Understanding the Relationship Between Mood Symptoms and Mobile App Engagement Among Patients With Breast Cancer Using Machine Learning: Case Study

Abstract

Background Health interventions delivered via smart devices are increasingly being used to address mental health challenges associated with cancer treatment. Engagement with mobile interventions has been associated with treatment success; however, the relationship between mood and engagement among patients with cancer remains poorly understood. A reason for this is the lack of a data-driven process for analyzing mood and app engagement data for patients with cancer. Objective This study aimed to provide a step-by-step process for using app engagement metrics to predict continuously assessed mood outcomes in patients with breast cancer. Methods We described the steps involved in data preprocessing, feature extraction, and data modeling and prediction. We applied this process as a case study to data collected from patients with breast cancer who engaged with a mobile mental health app intervention (IntelliCare) over 7 weeks. We compared engagement patterns over time (eg, frequency and days of use) between participants with high and low anxiety and between participants with high and low depression. We then used a linear mixed model to identify significant effects and evaluate the performance of the random forest and XGBoost classifiers in predicting weekly mood from baseline affect and engagement features. Results We observed differences in engagement patterns between the participants with high and low levels of anxiety and depression. The linear mixed model results varied by the feature set; these results revealed weak effects for several features of engagement, including duration-based metrics and frequency. The accuracy of predicting depressed mood varied according to the feature set and classifier. The feature set containing survey features and overall app engagement features achieved the best performance (accuracy: 84.6%; precision: 82.5%; recall: 64.4%; F1 score: 67.8%) when used with a random forest classifier. Conclusions The results from the case study support the feasibility and potential of our analytic process for understanding the relationship between app engagement and mood outcomes in patients with breast cancer. The ability to leverage both self-report and engagement features to analyze and predict mood during an intervention could be used to enhance decision-making for researchers and clinicians and assist in developing more personalized interventions for patients with breast cancer.

Machine Learning Approaches to Predict Symptoms in People With Cancer: Systematic Review

“…Of the 42 included studies, 10 unique symptoms were reported as outcome variables in the predictions. Those included were xerostomia (9/42, 14%) [ 27 , 30 , 34 , 49 - 51 , 55 , 58 , 60 ], depression (8/42, 13%) [ 22 , 33 , 37 , 41 , 43 , 45 , 52 , 59 ], pain (8/42, 13%) [ 20 , 25 , 26 , 35 , 37 , 40 , 42 , 56 ], fatigue (6/42, 10%) [ 23 , 24 , 29 , 38 , 42 , 44 ], anxiety (3/42, 5%) [ 33 , 46 , 52 ], sleep disturbance or insomnia (3/42, 5%) [ 26 , 28 , 52 ], nausea or vomiting (3/42, 5%) [ 17 , 26 , 29 ], weight loss (2/42, 3%) [ 47 , 53 ], cognitive impairment (2/42, 3%) [ 21 , 36 ], and diarrhea (2/42, 3%) [ 29 , 42 ].…”

Engagement With a Relaxation and Mindfulness Mobile App Among People With Cancer: Exploratory Analysis of Use Data and Self-Reports From a Randomized Controlled Trial

“…Data on macrolevel engagement are necessary to understand how engagement with an mHealth app changes over time and how these engagement patterns relate to the intended health outcomes. Baglione et al [ 17 ] found that high baseline distress was associated with initially higher engagement that declined over time, whereas the engagement of the group of participants with lower baseline distress increased over the course of a 7-week intervention, resulting in similar engagement levels in both groups at week 7. Siebenhüner et al [ 79 ] examined the associations between distress and adherence (ie, app use) in the CanRelax app 1.0 and showed that a decrease in the level of distress over time (ie, an improvement in health outcomes) was associated with lower adherence.…”

“…Data on macrolevel engagement are necessary to understand how engagement with an mHealth app changes over time and how these engagement patterns relate to the intended health outcomes. Baglione et al [17] found that high baseline distress was associated with initially higher engagement that declined over time, whereas the engagement of the group of participants…”

“…In studies using mHealth apps, poor engagement can also confound the outcome and impact the validity of the results as study dropouts may differ from completers [2,9]. While many mHealth apps have significant issues with sustained engagement [10][11][12][13][14][15][16][17], this is particularly true for fully automated mHealth apps without any human support, also termed unguided or self-guided mHealth apps. A high level of human support in guided mHealth apps is typically associated with better engagement rates but at the cost of reduced scalability [18,19].…”

“…However, although system use data undoubtedly provide valuable information on certain aspects of microlevel engagement, these data are not considered a valid measure of engagement on their own [ 24 ]. Greater efforts are needed to combine different data sources, such as pairing system use data with self-report data or qualitative methods, to better understand the user experience [ 5 , 6 , 17 , 24 ]. At the macro level, assessing engagement remains a challenge and is often neglected in mHealth studies.…”