A group of scientists from the Harvard Medical School tested and assessed the accuracy of 23 online symptom checkers that people usually use to learn whether they may have developed a certain disease or condition without paying their GP a visit.
The team found that the online tools provided an accurate diagnosis in only 34 percent of the cases, and after three attempts in only 51 percent of cases. The online symptom checkers included those run by top sites including the Mayo Clinic, WebMD, and the American Academy of Pediatrics.
Researchers concluded that most of those tools regardless of the site they were on were unbelievably inaccurate. Additionally, diagnosis and related advice varied a great deal from one site to another.
Scientists noted that the online diagnosis tools are “frequently wrong” so users should be extra cautious when taking for granted a piece of medical advice. Yet people are tempted to use online symptom checkers because they are very simple to use, interactive, and provide a quick diagnosis that can lower their anxiety.
You only need to type in your pain, aches, or rashes to learn what type of disease you may have. Most online diagnosis tools use some extra questions before providing a final diagnosis on a disease or condition. But the majority doesn’t stick to a single diagnosis. Instead they provide a list of diagnoses with the most likely disease on top of the list.
A few years ago, these tools didn’t even use algorithms to provide a diagnosis. They merely looked up for key words in their database. Yet, hopefully today the symptom checkers use complex algorithms based on either branching or Bayesian inference, which are supposed to be a lot more accurate.
Scientists published their findings in the BMJ. They wrote in their paper that they asked 45 volunteers to run diagnosis checks by providing their symptoms on various sites. Fifteen of the participants required a visit to the ER, 15 didn’t require ER care but they needed the advice of a doctor, while in 15 cases self care was sufficient.
Researchers awarded the highest score to sites that were able to provide the accurate diagnosis on top of the list of potential diagnoses. Yet, very few sites managed to do that. Most of them were able to list the right diagnosis within the first three positions. These sites were tagged as “potentially useful” by researchers. Two of the sites provided up to 99 potential diagnoses to some specific symptoms. These two sites were deemed “unlikely to be useful.”
Study authors also learned that online diagnosis tools’ accuracy varied from 29 percent to 79 percent.
Image Source: Alamy