BACKGROUND: Admission hypocalcemia predicts both massive transfusion and mortality in severely injured patients. However, the effect of calcium derangements during resuscitation remains unexplored. We hypothesize that any hypocalcemia or hypercalcemia (either primary or from overcorrection) in the first 24 hours after severe injury is associated with increased mortality. METHODS: All patients at our institution with massive transfusion protocol activation from January 2013 through December 2014 were identified. Patients transferred from another hospital, those not transfused, those with no ionized calcium (Ca²⁺) measured, and those who expired in the trauma bay were excluded. Hypocalcemia and hypercalcemia were defined as any level outside the normal range of Ca²⁺ at our institution (1–1.25 mmol/L). Receiver operator curve analysis was also used to further examine significant thresholds for both hypocalcemia and hypercalcemia. Hospital mortality was compared between groups. Secondary outcomes included advanced cardiovascular life support, damage control surgery, ventilator days, and intensive care unit days. RESULTS: The massive transfusion protocol was activated for 77 patients of whom 36 were excluded leaving 41 for analysis. Hypocalcemia occurred in 35 (85%) patients and hypercalcemia occurred in 9 (22%). Mortality was no different in hypocalcemia versus no hypocalcemia (29% vs 0%; P = .13) but was greater in hypercalcemia versus no hypercalcemia (78% vs 9%; P < .01). Receiver operator curve analysis identified inflection points in mortality outside a Ca²⁺ range of 0.84 to 1.30 mmol/L. Using these extreme values, 15 (37%) had hypocalcemia with a 60% mortality (vs 4%; P < .01) and 9 (22%) had hypercalcemia with a 78% mortality (vs 9%; P < .01). Patients with extreme hypocalcemia and hypercalcemia also received more red blood cells, plasma, platelets, and calcium repletion. CONCLUSIONS: Hypocalcemia and hypercalcemia occur commonly during the initial resuscitation of severely injured patients. Mild hypocalcemia may be tolerable, but more extreme hypocalcemia and any hypercalcemia should be avoided. Further assessment to define best practice for calcium management during resuscitation is warranted.

Each year, 150 000 patients undergo high-risk,^1-3 open cardiac valve or proximal aortic surgery in the US.⁴ Transesophageal echocardiography (TEE) is an ultrasonography-based, cardiac imaging tool used in cardiac surgical procedures to facilitate informed surgical decision making^5-7 and manage intraoperative complications.^5-9 However, the current American Heart Association (AHA) and American College of Cardiology (ACC) guidelines do not specifically recommend for or against the use of intraoperative TEE in the majority of cardiac surgical procedures^10-12 because prior to 2020, observational studies have not directly associated intraoperative TEE with improved clinical outcomes.^5-9 Recently, evidence has begun to accumulate on improved outcomes with TEE use after cardiac valve and coronary artery bypass graft (CABG) surgical procedures.^13-15 But there is no research directly comparing outcomes after proximal aortic surgery with TEE vs without TEE, and the existing study on improved outcomes with TEE after cardiac valve repair or replacement surgery used administrative claims data.¹⁵

Objective: This study aimed to develop and validate a claims-based, machine learning algorithm to predict clinical outcomes across both medical and surgical patient populations. Methods: This retrospective, observational cohort study, used a random 5% sample of 770,777 fee-for-service Medicare beneficiaries with an inpatient hospitalization between 2009–2011. The machine learning algorithms tested included: support vector machine, random forest, multilayer perceptron, extreme gradient boosted tree, and logistic regression. The extreme gradient boosted tree algorithm outperformed the alternatives and was the machine learning method used for the final risk model. Primary outcome was 30-day mortality. Secondary outcomes were: rehospitalization, and any of 23 adverse clinical events occurring within 30 days of the index admission date. Results: The machine learning algorithm performance was evaluated by both the area under the receiver operating curve (AUROC) and Brier Score. The risk model demonstrated high performance for prediction of: 30-day mortality (AUROC = 0.88; Brier Score = 0.06), and 17 of the 23 adverse events (AUROC range: 0.80–0.86; Brier Score range: 0.01–0.05). The risk model demonstrated moderate performance for prediction of: rehospitalization within 30 days (AUROC = 0.73; Brier Score: = 0.07) and six of the 23 adverse events (AUROC range: 0.74–0.79; Brier Score range: 0.01–0.02). The machine learning risk model performed comparably on a second, independent validation dataset, confirming that the risk model was not overfit. Conclusions and relevance: We have developed and validated a robust, claims-based, machine learning risk model that is applicable to both medical and surgical patient populations and demonstrates comparable predictive accuracy to existing risk models.