Wednesday, April 30, 2008

Decision Support Service Providers

I've recently written about decision support and speculated on the ways we can transform data to information to knowledge to wisdom.

Over the past few weeks, I've seen a convergence of emerging ideas that suggest a new path forward for decision support. Application Service Providers offer remotely hosted, high value Software as a Service applications at low cost. I believe we need Decision Support Service Providers (DSSP), offering remotely hosted, low cost knowledge services to support the increasing need for evidence-based clinical decision making.

BIDMC has traditionally bought and built its applications. Our decision support strategy will also be a combination of building and buying. However, it's important to note that creating and maintaining your own decision support rules requires significant staff resources, governance, accountability, and consistency. Our Pharmacy and Therapeutics Committee recently examined all the issues involved in maintaining our own decision support rules and you'll see that it's an extensive amount of work. We use First Data Bank as a foundation for medication safety rules. We use Safe-Med to provide radiology ordering guidelines based on American College of Radiology rules. Our internal committees and pharmacy create and maintain guidelines, protocols, dosing limits, and various alerts/reminders. We have 2 full time RNs just to maintain our chemotherapy protocols.

Many hospitals and academic institutions do not have the resources to create and maintain their own best practice protocols, guidelines, and order sets. The amount of new evidence produced every year exceeds the capacity of any single committee or physician to review it. The only way to keep knowledge up to date is to divide the maintenance cost and effort among many institutions.

A number of firms have assembled teams of clinicians and informatics experts to offer these kinds of knowledge resources. UptoDate maintains world class clinical information with thousands of authors reviewing literature and providing quarterly revisions. Safe-Med has a large team of experts codifying decision support rules and building the vocabulary tools needed to make them work with real world clinical data. Medventive provides the business intelligence tools needed to create physician report cards and achieve pay for performance incentives.

However, none of these firms can plug directly into an electronic health record in a way that offers clinicians just in time decision support.

Here's a strawman for the way a Decision Support Service Provider should work:
a. A hospital or clinic selects one or many Decision Support Service Providers based on clinician workflow needs, compliance requirements and quality goals
b. Electronic health record software connects to Decision Support Service Providers via a web services architecture, including appropriate security to protect any patient specific information transfered to remote decision support engines. For example, an EHR might transfer a clinical summary such as the Continuity of Care Document to a Decision Support Service Provider along with a clinical question to be answered.
c. A clinician begins to order a therapy or diagnostic test. The patient's insurance eligibility and formulary are checked via a web service. The patient's latest problem list, labs, and genetic markers are compared to best practices in the literature for treating their specific condition. A web service returns a rank ordered list of desirable therapies or diagnostics, based on evidence, and provides alerts, reminders, or monographs personalized for the patient.
d. Clinicians complete their orders, complying with clinical guidelines, pay for performance incentives and best practices.
e. The decision support feedback is realtime and prospective, not retrospective. Physicians get CME credit from learning new approaches to diagnosis and treatment.

In order to do this, EHR vendors must work with Decision Support Service Providers to implement the uniform architecture and interoperability standards needed to integrate decision support into EHR workflow. I would be happy to host a Harvard sponsored conference with all the stakeholder companies to kick off this work.

Of course, some may worry about the liability issues involved in using a Decision Support Service Provider. What if clinicians comply with flawed guidelines or fail to comply with suggested therapies and bad outcomes occur?

An excellent summary of Information-Based Liability and Clinical Decision Support Systems is available on the Clinical Informatics Wiki.

Based on my review of the literature, I believe decision support liability is a new area without significant case law. The good news is that there are no substantive judgments against clinicians for failing to adhere to a clinical decision support alert. As a licensed professional, the treating clinician is ultimately responsible for the final decision, regardless of the recommendations of a textbook, journal, or Decision Support Service Provider. However, as Clinical Decision Support matures and becomes more powerful and relevant, I believe that there could be greater liability for not using such tools to prevent harm.

This blog entry is a call to action for EHR vendors and emerging Decision Support Service Provider firms. It's time to align our efforts and integrate decision support into electronic health records. Working together is the only affordable way for the country to rapidly implement and maintain high quality decision support.


Joe said...

Great post!
With a significant number of community hospitals without the IT staff required for success, I dont know how success will be achived across the nation without such a service.

Also..check our
"Medicare has issued a proposal that will stop paying for avoidable hospital associated complications , and with it, major private insurance companies are expected to follow the lead and cut out payments for this common and expensive complications. The denial process will alert patients to the fact that their condition resulted from hospital error, and can be expected to increase malpractice claims for such complications and create a legal argument that these issues are automatically malpractice." - May 2007 the Clinical Decision Support question on top of that one.

John R. Christiansen said...

Oh, good lord, John.

I love your work and style. And I agree very much with your thoughts about where we need to be going.

The problem you have - and which I suffer from too, too often - is what Nathan Myrvhold characterized as "living in the future." In other words, you see further ahead than we can really execute. This is not good or bad, it's simply recognition of where the status quo appears to (and perhaps in fact is) at the time yo are responding to.

Me, I'm responding as a lawyer - my own blog being at, (you can look or not as you choose but it's my bona fides here) - I have to agree there's no significant case law in this area. There's *some* sort-of-relevant caselaw; I've done work in clinical decision support law; but calling it minimal is polite.

But in doing this work I've also seen a lot of risk-aversion, an unwillingness to assume liability by standing behind the decision-support system. Which makes a ton of sense; if you're the DSSP you probably aren't making *nearly* enough money to make the assumption of literally millions of dollars of risk exposure a prudent balance. I've talked to VC boards on this level - not gonna happen without *major* visionaries - with *major* reserves.

Nor are clinicians often going to be willing to take this on. They may not have a VC board, but hospital/clinical boards are not likely to be more risk-philiac than VCs. Less so? Yes, as a rule, I think.

And in this kind of space don't clinicians want insurance? Of course. And their willingness to step up? Hmmm . . .

As I said, I think I'm in the same place (the same quadrant? whatever) as you are on the benefits and where we ought to be. But I've also seen too many good - no, I mean really good, good to the point that I put prospects on the line myself - concepts about clinical decision support go aground on liability issues.

You imply but are absolutely right (I think, as a non-clinical care provider myself) that there are tremendous benefits to be gained from a DSSP model. No single provider organization can or should be expected to aggregate and analyze the data necessary to support DSSP rules, nor does any one of them have an incentive to do so for a community benefit. If community benefits were enough we'd all have RHIOs by now - or would have had a CHIN years ago.

So my question is: How do we enable a model which allows DSSPs to come into being, given the liabilities they would have to assume (unless they were relieved of them - don't rule that out if you're living in the future!) and the realities of financial incentives for healthcare organizations as they currently exist?

Unknown said...

Dr. Halamaka;
Great Blog!! From driveway vegetable washing (excellent perspective on child rearing) to eHx.

Congratulations on the 48. That is amazing!!

Most Cordially, Greetings to the emerging DSSP legal niche;
With respect to liability, SafeMed does not author rules. SafeMed provides a tool for clinicians that identifies and prioritize evidence based medicine guidelines from multiple sources, which are user selectable. These guidelines (that are available without SafeMed in desperate sources, semantic formats and difficult to access publications) are instantly provided to the clinicians with; why was the guideline triggered, what organization created the guideline and its reference, and why is the care to date compliant or non‐compliant with the guideline, all based on the inputs of the clinician (or patient/consumer). Advanced Clinical Decision Support from SafeMed is 'supporting' the physicians existing knowledge with a massive data base that has never before been assembled. This is not a liability, but an answer to safety, quality and cost in the healthcare ecosystem."

David Potter said...

Upfront disclosure: I am a stakeholder in an ASP based chronic disease decision support company whose technology readily integrates with leading EHR programs, so my comments are admittedly self serving. This technology is being used in multiple clinical settings, and among other things have yielded process efficiencies and annual treatment cost savings of more than $2,000 per patient. Moreover it "pushes" best practices data to clinicians without requiring their data entry, as envisoned in John's Model.
We've found three major obstacles to broader adaptation:
1.) CTO's are reluctant to replace their homegrown "pull" applications, despite the proven comparative advantages of the alternative, typically due to "sunk cost" or " Not Invented Here" syndromes.
2.) Because the application is a "niche" one, albeit an easy plug and play representing a large niche, some decision makers hold out for a "universal platform" encompassing domains that are in fact rarely encountered by clinicians
3.) Institutions with only a minority of providers currently using EHR/EMR's often want to wait for higher penetration...which means that the outcomes and process improvemnts and treatment cost savings become hostage, for years, to an admirable yet distant objective. Our experience is that clinicians readily embrace the fax-delivered version of
Bottom line point: John's Vision is very viable, but requires a willingness on the part of CTO's, CFO's, Medical Directors, and clinicans to be willing to adapt an incrementalist approach for interim benefits, i.e, be willing to, at least temporarily, weave in specialty applications, sometimes ones that involve relatively low- tech patches for the late clinical adapter.