Jefferson Health
Maximizing the impact of AI in healthcare
I led a lean team of designers and data analysts to drive increased traffic to the Ada AI-powered symptom checker tool on Jefferson Health’s site. By introducing targeted CTAs at key points in the user journey, we were able to capture organic traffic and drive a 169% increase to daily average page views of the symptom checker, increasing the value of the tool for both patients and the business.
Problem:
Jefferson Health invested in the Ada AI-powered symptom assessment for their site to help route patients to the correct points of care and manage volume for providers. They needed to organically drive more traffic to the symptom checker to begin realizing these benefits without excessive marketing spend.
Evaluate User Analytics
In reviewing search behaviors, navigation patterns, and user interactions with the AI symptom checker, we learned that:
Organic search engine queries for symptoms were currently driving to articles in Jefferson Health’s content hub, and to condition detail pages.
Usage patterns indicated that patients were using the AI symptom checker as a self-triage tool, to determine what level of care they needed, especially after hours.
Optimize Flows for Conversion
Instead of anchoring the experience in a homepage-first framework, we chose as our entry points the same searches that were driving traffic to the site already. Then, we identified the most likely path to the AI symptom checker based on observed clicks.
Some pages showed a large plurality of traffic from within the site. For these, we pulled additional analytics to identify the internal sources of page traffic, and mapped our experience accordingly. We then added CTAs into the journey at the points they would offer the greatest value to the user.
Rapid Prototype Testing
To get rapid results, we iteratively launched two wireframe prototypes representing different key user flows selected by the Jefferson Health team. Test objectives included answering questions like:
Is the user motivated to use the Symptom Checker, and if so, when?
What factors make the user more or less likely to use the Symptom Checker?
Does the Symptom Checker enhance or detract from the user’s experience?
Insights & Analysis
Through this rapid prototype testing, we learned that users preferred trying the AI symptom checker over reading copy about their symptoms, that they believed the AI symptom checker would save them time and money, and that the experience would be personalized.
In particular, participants emphasized how reassuring they found the experience when they felt anxiety about their symptoms.
“Introducing call-outs for the AI symptom checker in the places we’ve identified will lead to increased usage of the symptom checker.”
— Design team hypothesis at delivery
Results
With a great deal of confidence in our final designs, we invited the client to measure specific metrics against our hypothesis. Once the changes had been live for 11 weeks, we compared traffic data to the 11 weeks preceding the launch, and found:
23% increase in completed assessments
169% increase in daily average page views
New record set for assessments completed in a day
Page referral data clearly demonstrated that the new traffic was coming from our new CTAs, validating our targeted strategy.
Finally, we could also see that diagnoses consistent with searched symptoms were being suggested at a higher rate in completed assessments, indicating the AI Symptom Checker was now meeting the users’ immediate needs.
-
Atlantic Health Digital Transformation
-
Amwell Telemedicine Provider Dashboard
-
Optum for Business Experience