SMART on FHIR has had considerable early success in building an ecosystem of apps that can be plugged into a variety of EHR systems. It is lowering the barrier for innovation by making structured healthcare data available to third-party applications, and several major EHR vendors have implemented support for the protocol. But now informatics experts working on FHIR are tackling a more basic issue: how does a clinician know which app to use and when?
I had the chance today to speak with Josh Mandel, M.D., a research scientist in biomedical informatics at Harvard University and lead architect for SMART Health IT, about a new effort he is working on called CDS Hooks (CDS for clinical decision support.)
The SMART project has had a lot of success in getting different types of applications integrated into several EHR systems. One challenge, though, with integrating into clinician-facing EHRs is that in order to launch an app, the user has to decide to launch the app. “They have to know which app is going to be useful at which point in their workflow,” Mandel said. “If you want to run an app that is going to help you adjust the dosing of a drug based on a patient’s genotype, you might have to invoke that app while you are making a prescription. It would be much nicer if those services could run automatically, and the relevant advice presented to you only if and when it matters. CDS Hooks is an attempt to help clinicians know what they should be running by running checks automatically for them ahead of time, and then providing information within context within the EHR.”
Mandel described the basic approach of CDS Hooks: The EHR sends off notifications as events happen. For example, as a clinician is writing a prescription, the EHR might fire off a notification to an external decision support service. That service learns that the physician is in the process of writing a prescription, and it has the opportunity to return some information in the form of a “card” that will be displayed inside the EHR. It could offer up a new proposal. “If I am writing a prescription for a brand-name drug, the card that is returned by a CDS service might propose a drug that is going to be cheaper or more effective and present that in a way that I can just click a button and accept that proposal if I like it,” he explained.
CDS Hooks could also provide a link to an external app. “If I am at a point in my workflow where it might make sense to run an antibiotic selection application, a service might return a link to that app, Mandel said. “As the user, I can say, ‘yes, this does sound like it would be a useful tool for me right now.’ And even if I didn’t know about that app or wasn’t thinking about it, suddenly there is a card there reminding me that it exists, and offering to run it for me if I want. So CDS Hooks is really a very complementary technology to Smart Apps, because it makes it easier to run the right app at the right time and it can save explicit steps.”
I asked Mandel about the chief complaints physicians have about CDS — that the alerts are not relevant or too common.
He acknowledged that the concept of showing too many alerts or alerts that are too intrusive is a very real problem with EHR systems. “There are some things we do to provide to alleviate that pain,” he said. “We make the general assumption that these cards should not interrupt the work flow. They should be non-modal and shouldn’t be in the way.”
He added that because different users are going to find value in different services and cards, the EHR could have policies where the users have a say in the matter. “If you see a card with a kind of warning you just don’t care about or doesn’t apply because of your specialty, you could say, ‘don’t show me cards like this anymore.’ So EHRs have the option of getting that kind of feedback from users and taking it into account when they are displaying cards.”
I also asked what this might mean from a business model standpoint. Are they envisioning a broad ecosystem of decision support vendors?
Mandel said yes. They are trying to make it easy from the clinical perspective to call out to multiple different services all at the same time and integrate or aggregate results. “That means you can pick the services that are best in breed,” he said. “We think there are probably also opportunities for things like aggregation services that can handle some of the logic of collating those results for you, but it is still early in the ecosystem for those to emerge.”
I asked him who would vet the quality of CDS offerings. He said hospitals have to decide that a particular service is worth buying or using. “CDS Hooks is a technology,” Mandel said. “It doesn’t really solve that problem one way or the other.” As with any kind of service contract, you need to understand what they are giving you and whether it is appropriate.” He noted that because of the way they structure the technology, there are some things that are easy to do with CDS Hooks in terms of testing. “If you are a hospital and you are prescribing to a service that is going to provide drug-drug interaction warnings through a CDS Hooks interface, there is a simple API through which you call these hooks. You can write a test where you have two drugs that you know should have an interaction. You can routinely test to ensure that the results you get back really do flag that interaction. We know from experience that especially around times of system upgrades and changes, these types of issues tend to crop up,” he added. By building out tests like this that can be called automatically, the hospital can build up some assurance that they haven’t seen a change in the behavior of these CDS services.
Might EHR vendors see this as creating competition for their own CDS offerings?
“I suppose that is possible,” he replied, “but the commercial EHR vendors realize they can’t take on every aspect of care delivery. They all have some way to integrate external decision support. It just usually requires more engineering investment and more customization for each service you want to integrate, so this offers a way to do it in a pretty lightweight and consistent way.”
Does he envision a marketplace of CDS services for sale or an ecosystem where they would be shareable and free? Mandell said the answer is all of the above. Lowering the cost of the integration piece would allow people to focus on the value they are really adding, he stressed. “There are some hospitals — I would like to hope this applies to academic centers — where the kinds of advice they build up are things they consider a public good and they are willing to share, especially to the extent that any of these tools are built through government funding. It makes a lot of sense to share them pretty widely.” On the other hand, it is certainly a common trend among academic medical centers to figure out how to commercialize technology built in-house, he added. “We are seeing pretty strong pushes on both sides. But lowering the technology barriers to integration means we can actually focus on what is important: how to design services and market them.”
Mandel made clear that CDS Hooks is distinctly less mature than the SMART on FHIR protocol. “We have been working on SMART on FHIR for years. It answers question all the way up and down the stack from security to data to vocabulary. CDS Hooks is in a much earlier stage. We have the basic workflow lined up: the way a system asks for advice and the way advice is returned, but there are many details, including the way we do a security model and the way we launch apps where we are going to want to do more engineering work.”
He said the trick is to drive that work based on real requirements from CDS vendors and the EHRs. “It is tempting to sit down and make up a set of solutions on our own, but the truth is when you make solutions that way, they work for you, and in your demos, but they don’t necessarily work in the real world.”
Representatives from EHR vendors Epic and Cerner and from companies that build decision services are actively involved in the project. “These are businesses that clearly have a motivation to work together. And up until this point they have been doing it through a series of one-off integrations or proprietary vendor-specific integrations.”
The CDS Hooks group is meeting next week at an HL7 FHIR “Connectathon” in Baltimore to provide feedback about which features are the most important to prioritize next. Piloting CDS Hooks at a few hospitals is on the roadmap for the next year.