Fixing the background technologies on which health apps depend

Developers are flocking to health IT with the laudable goals of making a difference in people’s health (and earning some money in the process). The complex and balkanized field presents numerous barriers to entrepreneurs breaking into the space,   Here are a few of the dilemmas that health reformers face, and that bedevil efforts to provide useful apps and medical devices:

A lot of developers are offering apps or fitness devices. But without rigorous testing (double-blinded experiments) the FDA probably won’t let apps or devices be used for medical purposes, and insurers won’t cover them. Although people obsessed with good health are buying fitness devices and reporting that self-tracking is changing their lives, the people who need intervention the most (such as obese smokers with diabetes) aren’t widely using digital health products.

A full-press clinical research effort on every app or device would be absurd and would shut down innovation. But it’s not right for device manufacturers to bypass the FDA entirely, either. The FDA’s current position is reasonable, if vendors can understand it, but perhaps we can find a more modern approach to testing apps and devices. Crowdsourcing and data sharing may be the key to trusting apps and devices as well.

Lots of apps and devices come with web sites where patients can store and download their data, but the sites from different vendors don’t work together, and the doctors who work with those patients rarely accept the data. So the value for treatment is limited.

Several things must come together to achieve interoperability. APIs that offer windows into device data are a good start, but standards would be even more valuable—particularly to make sure the same vital sign is stored using the same format and units by different sites.

Next, electronic health records at the doctor’s office must open up to accept patient data. Stage 3 of Meaningful Use might address this.

All health reformers agree that improving quality and cutting costs requires health providers to collect and crunch data on patients.  Data crunching allows patient stratification (finding out who is most at risk), the reallocation of hospital resources where they are needed, and even uncovering new treatments.

Great idea—except who will do the data analysis? This is rocket science. Applying statistics casually to inappropriate data sets can yield execrable results. And the experts in statistics and data science who can carry out the crucial analysis are rare. Competition for them is fierce across all industries—health providers have to get in line with banks, retailers, governments, and everybody else looking for big data solutions.

Solutions will involve combining data, which is one reason so many health providers are merging. The popularity of Accountable Care Organizations (ACOs) also springs partly from the promise of data analysis—but forming an ACO or conglomerate doesn’t automatically make you a data guru.

For instance, institutions may find that lots of patient data has been entered incorrectly or inconsistently. Data entered for billing purposes may be misleading when used for clinical purposes. Maybe there’s a way out — statistical methods exist to correct for this kind of stuff too. But it takes deep expertise to extract the nuggets of wheat from the chaff.

Yes, there is a future for innovative developers in health IT. But we all have to work together to create this future. A just-released report I wrote goes into more depth on these topics, and hopefully we will all come to a fuller understanding of what’s needed over the next year or two. I’d like to hear from readers and see alternative viewpoints on the points in the report. Please comment or, to reach me privately, please write to infofix alias @ the domain.

Andy Oram is an editor at O’Reilly Media. An employee of the company since 1992, Andy currently specializes in open source technologies and software engineering. His work for O’Reilly includes the first books ever released by a U.S. publisher on Linux, the 2001 title Peer-to-Peer, and the 2007 best-seller Beautiful Code.