Determinants for successful de-implementation of low-value services: a systematic review

Background: Stopping proven ineffective medical practices is important for improving the quality of healthcare. These low-value services (LVS) have no added value for patients or have shown to be only effective for a limited group. Deimplementation of LVS is likely to face different challenges than implementation of new practices. Even with strong evidence against the use of an intervention or test, action is often required to restrict its use. Objectives: To investigate determinants for successful de-implementation strategies and to identify gaps in knowledge and areas for future research. Methods: MEDLINE, Embase, Cochrane, and Rx for Change databases were searched on 1 November 2015. Additional studies were found through checking references and healthcare websites. Studies of interest focused on the reduction or elimination of a LVS for clinical - rather than financial - reasons. Information on characteristics and effectiveness of de-implementation strategies, study design, and perceived/measured barriers and facilitators to these strategies were extracted. Results: About 120 studies were included: 65% on interventions (of which, drugs 80% vs non-drugs 20%); 25% on diagnostics; and 10% others (e.g. follow-up care or screening). Only 10% were randomized trials, most were before-after studies followed by interrupted time series. Most studies focused on adequate care or restricted use rather than total stoppage. About 70% claimed 'success' e.g. decreased use of LVS; 20% presented patient-health related outcomes. Only 1% considered the sustainability of the de-implementation. Most de-implementation strategies were multi-faced, with successful elements being patient education and empowerment, physician education and feedback, and organizational interventions. Serious barriers influencing the effectiveness of de-implementation were negative attitude towards change and continuing reimbursement. Strong facilitators were involvement of a medical leader and interaction with patients. Conclusions: We provide suggestions for quality improvement of future studies on de-implementation and give guidance for best practices to decrease LVS in health care