While we now have a large scientific evidence base about what care to provide, practice has lagged substantially behind,20
suggesting that better tools for translating evidence into practice are urgently needed. Electronic decision support represents one such tool and has been highly touted,1
but a series of recent studies have demonstrated no benefit.10,11,21,22,23
In this study, we found that substantial quality gaps exist for the management of diabetes and coronary artery disease. Electronic clinical reminders improved the chance that patients would receive recommended care beyond the effect of existing paper reminder systems, although there was variability by service and gaps persisted. We were encouraged by the finding that the reminders were well received by the physicians and that the majority preferred electronic reminders to paper-based reminders.
Our findings of suboptimal care for diabetes and coronary artery disease are consistent with those of previous studies. A recent analysis of diabetes management among commercial managed care organizations demonstrated an annual cholesterol screening rate of only 63% and an annual HbA1c screening rate of 83%.24
Among patients with coronary artery disease, only 40% receive beta-blocker therapy and 38% receive aspirin therapy.25
Our physician survey and other meta-analyses26
provide insight into why rates of guideline adherence remain so low. While external factors such as lack of visit time and patient noncompliance are perceived as important issues, it is important to note that physician-related factors continue to be an issue, including lack of familiarity with guidelines and lack of agreement with guideline recommendations.
Despite the support for our reminder system expressed by physicians, many of the reminders did not affect provision of services. This finding is similar to other electronic reminder systems. In a recent investigation, Tierney et al.11
studied an electronic reminder system for coronary artery disease and congestive heart failure and found no significant impact on the management of heart disease. A similar system of electronic reminders within the Veterans Affairs health care system did not increase rates of beta-blocker use or cholesterol screening for patients with coronary artery disease and had a variable impact on diabetes quality measures.27
Our study provides insight into the mechanisms behind the successes and limitations of electronic reminder systems. We were able to demonstrate improvements in areas such as medication initiation where other systems have not succeeded. We attribute these successes to both effective design and physician acceptance. It is important for electronic decision support systems to provide actionable recommendations in a simple format to maximize their effectiveness.28
Our reminders provided succinct messages generally shorter than ten words in length with an immediately actionable item. In addition, we achieved physician acceptance by including many of the primary care physicians in the development process to maximize the integration of the system into the existing workflow. We also limited our reminder system to providing recommendations for aspects of care in which there is very little disagreement on appropriate management and kept the recommendations somewhat conservative to avoid inappropriate recommendations (for example, using an LDL cholesterol goal of 130 mg/dL for coronary artery disease instead of 100 mg/dL). This strategy avoids the pitfall of generating physician distrust of the reminder system while also capturing those patients in most need of improved disease management.
We also learned that requiring physician acknowledgment of reminders may be a critical step in achieving success, as highlighted by the small proportion of physicians who reported noticing the reminders during office encounters. The impact of this limitation is highlighted by the finding that among physicians who report noticing reminders, nearly three fourths acted on the recommendations. This suggests that our reminders could have a much larger impact by requiring physician acknowledgment of the recommendation.29
This is a challenging area, and there is a major tension between making reminders more intrusive and generating resentment among physicians. Our study also suggests that reminder systems may exhibit differential effectiveness depending on the service being recommended and the particular disease. For example, reminders for annual cholesterol testing were effective for patients with diabetes, but not coronary artery disease. Similarly, reminders for statin use were effective for patients with coronary artery disease but not diabetes. Future work should focus on these subtle usability and workflow issues of electronic decision support systems.
Previous studies on the factors affecting the success of reminder systems shed additional light on why our reminder system was successful in some areas but not others. Our reminder system likely benefited from both our efforts to maximize the accuracy of the recommendations30
and the fact that decision support systems are more likely to be used for conditions that are the focus of performance measurement, such as for diabetes and coronary artery disease care.31
However, physicians report that reminder systems often lengthen office visits,32
and this likely limited the effectiveness of our intervention given that lack of time was reported by physicians in our study as a significant barrier to guideline adherence. In addition, the concurrent use of paper forms in our study clinics likely also limited the effectiveness of our electronic intervention, as physicians were possibly less focused on the electronic record during the office encounter.30
Our study has several limitations. To date, a minority of ambulatory practices in the United States are using electronic medical record systems with integrated laboratory and medication data; however, their use is routine in many other countries and there are active efforts to increase their use in the United States.4
In addition, we had to rely on physician data entry into the electronic record for some measures. The low rates of baseline performance of dilated eye examinations likely reflect deficiencies in documentation rather than abnormally low adherence for this measure. We did not assess outcomes of care, although most of the process measures that we assessed have been demonstrated to result in improved outcomes in controlled trials, and outcome differences may take years to identify.
We used physician report to assess how often our reminders were noticed, which would be more rigorously assessed by direct methods of studying eye movement tracking. We did not assess physician-reported barriers to guideline adherence for specific aspects of care, such as for appropriate medication initiation or laboratory testing; however, the general domains that we assessed have been validated in previous studies.26
Finally, our reminder system lacked direct integration with computerized ordering, which could have potentially increased its effectiveness.