Current Clinical Decision Support Systems Research Articles

The use of artificial intelligence to optimize medication alerts generated by clinical decision support systems: a scoping review.

Current Clinical Decision Support Systems (CDSSs) generate medication alerts that are of limited clinical value, causing alert fatigue. Artificial Intelligence (AI)-based methods may help in optimizing medication alerts. Therefore, we conducted a scoping review on the current state of the use of AI to optimize medication alerts in a hospital setting. Specifically, we aimed to identify the applied AI methods used together with their performance measures and main outcome measures. We searched Medline, Embase, and Cochrane Library database on May 25, 2023 for studies of any quantitative design, in which the use of AI-based methods was investigated to optimize medication alerts generated by CDSSs in a hospital setting. The screening process was supported by ASReview software. Out of 5625 citations screened for eligibility, 10 studies were included. Three studies (30%) reported on both statistical performance and clinical outcomes. The most often reported performance measure was positive predictive value ranging from 9% to 100%. Regarding main outcome measures, alerts optimized using AI-based methods resulted in a decreased alert burden, increased identification of inappropriate or atypical prescriptions, and enabled prediction of user responses. In only 2 studies the AI-based alerts were implemented in hospital practice, and none of the studies conducted external validation. AI-based methods can be used to optimize medication alerts in a hospital setting. However, reporting on models' development and validation should be improved, and external validation and implementation in hospital practice should be encouraged.

Comparison of Clinical Importance of Drug Interactions Identified by Hospital Pharmacists and a Local Clinical Decision Support System.

Drug-drug interactions (DDIs) may cause adverse drug events, potentially leading to hospital admission. Clinical decision support systems (CDSSs) can improve decision-making by clinicians as well as drug safety. However, previous research has suggested that pharmacists are concerned about discrepancies between CDSSs and common clinical practice in terms of severity ratings and recommended actions for DDIs. The primary objective was to characterize the level of agreement in terms of DDI severity ranking and actions recommended between the local CDSS and pharmacists. The secondary objectives were to determine the level of agreement among pharmacists concerning DDI severity, to determine the influence of the CDSS on clinicians' decision-making, and to review the literature supporting the severity rankings of DDIs identified in the study institution's database. This 2-part survey study involved pharmacists and pharmacy residents working at 1 of 4 health organizations within the Lower Mainland Pharmacy Services, British Columbia, who were invited to participate by email. Participants were first asked to rank the severity of 15 drug pairs (representing potential DDIs) on a 5-point Likert scale and to select an action to manage each interaction. Participants were then given the CDSS severity classification for the same 15 pairs and again asked to select an appropriate management action. Of the estimated 500 eligible pharmacists, a total of 73 pharmacists participated, for a response rate of about 15%. For DDIs of moderate severity, most participants chose to monitor. For severe and contraindicated interactions, the severity ranking and action proposed by participants varied, despite the same severity classification by the CDSS. There was poor agreement among respondents about the severity of the various DDIs. Moreover, knowledge of the CDSS severity ranking did not seem to change the actions proposed by most respondents. This study identified a gap between the local CDSS and clinical practice. There were discrepancies in terms of severity rankings and actions proposed to manage DDIs, particularly for severe and contraindicated DDIs. The current CDSS did not appear to have a large impact on clinical decision-making, which suggests that it may not be functioning to its full potential.

Looking through the eyes of the multidisciplinary team: the design and clinical evaluation of a decision support system for lung cancer care.

BackgroundDecision-making in lung cancer is complex due to a rapidly increasing amount of diagnostic data and treatment options. The need for timely and accurate diagnosis and delivery of care demands high-quality multidisciplinary team (MDT) collaboration and coordination. Clinical decision support systems (CDSSs) can potentially support MDTs in constructing a shared mental model of a patient case. This enables the team to assess the strength and completeness of collected diagnostic data, stratification for the right personalized therapy driven by clinical stage and other treatment-influencing factors, and adapt care management strategies when needed. Current CDSSs often have a suboptimal fit into the decision-making workflow, which hampers their impact in clinical practice.MethodsA CDSS for multidisciplinary decision-making in lung cancer was designed to support the abovementioned goals through presentation of relevant clinical data in line with existing mental model structures of the MDT members. The CDSS was tested in a simulated multidisciplinary tumor board meeting for primary diagnosis and treatment selection, based on de-identified primary lung cancer cases (n=8). Decision course analysis, eye-tracking data and questionnaires were used to assess the impact of the CDSS on constructing shared mental models to improve the decision-making process and outcome.ResultsThe CDSS supported the team in their self-correcting capacity for accurate diagnosis and TNM classification. It enabled cross-validation of diagnostic findings, surfaced discordance between diagnostic tests and facilitated cancer staging according the diagnostic evidence, as well as spotting contra-indications for personalized treatment selection.ConclusionsThis study shows the potential of CDSS on clinical decision making, when these systems are properly designed in line with clinical thinking. The presented setup enables assessment of the impact of CDSS design on clinical decision making and optimization of CDSSs to maximize their effect on decision quality and confidence.

The value of participatory development to support antimicrobial stewardship with a clinical decision support system.

Current clinical decision support systems (CDSSs) for antimicrobial stewardship programs (ASPs) are guideline- or expert-driven. They are focused on (clinical) content, not on supporting real-time workflow. Thus, CDSSs fail to optimally support prudent antimicrobial prescribing in daily practice. Our aim was to demonstrate why and how participatory development (involving end-users and other stakeholders) can contribute to the success of CDSSs in ASPs. A mixed-methods approach was applied, combining scenario-based prototype evaluations (to support verbalization of work processes and out-of-the-box thinking) among 6 medical resident physicians with an online questionnaire (to cross-reference findings of the prototype evaluations) among 54 Dutch physicians. The prototype evaluations resulted in insight into the end-users and their way of working, as well as their needs and expectations. The online questionnaire that was distributed among a larger group of medical specialists, including lung and infection experts, complemented the findings of the prototype evaluations. It revealed a say/do problem concerning the unrecognized need of support for selecting diagnostic tests. Low-fidelity prototypes of a technology allow researchers to get to know the end-users, their way of working, and their work context. Involving experts allows technology developers to continuously check the fit between technology and clinical practice. The combination enables the participatory development of technology to successfully support ASPs.

Development of a Standardized Rating Tool for Drug Alerts to Reduce Information Overload.

A well-known problem in current clinical decision support systems (CDSS) is the high number of alerts, which are often medically incorrect or irrelevant. This may lead to the so-called alert fatigue, an overriding of alerts, including those that are clinically relevant, and underuse of CDSS in general. The aim of our study was to develop and to apply a standardized tool that allows its users to evaluate the quality of system-generated drug alerts. The users' ratings can subsequently be used to derive recommendations for developing a filter function to reduce irrelevant alerts. We developed a rating tool for drug alerts and performed a web-based evaluation study that also included a user review of alerts. In this study the following categories were evaluated: "data linked correctly", "medically correct", "action required", "medication change", "critical alert", "information gained" and "show again". For this purpose, 20 anonymized clinical cases were randomly selected and displayed in our customized CDSS research prototype, which used the summary of product characteristics (SPC) for alert generation. All the alerts that were provided were evaluated by 13 physicians. The users' ratings were used to derive a filtering algorithm to reduce overalerting. In total, our CDSS research prototype generated 399 alerts. In 98 % of all alerts, medication data were rated as linked correctly to drug information; in 93 %, the alerts were assessed as "medically correct"; 19.5 % of all alerts were rated as "show again". The interrater-agreement was, on average, 68.4 %. After the application of our filtering algorithm, the rate of alerts that should be shown again decreased to 14.8 %. The new standardized rating tool supports a standardized feedback of user-perceived clinical relevance of CDSS alerts. Overall, the results indicated that physicians may consider the majority of alerts formally correct but clinically irrelevant and override them. Filtering may help to reduce overalerting and increase the specificity of a CDSS.

Design and Evaluation of an Ontology-based Drug Application Database

Several recently published cases of preventable adverse drug reactions were associated with flaws in drug application. However, current clinical decision support (CDS) systems do not properly consider drug application issues and thus do not support effective prevention of such medication errors. With the aim to improve CDS in this respect, we developed a comprehensive model precisely describing all aspects of drug application. The model consists of 1) a schema comprising all relevant attributes of drug application and 2) an ontology providing a hierarchically structured vocabulary of terms that describe the possible values of the schema's attributes. Finally, medical products were annotated by a semi-automatic term assignment process. For evaluation, we developed an algorithm that uses our model to compute a meaningful similarity between medicinal products with respect to their drug application characteristics. Our schema consists of 22 attributes. The ontology contains 248 terms, textual descriptions, and synonym lists. More than 58,700 medicinal products were automatically annotated with >386,600 terms. 2,450 drugs were manually reviewed by experts, adding >4500 terms. The annotation and similarity measure allow for (similarity) searches, clustering, and proper discrimination of drugs with different drug application characteristics. We demonstrated the value of our approach by means of a set of case studies. Our model enables a detailed description of drug application, allowing for semantically meaningful comparisons of drugs. This is an important prerequisite for improving the ability of CDS systems to prevent prescription errors.

Successful strategy to improve the specificity of electronic statin–drug interaction alerts

A considerable weakness of current clinical decision support systems managing drug-drug interactions (DDI) is the high incidence of inappropriate alerts. Because DDI-induced, dose-dependent adverse events can be prevented by dosage adjustment, corresponding DDI alerts should only be issued if dosages exceed safe limits. We have designed a logical framework for a DDI alert-system that considers prescribed dosage and retrospectively evaluates the impact on the frequency of statin-drug interaction alerts. Upper statin dose limits were extracted from the drug label (SPC) (20 statin-drug combinations) or clinical trials specifying the extent of the pharmacokinetic interaction (43 statin-drug combinations). We retrospectively assessed electronic DDI alerts and compared the number of standard alerts to alerts that took dosage into account. From among 2457 electronic prescriptions, we identified 73 high-risk statin-drug pairs. Of these, SPC dosage information classified 19 warnings as inappropriate. Data from pharmacokinetic trials took quantitative dosage information more often into consideration and classified 40 warnings as inappropriate. This is a significant reduction in the number of alerts by 55% compared to SPC-based information (26%; p < 0.001). This retrospective study of pharmacokinetic statin interactions demonstrates that more than half of the DDI alerts that presented in a clinical decision support system were inappropriate if DDI-specific upper dose limits are not considered.