HIV Patient Risk Management: The Use of Data Visualization and AI Prediction
#HCI #Data Visualization #Chornic Disease #Health informatics
1.Project
Collaborating with Drexel Medical school, we designed and developed a data visualization tool for HIV clinic professionals to faciliate the HIV risk assessment process. We expected this work will be used in real HIV clinical settings to achieve the ultimate goals, facilitate diagnosis process, improve quality of care delivered, reduce the barriers to HIV care and cost, and increase research participation, new technologies, such as big data and mobile health applications have a potential to address these programmatic needs.
My Role
User Researcher, UX Designer, Developer
Duration
10 Weeks
2. PROBLEM
Larg number of HIV patients with low rentation rate in care.
There is a large number of new HIV diagnoses in the US based on CDC’s report in 2018. An estimated 1.2 million Americans had HIV at the end of 2018. But only 65% of them received some HIV care, and 50% were retained in care. The newest National Strategic Plan A Roadmap the End the Epidemic for the United States (2021-2025), has listed the key focus areas on “Diagnoses people with HIV as early as possible and promptly like them to care the treatment” and “Support all people with HIV to achieve and maintain viral suppression.”
"Ultimately, what we are looking for is a method to visualize where our patients fall on the continuum of care so that we can more readily know how many patients are at risk, identify them, and adopt strategies for retention." —— Dr. Aloff, HIV Professionals in Drexel
3. Design Solutions
In our design, we have three main sections to achieve the clinican's needs:
(1) Visualization Strategies. This part includes graph design, interactions design, and data presentation in order to present the information in an intutive way.
(2) Dashboard Design. This part is related to create a functional dashboard that allow clincians to upload data, read data and download analysis results.
(3) AI Prediction Model. In this part, we exaimined the factors that contribute to HIV risk assessment, and we developed a AI that can predict patient's risk in 3 or 6 months.
Tools we used
Prototype: Miro, Figma
Development: React Library, JavaScript Library - Vega Lite API, Python
3.1 VISUALIZATION & INTERACTION DESIGN
Scatter Plot Chart shows a group patient's risk levels at one time.
(1) I use different colors to label people who are in different risk groups (Low, Middle, and High).
(2) The shape of dot presents patient's care status.
( ● = in care, ✕ = out of care)
(3) Exact risk factor can be seleted by HIV professionals through the operation panel on the left.
Bar chart (left) shows the number of patients in each risk group.
Bar chart (right) shows the number of patients who are out of care in each risk group.
Below: Some Early Ideas of Visualization Strategies for Presenting HIV Risk Factors
3.2 INTERACTIONS
1. Individual Prediction Feature
Based on individual's visiting history, CD4 Cell, Viral Load, and other risk factors, our model will show the possible negative outcomes of patients. The factors which contribute to the prediction results are shown on the bottom to make the AI more explainable and transparency.
2. Filter feature to select the patient group based on Clinican's needs.
3. Select Range.
Bar charts will response to user's operations on the main chart (scatter plot chart)
I used Viga-Lite-API to do the initial development on visualization. Vega-Lite is a high-level grammar for visual analysis that generates complete Vega specifications. It has all features that we need for the application. CHEK MY CODES HERE