or residential settings may be at higher risk for adverse outcomes such as suicide [2,3,38], our findings highlight the need for an improved understanding of modifiable factors that contribute to irregular discharge in these settings. These data could assist in developing effective interventions to reduce this risk. Our findings emphasize a critical need for clinicians (and especially those in residential settings) to be aware of the high risk for irregular discharge among younger patients. These patients will benefit from interventions to improve treatment engagement and mitigate harms associated with irregular discharge. These patients may benefit from interventions that increase social support [39]. . In a randomized controlled trial of the Community Reinforcement Approach and Family Training for Treatment Retention (CRAFT-T) intervention in patients who received treatment for opioid use disorder in a detoxification program, Brigham et al (2014) found that CRAFT-T resulted in significantly longer times to dropout from treatment (HR = 0.4) when family were involved as part of the intervention [40]. Patients may also benefit from interventions to improve the patient experience during hospital admission [35]. Feeney et al. (2007) found that after redesigning an inpatient mental health unit to improve patient experience, patients were significantly less likely to leave AMA (OR = 0.35) [41]. Of course, patients with an irregular discharge may also require closer follow-up after discharge because they are at much higher risk for no showing their outpatient appointments [42].
4.1. Strengths and limitations
Our study has strengths and limitations. We included a robust dataset that spans 2003 and 2019 and covers the largest, integrated healthcare system in the US. To the best of our knowledge, our study is the first to report on irregular discharge rates in acute inpatient and residential settings. Our study, however, focused solely on VA-provided care. VA users share several characteristics that distinguish them from the general US population including greater comorbidity, higher rates of homelessness and worse socioeconomic status [43]. Although the features of VA users are very similar to psychiatrically hospitalized patients [44], VA users may nonetheless be a unique population. For example, VA users have access to a broad range of mental health treatments [45]. As such our findings may not necessarily be generalizable to other populations. Because our study relied entirely on administrative data, our analysis is limited by the quality and availability of the data in these resources. In particular, race data is imperfect, and the methods to identify these data vary. We encountered some missing race data, but the proportion of observations with missing data was reassuringly small (<1.5%). It is possible that our observation of a temporal effect on adjusted rates of irregular discharge over time could have been partially confounded by a simultaneous increase in diagnosed mental health conditions. We noted in our study population that the prevalence of mental health conditions did increase during the study period.
Our Cox analysis censored admissions after 90 days, while our logistic analysis characterized all stays>90 days as regular. There were large discrepancies in the proportion of admissions with LOS > 90 days in acute inpatient versus residential settings (0.7% vs 22%). Fortunately, our sensitivity analysis found very similar results whether we shortened the maximum follow-up time to 30 days or lengthened it to 182 days. Our dual faceted approach to the analysis of a binary outcome, in fact, is novel. This strategy offers greater insights than could not have been reached had we relied solely on logistic regression or time to event analysis. Researchers should consider this approach when evaluating events that can have greatly different risk time exposure, especially if the risk time is associated with covariates of interest. Using logistic regression in this context would result in effect estimates that could be considered biased. For example, in hospital settings complications that follow a procedure such as an infection or a fall may appear to be associated with increasing age, but if older patients need a greater amount of time to recover from a procedure than these patients will have, on average, longer LOS. As such, they will also acquire more risk time to experience the adverse event. We also found that irregular discharge was much more frequent in residential settings (as compared to acute inpatient settings). In our time to event analysis, however, the risk of irregular discharge in the first week or so of admission was very similar in both settings. It was only after this initial period that rates became much higher among those in residential settings versus those in acute inpatient settings.
Finally, we did not examine regional predictors of irregular discharge or adverse outcomes such as mortality. Future work should look to determine whether there is regional variation in irregular discharge rates and evaluate the relationship between irregular discharge and post-discharge mortality.
4.2. Conclusions
Our study highlights that younger patients with an acute inpatient or residential stay are at particularly high risk for irregular discharge. Although irregular discharge rates are initially similar across settings, the risk for irregular discharge after a residential stay remains elevated over time. Notably, 20% of residential admissions end in an irregular discharge. Given that irregular discharge is associated with numerous adverse outcomes including death by suicide, future research should evaluate interventions that may address individual- and system-level drivers of irregular discharge especially in the residential setting.
Supplementary Material
Acknowledgements
No additional individuals were involved in this work.
Funding
This study was funded by the VA National Center for Patient Safety Center of Inquiry Program, Ann Arbor MI (PSCI-WRJ-SHINER) and the VA Office of Rural Health, Veterans Rural Health Resource Center, White River Junction VT (ORH: 15533). The supporters had no role in the design, analysis, interpretation, or publication of this study. Dr. Riblet is the recipient of the VA Clinical Science Research and Development Career Development Award Program (MHBC-007–19F). Dr. Levis is the recipient of a VA New England Early Career Development Award (V1CDA-2020–60). The views expressed in this article do not necessarily represent the views of the Department of Veterans Affairs or of the United States government.
Footnotes
Conflicts of Interest
The authors have no conflicts of interest to report.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.org/10.1016/j.genhosppsych.2021.06.009.
References
- [1].Valevski A, Zalsman G, Tsafrir S, Lipschitz-Elhawi R, Weizman A, Shohat T. Rate of readmission and mortality risks of schizophrenia patients who were discharged against medical advice. Eur Psychiatry Oct 2012;27(7):496–9. 10.1016/j.eurpsy.2011.04.009. [DOI] [PubMed] [Google Scholar]
- [2].Kuo CJ, Tsai SY, Liao YT, Lee WC, Sung XW, Chen CC. Psychiatric discharge against medical advice is a risk factor for suicide but not for other causes of death. J Clin Psychiatry 2010;71(6):808–9. 10.4088/JCP.09l05404blu. [DOI] [PubMed] [Google Scholar]
- [3].Riblet N, Richardson JS, Shiner B, Peltzman TR, Watts BV, McCarthy JF. Death by suicide in the first year after irregular discharge from inpatient hospitalization. Psychiatr Serv 2018;69(9):1032–5. 10.1176/appi.ps.201800024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [4].Onukwugha E, Alfandre D. Against medical advice discharges are increasing for targeted conditions of the medicare hospital readmissions reduction program. J Gen Intern Med 2019;34(4):515–7. 10.1007/s11606-018-4765-3. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [5].Spooner KK, Salemi JL, Salihu HM, Zoorob RJ. Discharge against medical advice in the United States, 2002–2011. Mayo Clin Proc 2017;92(4):525–35. [DOI] [PubMed] [Google Scholar]
- [6].Tawk R, Freeds S, Mullner R. Associations of mental, and medical illness with against medical advice discharges: the National Hospital Discharge Survey, 1988–2006. Adm Policy Ment Health 2013;40:124–32. [DOI] [PubMed] [Google Scholar]
- [7].Brook M, Hilty DM, Liu W, Hu R, Frye MA. Discharge against medical advice from inpatient psychiatric treatment: a literature review. Psychiatr Serv 2006;57(8): 1192–8. 10.1176/ps.2006.57.8.1192. [DOI] [PubMed] [Google Scholar]
- [8].Sclar DA, Robison LM. Hospital admission for schizophrenia and discharge against medical advice in the United States. J Clin Psychiatry 2010;2(2):e1–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [9].Choi M, Kim H, Qian H, Palepu A. Readmission rates of patients discharged against medical advice: a matched cohort study. PLoS One 2011;6(9):e24459. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [10].Olufajo OA, Metcalfe D, Yorkgitis BK, et al. Whatever happens to trauma patients who leave against medical advice? Am J Surg 2016;211(4):677–83. [DOI] [PubMed] [Google Scholar]
- [11].Tan SY, Feng JY, Joyce C, Fisher J, Mostaghimi A. Association of hospital discharge against medical advice with readmission and in-hospital mortality. JAMA Netw Open 2020;3(6). 10.1001/jamanetworkopen.2020.6009. e206009. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [12].Onukwugha E, Nagarajan M, Offurum A, Gulati M, Alfandre D. Multi-Level predictors of discharges against medical advice: Identifying contributors to variation using an all-payer database. J Healthc Qual Jan-Feb 01 2021;43(1). 10.1097/jhq.0000000000000252. e8–e19. [DOI] [PubMed] [Google Scholar]
- [13].Ibrahim SA, Kwoh CK, Krishnan E. Factors associated with patients who leave acute-care hospitals against medical advice. Am J Public Health 2007;97(12): 2204–8. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [14].Franks P, Meldrum S, Fiscella K. Discharges against medical advice: are race/ethnicity predictors? J Gen Intern Med 2006;21(9):955–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [15].Healthcare cost and Utilization Project (HCUP). Nationwide Inpatient Sample Available at, https://hsrr.nlm.nih.gov/hsrr_search/record_details/1525. Revised March 27, 2019.
- [16].Glick ID, Sharfstein SS, Schwartz HI. Inpatient psychiatric care in the 21st century: the need for reform. Psychiatr Serv 2011;62(2):206–9. [DOI] [PubMed] [Google Scholar]
- [17].Commission on the Accreditation of Rehabilitation Facilities (CARF). Behavioral Health Program Descriptions 2020. p. 13–4. Available at, http://www.carf.org/programdescriptions/bh/2020. [Google Scholar]
- [18].Research Data Assistance Center (ResDAC). Patient discharge status code (FFS) Available at, resdac.org/cms-data/variables/patient-discharge-status-code-ffs; 2021. [Google Scholar]
- [19].Peterson R, Gundlapalli AV, Metraux S, et al. Identifying homelessness among veterans using VA administrative data: opportunities to expand detection criteria. PLoS One 2015;10(7):e0132664. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [20].Shiner B, Pelztman T, Cornelius SL, Gui J, Forehand J, Watts BV. Recent trends in the rural-urban suicide disparity among veterans using VA health care. J Behav Med 2020. 10.1007/s10865-020-00176-9. Online ahead of print. [DOI] [PubMed] [Google Scholar]
- [21].American Psychiatric Association. Diagnostic and statistical manual of mental disorders 5th ed. 2013. 10.1176/appi.books.9780890425596. Available at. [DOI] [Google Scholar]
- [22].Quan H, Sundararajan V, Halfon P, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005;43: 1130–9. [DOI] [PubMed] [Google Scholar]
- [23].West AN, Lee RE, Shambaugh-Miller MD, et al. Defining “rural” for veterans’ health care planning. J Rural Health 2010;26:301–9. [DOI] [PubMed] [Google Scholar]
- [24].Shiner B, Westgate CL, Simola V, Thompson R, Schnurr P, Cook JM. Measuring use of evidence-based psychotherapy for PTSD in VA residential treatment settings with clinician survey and electronic medical record templates. Mil Med 2018;183: e539. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [25].Huber PJ. The behavior of maximum likelihood estimates under non-standard conditions. In: Proceedings of the Fifth Berkley Symposium on Mathematical Statistics and Probability Berkley CA: University of California Press; 1967. p. 221–33. [Google Scholar]
- [26].Veterans Health Administration, Office of Mental Health and Suicide Prevention (OMHSP). Office of Mental Health and Suicdie Prevention Fact Sheet: Guidance on enhancing acute inpatient mental health and residential rehabilitation treatment program discharge planning and follow-up Washington, DC: Department of Veterans Affairs; 2017. [Google Scholar]
- [27].Veterans Health Administration. Mental Health Residential Rehabilitation Treatment Program (VHA memo 1162.02) Washington, DC: Author. (2019, July 15). [Google Scholar]
- [28].Brorson HH, Ajo Arnevik E, Rand-Hendriksen K, Duckert F. Drop-out from addiction treatment: A systematic review of risk factors. Clin Psychol Rev 2013;33 (8):1010–24. 10.1016/j.cpr.2013.07.007.2013/12/01/. [DOI] [PubMed] [Google Scholar]
- [29].Choi S, Adams SM, MacMaster SA, Seiters J. Predictors of residential treatment retention among individuals with co-occurring substance abuse and mental health disorders. J Psychoactive Drugs 2013;45(2):122–31. 10.1080/02791072.2013.785817. [DOI] [PubMed] [Google Scholar]
- [30].Stahler GJ, Mennis J, DuCette JP. Residential and outpatient treatment completion for substance use disorders in the U.S.: Moderation analysis by demographics and drug of choice. Addict Behav 2016;58:129–35. 10.1016/j.addbeh.2016.02.030. [DOI] [PubMed] [Google Scholar]
- [31].Smith NB, Sippel LM, Rozek DC, Hoff RA, Harpaz-Rotem I. Predictors of dropout from residential treatment for Posttraumatic Stress Disorder among military veterans. Front Psychol 2019;10(362). 10.3389/fpsyg.2019.00362. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [32].Stack K, Cortina J, Samples C, Zapata M, Arcand LF. Race, age, and back pain as factors in completion of residential substance abuse treatment by veterans. Psychiatr Serv 2000;51(9):1157–61. 10.1176/appi.ps.51.9.1157. [DOI] [PubMed] [Google Scholar]
- [33].Gundel R, Allen N III, Osborne S, Shwayhat S. Risk factors for early discharge from a residential addiction treatment program. J Addiction Res & Ther 2017;8(4). [Google Scholar]
- [34].Broome KM, Flynn PM, Simpson DD. Psychiatric comorbidity measures as predictors of retention in drug abuse treatment programs. Health Serv Res 1999;34 (3):791–806. [PMC free article] [PubMed] [Google Scholar]
- [35].Simon R, Snow R, Wakeman S. Understanding why patients with substance use disorders leave the hospital against medical advice: a qualitative study. Subst Abus 2019:1–7. [DOI] [PubMed] [Google Scholar]
- [36].Arnaudova I, Jin H, Amaro H. Pretreatment social network characteristics relate to increased risk of dropout and unfavorable outcomes among women in a residential treatment setting for substance use. J Subst Abuse Treat 2020;116:108044. 10.1016/j.jsat.2020.108044. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [37].McKellar J, Kelly J, Harris A, Moos R. Pretreatment and during treatment risk factors for dropout among patients with substance use disorders. Addict Behav 2006;31(3):450–60. 10.1016/j.addbeh.2005.05.024. [DOI] [PubMed] [Google Scholar]
- [38].Decker KP, Peglow SL, Samples CR, Cunningham TD. Long-term outcomes after residential substance use treatment: relapse, morbidity, and mortality. Mil Med 2017;182(1):e1589–95. [DOI] [PubMed] [Google Scholar]
- [39].Chan AC, Palepu A, Guh DP, et al. HIV-positive injection drug users who leave the hospital against medical advice: the mitigating role of methadone and social support. J Acquir Immune Defic Syndr 2004;35(1):56–9. 10.1097/00126334-200401010-00008. [DOI] [PubMed] [Google Scholar]
- [40].Brigham GS, Slesnick N, Winhusen TM, Lewis DF, Guo X, Somoza E. A randomized pilot clinical trial to evaluate the efficacy of community reinforcement and family training for treatment retention (CRAFT-T) for improving outcomes for patients completing opioid detoxification. Drug Alcohol Depend 2014;138:240–3. 10.1016/j.drugalcdep.2014.02.013. [DOI] [PMC free article] [PubMed] [Google Scholar]
- [41].Feeney L. Moving to a purpose built acute psychiatric unit on a general hospital site - does the new environment produce change for the better? Ir Med J 2007;1003(3): 391–3. [PubMed] [Google Scholar]
- [42].Miller MJ, Ambrose DM. The problem of missed mental healthcare appointments. Clin Schizophr Relat Psychoses 2019:177–84. [PubMed] [Google Scholar]
- [43].Eibner C, Krull H, Brown KM, Cefalu M, Mulcahy AWP MS. Current and projected characteristics and unique health care needs of the patient population served by the Department of Veterans Affairs. RAND Health Quarterly 2016;Vol 5(4):13. Available at, https://www.rand.org/pubs/periodicals/health-quarterly/issues/v5/n4/13.html. [PMC free article] [PubMed] [Google Scholar]
- [44].Lorine K, Goenjian H, Kim S, Steinberg AM, Schmidt K, Goenjian AK. Risk factors associated with psychiatric readmission. J Nerv Ment Dis 2015;203(6):425–30. 10.1097/NMD.0000000000000305. [DOI] [PubMed] [Google Scholar]
- [45].National Academies of Sciences, Engineering, and Medicine; Health and Medicine Division. Board on Health Care Services; Committee to Evaluate the Department of Veterans Affairs Mental Health Services. Evaluation of the Department of Veterans Affairs Mental Health Services3 Washington (DC): National Academies Press (US); 2018. Jan 31. The Veterans Health Administration’s Mental Health Services. Available at, https://www.ncbi.nlm.nih.gov/books/NBK499499/. [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.