Tuesday, September 16, 2008

Light Registries: All the taste -- none of the data entry

Recently Healthcare IT News published a story on the positive effects of a diabetes registry on the quality of care at Duke. They report that Duke has used DocSite, a registry product, for about a year and has improvements in the achievement of “perfect care”(in which the patient is on time or on target for every guideline). They report a "doubling" of this rate (although they don't report the rate itself, which is notoriously low - often well below 5%.)

DocSite has been around a long time in various forms and has suffered in the past from a major limitation: somebody has to type in all the data: labs, visits, vitals, smoking status, you-name-it. This limitation has caused many practices to give up on chronic disease registries. However, there are two strategies that can help: one is to use the registry as a back end to an EMR that already has all the interesting data in machine-readable form. I suspect this what Duke is doing. Essentially, the DocSite decision support tools are bolted on to the EMR.

The alternative is to develop a lighter registry that contains only those data that are easily accessible. For instance, Vermedx depends primarily on laboratory data because they are almost always electronic from the moment they are generated. As other data types become available (vaccine usage, medications, vitals, etc.), they can be added to the system with a concomitant expansion of decision support capability. Until then (which still looks like a long way off for most Americans), a registry like Vermedx can provide huge benefits without requiring a tremendous effort in data entry. It can:
  • allow population views at various levels (practice, provider, geographic, age-based, etc.)
  • drive communications to the practice about exactly what to do next
  • drive communications to the patient about what to expect and how to get it
  • stimulate improvements in the essential office micro-systems that are at the root of Primary Care quality
  • identify patients who need extra attention from specialists, case-managers or other resources
  • connect the practice to lab results from outside institutions - even if the two institutions don't share any infrastructure
These advantages, which translate into huge savings for the system as well as improved quality of life for the patient, are how, as Blackford Middleton of the Center for Information Technology Leadership puts it,
“...diabetes registries are the only form of information technology-enabled diabetes management we found to be cost-beneficial when adopted for patients with Type 2 diabetes.”
The challenge now is how to get light registries in use now while everyone is so busy installing (or planning to begin to think about installing) large, complex EMRs. We need the savings and the quality improvement now! This is a case of the perfect being the enemy of the good.

Ben

Saturday, September 13, 2008

Health Informatics Discussion Forum

You can't do it alone and the web gives us some great opportunities to get help from the pros around the world! Many thanks to Abbas Shojaee and Dr. Gil (Gunter) Pollanz from Chris Paton's Health Informatics Discussion Forum for the excellent advice and insight they provided on Vermedx. You can see the whole thread here.

If you have ideas on how to make personalized clinical decision support better, faster, cheaper, higher, louder or just plain cool, please join the discussion either here or at the Health Informatics Discussion Forum.

Thanks

Ben

Saturday, September 6, 2008

More Diabetes Maps

We love graphics of all kinds and can use Vermedx® to make some pretty cool maps. Now the folks at The Center for Public Health Informatics at the University of Washington in Seattle have developed an open-source GIS tool for public health that takes up your data set and makes a Google Map out of it. They call it EpiVue and it has a web page and has been published in The International Journal of Health Geographics (Hey, I'm a subscriber!).

I loaded up the CDC Diabetes Prevalence Data and gave it a whirl. I had some trouble getting the data formatted just right. And, it wouldn't take the whole dataset at once. But, I did get a pretty nice map of Texas showing the prevalence of diabetes by county. However, I couldn't for the life of me figure out how to save it so I could post it here!

Overall, it's a pretty grand idea - make a free tool to advance public health epidemiology using off-the-shelf parts (Java, Google Maps, JFreeChart, R, etc.). However, its going to need a better user interface to make it a real winner. I hope they keep going with it!

Ben

Wednesday, September 3, 2008

CDC Assessment Initiative Conference

Dr. Littenberg organized a panel discussion at the annual Centers for Disease Control and Prevention Assessment Initiative Conference. This conference brings together public health officials from around the country to share ideas and accomplishments regarding the assessment of public health. There are a variety of tools for this purpose including the Behavioral Risk Factor Surveillance System (BRFSS), which is the world’s largest ongoing telephone health survey. It has been continuously administered in every state since 1994, and in some states as far back as 1984. States use BRFSS data to identify emerging health problems, establish and track health objectives, and develop and evaluate public health policies and programs. Many states also use BRFSS data to support health-related legislative efforts. Presenters at the Assessment Initiative Conference discussed ways to make this information more available to policymakers, researchers and the public. Other conference presenters discussed topics such as novel ways to assess hard to reach populations, the role of assessment in promoting community action, and ways to link data from various sources in order to paint a more complete picture of the health of the population.

Dr. Littenberg’s panel was entitled Using the Clinical Laboratory for Public Health Surveillance of Chronic Disease. Dr. MacLean discussed the Vermont Diabetes Information Study and described the collection of lab data from disparate sources, creating a chronic disease registry, and providing decision support. Dr. Fernando Guerra, Directory of Public Health, San Antonio Metro Health District discussed the use of the Vermedx technology to create an A1C registry for San Antonio. Dr. MacLean then discussed the ethical, legal, and social implications of chronic disease surveillance, highlighting the experience in Vermont, New York City and San Antonio. Edward W. Gregg, PhD, acting chief, Epidemiology and Statistics Branch of the Division of Diabetes Translation at the CDC presented on the CDC’s diabetes surveillance system and the potential for adding laboratory based surveillance to this system.

This was a great opportunity to think about the bigger picture of how we assess the health of our various communities, ranging from the local to the national level. What would a network of labs contributing results to a central database be able to tell about regional variation and quality of care for chronic disease such as diabetes?

Tuesday, September 2, 2008

Looking at the outliers

This graph is a little unfamiliar, but it has lots of power to show what's going on behind the data. 71 small community practices are laid out on the horizontal in order of their average LDL-cholesterol. The vertical axis show just what that level was.



The red lines at 100 and 130 mark the guideline-recomended targets. Although most diabetic patients should be below 100 mg/dl, the average patient in most practices is getting there.

Pay extra attention to the right hand end of the line. The last 4 practices have very high levels. In fact, the highest 11 practices seem to have a bit of space between them and the regular run of practices. Perhaps these are the practices that need some extra help in taking care of cholesterol.

We something similar on the left hand side: the lowest ("best") 4 practices seem to be standouts. Maybe we should visit those places and see what they've got going on - so we can bottle it for others to use!

Of course, the graph is just suggestive. There may be other, less interesting, reasons for outliers such as small sample sizes, data errors, miscalibrated lab results, etc. But, the picture does seem to tell us where to look next.

One more example of what you can do from a population view that you can't do one doctor at a time!

Ben