Background: The optimal threshold to blood transfusion in patients after hip fracture surgery is unclear. This study leveraged regression discontinuity to quantify blood transfusion practices across haemoglobin thresholds of 7.0, 8.0, and 10.0 g/dL. Methods: This study used the enhanced claims-based Premier Inc. database and included inpatients, aged 50+, with an ICD-10 code for hip fracture and hip surgery, and at least one haemoglobin following surgery. For each patient-day following surgery, we determined the lowest haemoglobin level and whether blood transfusion was administered. Regression discontinuity (RD) models with local linear regression were used to calculate risk differences, mean differences, and 95% confidence intervals (CIs) for blood transfusion use. Results: There were 320,194 patient-days (among 93,111 patients) included in analyses. Transfusion occurred on 29,425 patient-days (9.2%). The median haemoglobin on patient-days with a transfusion was 7.0 (interquartile range IQR 6.7–7.6) g/dL and the median haemoglobin on patient-days without a transfusion was 9.3 (IQR 8.4–10.4) g/dL. There was a statistically significant increase in transfusion use crossing a haemoglobin threshold of 7.0 g/dL (19.0; 95% CI, 13.5–22.5 absolute percentage points) and a small, statistically significant increase in transfusion use crossing an 8.0 g/dL haemoglobin threshold. (3.2; 95% CI, 0.9–6.2 absolute percentage points). There was no discontinuity in transfusion use across a haemoglobin concentration threshold of 10.0 g/dL (-0.2; 95% CI -0.7–0.3 absolute percentage points). Conclusions: After hip fracture surgery, transfusion based on a haemoglobin threshold of 7.0 g/dL was more common than at guideline recommended 8.0 g/dL or at the liberal threshold of 10 g/dL.
Berube et al. (Wed,) studied this question.