Tuesday, March 24, 2009

Does This Diabetes Registry Reduce Health Care Costs?

Sometimes the Disease Management Care Blog thinks healthcare is one big ‘journal club,’ where we get huddle over interesting policy and scientific reports from the medical literature. It’s not sure if this Vermedx article would pass muster as particularly worthy of a lot attention, but given colleagues’ emails about the media attention this has received (here and here for example), the DMCB decided to take a closer look at this paper and the technology it represents.

Here’s the facts: Vermedx is an privately owned information system that receives lab results on a daily basis, stores the data in a registry and produces physician reports that include flow sheets and recommendations along with letters to patients. Overdue labs generate patient and physician reminders.

Vermedx was subjected to a ‘randomized clinical trial’ involving over 55 primary care practice sites caring for 7,000 patients across New York and Vermont. Primary care sites, not patients, were randomized to either use of Vermedx or to usual care. While the results were never published, it was observed that there was no impact on ‘glycemic control, cholesterol level, blood pressure and self-care behaviors’ for the patients in the Vermedx sites. Yet, patients’ self reports seemed to indicate that health care utilization had declined. If true, this would suggest Vermedx ‘saved money.’

That brings us to this paper, which was published in the American Journal of Managed Care. The authors sought to confirm that there were savings by looking at the claims experience of a subgroup of patients in “Vermont Managed Care,” a physician hospital organization (PHO) that apparently participated in the randomized trial described above. This PHO has just over 31,000 covered lives. 153 patients from the primary care sites that used Vermedx were compared to 870 persons with diabetes that were cared for by the PHO physicians without Vermedx. A comparison of ‘trend’ (the rate of increase in health care costs) after Vermedx was implemented favored the Vermedx primary care site patients: trend was observed to go down in the Vermedx patients so that at the end of one year, the per member per year difference in claims expense was just over $500 and at four years it was just over $3500. The study was financed ‘in part’ by an external grants (making the DMCB wonders if the rest was financed by Vermedx).

The DMCB thinks this is instructive because it helps us remember to ask certain fundamental questions when reading studies like this:

Was it the intervention (in this case Vermedx) or could something else have accounted for the observed improvement (in this case, claims expense)? Since these sites had to voluntarily agree to use the system ahead of time, the DMCB wonders if they may have already been inclined to take good care of their patients with diabetes. Readers may ask ‘who cares?’ so long as it worked, but the DMCB thinks it’s important to know if it was Vermedx or if it was Vermedx combined with especially motivated physicians.

Is the (870 person) comparison control population adequate? In this case, the patients that served as the comparator were drawn from clinics staffed by other physicians in the PHO. There is no description of the other physicians, many of who may not have been primary care physicians or had patients interested in a primary care-centered relationship. In other words, it may not have been the absence of Vermedx in this control group that drove up costs, but the kinds of patients and their physicians - which would be unaffected by Vermedx. Why didn’t the authors report claims from a group of patients from the PHO clinics that agreed to participate in the original randomized clinical trial but were not assigned Vermedx?

Are the results generalizable? In other words, readers need to know if the results are likely to be replicated in their clinical settings. Note that in this Vermont PHO, there are about 31,000 PHO enrollees, yet only about a thousand persons with diabetes were included in this study. Given the overall prevalence of diabetes, the DMCB thinks that may be low. The DMCB wonders how a ‘diabetic’ was defined in this study (in fact, the percent of Type 1 and Type 2 are not disclosed) and if this represents the definition of a person with diabetes in say, Omaha. This is especially true because the savings were based on a relatively small number of persons. The average health plan has tens of thousands of persons with diabetes. In addition, there is no information in this report about the source of the savings: inpatient vs. emergency room vs. other types of healthcare utilization were not reported.

The authors deserve credit for being open to the possibility of unmeasured differences between the two groups. This was brought up in their discussion of the results.

Conclusion? There isn’t enough here for the DMCB to believe the press reports. If these results can be repeated and more transparently replicated as well as more completely described in other settings, there may be merit to this information system. As noted before, the DMCB believes the solution to chronic illness care management ultimately resides in an interlocking system that includes electronic records, the medical home, disease management, smarter physician reimbursement, better insurance benefit designs, consumerism and registries. Vermedx and the approach it represents may - or may not - be a step in that direction.

Note: The DMCB emailed 'infoATvermedx' as well the principal author of the published study asking for more information about the possible limitations described above. After one day, there was no reply.

No comments: