Postoperative complications are an important consideration for patients and the surgical team. Ranging from protracted pain, infection, and even end-organ damage, the overall complication rate may be anywhere from 10% to 60% of all operations performed worldwide [1]. This article will discuss methods of predicting complications, ranging from opioid intake to machine-learning technologies, that can guide medical teams in assessing their patients’ risks before surgery.
A patient’s preoperative opioid use can be a strong indicator of their likelihood of experiencing postoperative complications. Jain and colleagues conducted a study focused on major joint replacement and lumbar fusion patients [2]. They found that total knee arthroplasty (TKA) patients who had chronically used opioids at least three months before surgery were more likely to visit the emergency department (ED) within 90 days and be readmitted to the hospital after their operations [2]. Meanwhile, TKA, total hip arthroplasty (THA), and posterior lumbar fusion (PLF) patients with at least a six-month history of opioid use were more likely to experience wound infection, pain-related and all-cause ED visits, and revision surgery within a year [2]. Similar results were reported by Wilson et al., demonstrating how opioid addiction can imperil surgical success [3].
Beyond tracking patients’ opioid use, medical teams can turn to increasingly sophisticated technologies, such as linear risk calculators, to determine how likely postsurgical complications are to occur. One such technology is the American College of Surgeons National Surgical Quality Improvement Program (ACS-NSQIP) [4]. It is described as “a universal surgical risk calculator intended to improve the informed consent process and surgical decision-making” [4]. By integrating patient information from more than 585 hospitals, the calculator has created a sizable dataset in hopes of predicting individual patients’ risk of certain postoperative outcomes [4]. Although the calculator has enjoyed some success in predicting patients’ outcomes when analyzed retrospectively, a 2018 study concluded that the tool was nonpredictive unless judicious parameters are applied [4].
The inconsistency of linear models has encouraged researchers to look into machine learning (ML) technologies to improve risk assessment. They claim that ML software could lead to improved prediction because of its ability to recalibrate itself with each variable, better mimicking the interaction between risk factors in the human body [5]. Already, these technologies have produced promising results. A pilot usability study focused on MySurgeryRisk, developed by University of Florida researchers, found that it could calculate patients’ risk of suffering from 8 complications more accurately than physicians [7]. Another ML technology, POTTER, accurately predicted 18 postoperative complications, morbidity, and mortality [5]. As these technologies continue to improve, medical teams and patients will enter surgery more informed than ever before.
Lastly, a new set of predictive technologies uses biological markers to develop risk assessments [1]. Recently, Rumer and colleagues developed a high-content proteomic tool to determine the risk of surgical site complications following major abdominal surgery [1]. After collecting patients’ blood samples before and after surgery, this tool uses mass cytometry to analyze immune cell subsets, with the ultimate goal of quantifying the 2,388 dynamic changes that cells undergo throughout surgery [1]. Their goal is to quantify these changes and identify any correlations with adverse postoperative events to create a new form of risk assessment for postoperative complications [1]. In the future, biological markers may serve as a robust foundation for surgical predictions.
There are many diverse risk assessment technologies that medical teams can turn to when advising their patients. While no one tool is perfect, the great amount of innovation that has occurred over the last decade certainly suggests that surgery-related complications will become easier to predict in the coming future.
References
[1] K. K. Rumer et al., “Integrated Single-cell and Plasma Proteomic Modeling to Predict Surgical Site Complications: A Prospective Cohort Study,” Annals of Surgery, vol. 275, no. 3, p. 582-590, March 2022. [Online]. Available: https://doi.org/10.1097/SLA.0000000000005348.
[2] N. Jain et al., “Prediction of Complications, Readmission, and Revision Surgery Based on Duration of Preoperative Opioid Use,” The Journal of Bone and Joint Surgery, vol. 101, no. 5, p. 384-391, March 2019. [Online]. Available: https://doi.org/10.2106/JBJS.18.00502.
[3] J. M. Wilson et al., “The Impact of Preoperative Opioid Use Disorder on Complications and Costs following Primary Total Hip and Knee Arthroplasty,” Advances in Orthopedics, vol. 2019, no. 9319480, p. 1-7, December 2019. [Online]. Available: https://doi.org/10.1155/2019/9319480.
[4] P. S. Vosler et al., “Predicting complications of major head and neck oncological surgery: an evaluation of the ACS NSQIP surgical risk calculator,” Journal of Otolaryngology, vol. 47, no. 21, p. 1-10, March 2018. [Online]. Available: https://doi.org/10.1186/s40463-018-0269-8.
[5] D. Bertsimas et al., “Surgical Risk Is Not Linear: Derivation and Validation of a Novel, User-friendly, and Machine-learning-based Predictive OpTimal Trees in Emergency Surgery Risk (POTTER) Calculator,” Annals of Surgery, vol. 268, no. 4, p. 574-583, October 2018. [Online]. Available: https://doi.org/10.1097/SLA.0000000000002956.
[6] A. Bihorac et al., “MySurgeryRisk: Development and Validation of a Machine-Learning Risk Algorithm for Major Complications and Death after Surgery,” Annals of Surgery, vol. 269, no. 4, p. 652-662, April 2019. [Online]. Available: https://doi.org/10.1097/SLA.0000000000002706.
[7] M. Brennan et al., “Comparing clinical judgment with the MySurgeryRisk algorithm for preoperative risk assessment: A pilot usability study,” Surgery, vol. 165, no. 5, p. 1035-1045, May 2019. [Online]. Available: https://doi.org/10.1016/j.surg.2019.01.002.
Recent Comments