Predicting postoperative day 1 hematocrit levels after hysterectomy for malignant indication: Validating a previously published model

Abstract

Objective To determine whether the Swenson model of postoperative day 1 (POD1) hematocrit after benign hysterectomy is applicable to gynecologic oncology hysterectomies. Methods Data were retrospectively collected from cases of hysterectomy with malignant pathology in Hartford, USA, from 2014 to 2016. Predicted POD1 hematocrit was compared with actual hematocrit. ROC curve analysis was used to determine the optimal cut‐off point for predicting hematocrit levels of 30% or less. Results Among 107 women, mean age was 62.9 years and body mass index was 34.0. Most underwent robotic (44.9%) or abdominal (43.9%) hysterectomy. The published equation correctly predicted hematocrit to within ±5% for 83.2% of women, which was less accurate than observed in the original validation set. The equation was more likely to underestimate lower hematocrit levels, adding safety to its use. By ROC curve analysis, the best cut‐off point for predicting actual hematocrit above 30% was predicted hematocrit 32.3% (100% specificity). Conclusion The Swenson equation predicted POD1 hematocrit less accurately in the current dataset. As a screening tool for hematocrit below 30%, however, ordering postoperative hematocrit is probably unnecessary if the predicted value is 32.3% or higher. This equation should be used as a screening tool to reduce unnecessary laboratory tests.