22 May The human-centered approach is critical in clinical decision support system development, implementation, and utilization. Often the gap between de

Case study cdss implantation

The University of Illinois Hospital and Health Sciences System (UI Health) has been a leader in adopting health information technology. When the hospital was built in 1982, a light pen–based computerized physician-order entry (CPOE) system was deployed—decades before CPOE use was incentivized through the Centers for Medicare & Medicaid Services’ meaningful use program. UI Health was one of the first to deploy a modern commercial electronic health record (EHR) system in 1997, for which it won the Healthcare Information and Management Systems Society's Nicholas E. Davies Award of Excellence. Work on decision support rules began shortly after the EHR was deployed and, over the ensuing decades, resulted in hundreds of custom-built decision support rules. Creation of so many rules can lead to challenges in maintenance requirements and unintended consequences if not managed properly. Over the years, a clinical decision support system (CDSS) governance committee structure was formed at UI Health (exhibit 6.4) to help ensure the creation of evidence-based, highly effective CDSS rules. Researchers at UI Health have published many articles on the effectiveness of such rules in facilitating problem identification, reducing contraindicated medications, and improving venous thromboembolism prophylaxis and warfarin dosing, among other findings (Falck et al. 2013; Galanter, Didomenico, and Polikaitis 2002, 2004, 2005; Galanter, Liu, and Lambert 2010; Galanter, Thambi, et al. 2010; Galanter, Heir, et al. 2010 Nutescu et al. 2013). Despite all the research and expertise, not every rule gets implemented smoothly. For example, a study by Lui and colleagues (2016) pointed out the need to appropriately identify psychiatric patients who are at risk for metabolic syndrome, which can be exacerbated by second-generation antipsychotic medications. The psychiatry department requested a new CDSS rule around the use of antipsychotics and the potential development of metabolic syndrome. This rule was then published into the UI Health production environment. After this CDSS rule was created, a primary care provider (PCP) updated a patient's medications list because the patient's outside psychiatrist changed her antipsychotic medication dose. The patient was a 50-year-old woman who had long-standing diabetes, hypertension, obesity, and bipolar disorder. A CDSS rule fired, indicating that the patient was on an antipsychotic and qualified for metabolic syndrome because of her abnormal girth, her systolic blood pressure (which was greater than 130), her triglyceride level (which was greater than 150), her HDL cholesterol level (which was less than 40), and her glucose level (which was greater than 110). Clinically, most consider metabolic syndrome to be a precursor to the diagnoses of hypertension, obesity, and diabetes, all of which the patient already had—and all were documented in her medical record. The rule suggested that the PCP add “metabolic syndrome” to the patient's problem list, but no link to do so was present. Recognizing that the rule was not firing as intended, the CDSS governance committee performed a root-cause analysis. The following lessons learned and opportunities for improvement were identified: The rule was firing every time a medication reconciliation (a required task for any care transition) was performed on patients who had an antipsychotic on their medication list and met the criteria for metabolic syndrome. As a result, an alert went out to providers even when they were just acknowledging a patient's current medication regimen. The rule was firing in all clinical settings—inpatient, outpatient, surgicenter, urgent care, and emergency department—and for providers who did not typically address conditions such as metabolic syndrome. The rule was designed by a psychiatrist and the CDSS committee's information systems staff and was not vetted by the committee's other clinical members. The creation and maintenance of the rule was outsourced to the EHR vendor and staffed by technical specialists who had little to no clinical expertise. All of this contributed to the confusion when reports of problems with the rule were being submitted. The rule was firing for all medications in the antipsychotic drug class, not just second-generation antipsychotics shown to significantly contribute to metabolic syndrome. Despite a lot of testing in nonproduction domains, creating a CDSS rule takes a lot of time and requires an iterative agile process to adequately address problems and improve the rule. Acknowledging that a patient may already have clinical conditions (e.g., diabetes, hypertension, obesity) documented in the problem list can help make a CDSS rule “smarter” and increase its level of credibility—and thus increase the receiving clinician's trust in the rule and its purpose. Attaching guidelines and literature to a CDSS rule request can help the development team ensure the rule is evidence based. A rule that recommends an action can be made more convenient if the action is available inside the alert. The information provided is then a convenience rather than a burden.