Archive

Monthly Archives: September 2015

Dr. Richard ValachovicIn this month’s Charting Progress, Dr. Rick Valachovic examines the promise of Big Data and its potential to enhance dental diagnosis, treatment and research.

In 2011, IBM’s Watson computer captivated the public when it challenged two legendary Jeopardy! champions on the iconic television quiz show. With access to more than 200 million pages of information and algorithms for sorting through this enormous quantity of data to identify the most likely responses to Jeopardy! clues, Watson outscored both human competitors.

Since this public demonstration of the power of what is referred to as “Big Data,” IBM engineers have focused on honing Watson’s skills in several new realms, including medical diagnosis. I don’t have to tell you that the amount of information available to clinicians now grows at a rate that far outstrips an individual’s ability to absorb it. A computer, on the other hand, is ideal for sifting through large amounts of data and looking for patterns. Watson takes this power one step further with a unique capacity for natural language processing and applying cognitive reasoning to analyze information and work with the clinician. In other words, Watson can understand a question posed in plain English, sort through massive amounts of data for potentially relevant answers, and communicate these to a clinician—in a matter of seconds—using what sounds like human speech.

Collaborators at the Cleveland Clinic and elsewhere have already provided Watson with a foundational medical “education.” This year, IBM announced that it is partnering with 14 major cancer centers to train Watson to analyze genetic data that can guide cancer therapy for individual patients. IBM has also invested heavily in acquiring health-related data sets to enhance Watson’s knowledge. These include large banks of images, which are pushing Watson’s programmers to equip “him” with a new set of skills in visual analysis.

A clinical world in which Watson and his successors provide seamless automated decision-support to clinicians may be some years down the road, but such a world no longer seems like the stuff of science fiction. Whether dentistry will benefit from these developments depends on our willingness to adopt a tool that has been remarkably controversial: dental diagnostic codes. In 2006, the ADEA House of Delegates passed a resolution declaring its support for the development and implementation of such codes to facilitate clinical research and assist in developing best practices for dental care delivery. Dentistry as a whole has been historically slow to take up this charge, and most dental schools are no exception. Fortunately, that situation has started to change.

Today, the Big Data revolution has finally given our community the incentive it needed to adopt diagnostic coding. Nearly all ADEA member schools now use electronic health records (EHRs); several dozen schools have introduced diagnostic codes into their EHR systems; and a small but growing subset of our institutions has banded together to create the first oral health database—BigMouth Dental Data Repository—developed from partially de-identified EHR data.

Six dental schools are currently participating in BigMouth, which resides on secure servers at the University of Texas School of Dentistry (UTSD) at Houston. BigMouth is a project of the Consortium for Oral Health Research and Informatics (COHRI), and researchers who want to query the data can submit a project proposal to a COHRI review committee for consideration. The database currently holds more than 2 million records, and already a few researchers have accessed the data to examine adherence to treatment protocols and the associations between several systemic and oral health conditions.

Muhammad Walji, Ph.D., Associate Dean for Technology Services & Informatics at the University of Texas School of Biomedical Informatics at Houston, leads the project. When we spoke recently, Muhammad told me that the project is now looking to expand the number of participating schools, standardize the way they collect data and, now that initial government grants have been spent, develop a financial sustainability plan. BigMouth is currently able to accept data from any institution that uses an axiUm EHR, but organizers are discussing plans to incorporate data from clinics using other systems in the future.

“We want the database to be as diverse as possible,” Muhammad told me, “especially geographically, so we can get a better understanding of what’s happening to patients throughout the country. We also want to reach outside of dental schools. We’re interested in having other institutions—such as large group practices—participate because they may be serving different types of patients.”

The creation of BigMouth was one of the driving forces behind the creation of the Dental Diagnostic System (DDS), previously known as EZCodes. DDS is currently in use at 16 dental schools and one dental support organization. An additional 16 dental institutions have loaded the DDS in their EHRs as a first step toward implementation of the terminology. Elsbeth Kalenderian, D.D.S., M.P.H., Ph.D., Chair, Oral Health Policy and Epidemiology, Harvard School of Dental Medicine, working with other Harvard and UTSD colleagues along with the University of California, San Francisco, School of Dentistry and ACTA (Academisch Centrum Tandheelkunde Amsterdam1) in the Netherlands, developed the DDS terminology in 2009.

More than two decades ago, the American Dental Association (ADA) recognized the need for a dental diagnostic coding system and began working to develop a separate system known as the Systematized Nomenclature of Dentistry (SNODENT®). Its codes are an official subset of the Systematized Nomenclature of Medicine – Clinical Terms (SNOMED CT®), a comprehensive set of medical terms that are recognized around the globe. SNODENT uses the same format as the International Classification of Diseases (ICD) codes that are the standard for health care reimbursement. The system offers a high level of detail; for example, it contains 25 ICD codes for different types of tooth decay. Yet the rollout of SNODENT has been periodically interrupted, and the dental community has been slow to adopt it.

“The initial development of SNODENT wasn’t ready for prime time,” explains David Preble, D.D.S., J.D., Vice President of the recently created ADA Practice Institute, “and the EZCodes (now DDS) were created to fill the gap.”

I called David to get an update on SNODENT and the acceptance of dental diagnostic coding generally. “People who are not informed still talk as though dentistry doesn’t have diagnostic codes,” David told me. “We do have codes; they’re just not widely implemented,” he emphasized.

Why has dentistry as a whole been resistant to diagnostic coding? In David’s view, the reticence is related to cost and culture. In an environment of decreasing reimbursement from insurers and fewer self-paying patients coming in for care, dental offices already feel under pressure to create efficiencies. Practitioners may see the introduction of EHRs and diagnostic codes as costly disruptions to the work flow that bring few benefits. But as he points out, “In the academic and large group practice environments, using diagnostic codes and electronic health records hasn’t turned out to be the workflow issue dentists fear.”

At dental schools, the desire to take part in Big Data initiatives such as BigMouth or others under way at the National Institutes of Health should serve as a motivator to adopt diagnostic coding. According to David, a handful of dental schools are already using SNODENT, and more are likely to join them. In the past year, the coding system was officially recognized by the American National Standards Institute (ANSI), and the ADA is now working with DDS code users and other stakeholders to develop an integrated coding system that can serve everyone’s needs.

David envisions that many future SNODENT users may want to begin by adopting smaller reference sets of SNODENT codes rather than working with the full 7,000+ terms contained in the system. While the specificity of SNODENT makes it an ideal system for research, David estimates that a clinician practicing general dentistry might only need 200 of those codes to document 99% of the diagnoses made in his or her practice.

Indeed, the more manageable size of the DDS system is one of its reported attractions. In the words of its creators, the DDS serves as an “interface terminology” whose terms are organized in a user friendly and meaningful manner for chairside use. SNODENT’s designers recognize the value of the DDS and have already incorporated about two-thirds of the DDS codes into the larger system.

Next month, the ADA will convene a meeting with stakeholders to address compatibility issues between the two coding systems. Then the ADA plans to put the revised version of SNODENT back on the ANSI ballot for approval. The goal is to ensure that the first ANSI-recognized version of SNODENT provides a solid foundation for future iterations of the nomenclature.

Elsbeth and Muhammad believe that delivering a unified diagnostic coding system for the dental profession by this fall will be a tall order; however, they are excited about working to harmonize the two systems. Meanwhile, everyone agrees that the adoption of diagnostic coding marks a turning point for our profession. How soon we will see universal adoption is harder to say, but there’s no doubt that it is the critical next step in fully realizing the potential of Big Data.

It’s also worth mentioning that Big Data is making some traditional researchers nervous. The National Institute of Dental and Craniofacial Research is trying to be sensitive to these concerns as it considers the creation of possible funding streams to support research based on EHR data. Where Big Data excels is in mining very large sets of existing data to establish correlations. A need will still exist for controlled trials that seek to establish causation.

Meanwhile, the desire to preserve today’s best quality research shouldn’t deter us from pursuing the new opportunities afforded by Big Data. As David points out, “Data alone will not replace traditional research, but there is a dearth of evidence for many things that happen in dentistry and a finite amount of money for controlled trials, and there are some trends you’re not going to see in a controlled trial no matter how many people you enroll.”

There’s no question that Big Data can enhance dental research and the care we provide, but even Big Data enthusiasts—such as Muhammad—caution against overzealousness.

“It’s not the answer to all of our questions,” he told me. “Big Data are inherently messy, so we need a team approach—clinicians, informaticians, researchers, statisticians, epidemiologists—and I think we have to be quite careful with what questions we ask of these systems as well.”

1 The English translation of the name is “Academic Center for Dentistry, Amsterdam.”