Clinical Decision Support is Cost-Effective

Clinical decision support systems incorporated within electronic medical records have the potential to save health costs. When not cost-saving, CDS systems demonstrate cost-effective ways to achieve improved clinical outcomes.

Only twenty percent of patients with diabetes in 2008 reached goals for HgbA1c, blood pressure, and LDL cholesterol. The top reasons include the lack of intensive pharmacotherapy by providers and medication non-adherence by patients. Electronic Medical Records (EMRs) that have Clinical Decision Support (CDS) programs to address intensive pharmacotherapy have now evolved from simple general prompts and medication reminders for physicians to sophisticated algorithms that take advantage of current and past clinical information so as to provide patient-specific timely reminders in line with evidence-based practice.

EMR and software evolution is beginning to improve diabetes health outcomes, but at what cost? The cost-effectiveness of the billions of dollars invested into EMR-based CDS for adults with diabetes was evaluated in this study using a predictive statistical model.

The data were extracted from a 2008-2009 study which showed that EMR-based diabetes-specific clinical decision support improved clinical outcomes of HgbA1c (lower by 0.26 percentage points) but had no effect on blood pressure or LDL cholesterol for patients with diabetes over a 1 year period. The specific EMR was EPIC and the specific CDS addition was Diabetes Wizard.

The initial study focused on 1,092 adults with poorly-controlled type 2 diabetes in a large Minnesota medical group composed of  11 clinics and 41 participating providers. The patients were divided into either the control group (n=621) which received the standard of care EMR without CDS prompts or reminders or the intervention group (n=471) where providers utilized EMR with the Diabetes Wizard CDS prompts integrated into the workflow of the clinic. Cost data for the health care system were derived from the initial study which accounted for estimates of implementation costs, annual algorithm programing costs, user training costs, and physician incentives. Future pharmacy costs and costs of diabetes complications were also estimated by the authors. The clinical outcomes data and cost for each group over a one year period were compared when each set of numbers were entered into a statistical prediction model called The UK Prospective Diabetes Study Outcomes Model (which was validated and created within a UK patient population). The initial data set did not include length of time of diabetes or measures for medication adherence so estimations were made based on other studies conducted in the same clinics.

The relatively low cost EMR-based CDS was found to be modestly cost-effective for the health care system according to commonly accepted standards, both in the base case and in multiple sensitivity analyses that describe a range of potential confounding assumptions.

The per capita cost of implemeting and maintaining the EMR-based CDS was $5. Overall, EMR-based CDS costs $3,017 per quality-adjusted life year (QALY). Since the accepted value for cost-effectiiveness of an additional QALY is agreed upon by most researchers as $50,000, this study demonstrated that EMR-based CDS system  are indeed cost effective.

Remaining life years, quality-adjusted-life-years, and cost-effectiveness ratios were all higher for EMR-based CDS in many varied sensitivity analyses.


This study helps to quantify the modest cost-avoidance benefit to health care systems employing EMRs with clinical decision support. However, the authors recognize that the data, derived from a research study that gave incentives to physicians to utilize the CDS, is a significant confounding factor when extrapolating these results to other health care systems. In fact, provider utilization of CDS decreased by 50 percent in the initial study when incentives were reduced. Utilization of the software by providers is a very important factor when making predictions for its effect on provider pharmacotherapy recommendations to patients, a major reason why few patients are following clinical guidelines. However, patient medication adherence, another equally important hindrance to following guidelines, is not explored in this study. All in all, a health care system that better guarantees physician adherence to pharmacotherapy and patient adherence to medication is key. These guarantees require more than just clinical decision support. Perhaps a platform which rewards patients and providers and subsequently engages them with feedback would work best.

Gilmer, TP, et al. “Cost-Effectiveness of an Electronic Medical Record Based Clinical Decision Support System.” HSR, 2012. Epub.


Jennifer Shine Dyer, MD, MPH, FAAP