The search for alternative care structures that improve quality and lower costs is intensifying. Alternative care organizations (ACOs) offer promise but must demonstrate their cost–effectiveness for widespread adoption to occur. Readmissions to the hospital are a prime target for ACOs, as they are common, costly, and will increasingly be penalized by Medicare. We undertook a review of risk factors for readmission and of interventions to reduce this risk to offer insights for ACOs interested in reducing readmissions and improving cost–effectiveness.


Electronic databases were systematically searched with keyword queries relevant to the topic. Relevant abstracts were reviewed in full. Inclusion criteria for papers regarding risk factors for readmission were publication in English between 1976 and 2010, size of at least 100 subjects, and hospital readmission as a primary outcome; they were excluded if there was no adjustment for baseline characteristics. Inclusion criteria for studies describing interventions to decrease hospital readmissions were prospective design and publication in English prior to 2011. Papers were excluded, if they had fewer than 40 patients, did not have a control group or had historical controls, or were primarily pharmaceutical treatment studies.


Forty–six articles identifying risk factors for readmission and 48 articles describing interventions to decrease readmissions were identified. System–level factors, such as the quality of inpatient care, number of inpatient beds available, quality of communication with outpatient providers, and quality of primary care had small effects on readmission risk. Patient–level factors were much more predictive, especially the amount of prior health care utilization and degree of underlying comorbidity. Risk models had poor accuracy for predicting likelihood of readmission. Less than half of published interventions achieved success in reducing readmissions and less than one quarter demonstrated cost savings. The three most successful trials all had a similar framework: intensive inpatient counseling, patient and caregiver education using teach–back methods, and medication reconciliation prior to discharge followed by short–term, frequent, in–home follow–up by nurses.


Existing published models largely fail to accurately predict readmission. Predictive models could be improved by inclusion of groups thus far excluded from this literature: patients with cognitive impairment, psychiatric comorbidity, limited social supports, poor literacy, and terminal illness. The lessons learned in the Table may inform future efforts.

Table 1Lessons Learned from Readmissions Studies