Using Electronic Health Records to Improve Outpatient Diabetes Care
Preface
- Patrick J. O'Connor, MD, MPH, Guest Editor
Early versions of electronic health records (EHRs) were developed in the 1980s. The potential to use such technology to improve care—especially preventive care and outpatient chronic disease care—has been widely assumed since the publication of two landmark Institute of Medicine reports in 1992 and 1994.1,2 However, nearly two decades later, the potential for EHR technology to improve chronic disease care—in particular, care of adults with type 2 diabetes—remains unrealized.
The unfulfilled potential is partially related to low levels of outpatient EHR use. As recently as 2008, only ~ 13% of primary care physicians (PCPs) used basic EHR systems, and only 4% reported using fully functional systems.3 Recent federal policies provide incentives for and subsidize the use of outpatient EHRs and are likely to accelerate the use of such systems in the near future. However, other concerns remain.4
Patients appear positively disposed to their providers' use of EHRs,5 but provider reactions have been less enthusiastic. Many early adopters of EHRs reported decreased practice revenue, disrupted clinic workflows, and transient deterioration in the office environment.6,7
Improvement of quality of care, a much anticipated result of the substantial financial investments needed to impel EHR use, has been slow to materialize. Some early reports even noted deterioration in care after EHR implementation.8–11 More recent reports reiterate that the effect of EHR systems on quality is marginal and assert that use of specific features of EHRs may be a better predictor of improved quality than simple implementation of basic systems.12,13
What about diabetes care specifically? What evidence do we have that EHR use or use of specific EHR features (e.g., clinical decision support) will improve diabetes care? A number of randomized trials have shown no improvement in diabetes care outcomes related to EHR use. In many of these negative studies, the EHRs being evaluated provided only rudimentary clinical decision support (e.g., prompts and reminders) and improved processes of care (e.g., test ordering) but not intermediate outcomes (i.e., glucose, blood pressure, and lipid control).14–17
Fortunately, some recent reports indicate that tracking systems to guide active outreach to diabetes patients may have some clinical benefit.18,19 In this Diabetes Spectrum From Research to Practice section, JoAnn M. Sperl-Hillen, MD, et al. (p. 149) report on an EHR-based system that included detailed patient-customized point-of-care clinical decision support. The system was used in a high proportion of visits by diabetes patients, was well accepted by PCPs, and had a positive effect on some intermediate outcomes of care.
In 1998, Greenfield et al.20 showed that encouraging diabetes patients to be actively involved in setting the agenda for an office visit led to better care and higher patient quality of life. Replication of these results using more efficient, information-driven interfaces is a much-sought-after goal of researchers.21 In a provocative article in this research section, Paris Roach, MD, et al. (p. 154) demonstrate the use of EHR-extracted data to provide patients with risk information using tablet computers before office visits. Patients who engaged in this form of visit planning reported more discussion of cardiovascular risks, ways to lower cholesterol, and smoking-cessation strategies. Although more work is needed to integrate this activity into clinic workflows and increase the efficiency of information transfer from the EHR to tablet computers, further development of such patient-activation strategies is promising because patient readiness to change is a strong predictor of favorable diabetes care outcomes.22
EHR clinical data can also be used to introspectively calculate a given patient's cardiovascular risk and display it to providers as an additional “vital sign” at the time of patient encounters. In his article on p. 160, Jeffrey J. VanWormer, MS, posits that timely display of such information may increase attention to cardiovascular risk factor reduction. In addition, VanWormer presents an innovative plan to consolidate EHR data from many providers to monitor cardiovascular events and cardiovascular risk factor reduction across an entire well-defined geographical community in rural Minnesota. This innovative approach to population health surveillance, if proven workable, could open up new possibilities for quality improvement and guide next-generation community-based health promotion and disease prevention efforts.
Another barely tapped potential of EHRs is to profile clinical care to identify strengths and areas for improvement at the individual provider level. Such an “audit and feedback” method has long been an effective learning method in residency and other training programs but was time-consuming and cumbersome when done by hand.
In this issue, Frederick J. Bloom, Jr., MD, MMM, et al. (p. 164) outline a method for mapping clinical performance across nine domains of diabetes care and using EHR data to gauge performance of clinic-based teams. This led to rapid improvement in documentation of immunization and smoking status and in process measures such as microalbuminuria screening. However, the intervention led to much slower improvement in control of glucose, blood pressure, and lipids. This study was limited by lack of a control group but supports the use of team-based diabetes care and points out some innovative EHR applications.
Health information technology (HIT) enables the development and widespread use of sophisticated performance and quality metrics. However, the design of measures of diabetes care quality that are meant to be used for purposes of accountability or pay-for-performance is a topic of great attention and concern. As Bloom et al. point out, there are some major drawbacks to and potential risks from the use of all-or-none measures of diabetes care quality such as the one they developed.
I would add that such measures, especially when they endorse aggressive clinical goals based primarily on observational data rather than randomized clinical trial evidence (such as the goals of A1C < 7% and systolic blood pressure < 130 mmHg) may provide an incentive for PCPs and health systems to over-treat patients. In addition to increased costs of care, increased hypoglycemia, increased weight gain, and increased risk of drug-drug and drug-condition interactions, overly aggressive treatment may significantly increase the risk of death based on results from the Action to Control Cardiovascular Risk in Diabetes trial.23
Fortunately, HIT technology can also be used to address this major issue. We now evaluate diabetes care by applying a single clinical goal to nearly all patients. In the future, EHR-based clinical decision support systems will be able to identify all open evidence-based treatment options and then prioritize them based on the marginal benefit to a particular patient at a particular time. Quality of care will ultimately be measured by how much we reduce complication risk—not by whether blood pressure, A1C, or LDL reach pre-defined and arbitrary threshold values applied indiscriminately to nearly all patients. Indeed, this transition will actualize the potential of “personalized medicine,” usher out current “goal-based” clinical guidelines, and usher in a more personalized and efficient “risk-based” approach to diabetes care. I estimate this transition will be in full swing 5–7 years from now.
Thomas Edison's first 2,000 attempts to develop the light bulb failed. These failures provided useful information, and Edison eventually succeeded in delivering a new technology that was truly revolutionary. We have two decades of failed EHR experiments under our belt, and, although we have often felt frustrated, we have learned a lot about what not to do. These new insights, coupled with new, more user-friendly technology, have brought us to the verge of success. Our cynicism is giving way to guarded hope that the informatics revolution in health care will, indeed, lead to measurably better care for patients with diabetes.
The articles in this research section sketch the future broadly. Many details are still to come. However, we are blessed to live in interesting times, and soon most of us will be using EHR-based clinical decision support, patient-centered records, and integrated health informatics systems to provide better diabetes care. To achieve this end, we must be willing to change, sometimes in radical ways, how we do our daily work. Change can be painful and awkward, and we often resist it. But maybe we can find the courage to change—for the sake of our patients.
Footnotes
-
Dr. O'Connor can be reached at patrick.j.oconnor{at}healthpartners.com. He would appreciate your responses to this or other articles in this From Research to Practice section.
- American Diabetes Association(R) Inc., 2010
