Risks for re-hospitalization of persons with severe mental illness living in rehabilitation care settings

Background The high rates of psychiatric re-hospitalizations (also termed “revolving door”) presents a “wicked problem” which requires a systematic and holistic approach to its resolution. Israel’s mental-health rehabilitation law provides a comprehensive set of services intended to support the ability of persons with severe mental illness to rely on community rather than in-patient facilities for their ongoing care needs. Guided by the Health Behavior Model, we examined the relationship between psychiatric re-hospitalizations and the three Health Behavior Model factors (predisposing factor: socio-demographic characteristics and health beliefs; enabling factor: personal and social/vocational relationships facilitated by rehabilitation interventions and services; and need factor: outcomes including symptoms, and mental health and functional status) among persons with severe mental illness receiving rehabilitation services. Methods Logistic regression models were used to measure the association between re-hospitalization within a year and variables comprising the three Health Behavior Model factors on the sample of consumers utilizing psychiatric services (n = 7,165). The area under the curve for the model was calculated for each factor separately and for all three factors combined. Results A total of 846 (11.8%) consumers were hospitalized within a year after the study began. Although multivariable analyses showed significant associations between re-hospitalization and all three Health Behavior Model factors, the magnitude of the model’s area under the curve differed: 0.61 (CI = 0.59–0.64), 0.56 (CI = 0.54–0.58), 0.78 (CI = 0.77–0.80) and 0.78 (CI = 0.76–0.80) for predisposing, enabling, need and the full three-factor Health Behavior Model, respectively. Conclusion Findings revealed that among the three Health Behavior Model factors, the need factor best predicted re-hospitalization. The enabling factor, comprised of personal relationships and social/vocational activities facilitated by interventions and services representing many of psychiatric rehabilitation’s key goals, had the weakest association with reduced rates of re-hospitalization. Possible explanations may be inaccurate assessments of consumers' personal relationships and social/vocational activities by the mental healthcare professionals, problematic provider-consumer communication on the consumers' involvement in social/vocational activities, or ineffective methods of facilitating consumer participation in these activities. Clearly to reduce the wicked “revolving-door” phenomenon, there is a need for targeted interventions and a review of current psychiatric rehabilitation policies to promote the comprehensive integration of community rehabilitation services by decreasing the fragmentation of care, facilitating continuity of care with other healthcare services, and utilizing effective personal reported outcomes and experiences of consumers with severe mental illness.


Background
Persons with severe mental illness (defined as diagnoses such as schizophrenia, schizoaffective disorder, bipolar disorder, and severe, chronic depressive disorder possessing psychotic presentations during which there is chronic pervasive impairment in all or most aspects of personal and social functioning [1] encounter the continuing problem of psychiatric re-hospitalizations, which has been referred to as the "revolving-door" phenomenon [2][3][4][5].One approach used to address this phenomenon has been to change the mental healthcare delivery system from the traditional medical model to a more holistic, biopsychosocial, recovery-oriented path [6].Using this biopsychosocial approach, mental healthcare professionals conduct a thorough assessment of the lives of persons with severe mental illness including their life goals, experiences and challenges, and based on this information, identify and implement the specific services and interventions that would best meet the needs of the individual and the individual's goals.Community-based rehabilitation, following psychiatric hospitalization discharge, provides mental health treatment, interventions, services and other resources that support daily living activities and improve personal, social and vocational skills, all of which optimally will promote societal integration and prevent re-hospitalizations [4,7].
Other factors influence the risk of re-hospitalization.For example, socio-demographic and clinical characteristics, previous use of psychiatric services, and prior hospitalization are associated with re-hospitalization [5,8,9].Perceptions of personal well-being, such as self-reported quality of life, self-assessment of mental health symptoms and functioning also are related to re-hospitalization; as a result, many researchers and rehabilitation programs have worked to improve quality of life [10][11][12].Nevertheless, despite these efforts, the wicked "revolving-door" phenomenon persists [2,3,13,14], presenting a hardship to persons with severe mental illness and their families, and a burden on the healthcare service system [8,12,14,15].
Although an extensive body of literature has investigated psychiatric re-hospitalizations, few studies have examined the impact of psychiatric rehabilitation services in reducing readmission rates [16].A major problem in this field, which was proposed more than two decades ago, is that although substantial information has been collected on the predictors of psychiatric rehospitalization, there lacks an overarching theoretical framework explaining re-hospitalization [17].Furthermore, there is a need for "principles based on a robust theory to guide practices that enhance participation in institutional settings" [18].Mental healthcare professionals must critically examine services that contribute to the best healthcare outcomes and revise those services that tend to produce undesirable healthcare outcomes.Understanding complex phenomena, such as re-hospitalization, can benefit from a broad theoretical conceptualization that identifies the wide range of health behaviors that potentially can lead to either desirable or undesirable outcomes [19].The current study thus utilizes the theoretical Health Behavior Model as a framework to guide the comprehensive assessment of community-based, psychiatric rehabilitation and its association with psychiatric re-hospitalization.

Health behavior model
The well-known Health Behavior Model [20] has been used by many researchers to examine different factors leading to re-hospitalization [19]; however, it most frequently has been applied to physical health conditions [21].Moreover, the Health Behavior Model contains feedback loops that indicate how health behaviors can influence predisposing, enabling and need factors [19,20,22,23] and consequently outcomes.
Substantial research has focused on the three Health Behavior Model factors.Systematic reviews have examined psychiatric re-hospitalization and found associations with several variables represented by the predisposing factor (e.g., age, sex, marital status and care management trajectories) [3,19,21].For example, studies show that younger age and being single are associated with greater readmission risk [3].The enabling factor includes using, accessing and possessing services or resources, such as possessing good family relationships or a routine source of care, which protect against health deterioration or rehospitalization [19,21,24,25].It is important to note that the purpose of community-based rehabilitation is to provide interventions and services that promote social/ vocational activities, improve personal and family relationships, and encourage activities facilitating daily functioning for persons with severe mental illness [1].These services, which can increase personal, vocational and other resources (the enabling factor), are designed and modified by the health professionals at the communitybased rehabilitation setting to promote social integration and reduce the risk for re-hospitalization.The third factor is need, which typically is represented by objective health measurements and includes the length of stay, and the number and total time of previous hospitalizations [3,8,19,23].
The Health Behavior Model is used to guide this research by providing a broad perspective on psychiatric re-hospitalizations among persons with severe mental illness and predicting healthcare utilization and health behavior outcomes through its three factors (i.e., predisposing, enabling and need factors).The predisposing factor includes socio-demographic variables; the enabling factor is represented by the access, use and possession of resources including personal relationships, and daily living and social/vocational activities that result from the use of community-based rehabilitation interventions and services; and the need factor refers to variables of health function and status.Table 1 describes the classification of the study variables using accepted definitions of the Health Behavior Model factors [19,20,22].For example, income, the availability of financial resources to pay for services, the effective price of healthcare, having insurance, personal relationships, social support and regular source of care have all been classified as enabling factors that ease or hinder access to services [19][20][21][22].To the best of our knowledge, this is the first study using the three Health Behavior Model factors to investigate the relationship of the broad range of need, enabling and predisposing risk factors to re-hospitalizations among persons with severe mental illness.

Mental health rehabilitation services in Israel
Israel 's national health service has provided universal coverage through its National Health Insurance Law since 1995.Yet, the provision of mental health services was not comprehensively addressed in this law.In response to the growing need for mental health reform and de-institutionalization (shifting the main focus of psychiatric care from hospitals to the community), Israel passed the progressive "Rehabilitation of the Mentally Disabled in the Community Act 2000" [26].The Rehabilitation Act provides a wide range of employment, housing, education, case-management, dental care, social-life and leisure activities (i.e., known as "rehabilitation-basket") to all residents, aged 18 or over, who were diagnosed with severe mental illness and pass the threshold of a 40% psychiatric disability [27,28].
The "rehabilitation-basket" of services is tailored to the specific needs, preferences, and goals of each psychiatric service consumer [29].Qualifying persons with severe mental illness may receive placements at hostels/ assisted living arrangement or adapted employment facilities.At these residences, mental healthcare physicians, nurses, social workers and other healthcare professionals not only provide therapy, but also assist persons with severe mental illness to improve their skills in daily living, home management, symptom self-management, physical health education, vocational rehabilitation, social skills, and development of community networks.A major goal of the Rehabilitation Act is to facilitate community participation and the recovery process [30].
Since 2012, as part of the national program, the Psychiatric Rehabilitation Outcome Measures project was initiated for persons with severe mental illness who were consumers of community-based rehabilitation services and who were either living in a hostel/assisted-living arrangement or were living at home and working at an employment facility.Participants were invited to complete an annual comprehensive assessment of their social ties, employment, illness literacy, coping with symptoms, wellbeing, functioning, and support that they received as well as challenges with which they were confronted in the rehabilitation setting [28].Although the Psychiatric Rehabilitation Outcome Measures project's database did not include collect information on the participants' psychiatric diagnoses, the cohort comprising the database is representative of persons with severe mental illness in community rehabilitation; and according to a report on this population, the majority of adults in community rehabilitation are diagnosed with schizophrenia, schizoaffective disorder, bipolar disorder and some chronic delusional disorders [31].Accordingly, these data provide a snapshot of the state of consumers with severe mental illness receiving community-based rehabilitation services.

Study aim
Using the Health Behavior Model theoretical framework, the purpose of this study was to measure the contributions of predisposing, enabling and need factors to the risk of psychiatric re-hospitalizations among persons with severe mental illness in community-based, psychiatric rehabilitation.An additional aim was to determine whether the enabling factor contributed to the prediction of re-hospitalization, above and beyond the predisposing and need factors.This last issue has important policy implications as the enabling factor comprises the use, access and possession of personal, social and vocation resources that are facilitated by engaging in rehabilitation interventions and services.Therefore, it is the enabling factor that mental healthcare professionals can modify to better serve consumers with severe mental illness.

Design and setting
This secondary data analysis used the prospective dataset of the psychiatric rehabilitation outcome measures project, collected between April 2014 and April 2017, comprising a cohort of mental health service consumers with severe mental illness who reside in community-based rehabilitation (N = 14,321).The psychiatric rehabilitation outcome measures dataset, prior to this secondary analysis, already were de-identified to preserve anonymity.Other aspects of the psychiatric rehabilitation outcome measures dataset collection are detailed elsewhere [28].The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation as dictated by the Helsinki Declaration of 1975 and its revision in 2008.We confirm that written informed consent was obtained from all participants, and that all experimental protocols and consenting procedures were approved As part of the psychiatric rehabilitation outcome measures project, participants completed either the regular or short form (for those consumers who were unable to complete the regular form) of the "consumers` questionnaire" annually.The questionnaire was available in six languages and could be completed by either computer or paper and pen.
We assessed all study participants for prior hospitalizations with data provided by the centralized data warehouse of the Ministry of Health.Previous reports show that non-consenting consumers of mental health services (due to cognitive impairment or difficulty in completion) differed from study participants, as they were approximately 3 years older, predominately men, and had a greater number of prior hospitalizations [10].
We used listwise deletion to handle missing data.A total of 2,158 (15.1%) mental health service consumers had no information on previous hospitalizations, and therefore, were not included in the current study.Additionally, we excluded observations of 2,338 (16.3%) mental health service consumers who completed only the short form (due to cognitive impairment or difficulty in questionnaire completion) and 2,660 (18.6%) mental health service consumers who responded to an earlier version of the questionnaire that did not include many items from the enabling and predisposing factors.The baseline sample included data from 7,165 (50%) consumers.

Outcome variable
The dichotomous dependent variable was hospitalization to a psychiatric ward within a year after enrollment into the psychiatric rehabilitation outcome measures project and completion of the study's questionnaire.This indicator was operationalized as "re-hospitalization" as all study participants had been hospitalized prior to admission to the community-based rehabilitation where data were collected.

Predictor variables
Independent variables were organized according to the Health Behavior Model factors (See Table 1).We explored all available variables within the database and mapped them according to the model factors.Most original variables used 4-point Likert scales: from 1 (not at all) to 4 (very good).Examination of these variables' distributions revealed that certain values were rarely selected, consequently, the 4-point Likert ratings were ordinal variables categorized to: 1 (not at/ all poor) to 0 (fair/ very good).
The enabling factor included the following: knowledge about illness, working status, education level, satisfaction with relationship with family of origin, involvement in an intimate relationship, number of close friends, social functioning, frequency of contact with friends, use of community resources, community functioning and use, and satisfaction with leisure activities (nominal variables).
The need factor included the following variables comprising objective measures (i.e., the number of days since last hospitalization and the number of hospitalizations prior to enrollment) (interval variables); and selfreported measures (i.e., quality of life, daily functioning and the effect of mental health symptoms on functioning) (ordinal variables).
Quality of life was assessed with a shortened version of the Manchester Assessment of Quality of Life scale.Seven items on the self-perception of physical and mental health, economic situation, residence, leisure and social activities domains were rated on a 5-point Likerttype scale from 1 (completely unsatisfied) to 5 (completely satisfied) [32].The summed score was used.The current scale was previously translated and validated in Hebrew and was found to have good psychometric properties [33,34].For this scale, the study's sample attained a Cronbach's α = 0.740.
Daily Functioning was assessed using a tool developed specifically for the project; it included 8 items taken from the Behavior and Symptoms Identification Scale [35,36] and the Role Functioning Scale [37].Internal consistency for these items in the study's sample attained a Cronbach's α = 0.834.
The influence of mental health symptoms on daily functioning (4 items) was assessed using a modified Sheehan Disability Scale, containing 4-point Likert-type scales from 1 (interferes strongly) to 4 (does not interfere) [38,39].This scale assessed the following four life domains: work and/or study, family relations, leisure and social activities.We used the summed score of the four items.The internal consistency for this scale in this study's sample attained a Cronbach's α = 0.845.

Analyses
Descriptive statistics were conducted to examine the distribution of all study variables.Bivariate analyses were conducted to compare re-hospitalized and nonre-hospitalized consumers: Chi-square tests were used to compare categorical variables and Student t-tests for continuous variables.
Multivariate logistic regression models were constructed for each of the factor models (predisposing, enabling and need) separately.Odds ratios (OR) and 95% confidence intervals (CI) are reported.Variables were included in the multivariate models if they were significantly associated with the outcome in the bi-variate analyses.All variables included in the specific factor models were included in the final 3-factor model (combining all enabling, predisposing, and need factors).Due to missing data on variables used in regression analyses, the final model represented 4,613 consumers.We compared the final and the baseline samples and found significant differences only with the variables: place of residence, education and mother's birthplace.There were no differences between the two groups in age, sex, degree of coping with psychiatric or emotional issues, work status, religion, degree of satisfaction with familial relationship, involvement in intimate relationship and degree of satisfaction with social ties.
We assessed the discriminatory power of the model (ability to discern among those re-hospitalized and not re-hospitalized) by measuring the area under the curve using the Receiver Operating Characteristic's C-statistic for each Health Behavior Model factor and for the full three-factor regression models.A C-statistic of 0.5 indicates model discrimination beyond what is achievable by chance.C-statistics of 0.5-0.6,0.6-0.7,0.7-0.8,0.8-0.9, and 0.9-1.0represent poor, fair, good, very good and excellent discrimination, respectively [40].Data were analyzed using SPSS ® Version 27.

Results
Of the total consumers (n = 7,165), 846 (11.8%) were hospitalized within a year after enrollment into the psychiatric rehabilitation outcome measures project and completion of the study's questionnaire.The bivariate analysis (Table 2) shows that consumers who were rehospitalized were significantly more likely to be younger (< 35 years of age), non-Jewish (vs.Jewish), use full residential services (hostel) and be without children, compared to non-re-hospitalized consumers.Re-hospitalized consumers also reported significantly poorer levels of coping with daily emotional issues, poorer levels of satisfaction with leisure activities, lower overall satisfaction with quality of life, and report moderate or much interference of mental health symptoms with their daily functioning, in comparison to those who were not rehospitalized.Also, those hospitalized within a year, compared to others, were more likely to have been recently (less than a year) discharged from the hospital and hospitalized prior to study enrollment (see Table 2).
The predisposing factor model showed that among persons with severe mental illness, those who were at significantly higher odds for re-hospitalization were less likely to be older, more likely to be non-Jewish, less likely to use full residential services compared to partial or no residential services, and less likely to report fair or very good coping with psychiatric or emotional issues compared to not at all or poor coping (Table 3).For the enabling factor, findings showed that those who were at significantly higher odds for re-hospitalization were more likely to report fair or very good social functioning, compared to not at all or poor functioning, and less likely to report fair or very good satisfaction with leisure activities compared to not at all or poor satisfaction (Table 3).For the need factor, variables indicated that those with significantly higher odds for re-hospitalization were less likely to report they had symptoms affecting or interfering with daily functioning, more likely to have been hospitalized during the year prior to completing the questionnaire, and more likely to have reported a higher number of hospitalizations prior to study enrollment (see Table 3).
The full three-factor Health Behavior Model regression model (i.e., predisposing, enabling and need factors) showed that those who had significantly lower odds for re-hospitalization were those who lived at home (OR = 0.72, CI = 0.56-0.92)or at assisted living facilities (OR = 0.69, CI = 0.51-0.93)compared to those who resided in hostels; who reported having others at least somewhat involved in their mental health treatment (versus others who were not at all involved or only involved in case of a serious problem) (p = 0.02); and who reported that their symptoms had no or little effect on their daily functioning (p < 0.001).Persons with severe mental illness who had significant higher odds for re-hospitalization were those with a higher number of hospitalizations prior to enrollment in the psychiatric rehabilitation outcome measures project (OR = 1.13,CI = 1.10-1.17)(see Table 4).Additionally, the odds of re-hospitalization were highest for those who were hospitalized up to a year before study enrollment (OR = 5.43, CI = 3.87-7.60)followed by those who were hospitalized 1-10 years prior to study enrollment (OR = 2.17, CI = 1.63-2.89)compared to those with no hospitalization in the previous 10 years.
The model for the predisposing factor was significant (p < 0.001) and yielded a C-statistic of 0.61 (CI = 0.59-0.64),indicating fair discrimination.The model for the enabling factor was significant (p < 0.001) and yielded a C-statistic of 0.56 (CI = 0.53-0.58),indicating poor discrimination.The model for the need factor was significant (p < 0.001) and yielded a C-statistic of 0.78 (CI = 0.76-0.80),indicating good discrimination.The full three-factor Health Behavior Model regression model was significant (p < 0.001) and yielded a C-statistic of 0.79 (CI = 0.76-0.80)(see Fig. 1).
To assess whether the enabling factor contributed to re-hospitalization prediction "above and beyond" the other two factors (predisposing and need), we performed C-statistics for the first and second models were similar indicating that neither the enabling factor alone nor the enabling and predisposing factors combined effectively discriminated between those who were versus those who were not rehospitalized.Only the integrative (third) model, which was strongly influenced by the need factor, demonstrated good discriminating power (Fig. 1).

Discussion
Among the three Health Behavior Model factors, we found that the need factor (representing the individual's general condition, e.g., the number of days since last hospitalization, quality of life, daily functioning and the effect of mental health symptoms on functioning) was the best predictor of re-hospitalizations among persons with severe mental illness using rehabilitation services.In fact, the model containing solely the need factor demonstrated virtually the same fit as the integrated model combining all three Health Behavior Model factors.The predisposing factor (including sociodemographic characteristics) contributed very little to the prediction of re-hospitalization; and the enabling factor represented by the access, use and possession of resources including personal relationships, and daily living and social/vocational activities that are facilitated by the use of community-based rehabilitation interventions and services, contributed virtually nothing to the re-hospitalization prediction.The robustness of the need factor variables as predictors of re-hospitalization is consistent with other research [5,21,41,42].The need factor in this study included both objective measures representing prior hospital service use (time since last hospitalization and number of prior hospitalizations) and the self-reported assessment of personal well-being, as assessed by the effect of the mental health symptoms on daily functioning.The importance of self-reported assessments related to health status and functioning as predictors for health outcomes has been noted in other studies examining early identification of illness and re-hospitalization prevention [6,23,43].Another study, using psychiatric rehabilitation outcome measures data, found that self-reports of the effect of symptoms on functioning significantly predicted 12-month hospitalization, beyond objective assessments of prior healthcare use, and thus highlighted the importance of self-reported outcome measures in rehospitalization prediction [10].
Our finding that the predisposing factor contributed very little to re-hospitalization prediction was somewhat surprising.Several prior studies have reported the association between psychiatric re-hospitalization and variables representing the predisposing factor, such as age, sex, marital status, ethnicity, place of residence and religious affiliation [3,14,15,44].However, the lack of association between re-hospitalization and the predisposing factor is consistent with other studies that found no correlation between re-hospitalizations and personal socio-demographic variables, especially after controlling for prior psychiatric history of healthcare service use [13].
Of all the predisposing variables, only the type of residence remained associated with re-hospitalization after all variables representing the three-factor Health Behavior Model were entered into the model.A potential explanation for the unique contribution of this variable is that residential type represented a proxy for the person's overall condition (in terms of functioning and medical needs), as those residing in hostels are persons with lower functioning or greater disability requiring more care and oversight than those who are independent or require only assisted living [29].Because assignment into a specific residential type is contingent upon the person's condition, the ability to directly assess the contribution of residential services on outcomes remains a challenge.
Further research should consider developing and using complex case-mix measures that more accurately capture the relationship between the personal health and functional status with the types of services consumed, including residential placements, and outcomes [45].
The third Health Behavior Model factor, the enabling factor in this study, comprising variables such

Previous hospitalizations
No hospitalization in previous 10 years 1.00 Previous hospitalization 1-10 years before enrollment 2.17 as personal and family relationships and engaging in social/vocational activities, which aim to enhance personal resources, increase social support and the ability to perform daily routines independently, demonstrated little contribution to the prediction of rehospitalization in this study.This finding was indeed disappointing, as psychiatric rehabilitation has the goal of increasing social capability and community integration (such as being employed and having various social relationships) [1].These services also help participants plan and carry out individualized rehabilitation plans and goals [27,46].Previous research on psychiatric re-hospitalization showed a host of enabling factor variables (i.e., contact with family members, family involvement, education, occupation, and social/emotional functioning and support) as related to reduced re-hospitalization [15,22,44,47].Our study observed that only involvement of others in the persons' mental health treatment was associated with reduced odds of re-hospitalization.
One enabling factor component that may play an important role in the prevention of re-hospitalization is post-discharge care; however, Donisi et al. (2016) showed that having post-discharge contact with a healthcare professional (i.e.psychiatrist, mental health nurse and psychologist) in the community service, after controlling for other predictors, did not serve as a protective factor for early re-hospitalization [41].This lack of association may indicate that mere post-discharge contact with a healthcare professional is insufficient.Perhaps, post-discharge care would be related to re-hospitalization if the mental healthcare professionals at the psychiatric inpatient unit strengthened the transfer process by both preparing the person with severe mental illness for the transition and ensuring that the healthcare staff at the rehabilitation setting received the essential information needed to facilitate a smooth transition.
Psychiatric rehabilitation strives to help persons with severe mental illness to re-enter the general population and remain de-institutionalized. Obtaining employment, therefore, is among the goals.Goldman and Frank [48] claimed that for persons with severe mental illness supported employment increased the ability to participate in competitive employment, although these employment interventions rarely evolved into a fulltime job.Goldman and Frank emphasized that "the implementation of an array of rehabilitation services encourages us to believe that we can alter the course of mental illness services delivery" [48].That is, no single service by itself is effective.Indeed, our study did not find a relationship between being employed and re-hospitalization risk, potentially indicating the complex nature of the definition of vocation within a mental-health rehabilitation context.
Our study contributes to the field by examining several different rehabilitation services and using a nationwide, self-reported data from a survey completed by persons with severe mental illness receiving rehabilitation services.Yet, generalizability to the general population of persons with severe mental illness may be limited.An estimated 15-20% of the eligible population apply for the services that they need [27].Amongst those, 25-30% did not use rehabilitation services, and therefore could not participate in the psychiatric rehabilitation outcome measures project [27,49].Moreover, populations of community-based rehabilitation may vary due to differences in criteria for admission, program specifications and other country-specific characteristics, so generalizations to other country's rehabilitation population of severe mental illness must be made with caution.For example, in this study, very few persons with severe mental illness reported illicit drug use; yet, it is well documented that comorbid substance abuse is quite prevalent among persons with severe mental illness [50][51][52][53].The low prevalence rate of substance abuse may indicate an unwillingness to report behaviors that might result in sanctions such as reducing benefits, including disability payments [54].
Additionally, persons with severe mental illness who responded to the short form or early version of the psychiatric rehabilitation outcome measures questionnaire, due to low cognitive abilities or inability to complete the full version, were not included in the final data sample.Additionally, data on psychiatric medical diagnosis was not available in the psychiatric rehabilitation outcome measures dataset which limited our ability to examine the medical predisposition.Moreover, the psychiatric rehabilitation outcome measures dataset did not include variables such as duration of hospitalizations, hospitalization's admission diagnosis or the dates when each hospitalization occurred; and these variables may have provided further detail to our findings.Nonetheless, all participants had a psychiatric disability level of 40%, indicating their need for post-acute rehabilitation services.The generalizability of this study's population was previously checked and reported [10].Additionally, this study's strength is that it comprised a nationwide dataset of persons with severe mental illness of persons of all sexes, religions and countries of birth, which allowed for a broad examination of personal and social factors.

Conclusions
By using the Health Behavior Model theory for understanding factors that can discriminate between persons with severe mental illness who require psychiatric re-hospitalization versus those who do not, we demonstrated the robustness of the need factor and the minor contributions of the enabling and predisposing factors.Characteristics which were shown to be related to the basic health needs of persons with severe mental illness included both objective (number of prior hospitalizations and time since last hospitalization) and subjective (selfreported effect of the symptoms on functioning) measures.These findings, coupled with the contribution of self-reported measures on the support dealing with the mental illness, exemplify the importance of understanding the personal reported outcomes and experiences, as part of the overall conditions and situations that put individuals at risk for psychiatric re-hospitalization.

Policy implications
Community-based rehabilitation aims to increase societal interactions for persons with severe mental illness; but to develop effective policies on community-based rehabilitation, a better understanding of the array of interventions and services that promote societal interactions is needed.Future studies, therefore, must determine ways to more precisely measure each type of intervention and service as well as the contribution of mental healthcare professionals.Also, since patterns of mental health treatment, hospitalization and rehabilitation may differ for persons in minority populations with severe mental illness, research is needed to focus specifically on minority persons with severe mental illness and assess both the trajectory of mental health care and the existence (or not) of the "revolving door" phenomenon in this population.In addition, as mental health rehabilitation services are still not a fully integrated part of Israel's healthcare system, a more comprehensive approach to mental healthcare is required.
Israel's recent Mental Health Care Reform, enables better integration of physical and mental medical services, under the auspices of Israel's four Health Funds, which serve as non-for-profit insurers and providers of healthcare services.Yet, community rehabilitation services operate separately, under the direct supervision of the Ministry of Health, increasing fragmentation and lack of continuity with all other healthcare services.Without a unifying approach to mental healthcare, the ability of psychiatric rehabilitation services to provide comprehensive ongoing support, which effectively reduces re-hospitalizations, remains partial, and therefore diminishes our ability to resolve the wicked "revolving-door" phenomenon.

Fig. 1 C
Fig. 1 C-statistic for the Health Behavior Model

Table 2
Model characteristics for predisposing, enabling and need factors

Need factors How is your satisfaction with quality of life?
sensitivity analysis, in a nested fashion, which tested the enabling model (first model), in comparison to the combined enabling and predisposing model (second model), and lastly the integrative model combining enabling, predisposing, and need model (third model).