Trigger warning: electronic medical records, blockchain, machine learning, AI, Microsoft Bob
Epic, Cerner, et al. should do to EMR what Apple will soon do to iTunes: break it up. What was once a Mac-only MP3 player with a killer feature — the store — had become a multi-platform mess of music, TV shows, movies, home videos, podcasts, apps, device syncing, streaming, and radio. Most of these features will soon become separate Mac apps.
To be clear, the dominant electronic medical records of today are much worse than iTunes ever was. They are worse than even Microsoft Bob — a misshapen chimera that tried to be a word processor, spreadsheet, presentation software, email client, calendar, AND a file system all at once — Clippy’s fever dream of becoming an operating system. EMRs are worse than that because they kill people: yes, metaphorically, as in the dead souls of nurses clicking through their ninetieth patient assessment of the day, but also physically, as in I can’t figure out what’s going on with this patient who has been seeing subspecialists from four different practices and admitted to three different hospital systems in the last year.
Step 1: modularity
Healthcare apps are already half-broken into two: EMRs and practice management software. But a half-done job is in this case worse than nothing at all — yes, it’s easy to decide that order entry, documentation, and result review all go into EMR and that scheduling should probably fall under practice management, but what about billing? And shouldn’t a patient portal include data from both? It’s as clear an example of a confusopoly as any.
The first instinct then is to lump everything together, which is what is in effect happening as purchases decide to get both pieces from the same vender for better interoperability. So let’s say the Big Healthcare Conglomerate’s software of choice has stellar billing features: maybe its algorithm parses through progress notes to determine appropriate ICD10 and encounter complexity codes. But the algorithm needs the documentation to be done just so, and because of that there are only check boxes and drop-down menus with dozens of items that you have to scroll and click through — no free texting allowed. Why should we allow Conglomerate to force MDs (who, and this is important, aren’t employees but have admitting privileges in one of its hospitals) to use the inferior note entry system, thereby decreasing their own productivity and increasing burnout? Isn’t capitalism all about the freedom of choice?
So that’s one way to break up the beast: into modular apps — more modular than the current EMR/practice management separation — that would each deal with a distinct domain that has its own target audience.
Step 2: blockchain
But wait, won’t this just exacerbate the problems with interoperability? Isn’t that why everyone is buying complete packages from the same 2-3 companies in the first place?
Well, here’s a more important decoupling: software from data. No worrying then whether modules can talk amongst themselves — more power to them if they can, but the ultimate way to communicate would be by reading and adding to the data, stored in a universal format, using distributed ledgers/blockchain (some day I’ll know the difference) to track changes and smart contracts to manage patient-provider relationships. Individual patient data are a perfect fit for blockchain: keep everything anonymous, give the key to the patient, and have them be in charge (and also watch for emerging companies to help them sync, store, and analyze their data). Goodbye, HIPAA, goodbye, corporate compliance, welcome back, family physicians, sorry you had to call yourself “boutique” for a while.
Step 3: mandate
I am optimistic up to this point. There is already a concrete blockchain-in-healthcare proposal from MIT that’s a step in the right direction, and it’s written by people who unlike me actually understand blockchain technology. It’s also close enough geographically to the best state where this could be implemented in terms of size and perceived independence of its citizens — New Hampshire (Massachusetts itself isn’t a stranger to state-run experiments in healthcare but is probably too big to be the first to try it out).
The parenthetical “it” is a mandate for 1) a universal healthcare data format that’s 2) stored on a blockchain (or a more appropriate form of a distributed ledger) which all EMRs must support and which institutions and individual providers must use in addition to or instead of their existing data management solutions. Those who decide not to keep the data permanently but rather retrieve it as needed (either directly from the patient or from the patient’s storage service of choice) would be rewarded by decreased costs of security and compliance.
So I’ve started with modularity, but for modularity to become possible there needs to be a mandated universal format. It need not and should not be a federal government mandate — an individual state or a large country (San Francisco comes to mind) can try it out, and if it works well others will copy. It is also enough for a single populous area to require it for all major EMR players to be forced to accommodate it; so, if the experiment is a success the price for any subsequent state to join will be much lower. Conversely, if it fails (and the ways in which can fail can be a post on its own), the damage is limited to that one county or state. If the federal government is to be involved in any way it could be to offset some of the initial costs of the mandate, but federal government encouraging states to experiment in that way should be a general policy.
Step 4: profit
What’s in it for each party?
- It is all downside and no upside for current major EMR vendors — this is why we need government intervention in the first place. I don’t think anyone will shed a tear for Cerner and EPIC.
- New and smaller EMR vendors have only upsides from this, as it will be enough for them to excel in a single area, let’s say inpatient dashboards or billing extraction algorithms, and still be a player.
- Insurers have mostly upsides: imagine a patient uploading their (anonymized but coded) data to the insurer’s portal, an algorithm goes through the encounter data and determines appropriate reimbursement to the patient and providers, providers’ algorithm does the same and if they match within X% everything is resolved on the spot and humans come into the picture only if there are significant conflicts.
- Large swaths of coders and other billing and administrative personnel have mostly downsides as their jobs will become obsolete and taken over by AIs, but that is bound to happen anyway and is only tangentially related to this proposal. NB: humans would still needed to resolve conflicts between opposing algorithms.
- Healthcare providers gain the option not to store identifiable private data; most small ones will probably exercise it and this is a big upside, larger ones may or may not and it is at worst neutral to them, and at best academic medical centers will have more anonymized data to work with so there will be fewer regulatory hurdles for some types of research.
- Patients take complete and full ownership of their data. This is mostly good, but with great power comes great responsibility: some will keep it, some will sell it, some will lose it, some will pay a third party to handle it for them, I don’t know the proportion of each.
- In the selling and the handling comes a large upside, as this would open up completely new markets in healthcare data management and analysis, not to mention integration with personal health devices and other quantified self instruments of which I’m not a big fan but they exist and many people of means buy into them.
- Finally, states and counties that start the process early and are among the pioneers will be the ones setting future standards in case this works. They could also mandate that any company working on the citizens’ data should relocate completely or in part to that state. (Controlled and ethical) opening up of healthcare data would start a new gold rush, and California didn’t do too badly in those in the past.
As I wrote, there is some work done on the blockchain portion of this proposal, but why aren’t we full steam ahead on all fronts, especially the political? Isn’t this an obvious solution?