Identifying keys to success in reducing readmissions using the ideal transitions in care framework

Unsafe transitions of care from the hospital to the community are common and are frequently associated with post-discharge adverse events, including hospital readmission [1]. While not all hospital readmissions are preventable, the volume of patients readmitted (nearly one in five Medicare patients by 30 days post-discharge) and costs associated with readmissions ($26-44 billion per year spent by Medicare) make remediating unsafe transitions essential [2].

However, best practices to cost-effectively reduce readmissions are not well-elucidated [3]. A previous systematic review of interventions to reduce hospital readmissions did not identify an intervention or bundle of interventions that reliably reduced readmissions, despite well-conducted individual trials that have reduced readmission rates [4]. In that review, the authors constructed a simple temporal taxonomy to categorize interventions into pre-discharge, post-discharge, and “bridging” interventions. We hypothesize that a taxonomy focused on individual activities that lead to safer transitions of care may provide new insights into why some interventions are successful and many others are not.

The Ideal Transition of Care (ITC) framework (Additional file 1: Figure S1) proposes 10 domains to consider in ensuring safe transitions of care, based upon expert guidelines, critical analysis of the literature, and clinical experience [5]. The ITC has been proposed as a method for analyzing failures and guiding new interventions in transitions of care, as well as creating process measures to monitor the quality of care transitions.

We had four related research aims in this study: 1) to establish how frequently each of the ten ITC domains have been utilized in prior interventions; 2) to discover how frequently prior interventions met with success in reducing readmissions; 3) to examine the relationship between each of the ten ITC domains individually with success in reducing readmissions; and 4) to evaluate the relationship between the total number of ITC domains included in an intervention and successful readmission reduction. Thus, we conducted a comprehensive review of the literature to identify prior interventions intended to reduce hospital readmission, and categorized them according to the ten ITC domains for our secondary analysis.

After application of inclusion and exclusion criteria, 66 articles were included in the final analysis (Additional file 2: Table S1). Median study size was 283 patients (interquartile range, 270). Thirty-five studies (53%) evaluated the primary endpoint at 30 or fewer days following hospital discharge; results of our statistical analyses were similar when comparing studies with primary endpoints of 30 or fewer days with those having endpoints greater than 30 days and thus all studies were analyzed as a single group. Interventions directed at all discharging patients accounted for 52% of included studies, while 41% were studies of heart failure patients exclusively. Overall, 42% of studies demonstrated a statistically significant reduction in readmissions between the intervention and control groups; 61% of these were studies of specific disease processes rather than all discharging patients.Prior interventions addressed 3.5 domains on average; only 23% addressed five or more (Figure 2). Monitoring and Managing Symptoms after Discharge (included as part of 74% of interventions), Educating Patients to Promote Self-Management (64%), and Coordinating Care among Team Members (55%) were the domains most frequently included as a part of the intervention. Conversely, Advance Care Planning was not included as a part of an intervention in any study, while the two domains concerning information transfer to receiving clinicians and the Medication Safety domain were rarely included (<20%, Figure 3).

In bivariate analysis, the Monitoring and Managing Symptoms after Discharge domain was significantly associated with success in reducing readmissions (OR 8.5 (95% CI 1.8 - 41.1), FDR-corrected p-value = 0.03). Two other domains, Enlisting Help of Social and Community Supports (OR 4.0 (1.3-12.6), FDR-corrected p = 0.07) and Educating Patients to Support Self-Management (OR 3.3 (1.1-10.0), FDR-corrected p = 0.09) showed relatively strong associations with reductions in readmissions (Table 2).

The number of domains included in an intervention was significantly associated with success in reducing readmissions, even after adjusting for study quality, duration, and size (OR per domain included 1.5, 95% CI 1.1-2.0).

The most important finding of our study for physicians charged with reducing readmissions is that increasing the number of targeted domains within the ITC was associated with significantly increased success in reducing readmissions. In addition, not all domains were associated with equal effect in reducing readmissions. Among the individual domains, systems for Monitoring and Managing Symptoms after Discharge were most associated with successful reduction in readmissions, while Enlisting Help of Social and Community Supports, and Educating Patients to Promote Self-Management may also be efficacious.

Categorizing prior studies in the ITC framework offered important insights into the “state of the science” of readmission reduction. We found most interventions targeting a reduction in hospital readmission published in the literature have not been successful. The 41% overall success rate of published interventions most likely reflects the fact that patients discharged from acute care settings exhibit multiple risk factors for readmission spanning the 10 domains of the Ideal Transition of Care. Since most interventions published targeted a few, similar domains, a correspondingly low success rate of an individual intervention may not be surprising, though our study design limits causal inference. While the ten domains of the ITC framework center on modifiable risk factors for admission, we did not assess how “preventable” readmissions were in included studies.

To the individual clinician, implementing these findings may seem daunting. However, effective multi-domain models exist [8, 14, 18, 39] and nearly all provide options for substantial training. A recurring characteristic of these models is provision of a single health care provider responsive to multiple patient needs, thereby targeting multiple domains of the Ideal Transition of Care. Jack et al. used a “discharge advocate” to provide intensive patient-centered education, discharge planning and post-discharge reinforcement [14]. Likewise, Coleman et al. implemented a “transition coach” to assist patients across health settings and encouraged patients to be active in their own care, while providing them the necessary tools to do so [8]. Similarly, Naylor et al. used an advance practice nurse to manage an individualized patient plan tailored to identified needs, with a focus on patient education and longitudinal collaboration of key providers from hospital admission through two weeks post-discharge [18]. Peikes et al. found success in local care coordination, effectively targeting multiple risk factors for readmission for enrolled patients, and changed their intervention from one that increased readmissions and cost to one that reduced both [39].

However, these models require substantial investment of resources. Clinicians and health care systems with limited resources (particularly those already penalized financially for elevated readmission rates) may struggle to implement these interventions. A key finding from our study is that one option for limiting costs- limiting the number of domains targeted- may not lead to success. A method to risk-stratify patients at the time of discharge, then selectively apply interventions based on this analysis, may maximize efficacy and minimize cost. [3] However, currently available risk prediction models lack accuracy and capture only a global assessment of risk that is difficult to apply to individual patients across highly variable delivery systems. [73] Frameworks similar to the ITC framework may hold promise as tools to better assess individual, modifiable risk factors for readmission of recently hospitalized patients, and design interventions to address these risk factors on a case-by-case basis in order to provide tailored, risk-stratified care.

Three domains within the ITC were most associated with success in reducing readmissions. Monitoring and Managing Symptoms after Discharge is plausible as an individual domain most strongly associated with success in reducing readmissions given post-discharge adverse events are common and frequently present with new symptoms. [1] Thus, close clinical monitoring of a recently discharged patient for active symptoms helps ensure effective post-hospital care. Home visits by health care professionals (rather than telemonitoring) appear to be a common theme in several successful interventions [8, 18, 40, 42].

Despite inclusion in fewer than one in four existing interventions, active integration of community and social support networks addressing needs of patients was also associated with success in reducing readmissions. Indeed, this is the intent of Medicare’s $500 million Community-Based Care Transitions Project, part of the Partnership for Patients instituted by the Affordable Care Act. A discharge planning protocol conducted by a social worker to assess living environment and social supports, then engaging community and social service referrals as needed, was the cornerstone of a successful intervention by Evans et al. [11] Several other successful interventions also addressed community supports as a component of a larger intervention [39, 40, 42], indicating the need to address this specific element of a patient’s care transition.

Patient education to promote active involvement in their own care has been a much more commonly targeted domain, though few interventions have assessed the efficacy of this education. Coleman’s transition coach taught patients how to self-manage and to interact with the health care system; benefits were found months after the intervention had concluded [8]. Providing patient education in isolation from other elements and without active patient involvement is likely insufficient to reduce readmissions [13]. Rather, successful interventions focus on engaging the patient to manage their chronic illnesses in an ongoing manner.

These results should be interpreted in the context of the literature reviewed. None of the interventions we reviewed were designed with the ITC framework in mind. As such, our evaluation of whether a domain was present or not represents our best assessment based on our review of these reports and understanding of the ITC. However, implementation of the intervention is infrequently described, and it is possible that the described and actual interventions varied significantly. No study included all ten domains, making our conclusions about the relative influence of inclusion of one domain versus another limited.

Publication bias may play a role in our findings, though we think the strong negative publication record in this regard limits its influence. Published reports may have other biases (academic settings, urban locations) that affect our findings, though analysis of these biases is beyond the scope of this analysis. Our measures of quality were limited to study size, general design, and duration. Other important methodologic constructs such as appropriateness of sampling, data collection, analysis plan, and generalizability were not captured and infrequently reported. While we note inclusion of these elements did not affect findings in prior systematic reviews [4], it is possible their inclusion could have affected our findings.

We excluded pediatric, obstetric, psychiatric, and surgical populations as their reasons for readmission may differ from medical patients. Domains in the ITC may not be independent of one another, but a formal principle components or factor analysis was beyond the scope of our review. We did not use the techniques of a meta-analysis, as the wide variability of the existing literature prevents this level of analysis. We also did not use the reporting standards of a formal systematic review, though we did search systematically for studies that met criteria for analysis. Our approach was necessarily more narrative and thus should be considered hypothesis-generating and requiring further prospective study.

Improving transitions of care from the hospital to the community requires multifaceted interventions targeting multidimensional risk factors present in patients discharged from the hospital. Until readmission risk factors- individually and collectively- are better understood and assessed, designing interventions to address these multifactorial risk factors using a framework like the ITC may be effective. In addition, incorporating systems actively involving patients in promoting self-management in their care, developing care processes to address active symptom development in the post-discharge period, and providing social and community support for this management merit special inclusion in any intervention. Future work evaluating the role of the Ideal Transition of Care framework in evaluating risk and designing interventions for individual patients may show benefit in providing cost-effective, safe transitional care.