Using machine learning techniques to develop forecasting algorithms for postoperative complications: protocol for a retrospective study

Abstract

IntroductionMortality and morbidity following surgery are pressing public health concerns in the USA. Traditional prediction models for postoperative adverse outcomes demonstrate good discrimination at the population level, but the ability to forecast an individual patient’s trajectory in real time remains poor. We propose to apply machine learning techniques to perioperative time-series data to develop algorithms for predicting adverse perioperative outcomes.Methods and analysisThis study will include all adult patients who had surgery at our tertiary care hospital over a 4-year period. Patient history, laboratory values, minute-by-minute intraoperative vital signs and medications administered will be extracted from the electronic medical record. Outcomes will include in-hospital mortality, postoperative acute kidney injury and postoperative respiratory failure. Forecasting algorithms for each of these outcomes will be constructed using density-based logistic regression after employing a Nadaraya-Watson kernel density estimator. Time-series variables will be analysed using first and second-order feature extraction, shapelet methods and convolutional neural networks. The algorithms will be validated through measurement of precision and recall.Ethics and disseminationThis study has been approved by the Human Research Protection Office at Washington University in St Louis. The successful development of these forecasting algorithms will allow perioperative healthcare clinicians to predict more accurately an individual patient’s risk for specific adverse perioperative outcomes in real time. Knowledge of a patient’s dynamic risk profile may allow clinicians to make targeted changes in the care plan that will alter the patient’s outcome trajectory. This hypothesis will be tested in a future randomised controlled trial.

Predicting intra‐operative and postoperative consequential events using machine‐learning techniques in patients undergoing robot‐assisted partial nephrectomy: a Vattikuti Collective Quality Initiative database study

“…Three models were built for each outcome, and the best model was selected based on the performance on an unseen test dataset. The variables for these models were selected based on the availability of data points in the VCQI dataset, taking a cue from the published research on the subject [12,32‐34]. We used AUC‐ROC and PR‐AUC to assess model performance.…”

Study protocol for the Anesthesiology Control Tower—Feedback Alerts to Supplement Treatments (ACTFAST-3) trial: a pilot randomized controlled trial in intraoperative telemedicine

“…Data collection for this study will utilize multiple sources to extract outcome measures 64 . All alert data generated by the AlertWatch Control Tower platform will be automatically logged to a secure database, including all responses by the providers in the ACT to individual alerts ( Figure 3).…”

“…Because the impact of a clinical intervention is dependent on the success of the process through which it is implemented 55 , we will also evaluate implementation outcomes that are relevant to the use of the ACT in the perioperative setting 56, 57 . In the second component of our approach, we will employ large-scale data analytics, integrating perioperative information in order to create forecasting algorithms for negative patient trajectories 58 . In the current manuscript, we describe the third element of our investigation: a pilot randomized controlled trial that aims to demonstrate the superiority of the ACT in improving adherence to best care practices when compared to enhanced usual care.…”

Perioperative clinical pharmacy practice: Responsibilities and scope within the surgical care continuum

“…This model allows the PCPh to be actively engaged in the care of multiple surgical patients in varying locations in real time. A clinical pharmacist can leverage the use of machine learning‐based forecasting algorithms and alerts to guide interventions such as timely antimicrobial infection prophylaxis administration and documentation, perioperative glycemic control, and neuromuscular blockade monitoring and reversal . The clinical pharmacist can have a direct impact on the quality of perioperative care provided, potentially resulting in improved clinical and safety outcomes.…”