Abstract
Introduction
One of the most common, but underreported, long-term effects after breast cancer diagnosis and treatment is cancer-related fatigue (CRF) [1]. An intervention for CRF can be started on time if patients with CRF or at risk for CRF are identified on time and can contribute to preventing CRF or preventing CRF from worsening and becoming chronic. The goal of this study is to build machine learning models and predict the individual risk of developing CRF.
Methods
The Primary Secondary Cancer Care Registry (PSCCR) was used as database to predict fatigue [2], defined as visits to the General Practitioner (GP) with fatigue complaints. A total of 12.813 breast cancer patients were included of whom 2.224 were fatigued. Predictors (n=64) were related to patient, tumour and treatment characteristics and GP visits before diagnosis. Missing data was imputed using Multiple Imputation by Chained Equations and Random Forest Classifier, Logistic Regression, Gaussian Naïve Bayes, K-Nearest Neighbours and Multi-Layer Perceptron were used as machine learning models. A nested 5-fold cross validation was used to optimize hyperparameters and the performance of the models was assessed using the area under the receiver operator characteristic curve (AUC-score).
Results
The best model, a Random Forest Classifier, performed only moderate with an AUC-score of 0.63±0.014. The other models performed even worse and their AUC-score ranged 0.54-0.62.
Conclusion
The individual risk of developing CRF cannot accurately be predicted using machine learning models on the PSCCR. Fatigue was defined as patients visiting GP with complaints, however, CRF is underreported at healthcare professionals. Additionally, more variables could be predictive of CRF next to patient, tumor and treatment characteristics. Therefore, in future studies, we hope to collect other data, related to other predictors and also clearer definition of CRF.
One of the most common, but underreported, long-term effects after breast cancer diagnosis and treatment is cancer-related fatigue (CRF) [1]. An intervention for CRF can be started on time if patients with CRF or at risk for CRF are identified on time and can contribute to preventing CRF or preventing CRF from worsening and becoming chronic. The goal of this study is to build machine learning models and predict the individual risk of developing CRF.
Methods
The Primary Secondary Cancer Care Registry (PSCCR) was used as database to predict fatigue [2], defined as visits to the General Practitioner (GP) with fatigue complaints. A total of 12.813 breast cancer patients were included of whom 2.224 were fatigued. Predictors (n=64) were related to patient, tumour and treatment characteristics and GP visits before diagnosis. Missing data was imputed using Multiple Imputation by Chained Equations and Random Forest Classifier, Logistic Regression, Gaussian Naïve Bayes, K-Nearest Neighbours and Multi-Layer Perceptron were used as machine learning models. A nested 5-fold cross validation was used to optimize hyperparameters and the performance of the models was assessed using the area under the receiver operator characteristic curve (AUC-score).
Results
The best model, a Random Forest Classifier, performed only moderate with an AUC-score of 0.63±0.014. The other models performed even worse and their AUC-score ranged 0.54-0.62.
Conclusion
The individual risk of developing CRF cannot accurately be predicted using machine learning models on the PSCCR. Fatigue was defined as patients visiting GP with complaints, however, CRF is underreported at healthcare professionals. Additionally, more variables could be predictive of CRF next to patient, tumor and treatment characteristics. Therefore, in future studies, we hope to collect other data, related to other predictors and also clearer definition of CRF.
[1] de Ligt et al., 2019, 10.1016/j.breast.2019.03.010
[2] Heins et al., 2018, 10.1016/S0959-8049(18)30344-7
| Original language | English |
|---|---|
| Publication status | Published - Jun 2022 |
| Event | 19th Bossche Mamma Congres 2022 - Ruwenberg, Sint Michielsgestel, Netherlands Duration: 14 Jun 2022 → 15 Jun 2022 Conference number: 19 https://www.bosschemammacongres.nl/ |
Conference
| Conference | 19th Bossche Mamma Congres 2022 |
|---|---|
| Country/Territory | Netherlands |
| City | Sint Michielsgestel |
| Period | 14/06/22 → 15/06/22 |
| Internet address |
