Development and multicentre validation of the FLEX score: personalised preoperative surgical risk prediction using attention-based ICD-10 and Current Procedural Terminology set embeddings

Ran Liu; Tom A D Stone; Praachi Raje; Rory V Mather; Laura A Santa Cruz Mercado; Kishore Bharadwaj; Jasmine Johnson; Masaya Higuchi; Ryan D Nipp; Hiroko Kunitake; Patrick L Purdon

doi:10.1016/j.bja.2023.11.039

Development and multicentre validation of the FLEX score: personalised preoperative surgical risk prediction using attention-based ICD-10 and Current Procedural Terminology set embeddings

Br J Anaesth. 2024 Mar;132(3):607-615. doi: 10.1016/j.bja.2023.11.039. Epub 2024 Jan 6.

Authors

Affiliations

¹ Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Boston, MA, USA; Harvard Medical School, Boston, MA, USA.
² Harvard Medical School, Boston, MA, USA; Department of Surgery, Massachusetts General Hospital, Boston, MA, USA.
³ Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Boston, MA, USA; Harvard Medical School, Boston, MA, USA; Harvard-MIT Program in Health Sciences and Technology, Cambridge, MA, USA.
⁴ Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Boston, MA, USA; Harvard Medical School, Boston, MA, USA; Department of Anesthesia, Critical Care and Pain Medicine, Beth Israel Deaconess Hospital, Boston, MA, USA.
⁵ Harvard Medical School, Boston, MA, USA; Department of Medicine, Division of Palliative Care and Geriatric Medicine, Massachusetts General Hospital, Boston, MA, USA.
⁶ Section of Hematology/Oncology, Department of Internal Medicine, University of Oklahoma Health Sciences Center, Stephenson Cancer Center, Oklahoma City, OK, USA.
⁷ Department of Anesthesia, Critical Care and Pain Medicine, Massachusetts General Hospital, Boston, MA, USA; Harvard Medical School, Boston, MA, USA. Electronic address: ppurdon@stanford.edu.

Abstract

Background: Preoperative knowledge of surgical risks can improve perioperative care and patient outcomes. However, assessments requiring clinician examination of patients or manual chart review can be too burdensome for routine use.

Methods: We conducted a multicentre retrospective study of 243 479 adult noncardiac surgical patients at four hospitals within the Mass General Brigham (MGB) system in the USA. We developed a machine learning method using routinely collected coding and patient characteristics data from the electronic health record which predicts 30-day mortality, 30-day readmission, discharge to long-term care, and hospital length of stay.

Results: Our method, the Flexible Surgical Set Embedding (FLEX) score, achieved state-of-the-art performance to identify comorbidities that significantly contribute to the risk of each adverse outcome. The contributions of comorbidities are weighted based on patient-specific context, yielding personalised risk predictions. Understanding the significant drivers of risk of adverse outcomes for each patient can inform clinicians of potential targets for intervention.

Conclusions: FLEX utilises information from a wider range of medical diagnostic and procedural codes than previously possible and can adapt to different coding practices to accurately predict adverse postoperative outcomes.

Keywords: adverse surgical outcomes; billing codes; machine learning; multicentre study; postoperative outcome; risk prediction; surgical comorbidities.

Publication types

Multicenter Study

MeSH terms

Adult
Current Procedural Terminology*
Humans
International Classification of Diseases*
Patient Readmission
Perioperative Care
Retrospective Studies

Grants and funding

F32 GM148114/GM/NIGMS NIH HHS/United States