Three Areas Where Health Information Technology Needs to Get its House in Order

Health reform is taking off, thanks to pressure from insurers, the promise with which innovative technologies tease us for low-cost treatments, and regulatory mandates dating back to the HITECH act of 2009. Recent hopeful signs for wider adoption of health technologies include FDA forebearance from regulating consumer health apps, calls for more support for telemedicine, and new health announcements from tech giants such as Apple and Google.

While technologists push forward in all these areas, we need to keep in mind that several big unsolved problems remain. Let's not get lost in the details--these major issues have to be tackled head on.

Health analytics to support smarter payment

Doesn't everyone love pay-for-value? It's universally understood to be a long overdue driver for improving the health care system. Paying doctors for each procedure is like paying policemen for handing out citations. Good police prevent crime rather than punish it--although they are prepared to respond to crimes as necessary--and good clinicians treat health problems the same way.

But without big data, pay-for-value is big trouble. How do we know whether a doctor offers efficient treatment for a patient with diabetes and heart disease? Suppose the patient also suffers from obesity and knee problems that make exercise difficult and raise costs? Or is depressed and can't stick to a treatment plan? In short, pay-for-value forces us to do patient stratification--to decide how much each patient costs.

Data is not yet collected in granular enough form, and with enough accuracy and consistency, to support stratification. Partly due to system design and partly due to organizational practice, doctors tend to restrict structured information to the miminum required for billing and public health reporting, and load their clinical insights into unstructured content. We can't even measure one of the most basic health metrics accurately: the rate of adverse effects in hospitals. The difficulty of assuring the accuracy and consistency of data goes beyond traditional EHRs; it extends also to consumer platforms such as the new Apple HealthKit and Google Fit.

Analytics are also too costly for most institutions, as I pointed out earlier this year. Recent news confirms that hospitals are holding off buying the very analytical tools that could improve patient care, overwhelmed as they are by other IT requirements like the benighted move to ICD-10.

Finally, data sharing runs into organizational barriers while raising ethical questions. Although good deidentifaction has proven resistant to breaches, poorly done deidentifaction is still common.

Reformers are moving from the current regime that places few restrictions on the selling of patient data--with dubious reliance on the promise of deidentification--to a future where patients decide who receives their data and agree to take on a reasonable degree of risk.  Researchers harbor understandable fears that patients will withhold their data under such a system--but it could equally well happen that patients open their hearts and their records to institutions they see as beneficial, particularly if researchers learn better how to engage the subjects of their work.

I've run into lots of developers working busily on systems handing control to patients and allowing them to selectively release data, perhaps for rewards. Two such systems that showcased at the recent Health Datapalooza were Medyear, which I highlighted elsewhere, and the Self-Generated Health Information exchange (SGHIx) prototyped by UNC Chapel Hill in partnership with RTI International. SGHIx is particularly interesting for its detailed sharing options: each type of data you collect can be withheld totally from researchers, shared for sums and averages only, released in deidentified form, or given upon request (perhaps in exchange for some benefit).

Record systems that doctors like

Knowledge workers in most fields are constantly seeking out new digital tools to streamline and augment their work. In health care, progress is more nuanced. Doctors load apps they like onto smart phones, but these don't have a wider impact on workflows or communications within the organization. Electronic records are almost universally hated. The health IT industry can't achieve its promise of altering behavior and lowering costs until it provides more useful tools and makes progress in addressing clinicians' complaints.

Admittedly, health care runs into problems that don't concern most industries when seeking technological innovation (although similar problems haven't held back the finance and defense industries). Errors can lead to deaths. Privacy concerns are heightened. (What does Google really learn about you from Google Docs, anyway?) And large organizations have many competing goals, along with a diversity of departments and actors.

I won't catalog the many criticisms of interfaces and behavior in EHRs (some of them are described in my health IT report and the complaints can vary based on the clinicians' general comfort and familiarity with computers, but reports suggest that all successful deployments spent considerable time getting clinicians to evaluate and approve the systems.

One company I have profiled, Modernizing Medicine, claims to have shaken off the detritus of older EHR systems. Using modern tablet-based data entry tools and ways of minimizing keystrokes--similar to the ones a search engine uses to autofill your search box--they claim to make the entry of structured data as easy as writing on paper. The rub is that they serve just a half dozen disciplines whose conditions and data needs are fairly circumscribed (ophthalmology, plastic surgery, dermatology, etc.). I have discussed with them how their model would be hard to extend to more complex areas of

There's no doubt that the application of modern user experience (UX) practices, or applying human factors engineering (HFE) to health systems, would improve electronic health records.

But the transformation of health care also requires something much bigger: new ways for team members to coordinate with each other. Technologists can't force institutions to make the needed changes, but they can be sensitive to creating the software and hardware that have to work in the new environments. They can also provide tools for exploring and choosing such environments.

Device validation

Patient monitoring could radically change the health equation. Wearing devices or placing sensors in their homes, as facilitated by companies such as Essence Smart Care,could keep people out of nursing homes, trigger warnings before expensive hospitalizations occur, and provide minute-by-minute data far more accurate than what clinicians can capture by interviewing patients. Some people expect patient communities--sucn as those who gather at PatientsLikeMe--to unveil a whole new style of medical treatment outside, or in parallel with, conventional office visits.

But monitoring rests on a long string of dependencies. The patients will (we hope) agree to if the doctors order it. But doctors need demonstrations of effectiveness before prescribing the use of devices, and so do payers, who can cover the costs if they feel confident the devices will save money. A lot rests on validation of devices--but the current gold standard of clinical trials is being found deficient. They're expensive, take a long time, use subjects who are unrepresentative of the general population, and (as we've discovered more and more) are often unreproduceable.

Can big data be used to validate sensors and devices, reducing the costs and time for validation? I raised this question in an earlier article. It is also possible that researchers could use clinical trials to validate a general practice--such as using text messages to send reminders to patients concerning their eating or exercise--and that payers could then entrust the details to the doctor and patient. Each could then choose the fitness device that's fits for him or her.

Tackling the big problems

Both technology and policy need to recognize the difficult problems that hold technology back from providing health solutions. A few approaches include:

  • Making electronic records better at capturing information for treatment, in addition to billing.
  • Following clinicians to determine how they are likely to use data, so that interfaces (whether on laptops, mobile devices, Google Glass, or operating room screens) can display what the clinician needs most at the moment, and facilitate data entry.
  • Providing flexible interfaces that clinicians can easily customize without hiring expensive consultants. Open source software can help. The best modern practice is to create separate software modules for data storage, data transmission, and interfaces, allowing third parties to create new interfaces or transmission mechanisms and giving users a choice.
  • Mimicking the oft-cited example of the aviation industry, which dramatically increased its safety through information sharing about accidents. Cybersecurity is taking a similar path. This always requires some tough policies, like an open database where users report all errors, not the Health IT Safety Center that ONC reviewers have already decided would be voluntary and unspecific.
  • Providing patient control over records, which not only allows them to become partners in their own care, but lets them correct errors and share data where they feel it appropriate. Apple HealthKit and Google Fit may be major steps forward, particularly because the availability of data through their APIs should prevent them from turning into new silos.
  • Creating new protocols for researchers to reach out to patients and request rich data sets--perhaps without deidentification, which reduces data accuracy--for use in seeking cures. Patients may well give their consent if licenses and acceptable use policies are in place.
  • Defining the fields in electronic records more clearly, so that staff always enters information of a certain type into a single correct field.
  • Creating derived fields from the data entered by clinicians, so that they don't need to re-enter data just to meet some reporting requirement.

While banging out the next great app, health care developers need to consider the big problems and the foundations for solutions. The result may be wider adoption of apps, and a contribution to making a real change in the field.



The Open Human Project

Andy, great article as usual. We just published an article on the Open Human project by Shauna Gordon-McKeon and titled Some Patients Are Eager to Share Their Personal Health Data. The article focuses on the Open Human project. This project is new to me and it should address some of the challenges that you raise. It turns out that they have recruited thousands to people to donate their personal medical/health data for research, including their DNA. The program requires "an entrance exam showing that they understand what will happen to their data, and the risks involved.” According to the article, “most participants are enthusiastic about sharing. One participant described it as 'donating my body to science, but I don't have to die first.’”