Beware of the order of symptoms

Using the internet to self-diagnose is becoming more and more common. Who hasn’t checked a symptom or two online? Virginia Kwan from Arizona State University and her colleagues from that and other institutions have found that people can come to very different conclusions about their health based on the way the online information is presented.  Specifically, the order of the listed symptoms seems to be critical.

If you search cancer information web sites like cancer.gov (the National Cancer Institute, NCI), you find a mix of symptoms, some mild and general, and some rare, severe or specific.  For example, among other symptoms for bladder cancer, the NCI includes finding blood in your urine and feeling an urgent need to empty your bladder.  How many people have never experienced the second symptom? Obviously the first symptom is much more specific and unusual.

Kwan and her colleagues gave students a checklist of symptoms about a fictitious type of thyroid cancer (dubbed isthmal cancer).  Each student was presented with six possible symptoms for the disease but in three different orders: general then specific, specific then general and alternating.  An example of a general symptom was shortness of breath whereas a specific symptom was pain in the throat or neck.  The students were asked to check off all the symptoms that applied to them, and based on their list, decide whether they might have the disease. I hope none of these students were already hypochondriacs, because if so, this study certainly didn’t do them any favors.

The students who received the alternating checklist were the least likely to think they had isthmal cancer.  The scientists hypothesized that the subjects with the other two checklists were more likely to check several items in a row (because the general symptoms were all clumped together). This in turn may have made them more likely to think they fit the pattern of having the illness. Interestingly, when the number of possible symptoms was increased to twelve, the order of the symptoms had little effect on whether people thought they were sick. Perhaps there were enough unchecked symptoms to make them realize they probably didn’t have cancer.

So how should online medical resources present their information?  According to the authors, that depends on the mission of the site.  If the goal is to get everyone in for a screening or vaccine, the website might want to make a short list of symptoms with the most general ones grouped together.  If, on the other hand, the goal is to discourage unnecessary medical interventions, websites should have long lists of symptoms or alternate general and specific symptoms.