Risk assessment tools for predicting transfusion in surgery: a systematic review and meta-analysis

United Kingdom blood shortages necessitate better prediction of surgical blood requirement. We sought to assess the predictive accuracy of tools designed to identify those patients requiring blood transfusion within the perioperative period.

We searched the Cochrane library, EMBASE, MEDLINE, ClinicalTrials.gov and WHO trials portal, 2000-July 2023. We included studies that developed and/or validated prediction tools for blood requirement during the early perioperative period (48 hours). Risk of bias was evaluated using the Prediction model Risk Of Bias Assessment Tool. We pooled area under receiver operating curve and calibration data via random effects meta-analysis. We evaluated certainty of evidence of any estimates using the Grading of Recommendations Assessment, Development and Evaluation framework. We used meta-regression to describe associations between included variables/tool characteristics with tool accuracy.

We included 50 papers, describing 67 unique prediction tools. Most tools were at high risk of bias, with limited external validation. Discrimination (area under receiver operating curve) of prognostic models ranged from 0.49 to 0.96. Only two surgery-specific tools, the McClusky Index (liver transplant surgery) and Papworth Bleeding Risk Score (cardiothoracic surgery), had sufficient data to enable pooling of discrimination measures. The McClusky Index's pooled area under receiver operating curve: 0.74 (95% CI 0.61 to 0.84) and Bleeding Risk Score's area under receiver operating curve: 0.68 (95% CI 0.49 to 0.82) were both rated 'very low' certainty by Grading of Recommendations Assessment, Development and Evaluation. Pooling calibration data was not possible for any prediction tools. Meta-regression suggested that fewer included variables, longer time from surgery and independent validation studies were all associated with lower accuracy.

There were insufficient studies to assess overall tool performance via meta-analysis in other surgical subgroups beyond cardiothoracic surgery and liver transplant. Our study population is also predominantly made up of elective surgeries which may make our results less generalisable to emergency settings.

Implementation and cost-effectiveness studies are needed to evaluate how promising tools could be applied to clinical practice and the economic impact such tools could have upon the service.

Despite the availability of multiple potential tools, available data suggest none are currently suitable for predicting blood transfusion in surgical practice. Our summary of the data comes with caveats around the quality of the included papers and the limited number of tools with more than one reported external validation.

This article presents independent research funded by the National Institute for Health and Care Research (NIHR) Health Technology Assessment programme as award number NIHR159933.