Table 1.

Benefits of clinical decision support systems (CDSS), possible harms, and evidence-based mitigation strategies.

1. Functions and advantages of CDSS	2A. Potential harm of CDSS	2B. Solution(s) to mitigate harm	2C. Explanation of solution(s)
Patient Safety Reducing incidence of medication/prescribing errors and adverse events.	Alert fatigue A phenomenon where too many insignificant alerts or CDSS recommendations are presented, and providers start to dismiss them regardless of importance.	Prioritize critical alerts, minimize use of disruptive alerts for non-critical indications.	Alert fatigue might be thwarted by prioritizing and selecting alerts that are critically important, that will have the greatest impact, and by tailoring alerts to specific specialties and severities (personalization).109 DDI testing software should ideally be programmed with an algorithm that incorporates concomitant medication, lab values, patient demographics, and administration times, to be as specific as possible.18
Clinical management Adherence to clinical guidelines, follow-up and treatment reminders, etc.	Negative impact on user skills One example is reliance on, or excessive trust in the accuracy of a system.	Avoid prescriptiveness in system design. Evaluate system impact on an ongoing basis.	Systems should be set up to be useful to clinicians, without jeopardizing autonomy or being too ‘prescriptive’ and definitive. It is important to conduct analysis to see how the system is being used in the long term, after implementation. If accuracy is an issue, design changes might need to be taken to prompt extra checks or confirmation of orders.85
Cost containment Reducing test and order duplication, suggesting cheaper medication or treatment options, automating tedious steps to reduce provider workload, etc.	Financial challenges Setup can be expensive (capital or human resource), and long-term cost-effectiveness is not guaranteed.	Design and plan for longitudinal cost analysis at the outset. Specify measurements for non-financial benefits where possible.	An analysis should be done to determine if the costs are justified and if there is a good return on investment.110 Cost analysis is notoriously missing in the literature, but examples can be found.107,111,112 Payers may be more willing to support CDSS if cost-savings can be shown elsewhere in the system / process. This means looking at more than just direct costs a using metrics such as patient outcomes or quality-adjusted life years (QALY).
Administrative function/automation Diagnostic code selection, automated documentation and note auto-fill.	System and content maintenance challenges As practice changes, there can be difficulty keeping the content and knowledge rules that power CDSS up to date.	(1) Knowledge Management (KM) Service in place, with a focus on translation to CDSS systems. (2) System for measurement and analysis of CDSS performance.	(1) Facilitates scheduled review, methods for acquiring and implementing new knowledge, and streamlined processes for gathering physician feedback on the system as well as training users on why certain data entry and standardization of data entry practices. Standards for organizing KM management have been published.113,114 (2) It is important to identify changes in performance and use over time. In addition, the quality of the data repository should be monitored and it is also important to ensure that conclusions are not being made on corrupted or poor quality data beforehand.115
Diagnostics support Providing diagnostic suggestions based on patient data, automating output from test results.	User distrust of CDSS Users may not agree with the guideline provided by the CDSS.	Reference expert knowledge—include scientific references in messages where appropriate.	To provide a verifiable source of information to the user on why the recommendation exists.116 In addition to increasing trust, this may provide direction for users to update their knowledge in case they were not aware of the recommendation. Many systems also query reasons for not following a recommendation in order to elucidate the source of mistrust.117 This is a good idea, but should not be mandatory or ‘bulky’ in design.
Diagnostics Support: Imaging, Laboratory, and Pathology Augmenting the extraction, visualization, and interpretation of medical images and laboratory test results.	Transportability/interoperability CDSS face challenges regarding integration with other hospitals or systems, making it inefficient for otherwise high-quality systems to be disseminated and scaled.	(1) Adoption of industry standards. (2) Secure cloud services and blockchain.	Major open standards for structural and semantic interoperability and exchange continue to be developed and improved by organizations such as Health Level 7 International (HL7),118 SNOMED International,119 Digital Imaging and Communications (DICOM) for imaging standards, and many others.120 As much as possible, these standards should be adopted at all levels within the healthcare organization, and with the external systems being used. Cloud-based EHR architecture allows for more open architecture, and flexible connectivity between systems. As with any medical system, security must be assured through compliance with legislation such as Health Insurance Portability and Accountability Act (HIPAA) in the USA, Personal Information Protection and Electronic Documents Act (PIPEDA) in Canada, and the Data Protection Directive and General Data Protection Regulation (GDPR) in Europe. In the future, we may also see blockchain used to enable greater interoperability and improve security for health information exchange (HIE).121,122
Patient decision support Decision support administered directly to patients through personal health records (PHR) and other systems.	Dependency on computer literacy CDSS may require a very high technological proficiency to use	(1) Conform to existing functionality. (2) Adequate training made available at launch.	(1) Maintaining consistency with the user interface of the pre-existing system (if there is one) is crucial to ensure users don’t have a steep learning curve to use the system. (2) Adequate training should be available and easily accessible for users. Training should ideally be done in person by a clinician leader with vast EHR experience to generate buy-in.123 Training needs to be available on an ongoing basis, as new staff and users join. One strategy is to have on-site team members designated as elite users, and capable of providing training sessions.
Better Documentation	Inaccurate and poor-quality data/documentation CDSS may aggregate data from multiple sources that are not synced properly. Users may develop manual workarounds that compromise data.	(1) Expert Knowledge of interlinked systems. (2) IT testing/debugging during development and implementation stage.	The team needs to be familiar and have expert knowledge of all external systems that feed data into the database used by the CDSS. Experts recommend testing clinical rules for PPV and NPV during the process of development and implementation.109 If user generated data is an issue, it may be that physicians have not received the proper training on how to read, interpret and respond to alerts, or are depending on pharmacists to check medication orders before dispensation.124,125
Workflow improvement CDSS can improve and expedite an existing clinical workflow in an EHR with better retrieval and presentation of data.	Disrupted/fragmented workflow CDSS can also disrupt existing workflows if they require interaction external to the EHR, or don’t match the providers’ real world information processing sequences.	(1) Usability evaluation. (2) Workflow modeling.	(1) Rigorous and iterative usability evaluations and pilot testing should be conducted on CDSS before using them in clinical settings. Many usability assessment tools are available, along with other quantitative methods and frameworks.126–129 (2) Unless a goal of the CDSS is to change the care process, the CDSS should be designed to fit within or conform to the current user workflows.