Summary: A new natural language processing algorithm is able to sift through doctors’ notes and predict a cancer patient’s survival rate over the next 60 months with 80% accuracy.
Source: University of British Columbia
A team of researchers from the University of British Columbia and BC Cancer has developed an artificial intelligence (AI) model that predicts cancer patient survival more accurately and with more readily available data than previous tools.
The model uses natural language processing (NLP) – a branch of AI that understands complex human language – to analyze oncologist notes after a patient’s first consultation visit – the first step in the patient’s journey. cancer after diagnosis.
By identifying patient-specific characteristics, the model was shown to predict survival at six months, 36 months, and 60 months with greater than 80% accuracy.
The results were published today in Open JAMA Network.
“Cancer survival prediction is an important factor that can be used to improve cancer care,” said lead author Dr. John-Jose Nunez, a psychiatrist and clinical researcher at the UBC Mood Disorders Center and the BC Cancer.
“This could suggest health providers make an earlier referral to support services or offer a more aggressive treatment option upfront. Our hope is that a tool like this can be used to personalize and optimize the care a patient receives immediately, giving them the best possible outcome.
Traditionally, cancer survival rates have been calculated retrospectively and categorized by only a few generic factors such as cancer site and tissue type. Despite being familiar with these rates, it can be difficult for oncologists to accurately predict a patient’s survival due to the many complex factors that influence patient outcomes.
The model developed by Dr. Nunez and his collaborators, which includes researchers from BC Cancer and UBC’s computer science and psychiatry departments, is able to pick up unique clues in a patient’s initial consultation document to provide a more nuanced assessment. It also applies to all cancers, whereas previous models were limited to certain types of cancer.
“The AI essentially reads the consultation document the same way a human would read it,” Dr. Nunez said. “These documents contain many details such as the patient’s age, type of cancer, underlying health conditions, previous substance use and family history. The AI brings it all together to paint a more complete picture. results for patients.”
The researchers trained and tested the model using data from 47,625 patients at the six BC Cancer sites located in British Columbia. To protect confidentiality, all patient data remained securely stored at BC Cancer and was presented anonymously. Unlike chart reviews by human research assistants, the new AI approach has the added benefit of maintaining complete confidentiality of patient records.
“Because the model is trained on data from British Columbia, this makes it a potentially powerful tool for predicting cancer survival here in the province,” Dr. Nunez said.
In the future, the technology could be applied in cancer clinics across Canada and around the world.
“The advantage of neural NLP models is that they are highly scalable, portable, and don’t require structured data sets,” Dr. Nunez said. “We can quickly train these models using local data to improve performance in a new region. I would suspect that these models provide a good baseline anywhere in the world where patients can see an oncologist.
In another area of work, Dr. Nunez examines how to facilitate the best possible psychiatric care and counseling for cancer patients using advanced AI techniques. He envisions a future where AI is integrated into many aspects of the healthcare system to improve patient care.
“I see AI almost acting like a virtual assistant for doctors,” Dr. Nunez said. “As medicine becomes more and more advanced, having AI to help sort and make sense of all the data will help inform decisions for doctors. Ultimately, it will help improve quality of life and patient outcomes.
About this news about AI and cancer research
Author: Press office
Source: University of British Columbia
Contact: Press Office – University of British Columbia
Picture: Image is in public domain
Original research: Free access.
“Predicting cancer patient survival from their initial oncology consultation document using natural language processing” by John-Jose Nunez et al. Open JAMA Network
Abstract
Predicting cancer patient survival from their initial oncology consultation document using natural language processing
Importance
Predicting the short- and long-term survival of cancer patients can improve their care. Previous predictive models use data with limited availability or predict the outcome of a single type of cancer.
Objective
See also
To examine whether natural language processing can predict survival in general cancer patients based on an oncologist’s initial consultation document.
Design, framework and participants
This retrospective prognostic study used data from 47,625 of 59,800 patients who entered cancer care at one of 6 BC Cancer centers located in the province of British Columbia between April 1, 2011 and December 31. 2016. Mortality data was updated to April 6, 2022 and data was analyzed from the update to September 30, 2022. All patients with a medical or radiation oncology consultation document generated within 180 days of diagnosis were included; patients seen for multiple cancers were excluded.
Exhibitions
Oncologist initial consultation documents were analyzed using traditional and neural language models.
Main results and measures
The primary outcome was the performance of the predictive models, including balanced accuracy and receiver operating characteristic area under the curve (AUC). The secondary result was to search for the words used by the models.
Results
Of the 47,625 patients in the sample, 25,428 (53.4%) were female and 22,197 (46.6%) were male, with a mean age (SD) of 64.9 (13.7) years. A total of 41,447 patients (87.0%) survived 6 months, 31,143 (65.4%) survived 36 months, and 27,880 (58.5%) survived 60 months, calculated from their initial consultation in oncology. The best models achieved a balanced accuracy of 0.856 (AUC, 0.928) for predicting 6-month survival, 0.842 (AUC, 0.918) for 36-month survival, and 0.837 (AUC, 0.918) for 60-month survival, on an exclusion test. together. Differences in words important to predict survival at 6 months and 60 months were found.
Conclusions and relevance
These results suggest that the models have comparable or better performance than previous models predicting cancer survival and that they may be able to predict survival using readily available data without focusing on 1 type of cancer.