• Waldenström Macroglobulinemia
  • Melanoma
  • Bladder Cancer
  • Brain Cancer
  • Breast Cancer
  • Childhood Cancers
  • Gastric Cancer
  • Gynecologic Cancer
  • Head & Neck Cancer
  • Immunotherapy
  • Kidney Cancer
  • Leukemia
  • Liver Cancer
  • Lung Cancer
  • Lymphoma Cancer
  • Mesothelioma
  • MPN
  • MDS
  • Myeloma
  • Prostate Cancer
  • Rare Cancers
  • Sarcoma
  • Skin Cancer
  • Testicular Cancer
  • Thyroid Cancer

Artificial Intelligence May Accurately Predict Short-Term Mortality in Patients with Cancer


An artificial intelligence-based algorithm within the electronic medical record provided real-time and accurate predictions of the short-term mortality rate of patients with cancer, outperforming all other prognostic incidences commonly used.

A machine learning algorithm, or artificial intelligence, accurately predicted the 180-day morality rate in real time of patients with cancer, according to a study published in JAMA Oncology.

“In this prognostic study our results suggest that an (machine learning) algorithm can be feasibly integrated into the (electronic health record) to generate real-time, accurate predictions of short-term morality risk for patients with cancer that outperform traditional prognostic indices,” the study authors wrote.

The objective of this study was to validate that the machine learning algorithm can predict the 180-day morality rate in patients, giving them answers and hopefully a better quality of life.

Researchers followed 24,582 patients (median age, 64.6 years; 62.3% women) between March 2019 and April 2019. Among those patients, 1,022 died within the 180 days of their first encounter with a medical or gynecologic oncology practice. The authors note that there was a large difference between the patients alive after 180 days compared with those who were not. Those who died were more likely to have stage 4 cancer, Eastern Cooperative Oncology Group greater than or equal to two (indicating limits to performance such as work activities, limited self-care or completely disabled) and Elixhauser comorbidity index greater than or equal to three (indicating a greater presence of two or more medical conditions or diseases).

Although the machine learning algorithm was accurate in the full group of patients, it varied across several disease types within tertiary (specialized care) practices. Despite this, the accuracy of the algorithm was similar between general oncology and tertiary practices.

Researchers also differentiated patients as high risk or low risk for mortality using a 40% mortality risk threshold. With this, mortality at 180 days was predicted in 45.2% of patients with high risk compared with 3.1% of those with low risk.

The machine learning algorithm, when integrated into other indices commonly used in clinical practice, helped reclassify patients into their appropriate risk profiles.

Machine learning may improve the prediction of mortality risk, in addition to better clinician and patient decision making. Machine learning predictions of mortality may better identify patients who have a short-term mortality, improve care discussions and meet metrics forquality end-of-life care.

“Such an automated tool may complement clinician intuition and lead to improved targeting of supportive care interventions for high-risk patients with cancer,” the study authors wrote.

For more news on cancer updates, research and education, don’t forget to subscribe to CURE®’s newsletters here.

Related Videos
Cancer survivor, Frank J. Peter, playing an original song on the piano
Brandi Benson, sarcoma survivor and military veteran, in an interview with CURE