Summary of findings

The purpose of this systematic review was to describe and evaluate the efficacy of interventions designed to improve the transition from in-patient to out-patient care in reducing early psychiatric readmission among adults with mental illness. The study is unique in its detailed characterisation of interventions with pre-discharge, post-discharge and bridging components and rigorous quality review. Our systematic search identified 15 studies evaluating a heterogeneous group of interventions that included 15 distinct intervention components. Interventions had a statistically significant impact on readmission in seven of the studies and the effect sizes in these studies were relatively large (13.6–37.0% absolute reduction in readmission rates). We found evidence that, on their own, psychoeducation interventions targeting disease management and living skills reduce readmission rates, as do structured assessments of patients' discharge needs. Effective components identified within the context of multicomponent interventions were pre-discharge medication education/reconciliation; post-discharge telephone follow-up, efforts to ensure timely follow-up appointments, home visits and peer support as well as the bridging components of transition manager and timely communication by in-patient staff with an out-patient care or community service provider during the transition.

Context within existing literature

Our results are consistent with the findings of a previous review of pre-discharge interventions in psychiatry where the pooled risk of readmission ratio for discharge planning interventions using data from only four studies was 0.66 (95% CI 0.51–0.84), suggesting that pre-discharge interventions are potentially effective, although there are a limited number of studies. ^{Reference Steffen, Kosters, Becker and Puschner10} There is a larger body of literature evaluating the delivery of ongoing community-based care for patients who are at very high risk of readmission (for example, assertive community treatment or other forms of intensive case management). Although not transitional care per se, some of the intervention components are similar to those identified in our review. In recent systematic reviews and meta-analyses comparing intensive case management to standard care, increased numbers of hospital readmissions were observed among individuals receiving intensive case management (odds ratio (OR) = 1.84, 99% CI 1.33–2.57), ^{Reference Marshall, Gray, Lockwood and Green31} although length of admission appears to be significantly reduced (mean difference −0.86, CI −1.37 to −0.34). ^{Reference Dieterich, Irving, Park and Marshall32} This likely reflects improved retention in care and enhanced quality of community care (as shown by the reduced number of days spent in hospital). None of our studies found an increased rate of readmission in the intervention group, suggesting that the interventions were not functioning to identify patients requiring urgent readmissions due to illness instability. This is perhaps not surprising considering that despite relatively high rates of repeat admissions in some of the samples that we studied, most studies explicitly excluded individuals who were receiving ongoing psychiatric case management services. Interestingly, our results contrast with a similar review of interventions to reduce 30-day readmission in medical in-patient settings where no distinct intervention component or combination of components was identified as effective in reducing early readmission rates. ^{Reference Hansen, Young, Hinami, Leung and Williams11} It has been argued that medical readmission may actually represent improved quality of care, potentially reflecting the fact that better hospitals keep sicker patients alive and that these patients appropriately require more readmissions. ^{Reference Joynt and Jha33–Reference van Walraven, Bennett, Jennings, Austin and Forster35} These same arguments may not be applicable to psychiatric admissions, however. Mortality is not as common an outcome for psychiatric in-patients compared with medical populations. Furthermore, major modifiable risk factors such as difficulty with medication use and lack of follow-up care explain significantly more of the variance in psychiatric readmission rates than for medical and surgical readmissions where acuity of the admission and medical comorbidity are more important predictors of readmission. ^{Reference van Walraven, Dhalla, Bell, Etchells, Stiell and Zarnke36} It follows that if hospitals were to focus on improving the quality of care during the transition from acute psychiatric admission to ambulatory treatment, there would likely be potential to reduce psychiatric readmission rates. Although relatively few in number and of moderate quality and small sample size, the studies described in our review do suggest that this is the case.

Explanations for findings

Multiple factors may contribute to high early readmission rates. Acute psychiatric admission plays an important role in the mental healthcare system in the treatment of individuals with serious mental illness. ^{Reference Tulloch, Fearon and David37,Reference Wickizer and Lessler38} Yet, financial pressure to reduce in-patient length of stay and realignment of mental health resources to community settings in current psychiatric practice has shifted the role of in-patient admission to stabilisation rather than complete recovery. ^{Reference Tulloch, Fearon and David37,Reference Wickizer and Lessler38} Incomplete or partial symptomatic recovery places individuals at risk for relapse of symptoms and subsequent readmission. ^{1,Reference Zhang, Harvey and Andrew7} However, other risk factors for early readmission may be more amenable to intervention during the transition to out-patient treatment. These include medication management difficulties, impairment in self-care and absence of aftercare (i.e. referrals and timely follow-up with supports and services). ^{1,Reference Zhang, Harvey and Andrew7,Reference Lien39–Reference Walker, Minor-Schork, Bloch and Esinhart43} Most successful interventions included in our review targeted these known major modifiable risk factors for readmission such as difficulty with medication use, knowledge about disease management and coordination of care after discharge. ^{1,Reference Zhang, Harvey and Andrew7,Reference Lien39–Reference Walker, Minor-Schork, Bloch and Esinhart43} Interventions identified in our review that did not generate statistically significant results also targeted these same modifiable risk factors. Several of these studies may have been underpowered to detect effects because they were pilot studies or because readmission rates were not the primary outcome. However, it is notable that in two cases there were conflicting findings with regard to very similar interventions. Only one of two studies evaluating the ‘transitional discharge model’ and one of two studies evaluating the ‘critical time intervention’ had positive findings. Although the studies with the positive findings also had substantially larger sample sizes, both interventions with positive findings also had a flexible duration of transitional care lasting up to 1 year where the termination of the intervention was determined by clinical decision or information regarding successful transfer of care to a community provider. In both of the studies, the median length of the intervention was between 6 and 7 months. The corresponding negative studies had time limits of 5 months for the transitional discharge model and 3 months for the ‘brief’ critical time intervention. These findings underscore the potential importance of flexibility in the length of transitional care programmes.

Study limitations

Despite the promising preliminary findings of our review, our results highlight that there is a great need for additional high-quality evidence to inform best practice. Given the large body of literature focused on risk factors for psychiatric readmission, ^{Reference Lien39–Reference Walker, Minor-Schork, Bloch and Esinhart43} it is surprising that only 15 studies evaluating transitional interventions to reduce readmission rates were identified. The majority of studies were conducted in the USA and therefore the results may not generalise to other health systems where funding and organisational characteristics may differ. Most studies had small sample sizes and were likely underpowered to detect clinically relevant effect sizes. Furthermore, although we did not directly assess the costs of interventions, only two of the included studies provided a cost-effectiveness analysis of the evaluated intervention. In addition, there were significant limitations in study quality. Few studies adequately assessed baseline differences between groups that might be important to readmission prognosis and missing data were often not adequately addressed, raising concern that either selection bias or differential loss to follow-up rates could be responsible for differences in readmission between groups. Few studies clearly demonstrated objective outcome reporting of readmissions and, even in those studies, it was unclear whether it was possible for outcomes to be captured if patients were readmitted to other institutions. Up to 20% of medical hospital readmissions are missed by failing to account for admission to other institutions and this may be an issue for mental health readmissions as well. ^{Reference Nasir, Lin, Bueno, Normand, Drye and Keenan44}

The clinical heterogeneity of the studies also limits our ability to generate practice and policy implications from the data. The wide variation in readmission rates between studies suggests that these interventions were applied in populations with different levels of risk. This limits our ability to comment on how effective some of these interventions might have been in populations with different baseline levels of risk for readmission. For example, some interventions may have been applied either in populations where the control group was very well resourced and there was no additional benefit from the interventions (i.e. populations with low comparison group readmission rates) or where participant illness acuity was too severe for time-limited, focused discharge interventions to be effective. To mitigate this limitation, we provided detailed information regarding the samples in which interventions were evaluated such that knowledge-users would more readily be able to determine in which populations effective interventions might be implemented (and in which populations the said interventions have not yet been evaluated). Another limitation of the included studies is that the majority tested multiple intervention components together, making it difficult to comment about the relative contribution of any of the individual components of the interventions. This leaves unanswered questions as to what specific components of transitional care interventions are necessary to effect results. However, given the promising nature of some of these multicomponent interventions (for example critical time intervention and transitional discharge model, among others), it is likely that certain combinations of components may be more effective than the sum of their parts. A study using critical time intervention in psychiatric populations published subsequent to the submission of this manuscript for publication adds further support to this concept. ^{Reference Tomita and Herman45}

Implications

We observed that transitional care interventions with pre-discharge, post-discharge and/or bridging components may reduce early psychiatric readmission – and that when they are effective, the magnitude of effect appears to be clinically meaningful. This supports the concept that decreasing early readmission rates in patients with psychiatric illness is an attainable goal. Many potentially effective intervention components described in this study, particularly patient psychoeducation, needs assessments and communication between providers, are standard components of best practice. Structured implementation of these intervention components in discharge planning might ensure that patients receive optimal quality of care, with resultant reduction in readmission risk among other positive outcomes. Other observed effective interventions do not necessarily require significant resources either, and it is feasible to evaluate them using readmission rate as one of several metrics. With such a high rate of psychiatric readmission and a paucity of high-quality evidence to inform best practice, additional evaluation of interventions to reduce early readmission in psychiatry is clearly needed. Policy-makers have shifted to the use of early hospital readmission as an indicator of the quality of care. Clinicians and researchers need to further develop the evidence base for clinical interventions that support more seamless transitions from in-patient to out-patient care. Future research can overcome the limitations of previous work to provide direction about specific components that are necessary and/or sufficient for preventing early psychiatric readmission. The taxonomy of classifying interventions into pre-discharge, post-discharge and bridging components may be a useful framework for better understanding which early readmissions are preventable in mental health populations and how to use transitional care interventions most effectively.