Machine learning models were applied to identify predictors of distress in breast cancer survivors, and support vector machines, XGBoost, and CatBoost showed good predictive performance in terms of AUC scores. Factors determined to be significant predictors were depression for emotional issues, relationship with partner, housing, and work/school for practical issues, and fatigue for physical issues.
The results of this study showed that after completing primary treatment, breast cancer patients had a mean distress level of 4.35 points, with 57.7% of participants classified as having severe distress based on the 4-point score defined in the NCCN guidelines. Distress in breast cancer patients occurs from the time of diagnosis to the completion of treatment, and persists in approximately one-third to one-half of patients after completing primary breast cancer treatment.8The time after treatment ends is crucial for cancer survivors' adjustment. However, at this point, patients who have experienced distress may experience slower recovery and persistent physical and psychological symptoms.20During this period, breast cancer patients who experience ongoing distress not only experience delayed adjustment to daily life and recovery;21,22It is important to note that many breast cancer patients do not receive any special care beyond regular visits after completing treatment, so it is necessary to assess the level of distress in breast cancer patients after primary treatment and take concrete and practical measures to alleviate it.
High-performance machine learning models, support vector machine, XGBoost, and CatBoost, were applied to identify predictors of distress in breast cancer survivors. Results showed that the accuracy, F1 score, and AUC score of these models were considerably high, exceeding 0.70. Support vector machine, XGBoost, and CatBoost models have been reported to show good predictive performance in multiple studies on predicting psychological symptoms such as distress.23,24,25A support vector machine is an algorithm that identifies the boundary with the maximum margin by setting a hyperplane between the data, and it has less overfitting and excellent classification performance.26The XGBoost model is an ensemble of decision trees that uses parallel processing to achieve fast training and classification speeds, and provides excellent predictive performance in classification and regression.27Furthermore, the CatBoost model uses ordinal boosting to achieve high accuracy for categorical variables.28This study demonstrates that compared with traditional binary classification methods of logistic regression, machine learning models not only have improved overall performance indicators, but also provide a more intuitive understanding of the relationship between multiple variables and their feature importance. In particular, support vector machine models showed superior classification performance compared to ensemble models such as XGBoost and CatBoost. This is due to the parallel combination of single models in ensemble methods, which can lead to problems such as increased computation time and overfitting.19Support vector machines alleviate these problems through an iterative learning process, and also exhibit good generalization to new data and robustness to outliers.19This makes it extremely useful in clinical settings where identifying and understanding a multitude of contributing factors is crucial.
Regarding the importance of variables predicting distress in breast cancer survivors, the results of support vector machine, XGBoost, and CatBoost models showed that emotional problems such as depression, fear, worry, loss of interest in daily activities, and nervousness were significant predictors. Emotional symptoms such as depression, fear, and anxiety have been reported to occur in a complex and clustered manner in breast cancer patients.29These symptoms may be present from the time of diagnosis and may persist for 10 or more years after treatment has ended.30. These emotional issues have been identified as the most important variables influencing the quality of life and adaptation of breast cancer patients after primary treatment.22,31Given the established link between these emotional symptoms and long-term survival in cancer patients32Emotional symptoms should be continually monitored and a comprehensive approach to mental health promotion should be implemented, including psychological support and counseling specifically designed for breast cancer survivors.
Moreover, depression is more prevalent than any other emotional condition.29 It is thought to be the biggest factor preventing patients from returning to their daily lives.32In particular, breast cancer patients who have completed primary treatment experience less attention and support from family and friends compared to those during treatment, and are in a mentally and socially vulnerable state, leading to more severe depression. Such severe depression may hinder breast cancer patients' return to normal life, affect their adaptation and transition as survivors, and increase the risk of recurrence and death.33,34Therefore, efforts are needed to detect depression early and deal with it effectively.
Support vector machine, XGBoost, and CatBoost models showed good predictive performance, identifying relationships with partner, housing, and work/school as the most influential factors among practical issues. Results showed that breast cancer survivors who have problems with relationships with their partner experience higher stress. Close interaction with caregivers is an important source of emotional support for breast cancer patients not only during surgery and treatment, but also in the post-treatment period when returning to daily life and adapting to various changes. This interaction plays an important role in managing the physical and psychological problems caused by post-treatment symptoms, enhancing the overall health and adaptation of patients.35Previous studies have shown that when breast cancer patients experience high levels of intimacy with their partners, it has a positive impact on their psychological and social adjustment.36,37Therefore, it can be concluded that enhancing intimacy with one's partner is an important intervention factor for reducing distress.
With regard to housing, the precarious housing situation of breast cancer survivors may result from the financial risks associated with cancer diagnosis and treatment, especially among low-income populations.38Housing problems have a significant impact on household economies, reducing household standards of living, increasing the risk of illness, exacerbating suffering, and impeding adherence to treatment.39.
Another important influencing factor identified in the study is the distress associated with returning to work, which serves as a social and economic safety net.40“Returning to work” is an indicator of improved self-esteem for breast cancer survivors and signifies the return to society from being a patient to being a survivor.41Additionally, the increased income that comes with returning to work provides breast cancer patients with financial stability and a sense of security, an important aspect of cancer recovery.42However, many breast cancer patients experience difficulties in the process of returning to work due to a variety of physical, psychological and social problems caused by treatment.43After returning to work, cancer survivors face many challenges, including a decline in social status and status, unwanted job changes, problems with employers and coworkers, and reduced physical abilities. These difficulties make it difficult to continue working and cause stress.40,43Therefore, practical interventions must be established to enable breast cancer survivors to return to work and achieve economic stability.42.
Finally, the machine learning model confirmed that physical symptoms, including fatigue, sleep, and pain, were major predictors of distress in breast cancer survivors. Fatigue, sleep problems, and pain are frequently reported symptom clusters in breast cancer survivors, and can persist for 5 years or more.44Furthermore, symptom clusters in breast cancer survivors show different patterns depending on symptom severity and affect physical and social functioning, thereby causing distress and reducing quality of life.45Therefore, in order for breast cancer survivors to successfully return to normal life after completion of primary treatment, interventions are needed to effectively monitor and manage the symptoms experienced by survivors.
This study is important in that it used machine learning techniques to assess the level of distress in breast cancer survivors and identify factors that influence the widespread occurrence of distress. However, the results of this study should be interpreted with caution due to the following limitations. First, this study was conducted only on breast cancer survivors registered at the Cancer Survivor Integrated Support Center in Gyeonggi-do, South Korea, so the study findings cannot be generalized. Therefore, future studies should expand the survey of study participants to a nationwide scale to evaluate distress in breast cancer survivors in detail. Second, this study was limited in identifying changes in distress experienced during the transition from breast cancer patient to breast cancer survivor and determining factors that influence these changes. Therefore, longitudinal studies should be conducted to identify patterns of changes in distress and associated factors in breast cancer survivors.
