Early Postoperative Vital Signs Predict Subsequent 90-Day Mortality After Pancreaticoduodenectomy.

Publication/Presentation Date

8-1-2023

Abstract

BACKGROUND: While complication rates after pancreaticoduodenectomy (PD) have improved in recent decades, surgical-related death remains a possibility. Postoperative vital signs offer an untapped opportunity to identify predictors of 90-day mortality.

METHODS: We performed a retrospective chart review interrogating postoperative day (POD 0-7) vital sign measurements from patients undergoing a PD at Thomas Jefferson University Hospital, Philadelphia, PA (2009-2014). Five specific vital signs were examined as predictors of mortality: temperature, heart rate (HR), systolic blood pressure (SBP), diastolic blood pressure (DBP), and mean arterial pressure. Statistical analyses and logic algorithms were employed to rank vital sign parameters, with cut-points, to identify those associated with the highest risk of mortality and the most clinical relevance.

RESULTS: In our cohort, 11/750 patients (1.5%) died within 30 days of surgery, and 21/750 patients (2.8%) died within 90 days of surgery. Vital sign perturbations associated with the highest risk of mortality included mean SBP <  95 mmHg on POD 7 (odds ratio 51.46) and the mean temperature <  96.9℉ on POD 3 (odds ratio 22.63) with specificities exceeding 99%. The most clinically relevant predictor (i.e., a higher sensitivity) was DBP <  60.5 mmHg on POD 7 (odds ratio 12.45, sensitivity of 75%). These predictors remained statistically significant in a multivariable model.

CONCLUSIONS: Vital signs can be more effectively utilized to predict 90-day mortality after pancreaticoduodenectomy. Values beyond an informative threshold can potentially identify patients for more intensive monitoring with a goal of rescuing patients and preventing death.

Volume

27

Issue

8

First Page

1660

Last Page

1667

ISSN

1873-4626

Disciplines

Business Administration, Management, and Operations | Health and Medical Administration | Management Sciences and Quantitative Methods

PubMedID

37106207

Department(s)

Administration and Leadership

Document Type

Article

Share

COinS