An artificial intelligence (AI) model developed by the researchers can predict a patient’s likelihood of being admitted to an unplanned hospital during cancer radiotherapy. The machine learning model uses daily step counts as a proxy for monitoring patients’ health while they are undergoing cancer treatment, providing clinicians with a real-time way to provide personalized care. The results will be presented today at the annual meeting of the American Society of Radiation Oncology (ASTRO).
An estimated 10-20% of patients who receive outpatient radiotherapy or chemotherapy will 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 may affect clinical outcomes. Early identification and intervention of patients at risk of developing complications can prevent these events.
“If you can anticipate a patient’s risk of unplanned hospitalization, you can change how you support them with cancer treatments and reduce their likelihood of being admitted to the emergency department or hospital,” said Julian Hong, MD, senior author of the study. and 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.
Dr. Hong’s team previously showed that using a machine learning algorithm health data Such as cancer history and treatment plan can identify patients at higher risk of ED visits during Cancer treatmentand that additional monitoring from their providers reduced acute care rates for these patients.
For the current study, he and Isabel Frisner, lead author and clinical data scientist at UCLA, collaborated with Nitin Ohri, MD, and colleagues at Montefiore Medical Center in New York to apply machine learning approaches to data from consumer wearables. Dr. Ohri and his team previously collected data from 214 patients in three prospective clinical trials (NCT02649569, NCT03102229, NCT03115398).
In each of these trials, participants wore fitness trackers that monitored their activity over several weeks while they were receiving chemotherapy. The trial participants had different types of primary cancers, the most common being head and neck cancer (30%) or lung cancer (29%).
The number of steps and other data from the records of these patients were used to develop and test a structured elastic logistic regression model, which is a kind of machine learning model It can analyze a large amount of complex information. The goal of their model was to predict a patient’s likelihood of being hospitalized in the next week, based on data from the previous two weeks.
The researchers created the model first by examining how well different variables predicted hospital admission, using data from 70% of the trial participants (151 people). Potential indicators in the model included patient characteristics (eg, age and ECOG performance status), as well as activity data measured before and during treatment. In addition to daily step totals, the researchers calculated other metrics, such as relative changes in a person’s averages week by week or the difference in the minimum and maximum number of steps each week.
The research team then validated the model using the remaining 30% of patients (63 people). The model calculating the integrated steps was strongly predictive of hospitalization in the following week (AUC = 0.80, 95% confidence interval (CI)). [CI] 0.60–0.90), and significantly outperformed the model without counting the steps (AUC = 0.46, 95% CI 0.24–0.66, p
“The step immediately before the prediction window ended up being generally more predictive than the clinical variables. The dynamic nature of the steps, the fact that they change every day, seems to make them an especially good indicator of the patient’s health status,” said Dr. Hong.
The most important predictive variables in the model included the number of steps from each day of the past two days, as well as relative changes in the maximum number of steps and the step number range over the past two weeks.
The use of dynamic data distinguishes this model from that 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 running prediction,” explained Ms. Frisner. “You can run the algorithm on any given day and get an idea of the patient’s risk level after one week, giving you time to provide the additional support they need.”
Dr Hong explained that this extra support is key to reducing hospitalization, whether that’s by scheduling more frequent follow-ups, changing something in the patient’s treatment plan or another personalized approach. “The essence of what works is that this is an additional point of contact for the doctor to see a patient. It gives the patient the reassurance of knowing that we are monitoring them.”
“As more people start using wearable devices, the question is whether the data they collect is useful. Our study shows that there is value in having our patients collect their health data during their daily lives, and that we can use that data and then monitor and predict their health.” .
The investigators’ next steps include a more rigorous validation of the algorithm in the NRGF-001 trial (NCT04878952) led by Dr. Urey, which will randomly select patients undergoing CRT for lung cancer for treatment with or without daily step count monitoring. Patients’ physicians on the step counting arm will receive output from the model throughout the treatment process.
The researchers also plan other studies to examine additional metrics collected by the wearable devices, such as heart rate and their usefulness 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, such as electronic health record data, different data points may work better for different patients,” Ms Frisner said.
After widespread adoption of telemedicine and telecare over the past several years, the need for remote monitoring via patient devices may also increase. Dr. Hong said clinics and policy makers should consider accessibility of these devices as their popularity grows.
“One of the challenges when working with wearable data in the real world is the economic and racial disparities that affect who owns the 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 wider range of patients.”
Conference summary: plan.core-apps.com/myastroapp2… dd-b968-4c33a85991cf
American Society of Radiation Oncology
the quote: AI model using daily step count predicts unplanned hospitalization during cancer treatment (2022, Oct 24), retrieved Oct 24, 2022 from https://medicalxpress.com/news/2022-10-ai-daily- unplanned-hospitalizations-cancer.html
This document is subject to copyright. Notwithstanding any fair dealing for the purpose of private study or research, no part may be reproduced without written permission. The content is provided for informational purposes only.