Systematic review of interventions to improve safety and quality of anticoagulant prescribing for therapeutic indications for hospital inpatients.

Purpose: Anticoagulation-associated adverse drug events are common in hospitalised patients and result in morbidity, mortality, increased length of hospital stay and higher costs of care. Many are preventable. We reviewed the literature to identify and assess interventions intended to improve safety or quality anticoagulant prescribing.

Methods: A systematic search of EMBASE, MEDLINE, the Cochrane Library, Pretty Darn Quick-Evidence and Health Systems Evidence was undertaken to identify controlled studies assessing system-level interventions to improve prescribing of oral or parenteral therapeutic anticoagulation for any indication in hospitalised adults. Data were extracted for safety and quality outcomes, with studies grouped by intervention type for meta-analysis and narrative review.

Results: Of 10,640 records screened, 19 trials evaluating 12,742 participants were included for analysis. No study specifically evaluated prescribing of low molecular weight heparins (LMWHs) or direct acting oral anticoagulants (DOACs). Our findings suggest that physician-led anticoagulation consultation services may reduce bleeding rates in high-risk patients. On meta-analysis, decision supported warfarin dosing resulted in higher proportion of time with international normalised ratio in therapeutic range (p = 0.0007). Studies of other clinical decision support systems and heparin monitoring systems did not demonstrate improved safety, and quality findings were inconsistent. Systematic education and feedback programs were not efficacious.

Conclusions: There is currently insufficient high-quality evidence to recommend any reviewed intervention, though several warrant closer evaluation. Adequately powered controlled trials assessing safety outcomes and evidence-based quality markers in high-risk patient groups and studies of interventions to improve safety of LMWH and DOAC prescribing are needed.