What we’ve learned from 500,000 online symptom checkups

We collect tons of anonymous data every day. Storing information is easy, but what can we learn out of it? Keep reading to find out.

According to the Pew Internet Project, 72% of Americans search online for health information. Of these, 77% say they begin with a search engine such as Google, Bing, or Yahoo. As we all know, improper use of online health information may lead to problems such as anxiety, overuse of emergency departments (56% of all ER visits are avoidable) and misinformed decisions.

To help patients better assess their symptoms, we’ve developed a new symptom checker called Symptomate. Its goal is to provide more objective and relevant health information than what you can get from “Dr. Google”.

In this post we will explore who uses Symptomate and see what insightful information can be extracted from this data. The analysis was based on 0.5M anonymous user interactions (called checkups) with Symptomate (Android mobile app) and Doktor Medi (Polish language version).

Gender

Women use our symptom checker more often than men, 68% vs 32%. It might be partially caused by the default symptomate settings. However, this is not the only contributing factor, as we observe that symptoms related only to female occur more often than those related only to male.

Graph – Number of checkups by gender

There is also a less expected gender difference. We observed that women on average reported 9 symptoms, while men just 8. There seems to be no medical reason for this, so we can only speculate that men are less eager to share their symptoms than women.

Graph – Number of reported symptoms by gender

Naturally, we also observed a difference in the reported height and weight of female and male users. Just as before, one can spot the default values in our app. However most of the users change default values and provide their true height and weight.

Graph – Weight and height of the users

Age groups

We found that 85% of Symptomate users are under 30, which is not surprising as younger people are more likely to use mobile apps and on average spend more time doing that.

Graph – Number of checkups by age

Looking into symptoms reported by each age group gives us interesting insights. In the infographic below, we see how more (less) likely are given age groups to report certain symptoms. We selected few, that differentiate most between age groups. For example, people under 30 are more likely to have sex and less likely to experience reduced skin elasticity.

Graph – Likelihood of symptoms for a given age group

One might speculate whether hypersensitivity to sound might be related to having young children. Edema for people over 45 also makes sense. According to Mayo Clinic, among edema risk factors are drugs for hypertension and diabetes. Both of those conditions are related to age. People over 60 often report sunken eyeballs, which is again in accordance with common sense.

Risk factors

Our symptom checker allows the user to submit information about a few key risk factors, such as smoking, diabetes, high cholesterol and hypertension. Below we present how users responded to these questions.

Graph – Risk factors reported by users

Reported symptoms

The chart below shows the most frequently reported symptoms. Abdominal pain, nausea and various headache-related problems seem to be the primary issues of symptom checker users.

Graph – Most frequently reported symptoms

Commonly reported symptoms also vary depending on the age group. Young people seem more likely to report acute symptoms, while older users tend to focus more on chronic symptoms.

Co-occurrence of symptoms

Now let’s investigate which symptoms are related. To do this we’ll use something called association rules.

Let us briefly explain two terms that occur in the following visualization. For every pair of symptoms we check how often they are reported together. This value is called “support”. But this information is not enough, as symptoms may co-occur by pure chance. That’s why we report a second value, called “lift”, which represents the strength of the relation between the symptoms. Simply put, the bigger the lift, the stronger the connection between symptoms. The larger the support, the more users experience both symptoms.

With the interactive tool below you can play around with the graph and explore some of the most common patterns we identified.

Checkup duration

75% of checkups took less than 150 seconds. This is possible thanks to our intelligent algorithm, which optimizes the number of questions and only asks questions relevant to the reported problems.

Graph – Checkup duration

Time of a year

Finally, we compared the number of daily checkups with the intensity of Google searches for the term “sick”. While this is not a very precise measure, there seems to be a trend, as we observed more checkups during the spring and autumn.

Graph – Number of checkups daily

Conclusions

In this article we presented an analysis of a subset of patient checkups provided by Symptomate. Remember that the profile of an average user may vary depending on how a symptom checker is offered (mobile app, website, or a chatbot like the one we’ve recently presented).

We believe that a symptom checker integrated with existing channels such as patient portals, registration systems, hospital web sites or existing mobile apps would attract users of more uniform age groups.

We hope you’ve found this information useful!