When Health Technology Fails Us

The Line Between Being Well Informed and Becoming a Cyberchondriac

When Health Technology Fails Us

According to the Pew Research Centre, over a third of Americans use the internet when they believe they have a health issue. Their search results, however, are not always followed up with a visit to a doctor. Online self-diagnosing is becoming routine for Internet users who are increasingly aware of the vast amount of available online health resources and want to feel in control of their bodies and well-being. Instead of waiting for an appointment, having to discuss their symptoms with a doctor and occasionally pleading for additional diagnostic tests, potential patients now perform extensive searches of the Web and juxtapose different diagnoses with their symptoms until they discover the one that seems to fit best.

The Internet makes health-related information almost universally accessible. It helps educate people about their health and enables them to make informed decisions about their treatment options. There are examples of people diagnosing themselves correctly after years of misdiagnosis. A recent example is the unfortunate story of Bronte Doyne. Bronte was told by her doctors to stop self-diagnosing and ultimately died of a condition she had identified, but a condition that went unnoticed by the physicians treating her until it was too late.

On the other hand, Googling your medical symptoms does not necessarily end in a resolution and can in many cases bring out unnecessary anxieties, transforming former hypochondriacs into present-day cyberchondriacs. Some can even get addicted to constantly searching for health information online, examining themselves and looking for reassurance, as well as demanding tests and screenings that might not be appropriate.

Escalation of innocuous symptoms

Common symptomatology can prompt some users to start exploring rare and serious conditions that came up during their online searches. A large-scale survey completed in 2008 showed that Web search engines have the potential to escalate medical concerns of people who have little or no medical training. The study showed that escalation was influenced by the amount and distribution of medical content viewed by users, the use of alarming terminology on the sites they visited and the person’s predisposition to becoming anxious. In contrast, there are some people who can indeed diagnose themselves correctly, especially if what they are experiencing is very specific and atypical. For instance, in cases like Bronte’s, an outlier can sometimes get ignored or overlooked and treated by the medical team as a common medical condition when it is not.

However, health information found online is often incorrect or incomplete. When evaluating 23 symptom checkers for their diagnostic and triage accuracy, researchers from Harvard Medical School found some worrying deficits. Only a third (34 percent) managed to get the diagnosis right the first time, and just over half (57 percent) provided correct triage advice (e.g. recommended emergent or non-emergent care). Also, according to Mathew Chung of the University of South Carolina School of Medicine, the internet often provides recommendations that are not necessarily in line with up-to-date medical advice. Chung studied online recommendations for safe infant sleep. He found that out of the 1,300 websites, less than half (43.5 percent) provided accurate information on this health topic.

How to improve online symptom checkers?

When millions of users look for health information online, this creates a big pool of data. Researchers are now tapping into these datasets to test predictive algorithms that could make online symptom checkers better. The latest developments in machine learning are assisting their efforts to find patterns in online searches and diagnose a condition earlier. Doctoral student John Paparrizos teamed with Eric Horvitz and Ryen White, the authors of the 2008 report on cyberchondria, to design an algorithm that could identify people recently diagnosed with pancreatic cancer by looking at their previous online searches. Their study showed that a serious diagnosis could potentially be predicted by examining a person’s online queries. With an improved system of online tools, patients might be detected before it gets too late to treat them.

Preventing diagnostic mistakes

Clinical decision support systems (CDSSs) are interactive applications that can now help health-care workers make evidence-based decisions and can even predict treatment outcomes. Partially a response to the critique that physicians frequently misdiagnose, over or under-treat, and/or fail to refer to other medical specialties, CDSSs are considered a major form of artificial intelligence in medicine and are expected to become even more efficient and viable as we fully enter the digital revolution in health care.

CDSSs are increasingly used in triage, screening, risk assessment, diagnosing, treatment evaluation and monitoring. CDSSs can also be linked to patient data from electronic health records.

The preferred models of CDSSs rely on multiple sources of data such as genetic, clinical and socio-demographic information. CDSSs are a part of the so-called ‘personalized medicine’ movement that is not population-based, but instead focused around pharmacology and interventions tailored to an individual. A study led by Dr. Peter Elkin, who directs Mount Sinai’s Center for Biomedical Informatics, suggested that CDSSs can broaden the scope of differential diagnosis, which would make the correct diagnosis more likely, shorten hospital stays, save lives and provide economic value to both to the patient and the provider.

Widespread adoption of CDSSs has not occurred yet in routine practice, but many experts believe that such tools could help overcome idiosyncrasies that exist in health care today. Also, the value of CDSS is increasingly recognized in combination with electronic health records (EHR). This type of health technology could bridge the gap between theory and practice that often influences the diagnostic process and leaves patients dissatisfied. Patients and clinicians alike need to get familiar with the opportunities health technology affords us, while not losing site of the inherent challenges that come with technological disruption. As these tools evolve, the hope is users will be better equipped to make healthier, well-informed decisions about their own care and treatment options.

Chung, M., Oden, R. P., Joyner, B. L., Sims, A., & Moon, R. Y. (2012). Original Article: Safe Infant Sleep Recommendations on the Internet: Let's Google It. The Journal of Pediatrics, 161: 1080-1084

Paparrizos J, White R, Horvitz E. Screening for pancreatic adenocarcinoma using signals from web search logs: Feasibility study and results. Journal of Oncology Practice, 2016;12(8):737-744

White R, Horvitz E. Cyberchondria studies of the escalation of medical concerns in web search. ACM Transactions on Information Systems, 2009;(4):23

Semigran H, Mehrotra A, Linder J, Gidengil C. Evaluation of symptom checkers for self diagnosis and triage: Audit study, 2015;351

Was this page helpful?

Article Sources

  • Sources