BluBot: Integration of an mHealth Application and EMR System to Increase Frequency of Depression Screenings in Non-Clinical Settings Open Access

Olsen, Marisa (2017)

Permanent URL:


Introduction: Detecting and treating depressive disorders is a public health priority. About one in ten patients seen by primary care physicians has a depressive disorder, yet these disorders largely remain underdiagnosed. Post diagnosis, a study found that 67% of respondents with mental illness were interested in monitoring their symptoms through applications on their phones. With the wearable market expected to increase from 275 million devices in 2016 to 477 million devices in 2020, providers and researchers need a system to capture and use health data from these devices. The purpose of this thesis is to develop a prototype for integrating data from an mHealth application, collected near continuously, that screens for depression and translates information into an open electronic medical record system (EMR) for clinical and research analysis.

Methods: The BluBot app collects raw mobile phone sensor data and transforms it into location and mobile device use features significantly correlated with PHQ-9 scores. A predictive model is trained on user PHQ-9 responses to generate predictions. One user carried a mobile device that collected data points near continuously for 10 weeks. To compare accuracy of the BluBot model, 66 days of observations (collected from December 28, 2016 through March 3, 2017) were used to calculate the Root Mean Square Error (RMSE) and Mean Absolute Error (MAE).

Results:Infrastructure was created to inject app data into an OpenEMR system and to alert provider of depressive episodes. When compared to a running average of PHQ-9 responses (RMSE=1.99, MAE=1.52), the BluBot model did not perform as well (RSME= 2.48, MAE=1.97). This may be due to a lack of overall variation in the user scores and a series of spikes in the user scores during week 5 of data collection.

Conclusion: The BluBot prototype can be used for depression screenings in between doctor visits, providing intervention opportunities when depressive episodes occur. Steps should be taken to further validate the model and improve upon the application through user studies. This prototype, including its application, infrastructure, and repository, can all be repurposed and expanded upon to incorporate data from various wearables and mHealth applications into EMR/EHR systems for public health.

Table of Contents

Contents 1 Chapter 1: Introduction 1 1.1 Background and Rationale 1 1.1.1 Review of Literature 2 1.1.2 Diagnosis of Depression 4 1.1.4 The Role of Technology in Depression Screening 5 1.1.5 Automating Assessments 6 1.1.5 Integration of Mobile Device Data into EMR 8 1.2 Problem Statement 11 1.3 Purpose 13 1.4 Significance 14 2 Chapter 2: Methodology 14 2.1 Introduction 14 2.2 Data Types and Description 15 2.3 Project Design 19 2.4 Procedures 21 2.4.1 Ethics 21 2.4.2 Database and Server Procurement 21 2.4.3 App Development 22 2.4.4 Electronic Medical Record System 22 2.4.5 Data Standards 23 2.5 Instruments 24 2.6 Data Analysis 25 3 Chapter 3: Outcome 28 3.1 Introduction 28 3.2 Project Outcome/Deliverables 30 3.2.1 BluBot App 30 3.2.2 Application Server 35 3.2.3 Injecting Data into EMR 37 3.3 Model Fit 42 3.4 Summary 46 4 Chapter 4: Discussion 46 4.1 Introduction 46 4.2 Summary of Project 47 4.3 Implications 49 4.3.1 Clinical Implications 49 4.3.2 Implications for Public Health 51 4.3.3 Other Implications 51 4.4 Limitations 52 4.4.1 Incorporating PHQ-9 into OpenEMR 53 4.4.2 Model Performance and Validity 53 4.4.3 Interoperability 56 4.4.4 HIPAA Compliance 57 4.5 Recommendations and Next Steps 59 4.5.1 Recommendations 59 4.5.2 Next Steps 60 4.6 Conclusion 62 5 Chapter 5: Executive Summary 63 5.1 Overview 64 5.2 Innovation 64 5.3 Purpose 65 5.4 Findings 65 5.5 Incentives for Use 65 5.5 Recommendation 67 6. References 68 7. Appendix 75 Appendix A. List of Recognized Depression Symptoms 75 Appendix B. PHQ-9 Items: 76 Appendix C. BluBot prediction model data flow legend 77 List of Figures Figure 1. Technical Architecture Design for Park et al. Study 9 Figure 2. Data Flow: Transformation of raw data into PHQ-9 predictions 26 Figure 3. Overview of the BluBot and OpenEMR integration architecture 30 Figure 4. Data flow diagram for BluBot application 31 Figure 5. BluBot Graph: User and Model-Predicted PHQ-9 Scores 32 Figure 6. Itemized PHQ-9 scores: predicted and actual 32 Figure 7. Example of Device Location in BluBot user display 33 Figure 8. Screen State Display 34 Figure 9. BluBot predictive model settings 35 Figure 10. Data points within the BluBot repository 36 Figure 11. BluBot predictive model data flow 37 Figure 12. Health care provider view of Message and Reminder Center Dashboard 39 Figure 13. Mental Health Alert message contents 40 Figure 14. View of Patient Dashboard with Mental Health History embedded 42 Figure 15. BluBot graph showing one day training model performance 43 Figure 16. BluBot: one-week training and evaluation window model performance 43 Figure 17. Comparison of the user PHQ-9 scores to model predictions 44 Figure 18. Absolute errors: BluBot one-day model vs running average of PHQ-9 scores 45 Figure 19. User PHQ-9 scores: all dates with corresponding model predicted scores 48 Figure 20. RMSE One-day and one-week training models, evaluated over consecutive weeks 49 Figure 21. BluBot performance with and without circadian movement feature enabled 56 List of Tables Table 1. Statistical measures: BluBot Model vs running average of user PHQ-9 scores 45 Table 2. Healthy People 2020 Objectives potentially impacted 51 Table 3. BluBot model performance with and without the circadian movement feature enabled 55 Table 4. CMS incentives for capturing and using PHGD 61

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
  • English
Research Field
Committee Chair / Thesis Advisor
Committee Members
Partnering Agencies
Last modified

Primary PDF

Supplemental Files