AI model predicts likelihood of unplanned hospitalizations during cancer radiation treatments

AI model predicts likelihood of unplanned hospitalizations during cancer radiation treatments

An artificial intelligence (AI) model developed by researchers can predict the likelihood of a patient experiencing an unplanned hospitalization during their radiation treatment for cancer. The machine learning model uses daily step counts as a proxy to monitor patients' health during cancer therapy, providing physicians with a real-time method to deliver personalized care. The results will be presented today at the annual meeting of the American Society for Radiation Oncology (ASTRO).

An estimated 10-20% of patients receiving outpatient radiation or chemoradiotherapy require acute care in the form of an emergency department (ED) visit or hospitalization during their cancer treatment. These unplanned hospitalizations can be a major challenge for people undergoing cancer treatment, causing treatment interruptions and stress that can impact clinical outcomes. Early detection and intervention in patients at higher risk of complications can prevent these events.

If you can predict a patient’s risk of an unplanned hospitalization, you can change the way you support them with their cancer treatment and reduce the likelihood that they will end up in the emergency room or hospital.”

Julian Hong, MD, senior author of the study

Julian Hong is an Assistant Professor of Radiation Oncology and Computational Health Sciences at the University of California, San Francisco (UCSF), where he also serves as Medical Director of Radiation Oncology Informatics.

The team of Dr. Hong has previously shown that a machine learning algorithm can use health data such as cancer history and treatment plan to identify patients at higher risk of ED visits during cancer treatment and that additional monitoring by their providers reduces acute care rates for these patients.

For the current study, he and Isabel Friesner, lead author and clinical data scientist at UCSF, worked with Nitin Ohri, MD, and colleagues at Montefiore Medical Center in New York to apply machine learning approaches to data from wearable consumer devices. Dr. Ohri and his team previously collected data from 214 patients in three prospective clinical trials (NCT02649569, NCT03102229, NCT03115398). In each of these studies, participants wore fitness trackers that monitored their activity over several weeks while they received chemoradiotherapy. Study participants had various types of primary tumors, most commonly head and neck (30%) or lung cancer (29%).

Step counts and other data from these patients' records were used to develop and test an elastic net regularized logistic regression model, a type of machine learning model that can analyze a large amount of complex information. The goal of their model was to predict the probability that a patient would be hospitalized in the next week based on the previous two weeks of data.

Researchers first built the model by examining how well various variables predicted hospitalization, using data from 70% of study participants (151 people). Potential predictors in the model included patient characteristics (e.g., age, ECOG performance status) and activity data measured before and during treatment. In addition to daily step counts, researchers calculated other metrics, such as: B. Relative changes in a person's weekly averages or the difference between the minimum and maximum number of steps per week.

The research team then validated the model on the remaining 30% of patients (63 people). The model incorporating step counts was strongly predictive of hospitalization the following week (AUC = 0.80, 95% confidence interval [CI] 0.60-0.90), and it significantly outperformed the model without step counts (AUC = 0.46, 95% CI 0.24-0.66, p<0.001).

"Step counts immediately before the prediction window were generally more informative than clinical variables. The dynamic nature of step counts, the fact that they change every day, appears to make them a particularly good indicator of a patient's health status," said Dr. Hong.

Key predictor variables in the model included step counts from each of the past two days, as well as the relative changes in maximum step count and step count range over the past two weeks.

The use of dynamic data distinguishes this model from models based on clinical data such as performance status and tumor histology. “One of the unique parts of this model is that it is designed to be a rolling forecast,” Ms. Friesner explained. “You can run the algorithm on any day and have an idea of ​​a patient’s risk level a week in advance, giving you time to provide the additional support they need.”

This additional support is key to reducing hospital stays, explained Dr. Hong, whether it's scheduling more frequent follow-up visits, changing the patient's treatment plan, or taking another personalized approach. "The core of what works is that this is an additional touchpoint for a doctor to see a patient. It gives the patient peace of mind knowing we are looking out for them."

"As more people begin to use wearables, the question arises as to whether the data they collect could be useful. Our study shows that there is value in our patients collecting their own health data in everyday life, and that we can use this data to then monitor and predict their health status," added Ms. Friesner.

The next steps for the investigators include more rigorous validation of the algorithm in the manner described by Dr. Ohri-led NRGF-001 trial (NCT04878952), which will randomize patients undergoing CRT for lung cancer to treatment with or without daily step count monitoring. Physicians of patients in the Step Count arm receive data from the model throughout the treatment process.

The researchers also plan other studies to examine additional metrics collected by wearable devices, such as: B. heart rate and its use in the clinic.

"Wearable devices and patient-generated health data are still relatively new phenomena, and we are still learning how they can be useful. What other information can we get from the many sensors in our lives? How can these metrics complement each other and work with other types of data, like electronic health records? Different data points might work better for different patients," Ms. Friesner said.

Following the widespread adoption of telemedicine and remote care in recent years, the need for remote monitoring via patient devices may also increase. Clinicians and policymakers should keep access to these devices in mind as they become increasingly popular, Dr. Hong.

"One of the challenges of working with real-world wearable data is the economic and racial disparities that impact who owns devices that can capture this type of data. I think it's important to develop tools that are useful for the clinic but also accessible to a broader range of patients."

Source:

American Society for Radiation Oncology

.