ORIGINAL ARTICLE
Risk Factors Associated With Psychiatric Readmission Kim Lorine, PhD,* Haig Goenjian, MD,† Soeun Kim, PhD,‡ Alan M. Steinberg, PhD,§ Kendall Schmidt, BA,* and Armen K. Goenjian, MD*§ Abstract: The present study focused on identifying risk factors for early readmission of patients discharged from an urban community hospital. Retrospective chart reviews were conducted on 207 consecutive inpatient psychiatric admissions that included patients who were readmitted within 15 days, within 3 to 6 months, and not admitted for at least 12 months post-discharge. Findings indicated that a diagnosis of schizophrenia/schizoaffective disorder (OR = 18; 95% CI 2.70–117.7; p < 0.05), history of alcohol abuse (OR = 9; 95% CI 1.80–40.60; p < 0.05), number of previous psychiatric hospitalizations (OR = 2; 95% CI 1.28–3.73; p < 0.05), and type of residence at initial admission (e.g., homeless, OR = 29; 95% CI 3.99–217; p < 0.05) were significant risk factors for early readmission, where OR compares readmission group 1 versus group 3 in the multinomial logistic regression. Initial positive urine drug screen, history of drug abuse or incarceration, and legal status at initial admission did not predict early readmission. Reducing the risk factors associated with psychiatric readmissions has the potential to lead to the identification and development of preventative intervention strategies that can significantly improve patient safety, quality of care, well-being, and contain health care expenditures. Key Words: Psychiatric readmission, financial burden of readmission, readmission of schizophrenics, readmission of homeless (J Nerv Ment Dis 2015;203: 425–430)
I
n a study of 249 Florida hospitals of the 10 most common diagnoses that qualified for their criteria of potentially preventable readmissions, three were psychiatric diagnoses: schizophrenia was third, major depressive disorder (MDD) was fifth, and bipolar disorder was seventh (Goldfield et al., 2008). According to the 2013 Medicare Payment Advisory Commission Report to Congress (Hackbarth et al., 2013), 15% of Medicare beneficiaries discharged from hospitals in 2011 were readmitted within 30 days. A 2011 report by the Yale New Haven Health Services Corporation/Center for Outcome Research and Evaluation report (Horowitz et al., 2011) found that psychiatric readmission rates within 30 days of discharge for fee-for-service (FFS) Medicare beneficiaries 65 and older was 15.8%. In another study among Medicare beneficiaries, the cost of unplanned re-hospitalizations within 30 days of their discharge was estimated at $17.4 billion in 2004 (Jencks and Williams, 2009). The 30-day re-hospitalization rate for all illnesses was 19.6%, and for “psychosis” (the only psychiatric condition listed) it was 24.6%. Although these studies examined all causes of readmission, psychiatric readmission comprised a substantial proportion. Numerous studies have examined potential risk factors associated with psychiatric readmission. Some of these studies have been conducted in the United States (Jencks and Williams, 2009; Lyons et al., 1997; Olfson et al., 1999; Thompson et al., 2003), whereas other *Collaborative Neuroscience Network, LLC, Garden Grove; †Harbor UCLA Medical Center, Torrance, CA; ‡University of Texas Health Science Center, Division of Biostatistics, Houston, TX; and §UCLA/Duke University National Center for Child Traumatic Stress, Department of Psychiatry and Biobehavioral Sciences, University of California, Los Angeles, CA. Send reprint requests to Armen K. Goenjian, MD, UCLA/Duke University National Center for Child Traumatic Stress, Department of Psychiatry and Biobehavioral Sciences, University of California, 11150 W Olympic Blvd, Suite 650, Los Angeles, CA 90064. E-mail: [email protected]. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved. ISSN: 0022-3018/15/20306–0425 DOI: 10.1097/NMD.0000000000000305
studies were overseas (Bernardo and Forchuk, 2001; Loch, 2012; Silva et al., 2009; Zhang et al., 2011). These studies varied with regard to ethnic/racial mix, socioeconomic status, timing of assessments, and diagnostic categories evaluated. For example, some studies evaluated general psychiatric populations (Bernardo and Forchuk, 2001; Lyons et al., 1997; Thompson et al., 2003; Zhang et al., 2011), whereas others only evaluated patients with schizophrenia and schizoaffective disorder (SAD) (Fennig and Rabinowitz, 1999; Olfson et al., 1999; Schennach et al., 2012; Suzuki et al., 2003). Also, time of readmission varied across these studies from 30 days (Burke et al., 2013; Lyons et al., 1997), 3 months (Olfson et al., 1999), 6 months (Lyons et al., 1997; Thompson et al., 2003), and 1 year (Loch, 2012; Schennach et al., 2012; Zhang et al., 2011), or longer (Bernardo and Forchuk, 2001; Rosca et al., 2006). Even though these variations make generalizations across studies problematic, there are commonalities in many of the studies that may serve as guidelines to improve delivery of health care to severely ill psychiatric patients. Risk factors identified in studies among general psychiatric populations have included a diagnosis of schizophrenia (Rosca et al., 2006; Silva et al., 2009) and SAD (Thompson et al., 2003), presence of more severe symptoms and greater impairment in self-care (Lyons et al., 1997), younger age at first admission (Silva et al., 2009), use of restraints while hospitalized (Loch, 2012), number of previous admissions (Haywood et al., 1995; Silva et al., 2009; Zhang et al., 2011), alcohol intoxication, and posing a danger to others at the time of the initial admission (Zhang et al., 2011). In studies which included only patients with schizophrenia, risk for readmission included number of previous admissions (Olfson et al., 1999; Schennach et al., 2012), comorbid substance abuse (Olfson et al., 1999), medication noncompliance after discharge (Bernardo and Forchuk, 2001), not being in remission at time of discharge (Schennach et al., 2012), and noncompliance with outpatient clinic visits (Suzuki et al., 2003). The present study examined the association of sociodemographic, clinical, and post-discharge factors with readmission of psychiatric patients discharged from a general hospital located in an urban area of Los Angeles County. The objective was to identify risk factors associated with readmission within 15 days and within 3 to 9 months.
METHODS Participants For the present study, retrospective chart reviews were performed on 207 psychiatric inpatients from an urban community hospital. Patients were hospitalized for acute psychiatric illnesses. These patients were referred from local county and private emergency rooms, board and care facilities, residential facilities, mental health clinics, police, and private practitioners. Chart reviews were conducted on consecutive readmissions that fit in three time parameters: readmitted within 15 days (group 1, N = 62), readmitted within 3 to 6 months (group 2, N = 86), and not readmitted for at least 12 months post-discharge (group 3, N = 59). Patients within each group were independent and did not overlap (i.e., patients in group 1 were not included in groups 2 or 3). The admissions covered the period January through December 2010. Upon initial hospitalization, patients were stabilized (at a minimum, they no longer constituted a danger to themselves/others and
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015 Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jonmd.com
425
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Lorine et al.
were not gravely disabled). They were discharged with an appropriate medication prescription, aftercare instructions, and were referred back to their psychiatrist/therapist or to a new one if the patient did not have one or requested a new one. The majority returned back to their place of residence, which included board and cares, residential facilities, apartments, or homes. Some were referred to new facilities. For a subgroup of patients who were stabilized and refused placement, they were provided with medication prescriptions, follow-up care instructions, and were referred to a psychiatrist or a mental health clinic. IRB approval was received to conduct the current retrospective chart review.
Data Analysis Variables examined in the current study included age, gender, race, psychiatric diagnosis, history of incarceration, history of alcohol/ drug abuse, number of previous inpatient hospitalizations, type of insurance, length of stay of admission, residence admitted from (and whether patients lived alone or with family members), urine drug screen (UDS), legal status (voluntary/involuntary), medication compliance, and availability of outside support (e.g., family member, friend, church members, self-help groups). With regard to medications, those patients who were initially noncompliant for a few days but subsequently became compliant were classified as compliant. In building the regression model, we first compared demographic and potential risk factors of interest for the three groups. Next, a χ2 analysis was used to examine differences for each of the dichotomous variables (significance level was set at p < 0.05). We started with a full model including all significant variables from the univariate analysis and used backward elimination resulting in a final model where we included predictors of interest while controlling for significant confounders. Variables that were not statistically significant at the p < 0.05 level in this step were left out of the final model (e.g., gender, race). Although age showed slight significance, it was removed from the final model given that there was a heterogeneous pattern of direction between the groups and would not add benefit to the final analysis. Initial UDS, history of incarceration, and medication compliance, even though were not statistically significant, were added to the final model as they were considered clinically relevant. The final multinomial logistic regression calculated odds ratios (OR) and 95% confidence intervals (95%) CI for 10 predictors. SAS/STAT software was used to perform analyses.
RESULTS Table 1 presents the descriptive characteristics of patients included in the study. The majority of the sample (91%) was between ages 18 and 60. They were fairly evenly distributed between the age groups, with a slightly higher proportion in the 40–60 age range. Nine percent were over 60 years old. Male to female ratio was 3:2. Sixty-one percent of the patients were white, 31% were African-American, and 8% were Hispanic or Asian. Virtually all patients were unemployed, with 56% having a history of incarceration. Approximately 40% had no previous psychiatric hospitalization. Half of the patients had diagnosis of schizophrenia or SAD, 20% bipolar affective disorder (BAD), and 17% MDD. Almost half had a history of alcohol abuse and 63% drug abuse. Initial urine drug screen was positive for 34%. Thirty-one percent came from a board and care or a residential facility, 24% were homeless, and the remainder either lived alone or with someone else in an apartment or a house. Table 2 presents the comparison of variables within each group. The age distribution varied among the three groups, most notably for those over 60, where age was higher in group 3 (20%) versus groups 1 (7%) and 2 (2%). Diagnosis of schizophrenia/SAD was present with a higher proportion in group 1 (68%) and group 2 (50%), whereas major depression was higher in group 3 (31%). The proportion of patients with a history of alcohol abuse was higher in group 1 (61%) compared 426
www.jonmd.com
to the other two. Additionally, the proportion of patients with a history of drug abuse was higher in group 1 and group 2. The majority of patients in group 1 (40%) resided at board and care or residential facilities at the time of initial admission, whereas 50% of group 3 resided with family members versus 9% for group 1. There were higher proportions of homeless patients in both group 1 (36%) and group 2 (23%) versus group 3 (13%). Of interest was the initial UDS rate which was not significantly different across the three groups. Also UDS across diagnostic categories did not differ significantly among the three groups. Lastly, initial UDS by residence showed no significant difference among the groups. Medication compliance rates across the groups and for the different diagnostic categories across the group did not show significant difference. Also, medication compliance by place of residence across the groups was not significant. The average length of stay was 8.4 days. For the majority (97%), length of stay was between 1 and 15 days; for 27%, it was 1 to 5 days. Lastly, 85% of group 3 patients had no previous inpatient hospitalizations, whereas 44% of group 1 and 39% of group 2 patients had one to five previous inpatient hospitalizations. With regard to readmissions, the location where they came from was similar to the pattern noted at initial admission. The majority of patients in group 1 resided at board and cares or residential facilities, or were homeless. There was no significant difference between group 1 and group 2 with regard to readmission UDS and social support. Table 3 presents the OR and 95% CI of potential risk factors computed from the multinomial logistic regression with readmission group as the outcome variable with group 3 (no admission for a minimum of 12 months) as the reference group. The odds of being in group 1 relative to being in group 3 increased by a factor of 17.8 times given a diagnosis of schizophrenia/SAD. History of alcohol abuse increased the odds of being in group 1 by a factor of 8.5 and being in group 2 by a factor of 4. History of drug abuse did not increase the odds likelihood for early readmission. The odds ratio of being in group 1 was 27 for those admitted from board and care/residential facilities, 29-fold for the homeless, and almost 7.6-fold for those living in an apartment alone or with a non-family member. Each previous hospitalization increased the ratio of the odds of being in group 1 or 2 by a factor of 2. Lastly, length of stay at initial admit did not increase the odds likelihood for early readmission, but did increase slightly the odds likelihood for readmission within 3 to 9 months (1.26). History of incarceration, initial urine drug screen, medical compliance, and voluntary status were not significantly associated with being in the readmission groups.
DISCUSSION The present study evaluated the association of patient characteristics and clinical variables with readmission to a psychiatric unit of a general hospital located in an urban area. The study differed from previous studies in that it assessed readmission of patients within a 12-month window, including readmissions within 15 days, 3 to 9 months, and not readmitted for at least 12 months after discharge. The latter group served as the reference group. Risk factors that differentiated patients subsequently readmitted within 15 days of discharge compared to those not readmitted for at least 12 months included diagnosis of schizophrenia/SAD, which was associated with a high likelihood of being readmitted within 15 days (OR = 18) (Table 3). This finding supports and extends the findings of other studies. Two previous studies among patients discharged from a state hospital found SAD to be a risk factor for re-hospitalization (Thompson et al., 2003), whereas two studies from Brazil found a diagnosis of schizophrenia to be a risk factor (Gastal et al., 2000; Silva et al., 2009). However, these studies did not specifically assess early readmissions as times from discharge to readmission varied from 6 months to years (Haywood et al., 1995; Silva et al., 2009; Thompson et al., 2003). © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Factors Associated With Readmissions
TABLE 1. Descriptive Characteristics of Patients in the Readmission Study (N = 207) Variable
Category
Frequency
Percent
Age group
18–35 36–45 46–60 60+
54 58 77 18
26 28 37 9
Sex
M F
116 77
60 40
Race
White African-American Hispanic or Asian
126 64 17
61 31 8
Marital status
Single Divorced Widowed Married
169 11 4 13
86 6 2 7
Employment
Yes No
3 204
2 99
Schizophrenia/SAD BAD MDD Psychosis
97 39 33 24
50 20 17 12
History of incarceration
Yes No
115 91
56 44
History of alcohol abuse
Yes No
97 109
47 53
History of drug abuse
Yes No
130 76
63 37
Medicare/Medical or Medicare MediCal Uninsured “Cash”
56 135 16
27 65 8
Initial admit from
B/C or residential facility Residing alone or with non-family member Homeless Residing with family member
62 34 47 55
31 17 24 28
Initial discharge to
B/C or residential facilities Residing alone or with non-family member Residing with family member Own resources
86 18 47 56
42 9 23 27
Yes No
71 136
34 66
Medication compliance
Compliant Noncompliant
189 18
91 9
Legal status (voluntary)
Yes No
40 167
19 81
Danger to self (DTS) Danger to others (DTO) Gravely disabled (GD), GD/DTO, GD/DTS
111 32 37
62 18 21
Length of stay
1 to 5 6 to 10 11 to 15 16+
55 115 28 9
27 56 14 4
Previous hospitalizations
0 1 to 5 6 to 10 11+
83 69 26 29
40 33 13 14
Diagnosis
Insurance
Initial UDS
Involuntary status subtypes
B/C indicates board and care; UDS, urine drug screen.
© 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jonmd.com
427
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Lorine et al.
TABLE 2. Comparisons of Potential Risk Factors for Readmission Among the Three Study Groups (N = 207)
TABLE 2. (Continued)
Group 1 Group 2 Group 3 N = 62 N = 86 N = 59 Percent Percent Percent Age group***
Group 1 Group 2 Group 3 N = 62 N = 86 N = 59 Percent Percent Percent
18–35 36–45 46–60 60+
32 27 34 7
15 35 48 2
36 19 25 20
Medication compliance
Compliant Non-compliant
95 5
88 12
92 9
Legal status (voluntary)**
Yes No
31 70
15 85
14 86
Sex*
Male Female
69 31
61 40
47 53
Involuntary status subtypes
64
62
59
Race
White African-American Hispanic or Asian
53 37 10
65 31 4
63 24 14
14
17
25
23
22
16
Schizophrenia/SAD BAD Psychosis MDD
68 18 2 13
50 24 13 13
27 16 27 31
Danger to self (DTS) Danger to others (DTO) Gravely disabled (GD) GD/DTS, GD/DTO
46
36
–
12
23
–
History of incarceration
Yes No
61 39
57 44
49 51
History of alcohol abuse**
Yes No
61 39
41 59
42 58
Readmission from** B/C or residential facilities Residing alone or with non-family member Homeless Residing with family member
37 5
17 24
– –
History of drug abuse**
Yes No
67 33
72 28
46 54
Readmit UDS
Yes No
29 71
36 64
– –
Insurance**
Medicare/MediCal or Medicare MediCal Uninsured “Cash”
34
22
27
Outside support*
Yes No
60 40
71 29
– –
47 19
74 4
71 2
B/C or residential facility Residing alone or with non-family member Homeless Residing with family member
40
30
25
16
21
13
Values are column percentages within each group. The last three variables (readmission from, readmit UDS, and outside support) are not observed for group 3. The comparisons were done between group 1 and group 2. B/C indicates board and care; UDS, urine drug screen. *p < 0.10, **p < 0.05, ***p < 0.001.
36 9
23 26
13 50
B/C or residential facility Residing alone or with non-family member Own resources Residing with family member
42
48
32
5
11
10
45 8
21 21
17 41
Initial UDS
Yes No
42 58
34 66
27 73
Length of stay**
1 to 5 6 to 10 11 to 15 16+
29 55 8 8
17 54 24 5
37 59 3 0
24 44 16 16
21 38 19 22
85 15 0 0
Diagnosis***
Initial admit from***
Initial discharge to***
Previous 0 hospitalization*** 1 to 5 6 to 10 11+
428
www.jonmd.com
The high risk for readmission among patients with schizophrenia/ SAD may be caused by poorer insight into their illness, as a result these patients may be less aware of a pending exacerbation and need for intervention. Also, because of the nature of their symptoms, they may have alienated individuals who would otherwise have helped them obtain care when they began to decompensate. These patients may have required more time to stabilize, both as inpatients and outpatients. History of alcohol abuse was another risk factor for early readmission, increasing the odds ratio of being readmitted within 15 days by a factor of almost 9, and within 3 to 9 months by a factor of 4. These results extend previous findings indicating alcohol intoxication to be a risk factor among general psychiatric patients (Zhang et al., 2011), and comorbid alcohol abuse among schizophrenics (Haywood et al., 1995; Olfson et al., 1999). It is likely that alcohol abuse may have aggravated their primary psychiatric disorder and resulted in re-hospitalization. The finding indicates that clinicians may need to focus on alcohol relapse prevention when working with psychiatric patients at risk for readmission (e.g., schizophrenics). Unfortunately, during brief hospitalizations, only introductory work for a recovery plan can be initiated. In the absence of adequate outpatient resources to treat patients with alcohol and drug problems, ongoing abuse may continue to contribute to early readmission. Another factor significantly associated with early readmission was place of residence. The odds of early readmission was much higher © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Factors Associated With Readmissions
TABLE 3. Odds Ratios (OR) and Confidence Intervals (CI) of Risk Factors for Readmission of Psychiatric Patients (N = 207) Group 1 vs. Group 3
Diagnosis*
N = 59
N = 86
N = 59
OR
95% CI
OR
95% CI
Schizophrenia/SAD BAD Psychosis MDD
17.8 6.31 1.23
2.70 0.69 0.07
History of alcohol abuse*
Yes No
8.54
1.80
History of drug abuse
Yes No
0.93
0.22
B/C or residential facilities Residing alone or with non-family member Homeless Residing with family member
27.4 7.59 29.4
4.39 1.14 3.99
History of incarceration
Yes No
0.80
0.21
Initial UDS
Yes No
2.01
Compliant Noncompliant
Initial admit from*
Medication compliance
4.42 2.89 2.76
0.83 0.40 0.41
23.6 21.1 18.5
40.60
3.97
0.94
16.80
4.00
1.52
0.41
5.71
6.52 4.65 3.82
1.33 0.98 0.65
31.9 22.1 22.4
3.00
0.82
0.25
2.72
0.43
9.46
1.29
0.30
5.57
0.27
0.02
3.2
0.59
0.07
5.11
Yes No
2.61
0.53
12.9
1.12
0.24
5.16
Number of hospitalizations
2.18
1.28
3.73
2.28
1.34
3.90
Number of days
1.19
0.97
1.45
1.26
1.04
1.53
Voluntary Number of previous hospitalizations* Length of stay*
Group 2 vs. Group 3
N = 62
117.7 58.1 21.8
171 50.8 217
Group 3 is used as reference in the multinomial logistic regression. B/C indicates board and care; UDS, urine drug screen. *p < 0.05.
for those who were homeless (OR = 29), residing in board and care/ residential facilities (OR = 27), and, to a lesser degree, those residing alone or with non-family members (OR = 8). The homeless included patients who did not meet criteria to be placed on conservatorship and refused to be placed at a facility. The issue of refusing residence by patients is a daunting task beyond the control of physicians and hospital staff, and supports arguments that bearing the burden of readmission is multifactorial, often beyond the control of hospitals and doctors treating them during their hospitalization. One reason as to why a larger proportion of homeless or residents of board and care/residential facilities were represented in group 1 may be that these patients tend to be sicker compared to those living with family members. Additionally, the multiplicity of adversities and lack of support associated with homelessness can exacerbate illness that can lead to readmission. With regard to the difference in higher readmission rates among those residing at board and care/residential facilities versus those residing with families, it is possible that the support and supervision provided by family members may be effective in deterring early re-hospitalization. Future studies are needed to investigate the contribution of residence to risk for readmission. Number of previous hospitalizations was associated with increased odds of readmission for both group 1 and 2. Each previous hospitalization increased the odds ratio of being in group 1 or group 2 by a factor of 2. This finding adds to findings from other studies that
have shown previous hospitalization to be a risk for readmission (Haywood et al., 1995; Silva et al., 2009; Thompson et al., 2003; Zhang et al., 2011) suggesting that the number of previous hospitalizations should be considered by hospital staff as an important indicator for readmission and accordingly allocate clinical resources and plan for outpatient care. Consistent with previous studies (Bernardo and Forchuk, 2001; Thompson et al., 2003; Zhang et al., 2011), gender and race did not predict early readmission. Likewise, a history of incarceration was not a predictor. A similar finding was reported among patients with schizophrenia/SAD from a State Hospital (Haywood et al., 1995). Contrary to expectation, medication noncompliance for the whole sample and across diagnostic categories did not predict risk for early readmission. Previous studies among discharged patients have found inconsistent results (Silva et al., 2009; Sullivan et al., 1995). Larger controlled studies are needed to assess the effect of medication compliance on early re-hospitalization. We did not find an association between early readmission and being on involuntary status at the initial hospitalization. This finding is consistent with findings in a study among general psychiatric patients (Silva et al., 2009). Our study indicated that a higher percentage of patients in group 3 had been admitted on an involuntary basis compared to those in group 1 (86% vs. 69%). We anticipated that involuntary patients who may have poor insight into their condition will not be
© 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jonmd.com
429
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Lorine et al.
compliant after discharge and, as a result, be at risk. This was not substantiated. However, this finding supports the helpfulness of treating involuntary patients, as their rates of readmission are comparable to those of patients admitted on a voluntary basis. Positive UDS was not significantly different among the three groups. Further, the readmission UDS did not differ significantly between group 1 and group 2. Finally, rates of positive UDS across diagnostic categories did not differ significantly. These findings are contrary to the a priori notion of some mental health care providers that early readmission is often attributable to recent substance abuse. Replication studies with larger samples are needed to evaluate the specific contribution of drug abuse to early readmission. The cost of hospitalization for the majority of severely ill patients with similar profiles as in this study is subsidized by Medicare, Medicaid (MediCal in California), or both. These programs are regulated by the Centers for Medicare and Medicaid Services (CMS). The financial burden of CMS in subsidizing care for the mentally ill is passed on to health care providers. One of the reasons why mental health professionals refrain from treating this population is the imposition put on them by hospitals to discharge patients often times earlier than recommended by the treating psychiatrist (i.e., as soon as a patient is not in imminent danger to harm themselves or others) even though they may be unstable. This may be done for concerns of not getting reimbursed by CMS or relenting to the pressure exerted by hospitals to discharge patients as soon as possible for financial reasons (Dartmouth Atlas Project, 2013). This discharge practice may be justified if there were comprehensive and efficient outpatient networks available to manage these patients. In the absence of effective outpatient programs, which should include a substance abuse recovery component, high rates of early readmission are to be expected. Another deterrent for doctors to serve this population may be the stipulation imposed by the Affordable Care Act which established the Hospital Readmission Reduction Program requiring CMS to reduce payments to participating hospitals with excess readmissions. Despite the fact that most hospitals and doctors do not continue treating these patients once they are discharged, the stipulation by the Act will put the onus of preventing readmissions on them. Presently, this plan is being implemented in medicine and soon will include psychiatry. Previous studies have questioned the validity of the assertion that readmission rate is a quality indicator for hospital care (Lyons et al., 1997, Zhang et al., 2011).
LIMITATIONS There are several limitations to the present study. First, the study is limited by the modest sample size. Second, most of the patients in this study were severely ill and indigent; thus, the findings may not be generalized to patients with less severe illness and/or higher socioeconomic status. Third, intake assessments were done by clinical evaluations without the use of standardized assessments/interview questions. Thus, there may be variability in the results because of varying levels of knowledge and clinical experience of the clinicians conducting the evaluations.
CONCLUSION Risk factors associated with early readmission of severely mentally ill psychiatric patients included diagnosis of schizophrenia/SAD, history of alcohol abuse, place of residence at initial admission, and number of previous psychiatric admissions. These risk factors can help health care providers in psychiatric hospitals and outpatient settings adjust their management of patients accordingly. They may also help policymakers revisit guidelines for discharging patients and regulate the allocation of resources to improve the care provided to these patients in inpatient and outpatient settings. The cost of additional care may be offset by the cost of early readmission. 430
www.jonmd.com
DISCLOSURES The authors declare no conflict of interest. REFERENCES Bernardo AC, Forchuk C (2001) Factors associated with readmission to a psychiatric facility. Psychiatr Serv. 52:1100–1102. Burke RE, Donzé J, Schnipper JL (2013) Contribution of psychiatric illness and substance abuse to 30-day readmission risk. J Hosp Med. 8:450–455. Dartmouth Atlas Project (2013) A revolving door: A report on hospital readmissions. Robert Wood Johnson Foundation. Available at https://archive.org/stream/602813robert-wood-johnson-the-revolving-door-report-on/602813-robert-wood-johnson-therevolving-door-report-on_djvu.txt. Fennig S, Rabinowitz J (1999) Involuntary first admission of patients with schizophrenia as a predictor of future admissions. Psychiatr Serv. 50:1049–1052. Gastal FL, Andreoli SB, Quintana MIS, Gameiro MA, Leite SO, McGrath J (2000) Predicting the revolving door phenomenon among patients with schizophrenic, affective disorders and non-organic psychoses. Rev Saúde Pública. 34:280–285. Goldfield NI, McCullough EC, Hughes JS, Tang AM, Eastman B, Rawlins LK, Averill RF (2008) Identifying potentially preventable readmissions. Health Care Financing Review. 30:75–91. Hackbarth GM, Chernew M, Miller ME (2013) Medicare and the Health Care Delivery system. Medicare Payment Advisory Panel, Report to Congress, Available at http://www.medpac.gov/documents/Jun13_EntireReport.pdf. Haywood TW, Kravitz HM, Grossman LS, Cavenaugh JL, Davis JM, Lewis DA (1995) Predicting the revolving door phenomenon among patients with schizophrenia, schizoaffective and affective disorder. Am J Psychiatry. 152:856–861. Horowitz L, Partovian C, Lin Z, Herrin J, Grady J, Conover M, Montague J, Dillaway C, Bartczak K, Ross J, Bernheim S, Drye E, Krumholz H (2011) Hospital-Wide (All Condition) 30 day Risk-Standardized Readmission Measure. Prepared for Center for Medicare and Medicaid Services by Yale New Haven Health Services Corporation Center for Outcome Measures. Available at www.cms.gov. Jencks SF, Williams MV (2009) Rehospitalizations among patients in the Medicare Fee-for-Service Program. N Engl J Med. 360:1418–1428. Loch AA (2012) Stigma and higher rates of psychiatric re-hospitalization: São Paulo public mental health system. Rev Bras Psiquiatr. 34:185–192. Lyons JS, O’Mahoney MT, Miller SI, Neme J, Kabat J, Miller F (1997) Predicting readmission to the psychiatric hospital in a managed care environment: Implications for quality indicators. Am J Psychiatry. 154:337–340. Olfson M, Mechanic D, Boyer CA, Hansell S, Walkup J, Weiden PJ (1999) Assessing clinical predictions of early rehospitalization in schizophrenia. J Nerv Ment Dis. 187:721–729. Rosca P, Bauer A, Grinshpoon A, Khawaled R, Mester R, Ponizovsky AM (2006) Rehospitalizations among psychiatric patients whose first admission was involuntary: A 10-year follow-up. Isr J Psychiatry Relat Sci. 43:57–64. Schennach R, Obermeier M, Meyer S, Jager M, Schmauss M, Laux G, Pfeiffer H, Naber D, Schmidt L, Gaebel W, Klosterkotter J, Heuser I, Maier W, Lemke MR, Ruther E, Klingberg S, Gastpar M, Seemuller F, Moller HJ, Riedel M (2012) Predictors of relapse in the year after hospital discharge among patients with schizophrenia. Psychiatr Serv. 63:87–90. Silva NC, Bassani DG, Palazzo LS (2009) A case–control study of factors associated with multiple psychiatric readmissions. Psychiatr Serv. 60:789–791. Sullivan G, Wells KB, Morgenstern H, Leake B (1995) Identifying risk factors for rehospitalization: Case–control study of seriously mentally ill persons in Mississippi. Am J Psychiatry. 152:1749–1756. Suzuki Y, Yasumura S, Fukao A, Otani K (2003) Associated factors of rehospitalization among schizophrenic patients. Psychiatry Clin Neurosci. 57:555–561. Thompson E, Neighbors H, Munday C, Trierweiler S (2003) Length of stay, referral to aftercare, and rehospitalization among psychiatric inpatients. Psychiatr Serv. 54:1271–1276. Zhang J, Harvey C, Andrew C (2011) Factors associated with length of stay and the risk of readmission in an acute psychiatric inpatient facility: A retrospective study. Aust N Z J Psychiatry. 45:578–585.
© 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Risk Factors Associated With Psychiatric Readmission Kim Lorine, PhD,* Haig Goenjian, MD,† Soeun Kim, PhD,‡ Alan M. Steinberg, PhD,§ Kendall Schmidt, BA,* and Armen K. Goenjian, MD*§ Abstract: The present study focused on identifying risk factors for early readmission of patients discharged from an urban community hospital. Retrospective chart reviews were conducted on 207 consecutive inpatient psychiatric admissions that included patients who were readmitted within 15 days, within 3 to 6 months, and not admitted for at least 12 months post-discharge. Findings indicated that a diagnosis of schizophrenia/schizoaffective disorder (OR = 18; 95% CI 2.70–117.7; p < 0.05), history of alcohol abuse (OR = 9; 95% CI 1.80–40.60; p < 0.05), number of previous psychiatric hospitalizations (OR = 2; 95% CI 1.28–3.73; p < 0.05), and type of residence at initial admission (e.g., homeless, OR = 29; 95% CI 3.99–217; p < 0.05) were significant risk factors for early readmission, where OR compares readmission group 1 versus group 3 in the multinomial logistic regression. Initial positive urine drug screen, history of drug abuse or incarceration, and legal status at initial admission did not predict early readmission. Reducing the risk factors associated with psychiatric readmissions has the potential to lead to the identification and development of preventative intervention strategies that can significantly improve patient safety, quality of care, well-being, and contain health care expenditures. Key Words: Psychiatric readmission, financial burden of readmission, readmission of schizophrenics, readmission of homeless (J Nerv Ment Dis 2015;203: 425–430)
I
n a study of 249 Florida hospitals of the 10 most common diagnoses that qualified for their criteria of potentially preventable readmissions, three were psychiatric diagnoses: schizophrenia was third, major depressive disorder (MDD) was fifth, and bipolar disorder was seventh (Goldfield et al., 2008). According to the 2013 Medicare Payment Advisory Commission Report to Congress (Hackbarth et al., 2013), 15% of Medicare beneficiaries discharged from hospitals in 2011 were readmitted within 30 days. A 2011 report by the Yale New Haven Health Services Corporation/Center for Outcome Research and Evaluation report (Horowitz et al., 2011) found that psychiatric readmission rates within 30 days of discharge for fee-for-service (FFS) Medicare beneficiaries 65 and older was 15.8%. In another study among Medicare beneficiaries, the cost of unplanned re-hospitalizations within 30 days of their discharge was estimated at $17.4 billion in 2004 (Jencks and Williams, 2009). The 30-day re-hospitalization rate for all illnesses was 19.6%, and for “psychosis” (the only psychiatric condition listed) it was 24.6%. Although these studies examined all causes of readmission, psychiatric readmission comprised a substantial proportion. Numerous studies have examined potential risk factors associated with psychiatric readmission. Some of these studies have been conducted in the United States (Jencks and Williams, 2009; Lyons et al., 1997; Olfson et al., 1999; Thompson et al., 2003), whereas other *Collaborative Neuroscience Network, LLC, Garden Grove; †Harbor UCLA Medical Center, Torrance, CA; ‡University of Texas Health Science Center, Division of Biostatistics, Houston, TX; and §UCLA/Duke University National Center for Child Traumatic Stress, Department of Psychiatry and Biobehavioral Sciences, University of California, Los Angeles, CA. Send reprint requests to Armen K. Goenjian, MD, UCLA/Duke University National Center for Child Traumatic Stress, Department of Psychiatry and Biobehavioral Sciences, University of California, 11150 W Olympic Blvd, Suite 650, Los Angeles, CA 90064. E-mail: [email protected]. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved. ISSN: 0022-3018/15/20306–0425 DOI: 10.1097/NMD.0000000000000305
studies were overseas (Bernardo and Forchuk, 2001; Loch, 2012; Silva et al., 2009; Zhang et al., 2011). These studies varied with regard to ethnic/racial mix, socioeconomic status, timing of assessments, and diagnostic categories evaluated. For example, some studies evaluated general psychiatric populations (Bernardo and Forchuk, 2001; Lyons et al., 1997; Thompson et al., 2003; Zhang et al., 2011), whereas others only evaluated patients with schizophrenia and schizoaffective disorder (SAD) (Fennig and Rabinowitz, 1999; Olfson et al., 1999; Schennach et al., 2012; Suzuki et al., 2003). Also, time of readmission varied across these studies from 30 days (Burke et al., 2013; Lyons et al., 1997), 3 months (Olfson et al., 1999), 6 months (Lyons et al., 1997; Thompson et al., 2003), and 1 year (Loch, 2012; Schennach et al., 2012; Zhang et al., 2011), or longer (Bernardo and Forchuk, 2001; Rosca et al., 2006). Even though these variations make generalizations across studies problematic, there are commonalities in many of the studies that may serve as guidelines to improve delivery of health care to severely ill psychiatric patients. Risk factors identified in studies among general psychiatric populations have included a diagnosis of schizophrenia (Rosca et al., 2006; Silva et al., 2009) and SAD (Thompson et al., 2003), presence of more severe symptoms and greater impairment in self-care (Lyons et al., 1997), younger age at first admission (Silva et al., 2009), use of restraints while hospitalized (Loch, 2012), number of previous admissions (Haywood et al., 1995; Silva et al., 2009; Zhang et al., 2011), alcohol intoxication, and posing a danger to others at the time of the initial admission (Zhang et al., 2011). In studies which included only patients with schizophrenia, risk for readmission included number of previous admissions (Olfson et al., 1999; Schennach et al., 2012), comorbid substance abuse (Olfson et al., 1999), medication noncompliance after discharge (Bernardo and Forchuk, 2001), not being in remission at time of discharge (Schennach et al., 2012), and noncompliance with outpatient clinic visits (Suzuki et al., 2003). The present study examined the association of sociodemographic, clinical, and post-discharge factors with readmission of psychiatric patients discharged from a general hospital located in an urban area of Los Angeles County. The objective was to identify risk factors associated with readmission within 15 days and within 3 to 9 months.
METHODS Participants For the present study, retrospective chart reviews were performed on 207 psychiatric inpatients from an urban community hospital. Patients were hospitalized for acute psychiatric illnesses. These patients were referred from local county and private emergency rooms, board and care facilities, residential facilities, mental health clinics, police, and private practitioners. Chart reviews were conducted on consecutive readmissions that fit in three time parameters: readmitted within 15 days (group 1, N = 62), readmitted within 3 to 6 months (group 2, N = 86), and not readmitted for at least 12 months post-discharge (group 3, N = 59). Patients within each group were independent and did not overlap (i.e., patients in group 1 were not included in groups 2 or 3). The admissions covered the period January through December 2010. Upon initial hospitalization, patients were stabilized (at a minimum, they no longer constituted a danger to themselves/others and
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015 Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jonmd.com
425
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Lorine et al.
were not gravely disabled). They were discharged with an appropriate medication prescription, aftercare instructions, and were referred back to their psychiatrist/therapist or to a new one if the patient did not have one or requested a new one. The majority returned back to their place of residence, which included board and cares, residential facilities, apartments, or homes. Some were referred to new facilities. For a subgroup of patients who were stabilized and refused placement, they were provided with medication prescriptions, follow-up care instructions, and were referred to a psychiatrist or a mental health clinic. IRB approval was received to conduct the current retrospective chart review.
Data Analysis Variables examined in the current study included age, gender, race, psychiatric diagnosis, history of incarceration, history of alcohol/ drug abuse, number of previous inpatient hospitalizations, type of insurance, length of stay of admission, residence admitted from (and whether patients lived alone or with family members), urine drug screen (UDS), legal status (voluntary/involuntary), medication compliance, and availability of outside support (e.g., family member, friend, church members, self-help groups). With regard to medications, those patients who were initially noncompliant for a few days but subsequently became compliant were classified as compliant. In building the regression model, we first compared demographic and potential risk factors of interest for the three groups. Next, a χ2 analysis was used to examine differences for each of the dichotomous variables (significance level was set at p < 0.05). We started with a full model including all significant variables from the univariate analysis and used backward elimination resulting in a final model where we included predictors of interest while controlling for significant confounders. Variables that were not statistically significant at the p < 0.05 level in this step were left out of the final model (e.g., gender, race). Although age showed slight significance, it was removed from the final model given that there was a heterogeneous pattern of direction between the groups and would not add benefit to the final analysis. Initial UDS, history of incarceration, and medication compliance, even though were not statistically significant, were added to the final model as they were considered clinically relevant. The final multinomial logistic regression calculated odds ratios (OR) and 95% confidence intervals (95%) CI for 10 predictors. SAS/STAT software was used to perform analyses.
RESULTS Table 1 presents the descriptive characteristics of patients included in the study. The majority of the sample (91%) was between ages 18 and 60. They were fairly evenly distributed between the age groups, with a slightly higher proportion in the 40–60 age range. Nine percent were over 60 years old. Male to female ratio was 3:2. Sixty-one percent of the patients were white, 31% were African-American, and 8% were Hispanic or Asian. Virtually all patients were unemployed, with 56% having a history of incarceration. Approximately 40% had no previous psychiatric hospitalization. Half of the patients had diagnosis of schizophrenia or SAD, 20% bipolar affective disorder (BAD), and 17% MDD. Almost half had a history of alcohol abuse and 63% drug abuse. Initial urine drug screen was positive for 34%. Thirty-one percent came from a board and care or a residential facility, 24% were homeless, and the remainder either lived alone or with someone else in an apartment or a house. Table 2 presents the comparison of variables within each group. The age distribution varied among the three groups, most notably for those over 60, where age was higher in group 3 (20%) versus groups 1 (7%) and 2 (2%). Diagnosis of schizophrenia/SAD was present with a higher proportion in group 1 (68%) and group 2 (50%), whereas major depression was higher in group 3 (31%). The proportion of patients with a history of alcohol abuse was higher in group 1 (61%) compared 426
www.jonmd.com
to the other two. Additionally, the proportion of patients with a history of drug abuse was higher in group 1 and group 2. The majority of patients in group 1 (40%) resided at board and care or residential facilities at the time of initial admission, whereas 50% of group 3 resided with family members versus 9% for group 1. There were higher proportions of homeless patients in both group 1 (36%) and group 2 (23%) versus group 3 (13%). Of interest was the initial UDS rate which was not significantly different across the three groups. Also UDS across diagnostic categories did not differ significantly among the three groups. Lastly, initial UDS by residence showed no significant difference among the groups. Medication compliance rates across the groups and for the different diagnostic categories across the group did not show significant difference. Also, medication compliance by place of residence across the groups was not significant. The average length of stay was 8.4 days. For the majority (97%), length of stay was between 1 and 15 days; for 27%, it was 1 to 5 days. Lastly, 85% of group 3 patients had no previous inpatient hospitalizations, whereas 44% of group 1 and 39% of group 2 patients had one to five previous inpatient hospitalizations. With regard to readmissions, the location where they came from was similar to the pattern noted at initial admission. The majority of patients in group 1 resided at board and cares or residential facilities, or were homeless. There was no significant difference between group 1 and group 2 with regard to readmission UDS and social support. Table 3 presents the OR and 95% CI of potential risk factors computed from the multinomial logistic regression with readmission group as the outcome variable with group 3 (no admission for a minimum of 12 months) as the reference group. The odds of being in group 1 relative to being in group 3 increased by a factor of 17.8 times given a diagnosis of schizophrenia/SAD. History of alcohol abuse increased the odds of being in group 1 by a factor of 8.5 and being in group 2 by a factor of 4. History of drug abuse did not increase the odds likelihood for early readmission. The odds ratio of being in group 1 was 27 for those admitted from board and care/residential facilities, 29-fold for the homeless, and almost 7.6-fold for those living in an apartment alone or with a non-family member. Each previous hospitalization increased the ratio of the odds of being in group 1 or 2 by a factor of 2. Lastly, length of stay at initial admit did not increase the odds likelihood for early readmission, but did increase slightly the odds likelihood for readmission within 3 to 9 months (1.26). History of incarceration, initial urine drug screen, medical compliance, and voluntary status were not significantly associated with being in the readmission groups.
DISCUSSION The present study evaluated the association of patient characteristics and clinical variables with readmission to a psychiatric unit of a general hospital located in an urban area. The study differed from previous studies in that it assessed readmission of patients within a 12-month window, including readmissions within 15 days, 3 to 9 months, and not readmitted for at least 12 months after discharge. The latter group served as the reference group. Risk factors that differentiated patients subsequently readmitted within 15 days of discharge compared to those not readmitted for at least 12 months included diagnosis of schizophrenia/SAD, which was associated with a high likelihood of being readmitted within 15 days (OR = 18) (Table 3). This finding supports and extends the findings of other studies. Two previous studies among patients discharged from a state hospital found SAD to be a risk factor for re-hospitalization (Thompson et al., 2003), whereas two studies from Brazil found a diagnosis of schizophrenia to be a risk factor (Gastal et al., 2000; Silva et al., 2009). However, these studies did not specifically assess early readmissions as times from discharge to readmission varied from 6 months to years (Haywood et al., 1995; Silva et al., 2009; Thompson et al., 2003). © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Factors Associated With Readmissions
TABLE 1. Descriptive Characteristics of Patients in the Readmission Study (N = 207) Variable
Category
Frequency
Percent
Age group
18–35 36–45 46–60 60+
54 58 77 18
26 28 37 9
Sex
M F
116 77
60 40
Race
White African-American Hispanic or Asian
126 64 17
61 31 8
Marital status
Single Divorced Widowed Married
169 11 4 13
86 6 2 7
Employment
Yes No
3 204
2 99
Schizophrenia/SAD BAD MDD Psychosis
97 39 33 24
50 20 17 12
History of incarceration
Yes No
115 91
56 44
History of alcohol abuse
Yes No
97 109
47 53
History of drug abuse
Yes No
130 76
63 37
Medicare/Medical or Medicare MediCal Uninsured “Cash”
56 135 16
27 65 8
Initial admit from
B/C or residential facility Residing alone or with non-family member Homeless Residing with family member
62 34 47 55
31 17 24 28
Initial discharge to
B/C or residential facilities Residing alone or with non-family member Residing with family member Own resources
86 18 47 56
42 9 23 27
Yes No
71 136
34 66
Medication compliance
Compliant Noncompliant
189 18
91 9
Legal status (voluntary)
Yes No
40 167
19 81
Danger to self (DTS) Danger to others (DTO) Gravely disabled (GD), GD/DTO, GD/DTS
111 32 37
62 18 21
Length of stay
1 to 5 6 to 10 11 to 15 16+
55 115 28 9
27 56 14 4
Previous hospitalizations
0 1 to 5 6 to 10 11+
83 69 26 29
40 33 13 14
Diagnosis
Insurance
Initial UDS
Involuntary status subtypes
B/C indicates board and care; UDS, urine drug screen.
© 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jonmd.com
427
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Lorine et al.
TABLE 2. Comparisons of Potential Risk Factors for Readmission Among the Three Study Groups (N = 207)
TABLE 2. (Continued)
Group 1 Group 2 Group 3 N = 62 N = 86 N = 59 Percent Percent Percent Age group***
Group 1 Group 2 Group 3 N = 62 N = 86 N = 59 Percent Percent Percent
18–35 36–45 46–60 60+
32 27 34 7
15 35 48 2
36 19 25 20
Medication compliance
Compliant Non-compliant
95 5
88 12
92 9
Legal status (voluntary)**
Yes No
31 70
15 85
14 86
Sex*
Male Female
69 31
61 40
47 53
Involuntary status subtypes
64
62
59
Race
White African-American Hispanic or Asian
53 37 10
65 31 4
63 24 14
14
17
25
23
22
16
Schizophrenia/SAD BAD Psychosis MDD
68 18 2 13
50 24 13 13
27 16 27 31
Danger to self (DTS) Danger to others (DTO) Gravely disabled (GD) GD/DTS, GD/DTO
46
36
–
12
23
–
History of incarceration
Yes No
61 39
57 44
49 51
History of alcohol abuse**
Yes No
61 39
41 59
42 58
Readmission from** B/C or residential facilities Residing alone or with non-family member Homeless Residing with family member
37 5
17 24
– –
History of drug abuse**
Yes No
67 33
72 28
46 54
Readmit UDS
Yes No
29 71
36 64
– –
Insurance**
Medicare/MediCal or Medicare MediCal Uninsured “Cash”
34
22
27
Outside support*
Yes No
60 40
71 29
– –
47 19
74 4
71 2
B/C or residential facility Residing alone or with non-family member Homeless Residing with family member
40
30
25
16
21
13
Values are column percentages within each group. The last three variables (readmission from, readmit UDS, and outside support) are not observed for group 3. The comparisons were done between group 1 and group 2. B/C indicates board and care; UDS, urine drug screen. *p < 0.10, **p < 0.05, ***p < 0.001.
36 9
23 26
13 50
B/C or residential facility Residing alone or with non-family member Own resources Residing with family member
42
48
32
5
11
10
45 8
21 21
17 41
Initial UDS
Yes No
42 58
34 66
27 73
Length of stay**
1 to 5 6 to 10 11 to 15 16+
29 55 8 8
17 54 24 5
37 59 3 0
24 44 16 16
21 38 19 22
85 15 0 0
Diagnosis***
Initial admit from***
Initial discharge to***
Previous 0 hospitalization*** 1 to 5 6 to 10 11+
428
www.jonmd.com
The high risk for readmission among patients with schizophrenia/ SAD may be caused by poorer insight into their illness, as a result these patients may be less aware of a pending exacerbation and need for intervention. Also, because of the nature of their symptoms, they may have alienated individuals who would otherwise have helped them obtain care when they began to decompensate. These patients may have required more time to stabilize, both as inpatients and outpatients. History of alcohol abuse was another risk factor for early readmission, increasing the odds ratio of being readmitted within 15 days by a factor of almost 9, and within 3 to 9 months by a factor of 4. These results extend previous findings indicating alcohol intoxication to be a risk factor among general psychiatric patients (Zhang et al., 2011), and comorbid alcohol abuse among schizophrenics (Haywood et al., 1995; Olfson et al., 1999). It is likely that alcohol abuse may have aggravated their primary psychiatric disorder and resulted in re-hospitalization. The finding indicates that clinicians may need to focus on alcohol relapse prevention when working with psychiatric patients at risk for readmission (e.g., schizophrenics). Unfortunately, during brief hospitalizations, only introductory work for a recovery plan can be initiated. In the absence of adequate outpatient resources to treat patients with alcohol and drug problems, ongoing abuse may continue to contribute to early readmission. Another factor significantly associated with early readmission was place of residence. The odds of early readmission was much higher © 2015 Wolters Kluwer Health, Inc. All rights reserved.
Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Factors Associated With Readmissions
TABLE 3. Odds Ratios (OR) and Confidence Intervals (CI) of Risk Factors for Readmission of Psychiatric Patients (N = 207) Group 1 vs. Group 3
Diagnosis*
N = 59
N = 86
N = 59
OR
95% CI
OR
95% CI
Schizophrenia/SAD BAD Psychosis MDD
17.8 6.31 1.23
2.70 0.69 0.07
History of alcohol abuse*
Yes No
8.54
1.80
History of drug abuse
Yes No
0.93
0.22
B/C or residential facilities Residing alone or with non-family member Homeless Residing with family member
27.4 7.59 29.4
4.39 1.14 3.99
History of incarceration
Yes No
0.80
0.21
Initial UDS
Yes No
2.01
Compliant Noncompliant
Initial admit from*
Medication compliance
4.42 2.89 2.76
0.83 0.40 0.41
23.6 21.1 18.5
40.60
3.97
0.94
16.80
4.00
1.52
0.41
5.71
6.52 4.65 3.82
1.33 0.98 0.65
31.9 22.1 22.4
3.00
0.82
0.25
2.72
0.43
9.46
1.29
0.30
5.57
0.27
0.02
3.2
0.59
0.07
5.11
Yes No
2.61
0.53
12.9
1.12
0.24
5.16
Number of hospitalizations
2.18
1.28
3.73
2.28
1.34
3.90
Number of days
1.19
0.97
1.45
1.26
1.04
1.53
Voluntary Number of previous hospitalizations* Length of stay*
Group 2 vs. Group 3
N = 62
117.7 58.1 21.8
171 50.8 217
Group 3 is used as reference in the multinomial logistic regression. B/C indicates board and care; UDS, urine drug screen. *p < 0.05.
for those who were homeless (OR = 29), residing in board and care/ residential facilities (OR = 27), and, to a lesser degree, those residing alone or with non-family members (OR = 8). The homeless included patients who did not meet criteria to be placed on conservatorship and refused to be placed at a facility. The issue of refusing residence by patients is a daunting task beyond the control of physicians and hospital staff, and supports arguments that bearing the burden of readmission is multifactorial, often beyond the control of hospitals and doctors treating them during their hospitalization. One reason as to why a larger proportion of homeless or residents of board and care/residential facilities were represented in group 1 may be that these patients tend to be sicker compared to those living with family members. Additionally, the multiplicity of adversities and lack of support associated with homelessness can exacerbate illness that can lead to readmission. With regard to the difference in higher readmission rates among those residing at board and care/residential facilities versus those residing with families, it is possible that the support and supervision provided by family members may be effective in deterring early re-hospitalization. Future studies are needed to investigate the contribution of residence to risk for readmission. Number of previous hospitalizations was associated with increased odds of readmission for both group 1 and 2. Each previous hospitalization increased the odds ratio of being in group 1 or group 2 by a factor of 2. This finding adds to findings from other studies that
have shown previous hospitalization to be a risk for readmission (Haywood et al., 1995; Silva et al., 2009; Thompson et al., 2003; Zhang et al., 2011) suggesting that the number of previous hospitalizations should be considered by hospital staff as an important indicator for readmission and accordingly allocate clinical resources and plan for outpatient care. Consistent with previous studies (Bernardo and Forchuk, 2001; Thompson et al., 2003; Zhang et al., 2011), gender and race did not predict early readmission. Likewise, a history of incarceration was not a predictor. A similar finding was reported among patients with schizophrenia/SAD from a State Hospital (Haywood et al., 1995). Contrary to expectation, medication noncompliance for the whole sample and across diagnostic categories did not predict risk for early readmission. Previous studies among discharged patients have found inconsistent results (Silva et al., 2009; Sullivan et al., 1995). Larger controlled studies are needed to assess the effect of medication compliance on early re-hospitalization. We did not find an association between early readmission and being on involuntary status at the initial hospitalization. This finding is consistent with findings in a study among general psychiatric patients (Silva et al., 2009). Our study indicated that a higher percentage of patients in group 3 had been admitted on an involuntary basis compared to those in group 1 (86% vs. 69%). We anticipated that involuntary patients who may have poor insight into their condition will not be
© 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
www.jonmd.com
429
The Journal of Nervous and Mental Disease • Volume 203, Number 6, June 2015
Lorine et al.
compliant after discharge and, as a result, be at risk. This was not substantiated. However, this finding supports the helpfulness of treating involuntary patients, as their rates of readmission are comparable to those of patients admitted on a voluntary basis. Positive UDS was not significantly different among the three groups. Further, the readmission UDS did not differ significantly between group 1 and group 2. Finally, rates of positive UDS across diagnostic categories did not differ significantly. These findings are contrary to the a priori notion of some mental health care providers that early readmission is often attributable to recent substance abuse. Replication studies with larger samples are needed to evaluate the specific contribution of drug abuse to early readmission. The cost of hospitalization for the majority of severely ill patients with similar profiles as in this study is subsidized by Medicare, Medicaid (MediCal in California), or both. These programs are regulated by the Centers for Medicare and Medicaid Services (CMS). The financial burden of CMS in subsidizing care for the mentally ill is passed on to health care providers. One of the reasons why mental health professionals refrain from treating this population is the imposition put on them by hospitals to discharge patients often times earlier than recommended by the treating psychiatrist (i.e., as soon as a patient is not in imminent danger to harm themselves or others) even though they may be unstable. This may be done for concerns of not getting reimbursed by CMS or relenting to the pressure exerted by hospitals to discharge patients as soon as possible for financial reasons (Dartmouth Atlas Project, 2013). This discharge practice may be justified if there were comprehensive and efficient outpatient networks available to manage these patients. In the absence of effective outpatient programs, which should include a substance abuse recovery component, high rates of early readmission are to be expected. Another deterrent for doctors to serve this population may be the stipulation imposed by the Affordable Care Act which established the Hospital Readmission Reduction Program requiring CMS to reduce payments to participating hospitals with excess readmissions. Despite the fact that most hospitals and doctors do not continue treating these patients once they are discharged, the stipulation by the Act will put the onus of preventing readmissions on them. Presently, this plan is being implemented in medicine and soon will include psychiatry. Previous studies have questioned the validity of the assertion that readmission rate is a quality indicator for hospital care (Lyons et al., 1997, Zhang et al., 2011).
LIMITATIONS There are several limitations to the present study. First, the study is limited by the modest sample size. Second, most of the patients in this study were severely ill and indigent; thus, the findings may not be generalized to patients with less severe illness and/or higher socioeconomic status. Third, intake assessments were done by clinical evaluations without the use of standardized assessments/interview questions. Thus, there may be variability in the results because of varying levels of knowledge and clinical experience of the clinicians conducting the evaluations.
CONCLUSION Risk factors associated with early readmission of severely mentally ill psychiatric patients included diagnosis of schizophrenia/SAD, history of alcohol abuse, place of residence at initial admission, and number of previous psychiatric admissions. These risk factors can help health care providers in psychiatric hospitals and outpatient settings adjust their management of patients accordingly. They may also help policymakers revisit guidelines for discharging patients and regulate the allocation of resources to improve the care provided to these patients in inpatient and outpatient settings. The cost of additional care may be offset by the cost of early readmission. 430
www.jonmd.com
DISCLOSURES The authors declare no conflict of interest. REFERENCES Bernardo AC, Forchuk C (2001) Factors associated with readmission to a psychiatric facility. Psychiatr Serv. 52:1100–1102. Burke RE, Donzé J, Schnipper JL (2013) Contribution of psychiatric illness and substance abuse to 30-day readmission risk. J Hosp Med. 8:450–455. Dartmouth Atlas Project (2013) A revolving door: A report on hospital readmissions. Robert Wood Johnson Foundation. Available at https://archive.org/stream/602813robert-wood-johnson-the-revolving-door-report-on/602813-robert-wood-johnson-therevolving-door-report-on_djvu.txt. Fennig S, Rabinowitz J (1999) Involuntary first admission of patients with schizophrenia as a predictor of future admissions. Psychiatr Serv. 50:1049–1052. Gastal FL, Andreoli SB, Quintana MIS, Gameiro MA, Leite SO, McGrath J (2000) Predicting the revolving door phenomenon among patients with schizophrenic, affective disorders and non-organic psychoses. Rev Saúde Pública. 34:280–285. Goldfield NI, McCullough EC, Hughes JS, Tang AM, Eastman B, Rawlins LK, Averill RF (2008) Identifying potentially preventable readmissions. Health Care Financing Review. 30:75–91. Hackbarth GM, Chernew M, Miller ME (2013) Medicare and the Health Care Delivery system. Medicare Payment Advisory Panel, Report to Congress, Available at http://www.medpac.gov/documents/Jun13_EntireReport.pdf. Haywood TW, Kravitz HM, Grossman LS, Cavenaugh JL, Davis JM, Lewis DA (1995) Predicting the revolving door phenomenon among patients with schizophrenia, schizoaffective and affective disorder. Am J Psychiatry. 152:856–861. Horowitz L, Partovian C, Lin Z, Herrin J, Grady J, Conover M, Montague J, Dillaway C, Bartczak K, Ross J, Bernheim S, Drye E, Krumholz H (2011) Hospital-Wide (All Condition) 30 day Risk-Standardized Readmission Measure. Prepared for Center for Medicare and Medicaid Services by Yale New Haven Health Services Corporation Center for Outcome Measures. Available at www.cms.gov. Jencks SF, Williams MV (2009) Rehospitalizations among patients in the Medicare Fee-for-Service Program. N Engl J Med. 360:1418–1428. Loch AA (2012) Stigma and higher rates of psychiatric re-hospitalization: São Paulo public mental health system. Rev Bras Psiquiatr. 34:185–192. Lyons JS, O’Mahoney MT, Miller SI, Neme J, Kabat J, Miller F (1997) Predicting readmission to the psychiatric hospital in a managed care environment: Implications for quality indicators. Am J Psychiatry. 154:337–340. Olfson M, Mechanic D, Boyer CA, Hansell S, Walkup J, Weiden PJ (1999) Assessing clinical predictions of early rehospitalization in schizophrenia. J Nerv Ment Dis. 187:721–729. Rosca P, Bauer A, Grinshpoon A, Khawaled R, Mester R, Ponizovsky AM (2006) Rehospitalizations among psychiatric patients whose first admission was involuntary: A 10-year follow-up. Isr J Psychiatry Relat Sci. 43:57–64. Schennach R, Obermeier M, Meyer S, Jager M, Schmauss M, Laux G, Pfeiffer H, Naber D, Schmidt L, Gaebel W, Klosterkotter J, Heuser I, Maier W, Lemke MR, Ruther E, Klingberg S, Gastpar M, Seemuller F, Moller HJ, Riedel M (2012) Predictors of relapse in the year after hospital discharge among patients with schizophrenia. Psychiatr Serv. 63:87–90. Silva NC, Bassani DG, Palazzo LS (2009) A case–control study of factors associated with multiple psychiatric readmissions. Psychiatr Serv. 60:789–791. Sullivan G, Wells KB, Morgenstern H, Leake B (1995) Identifying risk factors for rehospitalization: Case–control study of seriously mentally ill persons in Mississippi. Am J Psychiatry. 152:1749–1756. Suzuki Y, Yasumura S, Fukao A, Otani K (2003) Associated factors of rehospitalization among schizophrenic patients. Psychiatry Clin Neurosci. 57:555–561. Thompson E, Neighbors H, Munday C, Trierweiler S (2003) Length of stay, referral to aftercare, and rehospitalization among psychiatric inpatients. Psychiatr Serv. 54:1271–1276. Zhang J, Harvey C, Andrew C (2011) Factors associated with length of stay and the risk of readmission in an acute psychiatric inpatient facility: A retrospective study. Aust N Z J Psychiatry. 45:578–585.
© 2015 Wolters Kluwer Health, Inc. All rights reserved. Copyright © 2015 Wolters Kluwer Health, Inc. All rights reserved.
