HIStalk Interviews Andrew Kanter, MD, Chief Medical Officer, Intelligent Medical Objects

Andrew Kanter, MD, MPH is chief medical officer of Intelligent Medical Objects and assistant professor in clinical biomedical informatics and clinical epidemiology at Columbia University.

Tell me about yourself and the company.

I’m an internal medicine physician and a global health specialist. I’ve been  with IMO since its founding in 1994. I’m the chief medical officer. I’ve also spent time as president and chief operating officer and I’m a previous board member for the company. I’ve been full time with IMO since the onset of ICD-10-CM and Meaningful Use.

In 2008, I joined Columbia University and the Earth Institute to help Jeffrey Sachs bring health information technology to less-developed countries. We’re trying to achieve the Millennium Development Goals.

IMO started as a computer science department. We evolved electronic health records, which is now being sold as Allscripts Professional. We’ve taken the company from primarily being a consulting firm and a terminology product company to now a solutions company. We recently received a major investment from Warburg Pincus, which is the fifth-largest private equity firm.

My job as chief medical officer is to ensure that the content and lessons that we’ve learned over the last two decades at IMO transfer to our customers and to our vendor partners. IMO partners with most of the EHR vendors. Our installed base covers about three-quarters of the acute and primary care sites in the United States. Over a half-million doctors use us in the US alone.

Our primary mission is to improve care by helping the health information systems capture clinical intent in the most accurate, specific way possible. We’re about paving the semantic highway and driving downstream workflows and secondary use of data.

Coding and terminology drives billing, but what are the patient and societal benefits?

It’s very much about capturing that clinical intent for these downstream processes, including things like population health, risk management, and predictive modeling. We’re trying to improve care as well as bend the cost curve. It’s through capturing the clinical descriptions — what physicians actually think about in their heads – that is so essential to teasing out that value proposition.

Many people think of it in terms of the big chronic disease areas like diabetes, coronary artery disease, and so on, which certainly does have a reimbursement or billing implication. But for trying to improve care, we’re trying to identify exactly the patients who can most benefit from therapy and the most accurate treatment possible.

Using the high level of granularity that clinicians have to take care of their patients can be used probably even more for driving the identification of high-risk patient groups or specific patients who will benefit from treatment, therefore significantly improving the quality of care.

Assuming that providers are willing and able to physically exchange information, what terminology and semantic interoperability problems remain?

That’s one of the areas that people have often missed in the strive to develop the electronic data interchange part, but not necessarily the semantic interoperability part.

What we’ve seen is that as information tends to move around the health information system ecosystems across enterprises, a lot of that semantic fidelity gets lost. Systems have been designed primarily to support single-term, single-code relationships. With FHIR and CCDA, where more information is going to be transferred, if that fidelity is not maintained, there’s a lot of loss of good data and the ability to act on it.

One of the things that IMO has worked pretty hard on is to first initially capture that clinical intent, but then ensure that it’s maintained as information moves around the health ecosystems. As a matter of fact, IMO has been working with the Structured Documents Group of HL7 to ensure that there’s an approved method of sending the IMO lexical identifier within all of those interoperability messages — whether it’s CCDA or FHIR — to ensure that the full color is not lost.

IMO’s terminology frequently has more than one reference map, whether it’s SNOMED, ICD, or LOINC. We have about 80 maps that come off of our clinical interface terms that ensure that all of those terms are maintained as information moves through the ecosystem.

What’s the best solution for codifying information for specific purposes while retaining the patient narrative to avoid losing the underlying context?

Over the last 20 years, there’s been a lot of lessons learned about forcing physicians into structured data collection. We have the scars on our backs to remember those lessons.

There is a balance between structured and unstructured, although it’s a little bit of a false dichotomy. If you focus on capturing clinical intent and using interface terminology, it should be possible — through structured data and unstructured data — to capture the right content in a clinically granular way.

We work closely with many natural language processing companies to ensure that those engines are able to identify these pre-coordinated or highly physician-friendly terms within both narrative, unstructured text as well as structured text. It’s a key thing to remember that it’s not always in the places that we expect to find that information, where key information will be determined.

There has to be a balance between maintaining the full clinical story in the narrative text as well as the highly codified structured text.

Should clinicians have the ability to electronically highlight the structured or unstructured information that they find most useful so that a colleague covering that patient or receiving a referral could make quick sense of a patient’s chart?

That’s a difficult question. That assumes that the information that clinicians are dealing with within the health information systems is hard to decipher and that their information is not clearly available or can be recognized quickly.

This is something that we’ve been working on a lot. How to organize the data now that we’ve collected it in such a clinically-friendly, granular way. How to then semantically group that to drive various kinds of workloads. How to visualize the problems or the information in the record in a way that’s most relevant to me as a provider, based on my specialty and my experience with that patient. Trying to not lose information that may not be as relevant, but organize it in such a way that it doesn’t distract me, but it also doesn’t hide things that I might be interested in in the medical record.

That’s a real challenge. We don’t always know what is going to be the most relevant for you in the record.

Also, as you start to semantically organize and group things together, you’re going to use that information to drive all sorts of downstream workflows. Things like clinical decision support, which hopefully will help prompt providers for things that they shouldn’t overlook or that would be most relevant to them. As well as driving quality improvement programs, population health, clinical research, and so on.

IMO is spending a lot of time in developing those services. It’s not just about tagging things in the medical record, but about using those tags meaningfully to help organize the data in a better way.

Are patients receiving the expected benefits from the migration to ICD-10?

ICD-10 certainly was a huge burden in the transformation from ICD-9. The 80,000 or so codes added would have normally been a disaster for many providers to deal with.

Most people don’t have to deal directly with the ICD-10-CM codes, so it’s not so much those codes that are improving patients. It’s the clinical concepts and the clinical terminology that are really most relevant to providers and t patients.

There’s no question that the evolution to more granular coding systems will benefit both patients and providers by giving more specific care and being able to perhaps group or subset patients based on more granular concepts. ICD-11 and SNOMED CT are looking to be much more closely integrated so that those two use cases — the billing use case and the secondary use case — will overlap.

That’s going to take a long time, for us in the US in particular, to move to. In the interim, it’s up to vendors like ourselves to try to focus on the level of granularity and information that is most needed by the providers and for the patients. That will improve the quality of care.

You have a background in public health and global social causes, having served as president of Physicians for Social Responsibility. What optimism can you offer in looking at the US healthcare system and our social policy when it’s so easy to find negatives?

There’s an interesting parallel. For me, if we want to get the most value from data — whether it’s big data or small data — it’s about accurately capturing what’s happening with the patient. What we currently suffer from is a distorted view of that reality. Everything that we’ve been trying to do — whether it’s our clinical decision support dashboards or health information technology in general – it’s just not going to perform well if we can’t see the world clearly.

IMO and other organizations are trying to facilitate patient-doctor communication to accurately capture and see the world. It’s through that joint solution that we can transform healthcare.

For me, for someone who’s so active in trying to save the world from global poverty, climate change, and nuclear war, healthcare is actually the easy part. If we establish that pattern of behavior — the ability to share information clearly and focus on solving the problems — we could use that technology and the lessons we’ve learned from working together globally to solve these other grave threats to our society.