Computer-Assisted Technology: Not If, Not When, But How

In his 2004 State of the Union Address, President Bush stated electronic health records (EHRs) will “avoid dangerous medical mistakes, reduce costs and improve medical care.”¹ The attractive prospect of using one intervention to simultaneously address the twin challenges of health care quality and cost was undoubtedly a major motivation behind Medicare's decision to offer a free version of an EHR to physicians nationwide.² This rationale has also been cited by several large physician organizations that have invested in these systems.^{3, 4} This focus is especially salient for primary care, where recent analyses of the EHR's impact on preventive and chronic condition management have shown the greatest opportunities for improvement.^{5, 6, 7} While the “core” documentation, retrieval, and billing functions of an EHR have some influence on physician and patient behavior, it is the luster of other information technology-based interventions that accompany the EHR that has added to the growing enthusiasm for computerization in every corner of clinical practice. These EHR add-ons include, for example, monitoring individual episodes of care, automation of patient reminders, physician decision support, and patient education.^8–10

In this issue of the Journal, Jackson et al. report what the medical literature has to say about the impact of these computer-assisted technologies in diabetes care. Diabetes is a useful domain to study this not only because of the increasing prevalence of type 2 diabetes and not uncommon failure of usual care to meet care standards,¹¹ but because of widespread consensus about the metrics that should be used to assess quality. These include the frequency of hemoglobin A1c (A1c) testing, and foot and eye screening examinations. Accordingly, any lessons learned for diabetes may be generalizable to other areas of primary care practice.

And what are the lessons learned? Based on 27 reports meeting a minimum standard of a randomized trial or observational study with sufficient detail, we learn the breadth of available information technology options includes interactive web information exchange as well as telephony designed to promote patient self-management. In fact, the “EHR” per se is not even mentioned in this report; instead the authors point to “computer-assisted integration of clinical management” interventions that are commonly packaged with an EHR, such as provision of practice guidelines, clinical reminders, and individualized feedback at the point of care or for later summary reporting.^{12, 13, 14, 15} Limited by a small number of reports and the heterogeneity of the methodologies, Jackson et al. found a pooled metaanalysis was not possible. Instead, their qualitative summary noted less than half (6 of the 14 studies that reported this) of all the interventions was associated with improved glycemic control, there was variable impact on blood lipid levels and “no significant difference” for weight, blood pressure, microalbumin, or creatinine levels. A smaller number of the studies showed an inconsistent impact on the frequency of hospitalizations, primary care visits, A1c testing, or foot and eye examinations. Information on cost was the least detailed and reported in a variety of ways such as spend per patient or per system; no information was provided on the impact on direct or indirect health care costs, health insurance claims expense, or return on investment.

While as a whole this report confirms there is emerging evidence that information technology can have an impact on quality and cost for diabetes care, readers are likely to be surprised at the dearth of high-quality studies, its spotty track record, and how little we truly know about its ultimate value. Considering the estimate of $17 billion necessary to install the EHR in physician practices nationwide,⁵ it would seem the disparate impacts on quality, utilization, and cost of computerized support reported by Jackson et al. may limit its ultimate contribution to the achievement of the president's vision—at least for the estimated 18.2 million U.S. residents with diabetes.¹⁶ While the initial purchase and maintenance costs of information technology are blamed as a major barrier to its widespread adoption in office practice,¹⁷ another important reason is lingering physician skepticism over its ultimate value.¹⁸ Jackson et al. also show we need to know more about the likelihood that information technology will further exacerbate the barriers experienced by medically underserved populations, who risk compounding their already poor health status by being further “shut out” by suboptimal access to the web, unreliable phone communication, or practice settings unable to adopt computerization.

Lacking certainty, readers of the Journal may conclude additional research is necessary before physicians adopt any of the interventions described in this report. I disagree. Many other interventions lacking evidence of efficacy are appropriately used in primary care settings because of other worthy reasons, including judgments about generalizability of evidence-based medicine to individual patients, local practice patterns, market demand, and societal need.¹⁹ I temper this apostasy by stating additional research is critically important while interactive computer-assisted technologies become more commonplace. If randomized clinical trials (RCTs) are not possible in busy practice settings, stakeholders can turn to alternate and acceptable study designs such as staggered roll out or propensity scoring.^20–22

While developers and suppliers of these technologies may be reluctant to share proprietary information or subject their business plans to scientific inquiry, the prospect of rushing into costly technologies that ultimately have no effect on health care quality is considerably worse. Accordingly, they—and the patients that are being subjected to these interventions—need the flexibility and expertise of the savvy health services researchers who regularly appear in the Journal or at the scientific sessions of the regional or national Society of General Internal Medicine (SGIM) meetings. Their talent is needed to efficiently define and measure not only the elements of web-based, telephonic, and computer-assisted integration that result in better outcomes at lower cost, but to tell us what are the most effective mutually supportive combinations of these interventions, with or without the other emerging initiatives in primary care settings, including the core EHR, pay-for-performance, the chronic care model, and disease management.²³ Indeed, the latter is especially vital, as many commercial disease management organizations are already using many of these same web, interactive voice response, and computer support technologies with considerable commercial success.²⁴

Despite the limited number of studies, it is not a question of “if,” or even “when” for computer-assisted interactive technology. Jackson et al. have raised the more important question of “how.” How should patient responses on a web site or during interactive voice response be summarized for physicians or “loaded” into an EHR? How should electronic patient education materials be provided during an office encounter? How should an EHR prompt busy clinicians with best practice alerts? How can these interventions improve access for the underserved? The opportunities for additional research are considerable, and readers of the Journal with the resources and expertise need to get to work.