Skip to main content

Predictive Scoring for Cancer Cases

By March 27, 2020March 28th, 2020Commentary

One of the problems in knowing how intensively to treat seriously ill patients is having a good sense of their prognosis.  For cancer in particular, there is often a time when further treatment is relatively futile, the patient is likely terminal and palliative care may be more appropriate.  A study in the Journal of the American Medical Association Network Open discusses a predictive algorithm that determines likely mortality for head and neck cancer patients.   (JAMA Article)   About 16,000 patients in Taiwan with an advanced form of the cancer were studied.  All had completed a course of chemoradiotherapy.  The 90 day mortality rate was measured after completion of this therapy.  The algorithm was designed to see if it could predict which patients were most likely to die.  Other studies had indicated that mortality rates in the near-term after therapy ranged from 5% to 18%.  Obviously for people who die within 90 days after the therapy, it was relatively futile to have had the treatment.  And these treatments are not without discomfort to the patient.  So being able to identify who is not likely to see longer term benefit allows for better patient counseling and decision-making.  A large number of patient demographic and health variables was included in the search for meaningful associations to add to a predictive algorithm.  Factors which appeared to have any meaningful association with 90-day mortality were selected into the algorithm and it was run against the study population.  The overall mortality rate was 6.66%.  The most significant factors associated with risk of dying appear to be being older and having certain comorbidities, such as pneumonia, heart disease and stroke.  The algorithm divided patients into very low risk of death, low risk of death, moderate risk of death and high risk of death.  There were significant differences in survival among these groups, indicating that the algorithm had a high level of success in predicting mortality risk and could be used to help identify patients who should be treated and those for whom treatment had a lower likelihood of success and for whom palliative care might be a more reasonable alternative.  Assuming they have a high level of accuracy, predictive algorithms can help guide care for people with serious illness.

Leave a comment