An unfathomably complex entity such as a health system grows over time like a city. Right now, communications and data usage in the US healthcare system is a bit like a medieval town, with new streets and squares popping up in unpredictable places and no clear paths between them. Growth in health information has accelerated tremendously over the past few years with the popularity of big data generally, and we are still erecting structures wherever seems convenient, without building codes.
In some cities, as growth reaches the breaking point, commissioners step in. Neighborhoods are razed, conduits are laid in the ground for electricity and plumbing, and magnificent new palaces take the place of the old slums. But our health infomation system lacks its Baron Haussmann. The only force that could seize that role–the Office ofthe National Coordinator–has been slow to impose order, even as it funds the creation of open standards. Today, however, we celebrate growth and imagine a future of ordered data.
The health data forum that started today (Health Datapalooza IV) celebrated all the achievements across government and industry in creating, using, and sharing health data.
Useful data, but not always usable
I came here asking two essential questions of people I met: “What data sources do you find most useful now?” and “What data is missing that you wish you had?” The answer to first can be found at a wonderful Health Data All-Stars site maintained by the Health Data Consortium,which is running the palooza.
The choices on this site include a lot of data from the Department of Health and Human Services, also available on their ground-breaking HealthData.gov site, but also a number of data sets from other places. The advantage of the All-Stars site is that it features just a few (fifty) sites that got high marks from a survey conducted among a wide range of data users, including government agencies, research facilities, and health care advocates.
Older forms of data resemble Haussmann’s Paris in being organized to some degree. Thus, KeonaHealth can reap the benefits of standards in the reporting of symptoms, diagnoses, medications, test results, and other well-understood medical terms over time. Their system helps triage patients over mobile devices, providing potentially life-saving information beforethe patient gets to a health center. To achieve this, they record the patients’ health conditions and tap into a learning system developed by the medical industry to do triaging.
On the other hand, a newer technology like Blue Button is less standardized, even though it’s an open format and dozens of organizations want to jump on its bandwagon. Humetrix, which has developed a mobile app to securely transmit and display patient data stored in Blue Button, found that every organization stores the data somewhat differently.
Some use an old plain-text format based on the original human-readable format developed by the Department of Veterans Affairs. Others use an XML format, but (like the much-cloned HL7 formats used throughout thehealth care field) they use it inconsistently, putting the same data in different tags. Some implementors use text for diagnoses (which leads to inconsistencies) while others use codes, but those codes may be from different standards. Humetrix has to attack these differing formats one by one. It provides translation between formats, Direct addresses for doctors and patients, and encryption to protect data as it travels over the network.
A company called Aware offers a system called Image Direct that provides acccess to medical images. Like Humetrix, it uses the Direct standard and Blue Button. Image Direct simply adds a field to Blue Button where a URL to an image can be stored. The system can be run on the health care provider’s site oroffered as a cloud service.
National Coordinator Farzad Mostashari announced a BlueBotton+ Co-Design Challenge today. Go to the site this week to submit ideas of apps that developers can create to use Blue Button data.
Open data and private data
Data, like pleasures, has greater value when it’s shared. A lack of standards restricts data to single institutions. The successes many institutions report, using just data they maintain privately, shows the incredible value of data (and the groaning inefficiencies of most current health care). One data effort, the Heritage Health
Network Prize Competition run by the Heritage Provider Network, made it into the morning keynotes today.
The Heritage Provider Network is a kind of managed care institution that has been acting like an Accountable Care Organization long before the term became popular. With about
150,000 patients, they promise to take care of the whole patient and make their money by keeping the patient healthy instead of charging for more care. They do such things as sending doctors on home visits and providing care managers to patients, all with the goal of reducing the patient’s need for expensive care.
They have even taken patients on bus trips to popular tourist destinations–partly to overcome the social isolation that degraded health, but also to create a captive audience for a talk from a dietician.
This year, they ran a challenge to find a way to predict patient hospital admissions. Because this proved to be too narrow a question (one that might be solved by luck), they extended the challenge to predict not only the admission but the length of stay. The winner, a team of data scientists called Powerdot, took home a $500,000 prize.
Jonathan Gluck of HPN told me that the input data to the challenge was nothing unusual for the health care system to collect. They do not have sensors tracking patient movements or checking on their behavior around the house, although they are looking forward to trying such things. Right now, the data they have for crunching include typical claims data, prescription orders, and labs. They do, however, have types of data that many providers lack, because HPN is both a payer and a provider. For a second challenge, they will anonymize and provide contestants with even more of this data.
What will they do with the results of the challenge? They will contact doctors to warn them that particular patients seem to be at risk. The doctors can then contact the patient to request an office visit, change medication, or urge behavior change.
Gluck says, “Behavior change is really hard, but if you tell a patient he’s headed for a hospital admission, he may say, gee, maybe it’s time to make a change.”
Large providers can streamline their systems through data use. The benefits of data use is one of the factors driving consolidation. We could extend similar benefits to all providers by increasing the number of publicly available data sets, such as those from HHS, and by adopting standards. Quality control is also critical.
Health care programmer Fred Trotter (author of the O’Reilly book Hacking Healthcare) alerted me to a recent bombshell in the health care industry that may radically increase the data that the public can use to judge doctor quality. On Friday, a federal judge lifted a ban going back to 1979 that kept Medicare claim information secret. Now, if the ruling stands, CMS will have to release enormous amounts of data that can be crunched and presented through apps and web sites to help patients choose their doctors.
But data is also unfortunately going in the opposite direction. Many public data sets get sucked into private apps or services run on private web sites, so that no one can build on the enhanced processing added by the developers.
I think too many developers harbor fantasies of making big bucks by providing a health care app that everybody feels they need touse. They may actually do better financially, as well as promoting better health systems, by creating something open. That seems to bethe approach taken by Microsoft, which is constantly developing opensource tools and partnering with organizations to provide access to HealthVault. This brings me back to the point I made at the beginning of the article about the complexity of health. Microsoft seems to realize that collaboration is the only way forward, and that there will be no Microsoft in health care–no proprietary winner that can become mainstream.
Andy Oram is an editor at O’Reilly Media, a highly respected book publisher and technology information provider. His work for O’Reilly includes the influential 2001 title Peer-to-Peer, the 2005 ground-breaking book Running Linux, and the 2007 best-seller Beautiful Code.