Q&A: Maine HIE works through a variety of challenges
 |
Where is HealthInfoNet in the implementation process? The demonstration phase began in late February 2008. We started clinical feeds and patient registration feeds in December 2008 with the complete clinical feed coming live by May 2009. Currently we have about 665,000 lives in the database—that’s about half the population of the state. Currently we service 15 hospitals and a large group practice that serves as a testbed for ambulatory care. Those represent about 52% of annual ER visits in the state and about 50% of discharges. About 2,000 providers are represented by these organizations, but we have just under 1,000 users on the system during this phase of the demonstration.
What information is included in the exchange? The demonstration data set is based on the Continuity of Care database: Information includes patient identification, allergies, prescription medications, lab results, diagnostic imaging results (text, not images), other diagnostic and/or visit summaries—ambulatory visit summaries, diagnosis, visit history, primary procedure.
For this phase, we chose to use portal-based strategy because we did not feel standards for CCD [Continuity of Care Document] or CCR [Continuity of Care Record] were mature enough to build a demonstration phase on them. We will begin to offer both that inbound and outbound strategy as we go into the full-fledged statewide implementation, but today, the portal is the strategy. The portal is actually managed inside our organizations’ EMRs, so they launch into the portal from their EMR, and they stay inside their EMR. We use a centralized data model, not federated.
We do a contextual launch: Once [providers] clear the security requirements and break the glass identifying their relationship with the patient, they are in that patient’s [record], they don’t have to search [for information]. That content appears on the screen in the context of their EMR, even though the data never moves into their EMR. We know docs will not entertain leaving their workspace and they will not do more than three clicks, so we tried very hard to make that part of the workflow so they don’t have to do that much other than go through some security steps to get where they’re going.
Can patients input data? Not yet, but we have a very active consumer advisory committee working on recommendations for delivering a consumer portal. (It was part of a commitment made in 2006.) We had to get the system up to begin to validate its effectiveness and usefulness with clinicians, but we’re targeting 2011 as the year we’ll begin to open up access to consumers, too.
What kind of feedback are you getting from providers? We seek out anecdotal reaction from providers; we’ll do our first qualitative survey of users soon. We hear a number of stories about decisions changed or studies not ordered as a result of using the system. The challenge they run into today is that about 13% of that 665,000 have been to two or more of the unaligned organizations -- and that’s where the real value is, being able to see someone else’s data -- so you have a 13% chance of hitting a patient that has been to multiple organizations.
But because we are also presenting a prescription medication history profile that is built on a commercial data set from RxHub and SureScripts, you have a 53% to 54% probability of finding a current medication profile. That’s proved to be a point of value, particularly in places like emergency rooms when you’re doing triage. We’re about to add the Medicaid prescription information, and that will bump up the probability of finding current medication data to about 64% to 65%.
What vendors are involved in the HealthInfoNet demonstration phase? The prime vendor is 3M Health Systems: they provide the back-end solutions including the data dictionary, master patient index, and clinical data repository infrastructure. The front end is Orion Health. They provide both inbound interface management through Rhapsody, and security too, and provide the presentation layer back out to end users through the Concerto product. Right now we’re using a hosted application service provider strategy hosted by Connectria in St. Louis. We work with five different EMR systems and will be expanding that out further as we add new organizations going forward.
What are some of the challenges for HealthInfoNet? State of Maine privacy and confidentiality laws did not recognize health [information] exchange, and we felt that the law as it was written covered us. But others, including our Attorney General, disagreed. The AG’s office decided that if Medicaid [MaineCare] was going to participate as a supplier of information, then the law needed to be changed, and HealthInfoNet, or an exchange, needed to be formally recognized in the law as being acceptable for support of treatment. So back in June of last year, the law was changed to recognize HealthInfoNet, not by name, but as an exchange.
Truly challenging are the concepts of audit and audit responsibility. When you go from an enterprise model to an integrated model, it’s been a fun, but challenging process of getting security officers comfortable with the concept that it’s not their patient anymore. It’s also raised some questions. The new law around breach and notification of breach, etc. is complicated to us because if you have an individual that has content from multiple organizations that are not aligned corporately and there’s a breach event, who notifies the patient? We’ve been struggling to come up with a workable solution that meets the letter of the law, but also recognizes the rather unique situation where you could have breach of information that’s from multiple organizations.
One of the objectives in the initial design and scope was to connect to public health structure. To do that, we decided to focus on automating lab results to Maine [office of the] CDC for the purposes of mandated disease reporting. I’ll call it a qualified success … it’s taught us a lot. We have five labs participating, from a very small hospital lab up through the two private reference labs in the state. These systems were not really set up to support automated rule logic. They were set up to produce results that are printed or read on a screen by a human.
We agreed to try to automate reporting for 30 of the 72 diseases that are mandated for reporting in the state of Maine. We’ve had a very wide range of success, but it’s been a matter of luck in terms of how the labs have set their systems up. So we now know what we think we need to do to go back and support those labs in actually changing how they report, but that’s going to [require] time and money to get that done.
We’re finding some good things: The labs that have been successful are finding that their results are getting there anywhere from four hours to a day or two days faster because there’s no human stuffing it in a fax machine. We’re finding a larger number of results being reported because you don’t have the human factor of oversight.
The other thing is that it’s been the first significant effort to standardize content. We do all the mapping as best we can to standards. All lab results--actually, all lab tests--are mapped to LOINC and we then present that content as LOINC-coded content. [This] allows the CDC to have no human intervention at all until an epidemiologist opens up that result within their system. And that result has already been parsed to be placed in a particular vector based on the disease. That’s been a qualified success. With some work and additional money, we can probably get this to go.
The other major area of challenge is engagement of the consumer. We are an opt-out state (codified in state law). HealthInfoNet serves as the agent for its customers, so we manage the opt-out process for the customer. Providers don’t know whether the patient has opted in or out until they attempt to access info within the system. If the patient has opted out, there’s no information in the system.
Getting consumers to understand and feel comfortable with the system has been a fairly significant challenge and continues as such. There’s a lot of misunderstanding and natural fear of computers and centralizing data. Our opt-out rate is fairly low—we’re running at about 0.7%. Interestingly, women [are more likely to] opt out than men, and older ages more likely to opt out.
What’s next? Now we’re in last six months of the demonstration phase, and we’re doing some evaluation both of the value of exchange itself and the current technology architecture. Both will be subject to some change. We have not accomplished everything we said we were going to try to accomplish in the demo phase in terms of impact, but we have a pretty good feel for where we have brought value.
Cost is an issue. It’s disproportionally unfair to smaller organizations, because you have much of the same up-front costs regardless of whether you’re a 25-bed hospital or 25-person practice as opposed to a 600- to 700-bed hospital. At the end of the day, those interfaces essentially cost the same regardless. We’re actively looking at how we can modify the design to modify the cost structure now.