Development and validation of a prediction model of poor performance status and severe symptoms over time in cancer patients (PROVIEW+)

Machine learning-derived model for predicting poor post-treatment quality of life in Korean cancer survivors

PURPOSE

A substantial number of cancer survivors have poor quality of life (QOL) even after completing cancer treatment. Thus, in this study, we used machine learning (ML) to develop predictive models for poor QOL in post-treatment cancer survivors in South Korea.

METHODS

This cross-sectional study used online survey data from 1,005 post-treatment cancer survivors in South Korea. The outcome variable was QOL, which was measured using the global QOL subscale of the European Organization of Cancer and Treatment for Cancer Quality of Life Questionnaire, where a global QOL score < 60.4 was defined as poor QOL. Three ML models (random forest (RF), support vector machine, and extreme gradient boosting) and three deep learning models were used to develop predictive models for poor QOL. Model performance regarding accuracy, area under the receiver operating characteristic curve, F1 score, precision, and recall was evaluated. The SHapely Additive exPlanation (SHAP) method was used to identify important features.

RESULTS

Of the 1,005 participants, 65.1% had poor QOL. Among the six models, the RF model had the best performance (accuracy = 0.85, F1 = 0.90). The SHAP method revealed that survivorship concerns (e.g., distress, pain, and fatigue) were the most important factors that affected poor QOL.

CONCLUSIONS

The ML-based prediction model developed to predict poor QOL in Korean post-treatment cancer survivors showed good accuracy. The ML model proposed in this study can be used to support clinical decision-making in identifying survivors at risk of poor QOL.

Predictive Models of Psychological Distress, Quality of Life, and Adherence to Medication in Breast Cancer Patients: A Scoping Review

Purpose

An interplay of clinical and psychosocial variables affects breast cancer patients' experiences and clinical trajectories. Several studies investigated the role of socio-demographic, clinical, and psychosocial factors in predicting relevant outcomes in breast cancer care, thus developing predictive models. Our aim is to summarize predictive models for specific psychological and behavioral outcomes: psychological distress, quality of life, and medication adherence. Specifically, we aim to map the determinants of the outcomes of interest, offering a thorough overview of these models.

Methods

Databases (PubMed, Scopus, Embase) have been searched to identify studies meeting the inclusion criteria: a breast cancer patients' sample, development/validation of a predictive model for selected psychological/behavioral outcomes (ie, psychological distress, quality of life, and medication adherence), and availability of English full-text.

Results

Twenty-one papers describing predictive models for psychological distress, quality of life, and adherence to medication in breast cancer were included. The models were developed using different statistical approaches. It has been shown that treatment-related factors (eg, side-effects, type of surgery or treatment received), socio-demographic (eg, younger age, lower income, and inactive occupational status), clinical (eg, advanced stage of disease, comorbidities, physical symptoms such as fatigue, insomnia, and pain) and psychological variables (eg, anxiety, depression, body image dissatisfaction) might predict poorer outcomes.

Conclusion

Predictive models of distress, quality of life, and adherence, although heterogeneous, showed good predictive values, as indicated by the reported performance measures and metrics. Many of the predictors are easily available in patients' health records, whereas others (eg, coping strategies, perceived social support, illness perceptions) might be introduced in routine assessment practices. The possibility to assess such factors is a relevant resource for clinicians and researchers involved in developing and implementing psychological interventions for breast cancer patients.

Development and validation of risk of CPS decline (RCD): a new prediction tool for worsening cognitive performance among home care clients in Canada

BACKGROUND

To develop and validate a prediction tool, or nomogram, for the risk of a decline in cognitive performance based on the interRAI Cognitive Performance Scale (CPS).

METHODS

Retrospective, population-based, cohort study using Canadian Resident Assessment Instrument for Home Care (RAI-HC) data, collected between 2010 and 2018. Eligible home care clients, aged 18+, with at least two assessments were selected randomly for model derivation (75%) and validation (25%). All clients had a CPS score of zero (intact) or one (borderline intact) on intake into the home care program, out of a possible score of six. All individuals had to remain as home care recipients for the six months observation window in order to be included in the analysis. The primary outcome was any degree of worsening (i.e., increase) on the CPS score within six months. Using the derivation cohort, we developed a multivariable logistic regression model to predict the risk of a deterioration in the CPS score. Model performance was assessed on the validation cohort using discrimination and calibration plots.

RESULTS

We identified 39,292 eligible home care clients, with a median age of 79.0 years, 62.3% were female, 38.8% were married and 38.6% lived alone. On average, 30.3% experienced a worsening on the CPS score within the six-month window (i.e., a change from 0 or 1 to 2, 3, 4, 5, or 6). The final model had good discrimination (c-statistic of 0.65), with excellent calibration.

CONCLUSIONS

The model accurately predicted the risk of deterioration on the CPS score over six months among home care clients. This type of predictive model may provide useful information to support decisions for home care clinicians who use interRAI data internationally.

Benchmarking machine learning approaches to predict radiation-induced toxicities in lung cancer patients

Background and purpose

Radiation-induced toxicities are common adverse events in lung cancer (LC) patients undergoing radiotherapy (RT). An accurate prediction of these adverse events might facilitate an informed and shared decision-making process between patient and radiation oncologist with a clearer view of life-balance implications in treatment choices. This work provides a benchmark of machine learning (ML) approaches to predict radiation-induced toxicities in LC patients built upon a real-world health dataset based on a generalizable methodology for their implementation and external validation.

Materials and Methods

Ten feature selection (FS) methods were combined with five ML-based classifiers to predict six RT-induced toxicities (acute esophagitis, acute cough, acute dyspnea, acute pneumonitis, chronic dyspnea, and chronic pneumonitis). A real-world health dataset (RWHD) built from 875 consecutive LC patients was used to train and validate the resulting 300 predictive models. Internal and external accuracy was calculated in terms of AUC per clinical endpoint, FS method, and ML-based classifier under analysis.

Results

Best performing predictive models obtained per clinical endpoint achieved comparable performances to methods from state-of-the-art at internal validation (AUC ≥ 0.81 in all cases) and at external validation (AUC ≥ 0.73 in 5 out of 6 cases).

Conclusion

A benchmark of 300 different ML-based approaches has been tested against a RWHD achieving satisfactory results following a generalizable methodology. The outcomes suggest potential relationships between underrecognized clinical factors and the onset of acute esophagitis or chronic dyspnea, thus demonstrating the potential that ML-based approaches have to generate novel data-driven hypotheses in the field.

Development and external validation of a machine learning-based prediction model for the cancer-related fatigue diagnostic screening in adult cancer patients: a cross-sectional study in China

PURPOSE

Cancer-related fatigue (CRF) is the most common symptom in cancer patients and may interfere with patients' daily activities and decrease survival rate. However, the etiology of CRF has not been identified. Diagnosing CRF is challenging. Thus, our study aimed to develop a CRF prediction model in cancer patients, using data that healthcare professionals routinely obtained from electronic health records (EHRs) based on the 3P model and externally validate this model in an independent dataset collected from another hospital.

METHODS

Between April 2022 and September 2022, a cross-sectional study was conducted on adult cancer patients at two first-class tertiary hospitals in China. Data that healthcare professionals routinely obtained from electronic health records (EHRs) based on the 3P model were collected. The outcome measure was according to ICD-10 diagnostic criteria for CRF. Data from one hospital (n = 305) were used for model development and internal validation. An independent data set from another hospital (n = 260) was utilized for external validation. logistic regression, random forest (RF), Naive Bayes (NB), and extreme gradient boosting (XGBoost) were constructed and compared. The model performance was evaluated in terms of both discrimination and calibration.

RESULTS

The prevalence of CRF in the two centers was 57.9% and 56.1%, respectively. The Random Forest model achieved the highest AUC of 0.86 among the four types of classifiers in the internal validation. The AUC of RF and NB were above 0.7 in the external validation, suggesting that the models also have an acceptable generalization ability.

CONCLUSIONS

The incidence of CRF remains high and deserves more attention. The fatigue prediction model based on the 3P theory can accurately predict the risk of CRF. Nonlinear algorithms such as Random Forest and Naive Bayes are more suitable for diagnosing and evaluating symptoms.

Physician home visits in Ontario: a cross-sectional analysis of patient characteristics and postvisit use of health care services

BACKGROUND

It is unknown how much of current physician home visit volume is driven by low-complexity or low-continuity visits. Our objectives were to measure physician home visit volumes and costs in Ontario from 2005/06 to 2018/19, and to compare patient characteristics and postvisit use of health care services across home visit types.

METHODS

This was a retrospective cross-sectional study using health administrative data. We examined annual physician home visit volumes and costs from 2005/06 to 2018/19 in Ontario, and characteristics and postvisit use of health care services of residents who received at least 1 home visit from any physician in 2014/15 to 2018/19. We categorized home visits as palliative, provided to a patient who also received home care services or "other," and compared characteristics and outcomes between groups.

RESULTS

A total of 4 418 334 physician home visits were performed between 2005/06 and 2018/19. More than half (2 256 667 [51.1%]) were classified as "other" and accounted for 39.1% ($22 million) of total annual physician billing costs. From 2014/15 to 2018/19, of the 413 057 home visit patients, 240 933 (58.3%) were adults aged 65 or more, and 323 283 (78.3%) lived in large urban areas. Compared to the palliative care and home care groups, the "other" group was younger, had fewer comorbidities, and had lower rates of emergency department visits and hospital admissions in the 30 days after the visit.

INTERPRETATION

About half of physician home visits in 2014/15 to 2018/19 were to patients who were receiving neither palliative care nor home care, a group that was younger and healthier, and had low use of health care services after the visit. There is an opportunity to refine policy tools to target patients most likely to benefit from physician home visits.