Depression Screening Using Daily Mental-Health Ratings from a Smartphone Application for Breast Cancer Patients

Junetae Kim, Sanghee Lim, Yul Ha Min, Yong‐Wook Shin, Byungtae Lee, Guiyun Sohn, Kyung Hae Jung, Jae‐Ho Lee, Byung Ho Son, Sei Hyun Ahn, Soo-Yong Shin, Jong Won LeeJournal of Medical Internet Research 2016 · DOI

TL;DR

This study explores the effectiveness of smartphone apps for daily mental health tracking to screen for depression in breast cancer patients. It demonstrates that these apps can achieve comparable accuracy to traditional screening methods, with higher patient adherence leading to significantly improved screening performance.

What Problem Does This Paper Address?

Detecting depression in cancer patients is challenging due to time constraints for both patients and clinicians, and patient reluctance to undergo screening. While mobile health trackers offer a promising solution by enabling daily self-reporting, their validity for depression screening and the impact of data processing methods and patient adherence remain underexplored. This paper aims to bridge this gap by evaluating the screening performance of a mobile mental-health tracker, proposing data processing strategies, and investigating the crucial role of patient adherence.

Methods

The researchers developed and evaluated a mobile mental-health tracker that collects daily ratings on sleep satisfaction, mood, and anxiety using facial emoticons. They proposed and compared three data-processing approaches to convert these daily ratings into indicator variables for depression screening:

  • Ratio approach: Calculates the ratio of days with a score indicating depression to the total reported days.
  • Frequency approach: Counts the number of days with a score indicating depression.
  • Average approach: Calculates the average of the daily ratings.

To assess depression status, they used the Patient Health Questionnaire-9 (PHQ-9) as the gold standard. They employed random-effect logistic regression and Receiver Operating Characteristic (ROC) analysis to evaluate the screening performance of the mobile tracker. Patient adherence was classified into "higher" and "lower" groups using k-means clustering based on activeness, timeliness, and persistence, and their respective screening accuracies were compared.

Datasets

The study utilized a dataset of 5792 daily mental-health ratings collected from 78 breast cancer patients over a 48-week period. These ratings were collected via the "Pit-a-Pat" smartphone application. The Patient Health Questionnaire-9 (PHQ-9) was administered bi-weekly to establish the true depression status.

Key Insights and Ideas

  • Smartphone apps are viable for depression screening: The study provides strong empirical evidence that mobile mental-health trackers can effectively screen for depression, achieving accuracy comparable to established methods like the PHQ-9.
  • Data processing matters: The "ratio" and "frequency" approaches for converting daily ratings into indicator variables showed promising results, with the ratio approach demonstrating the highest accuracy.
  • Adherence is critical for accuracy: Patients who consistently and diligently use the app (higher adherence) exhibit significantly better depression screening accuracy compared to those with lower adherence. This highlights the importance of patient engagement.
  • Multidimensional adherence is key: Defining adherence not just by response rate but by activeness, timeliness, and persistence provides a more nuanced and impactful understanding of its role in screening accuracy.

Key Results

  • The ratio approach yielded the highest Area Under the ROC Curve (AUC) of 0.8012, indicating good depression screening performance, comparable to PHQ-9 tests.
  • The frequency approach had a slightly lower AUC of 0.7635.
  • The average approach resulted in an AUC of 0.7867.
  • For patients with higher adherence, the AUC was significantly higher (0.8524 for the ratio approach) compared to those with lower adherence (0.7234 for the ratio approach), with a p-value of .002.
  • Robustness analyses confirmed these findings, showing consistent improvements in screening accuracy with higher adherence across different data processing methods.

Limitations and Future Work

  • Limited scope of mental health dimensions: The study focused on sleep, mood, and anxiety. Future research could explore other relevant mental health dimensions for daily tracking.
  • Missing data handling: The study did not explicitly account for methods of dealing with missing ratings. Further investigation into how missing values affect screening performance and how to impute them is needed.
  • Generalizability: The study was conducted in a breast cancer patient population in South Korea. Results may not be generalizable to other patient groups, mental illnesses, or geographical locations with different mobile technology adoption rates.
  • Future work: The authors suggest investigating optimal choices for mental dimensions to be tracked, developing better methods for handling missing data, and conducting validation studies in diverse patient populations and settings. They also highlight the need for further research on the impact of mobile mental-health trackers in different clinical contexts.

Summary updated

Abstract

Our results support the potential of a mobile mental-health tracker as a tool for screening for depression in practice. Also, this study provides clinicians with a guideline for generating indicator variables from daily mental-health ratings. Furthermore, our results provide empirical evidence for the critical role of adherence to self-reporting, which represents crucial information for both doctors and patients.
BibTeX
@article{kim2016depression,
  author = {Junetae Kim and Sanghee Lim and Yul Ha Min and Yong‐Wook Shin and Byungtae Lee and Guiyun Sohn and Kyung Hae Jung and Jae‐Ho Lee and Byung Ho Son and Sei Hyun Ahn and Soo-Yong Shin and Jong Won Lee and Jong Won Lee and Jong Won Lee},
  title = {{Depression Screening Using Daily Mental-Health Ratings from a Smartphone Application for Breast Cancer Patients}},
  journal = {Journal of Medical Internet Research},
  year = {2016},
  doi = {10.2196/jmir.5598},
}
Figure