December 10, 2005
Does Disease Management Work?
One of the realizations about healthcare in recent years is that the management of acute episodic care and chronic conditions may be different. And the care of chronic conditions is the real “growth area” of cost burden on the healthcare system. Translation: there are a lot of people getting older and sicker in a chronic way.
An approach called disease management developed in the past 15 years or so, with acceleration in the past 5 to 7 years, especially with the advent of connectivity via the Internet. What is disease management? Therein lies part of the problem. No one knows.
Broadly speaking, disease management has the potential of contributing significantly to The Abramson Family Center's principle of distributed healthcare. But does it? What's the “report card” to date?
Published in the October 31 HealthLeaders on-line newsletter by Scott MacStravic is a review of the question. With great conclusiveness he reports: “After reviewing 44 separate studies of the economic impact and ROI that DM programs targeting depression, diabetes and asthma, and other conditions had achieved, Cornell University and Thomson Medstat concluded in August 2005 that "the jury is still out."
“The study determined that too few rigorous scientific studies of financial impact had been conducted to make a sound business case for employers, insurers and governments to confidently invest in DM for its cost-savings potential.”
Most of the remainder of the article is a review on why disease management cannot be submitted to normal scientific analysis. The major argument closely follows the same one put forth by Michael DeBakey, MD, the innovator of cardiovascular surgery, when defending a new procedure, coronary artery bypass.
The argument essentially says that disease management cannot be studied scientifically because careful selection is an integral part of running disease management programs. Hence, randomized, placebo-controlled trials cannot be run to test the hypothesis, just like with coronary bypass before it.
One could argue that you could compare one treatment program against another and against “regular” community care. The thing that makes this difficult is the lack of any risk adjustment methodology to do the job that randomization would normally do.
Once again “induction” raises its ugly head. Just because you can characterize a population of patients with the same disease diagnosis doesn't mean there is applicability to the next individual patient. And we actually don't understand yet which factors or combination of factors predict risk of morbidity and cost with any degree of accuracy except at the extremes. So there is no valid way to compare groups as it now stands.
One answer from the Disease Management industry has been to use a statistical approach called Predictive Modeling. A couple of different approaches have been used. The most common is generally a regression-based model based on prior history of resource utilization and using that to predict future behavior or at least the impact of the intervention. Another approach has been to use neural network technology to identify populations at higher risk.
These are interesting approaches, but they have some serious flaws. First, almost all models are based on insurance claims data. Without getting into why this happens, the diagnoses on insurance claims carry a high degree of inaccuracy. Predicting the outcome of a population would be difficult, of an individual highly inaccurate. Except at the extremes--in which case no additional tool is needed.
Disease management is based upon the idea that there is a set of information and treatment that essentially all individuals with a given condition need or ought to do to ensure optimal medical outcome as defined by the medical or insurance community. The idea is that this leads to lower total medical consumption and hence cost.
From an insurance company or government perspective this can make some sense. It leaves out a few things like complexity, biologic variation, individual choice and individual decision making.
So, customization by nurse-coaches or computer-assisted customization to assist nurses to deliver the care has been a solution. Its problem is it is expensive. Hence, the problem with showing a positive return on investment. According to the HealthLeaders piece:
“…an intrinsic element of the American Healthways approach, and one that probably greatly adds to its effectiveness, is the customization of each intervention as part of the personal nurse-coach and patient relationship and each phone coaching interaction. Individual nurses use their own best professional judgment, not a uniform "script" when interacting with patients, and patients inject concerns and issues that must be addressed, meaning that the "treatment" given to each patient is not only different from that given other patients with the same condition, but varies from session to session. This same across-patient and across-visit variability is inevitable with the physician visits that patients include in their DM process.”
“HealthMedia, an Ann Arbor, Mich.-based DM vendor that manages most of the "Thrive" DM programs offered by Kaiser Permanente, customizes its DM treatments by computer. Every DM participant completes a 35- to 50-question survey, which computers use to tailor e-mail and post communications to them.”
So standardization and consistency may be just as difficult for these programs as for other practitioners.
This leads to the last challenge on my list: matching the appropriate level of care to the individual's needs and desires for treatment. Very inexpensive and subtle intervention almost like marketing approaches work for the least medically complex. This mainly constitutes informing and removing barriers. Companies are reluctant to go there because informing and removing barriers to care don't generate much revenue. For the most complex, good standard medical care by competent providers is what is needed. Some additional “standard” intervention may help with improving the outcomes. How best to deliver this type of intervention remains to be seen.
Technology may eventually help ensure that the care for chronic conditions occurs at the level where it is most effective—in the hands of the person with the condition. Connectivity, transparency of information, decisioning tools, and distributed healthcare are directionally where disease management, whatever it is, should head.
Robert B. Teague is a pulmonologist and business consultant who is based in Houston, Texas. E-mail him.
Read other blogs in this series.
