Bridging the dichotomy of actual versus aspirational digital health

The future of digital health is bright. Enthusiasm for smartphone apps, virtual reality, artificial intelligence, machine learning and more in health care is no longer a niche interest, but rather mainstream fascination. From patient groups creating technology solutions for long‐term conditions1, to large technology companies like Google entering the digital health space2, digital health enables new voices and perspectives to partake in advancing health care. But with such broad enthusiasm and so many voices active in digital health, it can also be overwhelming to separate out actual from aspirational claims of progress.

Attempting to delineate digital health fact from fiction is somewhat futile, given that the border is constantly shifting with new discoveries. Rather, it is more productive to consider a spectrum of actual to aspirational digital health claims. Here we outline two simple factors to facilitate decisions about where claims lie on that spectrum: a) the appropriateness for the target population, and b) the incentives used to obtain any given results.

When evaluating any digital health claim, it is critical to consider if the population the technology claims to help matches the actual population that was studied. While on the surface this sounds trivial, in the age of Internet crowdsourcing research, it is a growing concern. One of the most difficult aspects of traditional clinical research is recruiting a sufficient number of patients to partake in a study. The Internet offers a potential solution, where an ad on Facebook or Craigslist can bring in hundreds of potential research subjects willing to fill out a survey or try a new health app3. But who are these online subjects, that are never actually even seen by the research team?

In mental health research, there has been a trend to offer simple screening tools, such as the Patient Health Questionnaire‐9 (PHQ‐9), never intended to diagnose, as diagnostic inclusion criteria4. Without verification or ruling out of other physical or mental conditions, it is increasingly easy to join an online mental health study even if someone does not actually have a mental health condition. A similar concern is digital health claims related to symptoms of an illness obtained from those likely without the illness. What does it mean to study individual symptoms of post‐traumatic stress disorder from those who may not meet the diagnostic criteria of having the actual condition5?

While there is tremendous value in crowdsourced and online‐based health research as well as in studying non‐traditional metrics or classifications of illness, it is critical to be cognizant of the differences from traditional research. In some cases these online methods may actually be superior to traditional face‐to‐face research. However, determining if these results are actually applicable to established definitions of illness is important to consider when deciding if these novel approaches can genuinely improve patient outcomes in a clinical setting. Novel approaches creating new definitions of illness or identifying new populations at risk of illness are equally important, but by their nature require further validation and are less actionable today.

Another factor to consider when evaluating any digital health claim is the role of incentives in obtaining that result. Again, on the surface this sounds trivial, but the complexities of the digital landscape add new challenges. Digital health research offers participants incentives to partake that can range from new smartphones, money for using of the device, extra coaching sessions, and more. But what happens when the digital health platform enters the real world, when the incentives disappear and when there is no external attention or interest in one's use of the technology? A recent study of an asthma‐monitoring app reported that, while over 49,000 people downloaded the study's app, only 175 (0.35%) were actively engaged with it at six months5. A successful study of an app for alcohol use disorder6 reported less success when later deploying without broad incentives7.

Thus, understanding the context and incentives used to generate a positive outcome from a novel digital health intervention is important in deciding if that intervention can realistically be implemented today, or if it is more aspirational, with new resources and efforts required to sustain engagement. That is not to say that outcomes from research projects which incentivize participants are invalid or redundant – rather, these findings provide valuable insights into what is necessary to make digital health work and how the health care system may have to evolve to support it. However, these factors must be considered when deciding what is immediately implementable, versus that which requires a supportive framework which has yet to be created.

All digital health research and claims are informative. Some offer immediate solutions to health care that should be implemented today and others highlight the potential of what may be possible. However, blurring the line between actual and aspirational can be counterproductive. Claiming that aspirational digital health research is ready for immediate use can lead to immediate negative results and broad disappointment. It may even inadvertently contribute to digital health “hype” and foster undue skepticism for the field.

However, ignoring digital health technologies with good evidence for real‐world implementation is a missed opportunity for improving patient outcomes. Appreciating how aspirational research can guide, inform, and inspire current efforts is also important. Likewise, appreciating the real world success of actualized efforts can help guide aspirational research to be more translatable into health care systems.

There is no superior designation, as both ends of the actual and aspirational spectrum have critical roles that cannot be separated. However, the value of both depends upon correct identification of where any given project lies on this spectrum – and further consideration of populations sampled and incentives used are critical to determining this.

John Torous1, Joseph Firth2
1Department of Psychiatry and Division of Clinical Informatics, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, MA, USA; 2NICM, School of Science and Health, Western Sydney University, Sydney, Australia