New Machine Learning Method Better Predicts Spine Surgery Outcomes

Surgical recovery is impacted by the patient’s preoperative physical and mental health. Additionally, some individuals may experience heightened anxiety or physiological issues that exacerbate pain and impede recovery. If doctors can identify potential challenges a patient may face, they can better customize treatment plans. Researchers have been utilizing mobile health data from Fitbit devices to monitor and measure recovery, comparing activity levels over time. Now, these researchers, using Fitbit data to predict surgical outcomes, have a new method to more accurately assess how patients may recover from spine surgery.

Researchers at Washington University in St. Louis (St. Louis, MO, USA) employed machine-learning techniques to develop a method that can more precisely forecast recovery from lumbar spine surgery. Their earlier research demonstrated that combining patient-reported data with objective wearable measurements enhances predictions of early recovery compared to traditional patient assessments. They showed that Fitbit data could be correlated with various surveys evaluating a person’s social and emotional state. This data was collected through “ecological momentary assessments” (EMAs), using smartphones to prompt patients frequently throughout the day to assess mood, pain levels, and behavior.

In the most recent study, the researchers combined all these factors to develop a new machine-learning technique called “Multi-Modal Multi-Task Learning” to effectively integrate different types of data to predict multiple recovery outcomes. This approach allows the AI to understand the relationships among different outcomes while recognizing their distinct differences from the multimodal data. The method utilizes shared information on interrelated tasks of predicting different outcomes and leverages this shared information to improve the accuracy of predictions. The final result is a predicted change in each patient’s post-operative pain interference and physical function score. The study is ongoing, with researchers continuing to refine their models to perform more detailed assessments, predict outcomes, and, most importantly, identify modifiable factors to enhance long-term outcomes.

“We combine wearables, EMA and clinical records to capture a broad range of information about the patients, from physical activities to subjective reports of pain and mental health, and to clinical characteristics,” said WUSTL Professor Chenyang Lu.

“By predicting the outcomes before the surgery, we can help establish some expectations and help with early interventions and identify high risk factors,” added Ziqi Xu, a PhD student in Lu’s lab.

Related Links:
Washington University in St. Louis