JMIR Publications

ABSTRACT

Background:

The importance of classifying cancer patients into high or low risk groups has led many research teams, from the biomedical and the bioinformatics field, to study the application of Machine Learning (ML) algorithms. The International Society of Geriatric Oncology (SIOG) recommends the use of the Comprehensive Geriatric Assessment (CGA), a multidisciplinary tool to evaluate health domains, for the follow-up of elderly cancer patients. However, no ML applications have been proposed using CGA to classify elderly cancer patients.

Objective:

To propose and develop predictive models, using ML and CGA, to estimate the risk of early death in elderly cancer patients.

Methods:

The ability of ML algorithms to predict early mortality in a cohort involving 608 elderly cancer patients was evaluated. The CGA was conducted during admission by a multidisciplinary team and included the following questionnaires: Mini-Mental State Examination, Geriatric Depression Scale, International Physical Activity Questionnaire, Timed Get-Up and Go, Katz Index, Charlson Comorbidity Index, Karnofsky Performance Scale, Polypharmacy, Mini Nutritional Assessment. The 10-fold Cross Validation algorithm was used to evaluate all possible combinations of these questionnaires to estimate the risk of early death, considered when occurring within six months of diagnosis, in a variety of ML classifiers, including Naive Bayes (NB), Decision Tree (J48) and Multilayer Perceptron (MLP). On each fold of evaluation, tie-breaking is handled by choosing the smallest set of questionnaires.

Results:

It was possible to select CGA questionnaire subsets with high predictive capacity for early death, either statistically similar (NB) or higher (J48 and MLP) compared to the use of all questionnaires investigated. These results show that CGA questionnaire selection can improve accuracy rates and decrease the time spent to evaluate elderly cancer patients. The only questionnaire selected in all folds was the Mini Nutrition Evaluation. The Karnofsky Performance Scale was selected in all folds by the NB and MLP, while the Mini Mental State Examination was selected in all folds by the NB.

Conclusions:

A simplified predictive model aiming to estimate the risk of early death in elderly cancer patients is proposed herein, minimally composed by the Mini Nutrition Evaluation accompanied or not by the Karnofsky Performance Scale and/or the Mini-Mental State Examination.