Aging Mobility

[zaouache]

Source Code

The complete source code is available here: https://github.com/master-csmi/2024-m2-project-aging-mobility

Team

This project was carried out as a pair by:
Narimane ZAOUACHE
Helya AMIRI

Overview

The main objective of my Master's degree project, under the supervision of Laure Combourieu and Pierre-Yves Rohan, was to develop a digital tool for motor frailty screening in private physiotherapy clinics. The project transformed a paper-based questionnaire (SFDFM - Functional Screening Score for Motor Frailty) into an intuitive web application for tablets, ensuring faster, more reliable, and GDPR-compliant screening.

Project Objectives

The project had four main objectives:

1. Improve the existing questionnaire interface:

• Develop an interactive web interface compatible with tablets

• Implement a drawing component for geometric exercises (Dash-Canvas)

• Automate score calculations from responses

• Enable secure data storage with fine-grained access control (Girder)

• Implement API key authentication for sending files to Girder

2. Create a consultation interface:

• Develop an API-key authenticated interface to display all patient files

• Allow secure access to results outside the Girder environment

3. Deployment and website creation:

• Build a single website providing access to both interfaces

• Ensure multi-device compatibility (tablets, smartphones, computers)

4. Data analysis:

• Analyze existing data to identify variables influencing SFDFM scores

• Help clinicians adjust frailty care strategies based on findings

Technologies Used

The project leveraged several technologies:

Frameworks and Libraries:

• Dash: For creating interactive web interfaces in Python

• Pandas: For data management and analysis

• Dash-Canvas: Drawing component for the MOCA test

• Requests: For API communication with Girder

• Gunicorn + Heroku: For secure cloud deployment

Main Features:

• Digital questionnaire for clinical and cognitive data

• Dynamic alert and validation system

• Automatic Excel report generation

• Secure data transfer to Girder

• Independent consultation interface

Key Developments

1. Questionnaire Interface Redesign
The initial interface was completely redesigned to improve usability:

• Each test section presented in independent windows with "Next" buttons

• Larger fonts and enhanced contrast for better accessibility

• Clear error messages and automatic validations

• Optimized for tablet use in clinical settings

2. Secure Consultation Interface
A separate interface was developed for healthcare professionals to:

• Consult generated Excel files

• Download reports for analysis

• Access results without needing Girder expertise

3. Data Analysis Findings
The analysis of collected data revealed important insights:

• SFDFM Score Distribution: Most scores concentrated between 5-15 (moderate frailty)

• Fried Score Correlation: Strong correlation (r = 0.70) with SFDFM scores

• Care Setting Impact: Hospitalized patients showed higher frailty scores

• Gender Differences: Female patients had slightly higher frailty scores

• Age Factor: Moderate correlation (r = 0.38) between age and frailty

4. Modeling Results
A multiple linear regression model was developed to predict SFDFM scores:

• Explained 54% of score variance (R² test = 0.54)

• Confirmed Fried score as the main predictor

• Identified hospital care and female sex as significant factors
  

Challenges and Solutions

1. Girder API Authentication: Initially faced permission issues which were resolved by adjusting access rights and modifying Python request functions.

2. Multi-device Compatibility: Optimized the interface for tablets and smartphones while maintaining full computer functionality.

3. Data Cleaning: Standardized patient identifiers and handled missing values to ensure robust analysis.

Results and Impact

The project successfully delivered:

• A fully functional digital screening tool deployed at: https://depistage-fragilite-motrice-1363f3377112.herokuapp.com/

• Validated the SFDFM as a reliable clinical indicator

• Identified key factors influencing motor frailty

• Provided clinicians with tools for early detection and intervention

Future Perspectives

The work opens several promising avenues:

• National Data Collection: Extend beyond Paris to build a more representative database

• Longitudinal Studies: Track frailty progression over time

• Algorithm Improvements: Enhance predictive models with more data

• Clinical Integration: Wider deployment in physiotherapy practices

Conclusion

This project created an innovative, GDPR-compliant digital application for motor frailty screening that addresses the needs of healthcare professionals. By combining technical development with clinical data analysis, it provides both a practical tool and valuable insights into motor frailty detection and management.

The tool is now operational and ready for broader clinical deployment, offering significant potential to improve early detection and care for older adults at risk of motor frailty.