bs_bs_banner
Journal of Evaluation in Clinical Practice ISSN 1365-2753
Factors differentially associated with early readmission at a university teaching psychiatric hospital Jane E. Hamilton PhD MPH,1 Howard Rhoades PhD,4 Juan Galvez MD,1 Melissa Allen DO,2 Charles Green PhD,6 Mildred Aller5 and Jair C. Soares MD PhD3 1 Postdoctoral Research Fellow, 2Assistant Professor, 3Professor and Chairman, Department of Psychiatry, University of Texas Medical School at Houston, Houston, TX, USA 4 Senior Statistician, 5Research Coordinator, University of Texas Harris County Psychiatric Center, Houston, TX, USA 6 Associate Professor, Center for Clinical Research and Evidence-Based Medicine, University of Texas Medical School at Houston, Houston, TX, USA
Keywords evaluation, health services research, patient-centred care Correspondence Dr Jane E. Hamilton Department of Psychiatry University of Texas (UT) Medical School at Houston 2800 S. MacGregor Way, HCP 3-E50 Houston, TX 77021 USA E-mail: [email protected] Accepted for publication: 13 January 2015 doi:10.1111/jep.12335
Abstract Rationale, aims and objectives The rate of psychiatric readmissions within 30 days of discharge is a well-established behavioural health system performance measure linked to the quality of inpatient hospital care as well as to access to community-based aftercare services. The purpose of this study was to examine the factors differentially associated with earlier readmission among a sample of patients (n = 588) readmitted within 30 days of discharge to a university teaching psychiatric hospital from 2001 to 2010. Methods Quality assurance interviews were conducted with patients readmitted within 30 days of discharge. The interview data were merged with clinical symptom and electronic medical record data. Multinomial logistic regression analysis was used to examine readmission within 7 days and from 8 to 14 days compared with 15–30 days after discharge while controlling for socio-demographic and treatment variables previously associated with psychiatric readmission. Results Multiple clinical, treatment and patient-reported factors were differentially associated with earlier readmission. In particular, lack of engagement in post-discharge aftercare services was a strong predictor of earlier readmission. Conclusions Strategies are needed to improve patients’ transition from inpatient psychiatric hospitalization to aftercare services. Psychiatric hospitals attempting to reduce very early readmissions should seek to implement innovative transitional care initiatives targeting both patient and treatment factors.
Introduction In the era of health care reform, health care payers, policy makers and providers have become increasingly concerned about the high rates of patient readmissions following hospitalization [1]. Although the primary focus has centred on readmission to shortterm acute care hospitals, there is growing interest in understanding readmission following psychiatric hospitalization. Psychiatric readmission rates have historically been viewed as important indicators for health care planning due to their connection to the quality and continuity of care for patients as well as the high costs associated with additional inpatient care [2–6]. The rate of psychiatric patients readmitted within 30 days of discharge is an established behavioural health system performance measure [7] linked to the quality of inpatient hospital care [8] and to access to community-based aftercare services [9]. Although 572
30-day psychiatric readmissions have been a focus of research, studies have not examined differential influences of patient and treatment factors on critical time frames within a 30-day readmission period [10–16]. The imperatives of the Affordable Care Act to reduce 30-day readmissions in the United States provide an opportunity to develop and implement readmission reduction strategies. However, additional research is needed to elucidate how specific factors influence psychiatric readmissions at different time periods within 30 days of discharge. To extend current research, we linked and analysed patient interview, psychiatric symptom and medical record data obtained from our psychiatric hospital. The goal of our study was to examine readmission patterns among psychiatric patients and the predictors of 30-day readmission from 1 to 7 days and from 8 to 14 days after discharge compared with 15–30 days after discharge. Based on prior research, we hypothesized that
Journal of Evaluation in Clinical Practice 21 (2015) 572–578 © 2015 John Wiley & Sons, Ltd.
J.E. Hamilton et al.
patients would be at higher risk for readmission during the first 2 weeks following discharge [17–19]. Readmissions occurring very close to the time of discharge have been associated with factors related to hospital quality of care [20,21]. Additionally, suicidal patients requiring intensive aftercare services and family supports are at high risk for readmission during the first week following discharge [22,23]. Patient factors such as readiness for discharge and socio-economic status have been shown to affect readmission within 14 days of discharge [19,24,25]. In addition to hospital quality and patient factors, we wanted to explore additional factors predicting readmission from 8 to 14 days following discharge as patients transitioned into the community and attempted to access mental health services. Rather than examining time to readmission as a continuous variable, we established cut points for our outcome variable that reflect the 7-day supply of medication provided to patients at discharge and the time frame in which our discharged patients were scheduled for an aftercare appointment (within 10 days of discharge). We organized predictor variables conceptually into four areas: socio-demographic, clinical, treatment and patient interview information. The patient-reported information we collected has been previously found to influence psychiatric readmissions including education level, financial support, unemployment, homelessness, satisfaction with previous hospitalization, discharge readiness, medication adherence, aftercare services and aftercare attendance [15,19,25–32]. We are only aware of one prior study examining the relationship between patient-reported satisfaction with psychiatric care and 30-day readmissions [15], and we are not aware of any studies examining additional patient-reported information. Although several studies have examined differential factors associated with readmissions up to 5 years after discharge [19,33], no prior studies have examined differential factors associated with readmission within 30 days of discharge.
Differential factors
Measures Patient interviews Trained interviewers screened 30-day readmitted patients for participation in 10-minute interviews through a daily review of the hospital census. Prior to being interviewed, patients were identified as stable by unit staff and agreed to participation. The information obtained from the interviews has not been previously published. Interview questions are presented in Table 1. BPRS The BPRS is an 18-item clinician-rated measure of psychiatric symptoms [34] used to examine response to psychopharmacological treatment [35] and the link between psychiatric symptoms and psychiatric readmissions [13,36]. During the study period, 51% of 30-day readmitted patients were assessed with the BPRS by 53 different attending psychiatrists as part of their standard practice of care. Medical record data We extracted patient medical record information to control for socio-demographics (age, gender, race/ethnicity), DSM diagnosis and treatment characteristics [prior hospitalizations, length of stay
30-day readmiƩed paƟents January 2001 to November 2010 n = 5009 Excluded due to not being interviewed n = 3723
Methods Sample The sample included patients who participated in quality assurance interviews at our hospital from January 2001 to November 2010. Since 1986, our hospital has been the regional safety net provider of inpatient psychiatric services in Harris County, Texas. In the present study, we combined interview data with data from the Brief Psychiatric Rating Scale (BPRS) to examine the influence of psychiatric symptoms on 30-day readmissions, controlling for factors known to be associated with 30-day readmissions [10]. To examine how well our sample represented all our 30-day readmitted patients, we compared the demographic and clinical characteristics of the final sample (n = 588) with the demographic and clinical characteristics of a larger sample of all patients readmitted within 30 days of discharge during the same time period (n = 5009) as well as the entire sample of interviewed patients (n = 1286). The three samples did not differ proportionally by age, gender, race/ethnicity, Diagnostic and Statistical Manual of Mental Disorders (DSM) diagnosis or marital status (patient characteristics available for all three samples examined). A patient inclusion diagram is presented in Fig. 1. This study was approved by the University of Texas Health Science Center at Houston Institutional Review Board.
© 2015 John Wiley & Sons, Ltd.
ReadmiƩed paƟents interviewed n = 1286 Excluded due to incomplete BPRS data n = 617 ReadmiƩed paƟents with complete BPRS data n = 669 Excluded 2nd and subsequent interviews for paƟents with mulƟple readmissions n = 67 First-Ɵme interviewed paƟents n = 602 Excluded due to missing data n = 14 PaƟents included in final analysis n = 588 Figure 1 Patient inclusion diagram.
573
Differential factors
J.E. Hamilton et al.
Table 1 Patient interview questions Question
Response items
Marital status
0. Not married, divorced, widowed 1. Married Employment status 0. Unemployed 1. Part time 2. Full time Years of education 12 Arrest history 0. No prior arrest 1. Prior arrest Involuntary/voluntary status 1. Involuntary 2. Voluntary 0. No Since hospitalization, has the patient been 1. Part time 2. Full time employed? Sources of financial support: 0. None 1. Disability 2. Government 3. Family 4. Other sources 5. Employment 1. Homeless Where did the patient live 2. Home after the last hospitalization? 3. Hospital Patient’s belief as to why 1. Wasn’t ready to discharge s/he returned so quickly? 2. Medication not effective 3. Medication not taken 4. Stressful environment after discharge 5. Unsure Helpfulness of previous 1. Not helpful hospitalization: 2. Little 3. Somewhat 4. Very Compliance to psychiatric 1. Never taken medication: 2. Taken inconsistently 3. Taken consistently Experience with medication 1. Medication not effective effectiveness? 2. Medication was effective Experience with medication 1. Side effects side effects? 2. No side effects Aftercare referral: 0. None 1. Public system 2. Other 3. Day treatment 4. Facility/hospital Aftercare attendance: 0. Appointment in the future 1. Missed first appointment 2. Attended >1 appointment
(LOS), involuntary status] as performed in prior studies examining 30-day psychiatric readmissions [11–13,19]. Private health information including patient name, address, date of birth and social security number was not included in the dataset to protect patient confidentiality.
Statistical analysis We explored the effect of predictors on patient readmissions within 7 days and from 8 to 14 days of discharge, using 15–30 days 574
after discharge as the reference group in a series of univariate and multinomial logistic regression analyses. Data were analysed with IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp.
Results In the final sample of 588 adult patients (mean age 35.5 years; range 18–71 years), the majority of patients (372; 63%) were readmitted during the first 2 weeks following discharge. Readmitted patients were more likely to be African American, male, unemployed and diagnosed with schizophrenia. A full listing of the sample characteristics is presented in Table 2. The univariate (unadjusted) analyses showed that patients with a diagnosis other than depression, bipolar disorder or schizophrenia were less likely to return during the first week following discharge compared with days 15–30 [odds ratio (OR): 0.292, 95% confidence interval (CI): 0.093–0.924, P = 0.036]. Patients with elevated BPRS symptoms of grandiosity (OR: 1.442, 95% CI: 1.135–1.832, P = 0.003) and suspiciousness (OR: 1.252, 95% CI: 1.015–1.544, P = 0.036) were more likely to be readmitted during week 1 compared with 15–30 days after discharge. Patients with elevated BPRS symptoms of grandiosity (OR: 1.314, 95% CI: 1.017–1.697, P = 0.037) and suspiciousness (OR: 1.318, 95% CI: 1.063–1.633, P = 0.012) were also more likely to be readmitted during week 2 compared with 15–30 days after discharge. Patients with elevated symptoms of depression (OR: 0.757, 95% CI: 0.594–0.963, P = 0.024) had a decreased risk of readmission during week 1 compared with 15–30 days after discharge. The patient reported belief of not being ready for discharge during the previous hospitalization predicted readmission within 7 days of discharge compared with 15–30 days after discharge (OR: 2.849, 95% CI: 1.203–6.749, P = 0.017). A longer LOS also predicted readmission during 1–7 days after discharge (OR: 1.032, 95% CI: 1.007–1.058, P = 0.013). Patients referred to aftercare within our public mental health system were less likely to be readmitted during the first week following discharge (OR: 0.485, 95% CI: 0.235–0.893, P = 0.022). During the first week after discharge, being readmitted prior to the scheduled aftercare appointment (OR: 5.484, 95% CI: 3.188–9.435, P < 0.0001) or missing the first scheduled appointment (OR: 1.612, 95% CI: 1.016–2.558, P = 0.043) was associated with readmission. During the second week after discharge, patients readmitted prior to attending an aftercare appointment (OR: 2.033, 95% CI: 1.134– 3.645, P = 0.017) or reporting having no means of support after discharge were more likely to be readmitted (OR: 2.990, 95% CI: 1.155–7.742, P = 0.024). In the multivariate model, the likelihood ratio for the chi-square test with 92 degrees of freedom was equal to 157.20, and the probability of obtaining a chi-square statistic given the null hypothesis was true was less than 0.001. The pseudo r2 values were 0.24 (Cox and Snell) and 0.26 (Nagelkerke). Controlling for other factors in the model, a longer LOS during the prior hospitalization was associated with increased readmission risk during the first week compared with days 15–30 (OR: 1.041, 95% CI: 1.011–1.073, P = 0.008). BPRS-rated symptoms of grandiosity and suspiciousness were elevated among patients readmitted during the first 7 days following discharge. For each one point increase on the BPRS item grandiosity, the odds of readmission in days 1–7 relative to days
© 2015 John Wiley & Sons, Ltd.
J.E. Hamilton et al.
Differential factors
Table 2 Sample characteristics Variable Male Female African American White Hispanic Other race/ethnicity 12 years Unemployed Part time Full time Homeless Home Facility/hospital No financial support Government disability Prior arrest(s) Schizophrenia Bipolar Depression Other diagnosis Personality disorder Involuntary Age Prior LOS # HCPC hospitalizations
Total sample n = 588
1–7 days n = 206
8–14 days n = 166
15–30 days n = 216
363 225 267 228 79 14 134 228 226 476 45 67 126 327 135 145 163 186 245 216 102 25 82 312 Mean (SD) 35.1 (11.4) 11.8 (8.3) 5.5 (5.1)
129 78 95 81 23 8 47 81 79 161 20 26 41 109 56 51 61 69 81 80 41 4 33 109
99 68 77 67 20 3 41 71 55 140 12 14 35 97 34 37 44 54 80 55 24 7 21 88
135 79 95 80 36 3 46 76 92 175 13 27 49 123 44 56 57 64 84 81 37 14 28 115
35.3 (11.2) 12.9 (10.1) 5.5 (5.1)
35.1 (11.2) 11.3 (7.4) 5.2 (5.3)
34.5 (11.9) 10.9 (6.9) 5.6 (5.0)
HCPC, Harris County Psychiatric Center; LOS, length of stay; SD, standard deviation.
15–30 increased by a factor of 1.52 (OR: 1.515, 95% CI: 1.130– 2.031, P = 0.005). For each one point increase on the BPRS item suspiciousness, the odds of readmission in days 1–7 relative to days 15–30 increased by a factor of 1.40 (OR: 1.403, 95% CI: 1.067– 1.844, P = 0.015). Depression was associated with lower odds of readmission during the first week compared with days 15–30 (OR: 0.621, 95% CI: 0.840–0.841, P = 0.002). Additionally, patients with a diagnosis other than schizophrenia, depression or bipolar disorder (primarily substance abuse and anxiety disorders) had lower odds of being readmitted 1–7 days following discharge (OR: 0.207, 95% CI: 0.054–0.797, P = 0.022). Patients readmitted to the hospital prior to their first scheduled aftercare appointment were at an increased risk of readmission during 1–7 days compared with days 15–30 following discharge (OR: 10.230, 95% CI: 5.245–19.954, P < 0.001). Missing the first scheduled aftercare appointment was also associated with an increased risk of readmission during the first week after discharge (OR: 2.395, 95% CI: 1.338–4.285, P = 0.003). Patients reporting inconsistent sources of financial support were also at increased risk of readmission during the first 7 days after discharge (OR: 3.955, 95% CI: 1.088–14.370, P = 0.037). During 8–14 days following discharge, patients reporting 12 years of education were more likely to be readmitted (OR: 1.923, 95% CI: 1.140–3.245, P = 0.014). Additionally, patients readmitted before their first scheduled aftercare appointment were at increased risk of readmission during week 2 (OR: 2.489, 95%
© 2015 John Wiley & Sons, Ltd.
CI: 1.269–4.880, P = 0.008). Patients with a diagnosis of bipolar disorder had a decreased risk of readmission during the second week (OR: 0.564, 95% CI: 0.325–0.979, P = 0.042). Patients with a diagnosis other than schizophrenia, bipolar disorder or depression were also at a decreased risk of readmission during the second week following discharge (OR: 0.327, 95% CI: 0.109–0.981, P = 0.046). The complete results of the multivariate analysis are presented in Table 3.
Discussion Our study identified clinical, treatment and patient-reported factors associated with earlier readmission. Elevated symptoms of grandiosity and suspiciousness among earlier readmitted patients may indicate that patients did not adequately improve or needed further symptom stabilization prior to discharge during their previous hospitalization [12,14,33]. Additionally, patient externalization of these symptoms may have overwhelmed patient caregivers resulting in earlier readmission. In contrast, patients with elevated depression symptoms may have been less likely to readmit within 7 days of discharge because their internalizing behaviours were less problematic for caregivers. Attending psychiatrists may have also educated patients and caregivers about the longer length of time required for an antidepressant medication to be effective. With the knowledge that an antidepressant would not result in an immediate response, patients 575
Differential factors
J.E. Hamilton et al.
Table 3 Final model 1–7 days after discharge Predictors
OR
Intercept Socio-demographic Age Male (ref) Female African American (ref) White Hispanic Other Married (ref) Unmarried >12 years ed (ref)
Journal of Evaluation in Clinical Practice ISSN 1365-2753
Factors differentially associated with early readmission at a university teaching psychiatric hospital Jane E. Hamilton PhD MPH,1 Howard Rhoades PhD,4 Juan Galvez MD,1 Melissa Allen DO,2 Charles Green PhD,6 Mildred Aller5 and Jair C. Soares MD PhD3 1 Postdoctoral Research Fellow, 2Assistant Professor, 3Professor and Chairman, Department of Psychiatry, University of Texas Medical School at Houston, Houston, TX, USA 4 Senior Statistician, 5Research Coordinator, University of Texas Harris County Psychiatric Center, Houston, TX, USA 6 Associate Professor, Center for Clinical Research and Evidence-Based Medicine, University of Texas Medical School at Houston, Houston, TX, USA
Keywords evaluation, health services research, patient-centred care Correspondence Dr Jane E. Hamilton Department of Psychiatry University of Texas (UT) Medical School at Houston 2800 S. MacGregor Way, HCP 3-E50 Houston, TX 77021 USA E-mail: [email protected] Accepted for publication: 13 January 2015 doi:10.1111/jep.12335
Abstract Rationale, aims and objectives The rate of psychiatric readmissions within 30 days of discharge is a well-established behavioural health system performance measure linked to the quality of inpatient hospital care as well as to access to community-based aftercare services. The purpose of this study was to examine the factors differentially associated with earlier readmission among a sample of patients (n = 588) readmitted within 30 days of discharge to a university teaching psychiatric hospital from 2001 to 2010. Methods Quality assurance interviews were conducted with patients readmitted within 30 days of discharge. The interview data were merged with clinical symptom and electronic medical record data. Multinomial logistic regression analysis was used to examine readmission within 7 days and from 8 to 14 days compared with 15–30 days after discharge while controlling for socio-demographic and treatment variables previously associated with psychiatric readmission. Results Multiple clinical, treatment and patient-reported factors were differentially associated with earlier readmission. In particular, lack of engagement in post-discharge aftercare services was a strong predictor of earlier readmission. Conclusions Strategies are needed to improve patients’ transition from inpatient psychiatric hospitalization to aftercare services. Psychiatric hospitals attempting to reduce very early readmissions should seek to implement innovative transitional care initiatives targeting both patient and treatment factors.
Introduction In the era of health care reform, health care payers, policy makers and providers have become increasingly concerned about the high rates of patient readmissions following hospitalization [1]. Although the primary focus has centred on readmission to shortterm acute care hospitals, there is growing interest in understanding readmission following psychiatric hospitalization. Psychiatric readmission rates have historically been viewed as important indicators for health care planning due to their connection to the quality and continuity of care for patients as well as the high costs associated with additional inpatient care [2–6]. The rate of psychiatric patients readmitted within 30 days of discharge is an established behavioural health system performance measure [7] linked to the quality of inpatient hospital care [8] and to access to community-based aftercare services [9]. Although 572
30-day psychiatric readmissions have been a focus of research, studies have not examined differential influences of patient and treatment factors on critical time frames within a 30-day readmission period [10–16]. The imperatives of the Affordable Care Act to reduce 30-day readmissions in the United States provide an opportunity to develop and implement readmission reduction strategies. However, additional research is needed to elucidate how specific factors influence psychiatric readmissions at different time periods within 30 days of discharge. To extend current research, we linked and analysed patient interview, psychiatric symptom and medical record data obtained from our psychiatric hospital. The goal of our study was to examine readmission patterns among psychiatric patients and the predictors of 30-day readmission from 1 to 7 days and from 8 to 14 days after discharge compared with 15–30 days after discharge. Based on prior research, we hypothesized that
Journal of Evaluation in Clinical Practice 21 (2015) 572–578 © 2015 John Wiley & Sons, Ltd.
J.E. Hamilton et al.
patients would be at higher risk for readmission during the first 2 weeks following discharge [17–19]. Readmissions occurring very close to the time of discharge have been associated with factors related to hospital quality of care [20,21]. Additionally, suicidal patients requiring intensive aftercare services and family supports are at high risk for readmission during the first week following discharge [22,23]. Patient factors such as readiness for discharge and socio-economic status have been shown to affect readmission within 14 days of discharge [19,24,25]. In addition to hospital quality and patient factors, we wanted to explore additional factors predicting readmission from 8 to 14 days following discharge as patients transitioned into the community and attempted to access mental health services. Rather than examining time to readmission as a continuous variable, we established cut points for our outcome variable that reflect the 7-day supply of medication provided to patients at discharge and the time frame in which our discharged patients were scheduled for an aftercare appointment (within 10 days of discharge). We organized predictor variables conceptually into four areas: socio-demographic, clinical, treatment and patient interview information. The patient-reported information we collected has been previously found to influence psychiatric readmissions including education level, financial support, unemployment, homelessness, satisfaction with previous hospitalization, discharge readiness, medication adherence, aftercare services and aftercare attendance [15,19,25–32]. We are only aware of one prior study examining the relationship between patient-reported satisfaction with psychiatric care and 30-day readmissions [15], and we are not aware of any studies examining additional patient-reported information. Although several studies have examined differential factors associated with readmissions up to 5 years after discharge [19,33], no prior studies have examined differential factors associated with readmission within 30 days of discharge.
Differential factors
Measures Patient interviews Trained interviewers screened 30-day readmitted patients for participation in 10-minute interviews through a daily review of the hospital census. Prior to being interviewed, patients were identified as stable by unit staff and agreed to participation. The information obtained from the interviews has not been previously published. Interview questions are presented in Table 1. BPRS The BPRS is an 18-item clinician-rated measure of psychiatric symptoms [34] used to examine response to psychopharmacological treatment [35] and the link between psychiatric symptoms and psychiatric readmissions [13,36]. During the study period, 51% of 30-day readmitted patients were assessed with the BPRS by 53 different attending psychiatrists as part of their standard practice of care. Medical record data We extracted patient medical record information to control for socio-demographics (age, gender, race/ethnicity), DSM diagnosis and treatment characteristics [prior hospitalizations, length of stay
30-day readmiƩed paƟents January 2001 to November 2010 n = 5009 Excluded due to not being interviewed n = 3723
Methods Sample The sample included patients who participated in quality assurance interviews at our hospital from January 2001 to November 2010. Since 1986, our hospital has been the regional safety net provider of inpatient psychiatric services in Harris County, Texas. In the present study, we combined interview data with data from the Brief Psychiatric Rating Scale (BPRS) to examine the influence of psychiatric symptoms on 30-day readmissions, controlling for factors known to be associated with 30-day readmissions [10]. To examine how well our sample represented all our 30-day readmitted patients, we compared the demographic and clinical characteristics of the final sample (n = 588) with the demographic and clinical characteristics of a larger sample of all patients readmitted within 30 days of discharge during the same time period (n = 5009) as well as the entire sample of interviewed patients (n = 1286). The three samples did not differ proportionally by age, gender, race/ethnicity, Diagnostic and Statistical Manual of Mental Disorders (DSM) diagnosis or marital status (patient characteristics available for all three samples examined). A patient inclusion diagram is presented in Fig. 1. This study was approved by the University of Texas Health Science Center at Houston Institutional Review Board.
© 2015 John Wiley & Sons, Ltd.
ReadmiƩed paƟents interviewed n = 1286 Excluded due to incomplete BPRS data n = 617 ReadmiƩed paƟents with complete BPRS data n = 669 Excluded 2nd and subsequent interviews for paƟents with mulƟple readmissions n = 67 First-Ɵme interviewed paƟents n = 602 Excluded due to missing data n = 14 PaƟents included in final analysis n = 588 Figure 1 Patient inclusion diagram.
573
Differential factors
J.E. Hamilton et al.
Table 1 Patient interview questions Question
Response items
Marital status
0. Not married, divorced, widowed 1. Married Employment status 0. Unemployed 1. Part time 2. Full time Years of education 12 Arrest history 0. No prior arrest 1. Prior arrest Involuntary/voluntary status 1. Involuntary 2. Voluntary 0. No Since hospitalization, has the patient been 1. Part time 2. Full time employed? Sources of financial support: 0. None 1. Disability 2. Government 3. Family 4. Other sources 5. Employment 1. Homeless Where did the patient live 2. Home after the last hospitalization? 3. Hospital Patient’s belief as to why 1. Wasn’t ready to discharge s/he returned so quickly? 2. Medication not effective 3. Medication not taken 4. Stressful environment after discharge 5. Unsure Helpfulness of previous 1. Not helpful hospitalization: 2. Little 3. Somewhat 4. Very Compliance to psychiatric 1. Never taken medication: 2. Taken inconsistently 3. Taken consistently Experience with medication 1. Medication not effective effectiveness? 2. Medication was effective Experience with medication 1. Side effects side effects? 2. No side effects Aftercare referral: 0. None 1. Public system 2. Other 3. Day treatment 4. Facility/hospital Aftercare attendance: 0. Appointment in the future 1. Missed first appointment 2. Attended >1 appointment
(LOS), involuntary status] as performed in prior studies examining 30-day psychiatric readmissions [11–13,19]. Private health information including patient name, address, date of birth and social security number was not included in the dataset to protect patient confidentiality.
Statistical analysis We explored the effect of predictors on patient readmissions within 7 days and from 8 to 14 days of discharge, using 15–30 days 574
after discharge as the reference group in a series of univariate and multinomial logistic regression analyses. Data were analysed with IBM SPSS Statistics for Windows, Version 21.0. Armonk, NY: IBM Corp.
Results In the final sample of 588 adult patients (mean age 35.5 years; range 18–71 years), the majority of patients (372; 63%) were readmitted during the first 2 weeks following discharge. Readmitted patients were more likely to be African American, male, unemployed and diagnosed with schizophrenia. A full listing of the sample characteristics is presented in Table 2. The univariate (unadjusted) analyses showed that patients with a diagnosis other than depression, bipolar disorder or schizophrenia were less likely to return during the first week following discharge compared with days 15–30 [odds ratio (OR): 0.292, 95% confidence interval (CI): 0.093–0.924, P = 0.036]. Patients with elevated BPRS symptoms of grandiosity (OR: 1.442, 95% CI: 1.135–1.832, P = 0.003) and suspiciousness (OR: 1.252, 95% CI: 1.015–1.544, P = 0.036) were more likely to be readmitted during week 1 compared with 15–30 days after discharge. Patients with elevated BPRS symptoms of grandiosity (OR: 1.314, 95% CI: 1.017–1.697, P = 0.037) and suspiciousness (OR: 1.318, 95% CI: 1.063–1.633, P = 0.012) were also more likely to be readmitted during week 2 compared with 15–30 days after discharge. Patients with elevated symptoms of depression (OR: 0.757, 95% CI: 0.594–0.963, P = 0.024) had a decreased risk of readmission during week 1 compared with 15–30 days after discharge. The patient reported belief of not being ready for discharge during the previous hospitalization predicted readmission within 7 days of discharge compared with 15–30 days after discharge (OR: 2.849, 95% CI: 1.203–6.749, P = 0.017). A longer LOS also predicted readmission during 1–7 days after discharge (OR: 1.032, 95% CI: 1.007–1.058, P = 0.013). Patients referred to aftercare within our public mental health system were less likely to be readmitted during the first week following discharge (OR: 0.485, 95% CI: 0.235–0.893, P = 0.022). During the first week after discharge, being readmitted prior to the scheduled aftercare appointment (OR: 5.484, 95% CI: 3.188–9.435, P < 0.0001) or missing the first scheduled appointment (OR: 1.612, 95% CI: 1.016–2.558, P = 0.043) was associated with readmission. During the second week after discharge, patients readmitted prior to attending an aftercare appointment (OR: 2.033, 95% CI: 1.134– 3.645, P = 0.017) or reporting having no means of support after discharge were more likely to be readmitted (OR: 2.990, 95% CI: 1.155–7.742, P = 0.024). In the multivariate model, the likelihood ratio for the chi-square test with 92 degrees of freedom was equal to 157.20, and the probability of obtaining a chi-square statistic given the null hypothesis was true was less than 0.001. The pseudo r2 values were 0.24 (Cox and Snell) and 0.26 (Nagelkerke). Controlling for other factors in the model, a longer LOS during the prior hospitalization was associated with increased readmission risk during the first week compared with days 15–30 (OR: 1.041, 95% CI: 1.011–1.073, P = 0.008). BPRS-rated symptoms of grandiosity and suspiciousness were elevated among patients readmitted during the first 7 days following discharge. For each one point increase on the BPRS item grandiosity, the odds of readmission in days 1–7 relative to days
© 2015 John Wiley & Sons, Ltd.
J.E. Hamilton et al.
Differential factors
Table 2 Sample characteristics Variable Male Female African American White Hispanic Other race/ethnicity 12 years Unemployed Part time Full time Homeless Home Facility/hospital No financial support Government disability Prior arrest(s) Schizophrenia Bipolar Depression Other diagnosis Personality disorder Involuntary Age Prior LOS # HCPC hospitalizations
Total sample n = 588
1–7 days n = 206
8–14 days n = 166
15–30 days n = 216
363 225 267 228 79 14 134 228 226 476 45 67 126 327 135 145 163 186 245 216 102 25 82 312 Mean (SD) 35.1 (11.4) 11.8 (8.3) 5.5 (5.1)
129 78 95 81 23 8 47 81 79 161 20 26 41 109 56 51 61 69 81 80 41 4 33 109
99 68 77 67 20 3 41 71 55 140 12 14 35 97 34 37 44 54 80 55 24 7 21 88
135 79 95 80 36 3 46 76 92 175 13 27 49 123 44 56 57 64 84 81 37 14 28 115
35.3 (11.2) 12.9 (10.1) 5.5 (5.1)
35.1 (11.2) 11.3 (7.4) 5.2 (5.3)
34.5 (11.9) 10.9 (6.9) 5.6 (5.0)
HCPC, Harris County Psychiatric Center; LOS, length of stay; SD, standard deviation.
15–30 increased by a factor of 1.52 (OR: 1.515, 95% CI: 1.130– 2.031, P = 0.005). For each one point increase on the BPRS item suspiciousness, the odds of readmission in days 1–7 relative to days 15–30 increased by a factor of 1.40 (OR: 1.403, 95% CI: 1.067– 1.844, P = 0.015). Depression was associated with lower odds of readmission during the first week compared with days 15–30 (OR: 0.621, 95% CI: 0.840–0.841, P = 0.002). Additionally, patients with a diagnosis other than schizophrenia, depression or bipolar disorder (primarily substance abuse and anxiety disorders) had lower odds of being readmitted 1–7 days following discharge (OR: 0.207, 95% CI: 0.054–0.797, P = 0.022). Patients readmitted to the hospital prior to their first scheduled aftercare appointment were at an increased risk of readmission during 1–7 days compared with days 15–30 following discharge (OR: 10.230, 95% CI: 5.245–19.954, P < 0.001). Missing the first scheduled aftercare appointment was also associated with an increased risk of readmission during the first week after discharge (OR: 2.395, 95% CI: 1.338–4.285, P = 0.003). Patients reporting inconsistent sources of financial support were also at increased risk of readmission during the first 7 days after discharge (OR: 3.955, 95% CI: 1.088–14.370, P = 0.037). During 8–14 days following discharge, patients reporting 12 years of education were more likely to be readmitted (OR: 1.923, 95% CI: 1.140–3.245, P = 0.014). Additionally, patients readmitted before their first scheduled aftercare appointment were at increased risk of readmission during week 2 (OR: 2.489, 95%
© 2015 John Wiley & Sons, Ltd.
CI: 1.269–4.880, P = 0.008). Patients with a diagnosis of bipolar disorder had a decreased risk of readmission during the second week (OR: 0.564, 95% CI: 0.325–0.979, P = 0.042). Patients with a diagnosis other than schizophrenia, bipolar disorder or depression were also at a decreased risk of readmission during the second week following discharge (OR: 0.327, 95% CI: 0.109–0.981, P = 0.046). The complete results of the multivariate analysis are presented in Table 3.
Discussion Our study identified clinical, treatment and patient-reported factors associated with earlier readmission. Elevated symptoms of grandiosity and suspiciousness among earlier readmitted patients may indicate that patients did not adequately improve or needed further symptom stabilization prior to discharge during their previous hospitalization [12,14,33]. Additionally, patient externalization of these symptoms may have overwhelmed patient caregivers resulting in earlier readmission. In contrast, patients with elevated depression symptoms may have been less likely to readmit within 7 days of discharge because their internalizing behaviours were less problematic for caregivers. Attending psychiatrists may have also educated patients and caregivers about the longer length of time required for an antidepressant medication to be effective. With the knowledge that an antidepressant would not result in an immediate response, patients 575
Differential factors
J.E. Hamilton et al.
Table 3 Final model 1–7 days after discharge Predictors
OR
Intercept Socio-demographic Age Male (ref) Female African American (ref) White Hispanic Other Married (ref) Unmarried >12 years ed (ref)
