Clinical decision support systems: Potential with pitfalls

From Clinfowiki
Jump to: navigation, search

Eberhardt, J, Bilchick, A, Stojadinovic, A. (March, 2012). Clinical decision support systems: Potential with Pitfalls. Journal of Surgical Oncology, Vol#105, 502-510 DOI: 10.1002/jso.23053. [1]

The article revises the strength and weakness of CDSS and how these systems are developed and evaluated for clinical use.

Objective of the review

The objective of the article is to share experience acquired through several development and implementation processes, so that future developments and implementations apply best practices and avoid potential obstacles. The article focuses on an algorithmic clinical decision support system (ACDSS) that is used in diagnostic interpretation, treatment planning, and therapy recommendations. According to Eberhardt, Bilchick and Stojadinovic (2012), " One way to distinguish ACDSS from their broader environment is that an algorithmic system uses validated statistical algorithms to narrow decisions to an "acceptable" range prior to presentation" In reviewing ACDSS, the article attempts to address four points such as:

  1. Definition of CDSS;
  2. Common approaches to this system;
  3. The strength and weaknesses of CDSS,
  4. Evaluation and development of CDSS

A Definition of CDSS and ACDSS by Association

There is broadness and heterogeneity in defining CDSS; however, this review tried to provide clear definition by identifying current and widely accepted ones. In doing so, the article quotes the AHRQ definition as follows. "Common features of CDS systems that are designed to provide patient-specific guidance include the knowledge base (e.g., compiled clinical information on diagnoses, drug interactions, and guidelines), a program for combining that knowledge with patient-specific information, and a communication mechanism—in other words, a way of entering patient data (or importing it from the EMR) into the CDS application and providing relevant information (e.g., lists of possible diagnoses, drug interaction alerts, or preventive care reminders) back to the clinician" Further, the article described CDSS as a system that consists of three components such as: Knowledge, Program and User-interface. And, for this system to be useful the article quotes Osheroff et al’s "five rights:" "the right information, the right person, the right format through the right channel at the right time" must accompany the system.

Whiter Algorithmic Clinical Decision Support System (ACDSS)

According to the article, the ever-evolving clinical knowledge and diagnostic advance is calling for a system that assists clinician in developing personalized and evidence based care. The article noted, expansion in knowledge is causing reduced situational awareness and increased mental work load on clinicians. Here, ACDSS has showed great potential both in maximizing situational awareness and reduce irrelevant mental work load thereby allowing clinicians to focus on what is relevant in care delivery.

Issues in Decision Support

The article reported, even so it is widely accepted that ACDSS improves quality of care, there are mixed reports about the overall benefits of the system. And, these mixed reports have hampered the wide spread adoption of ACDSS. Consequently, the article enumerates four potential reasons that gave rise to mixed reports or limit the full capacity of ACDSS: "lack of focus on a specific clinical problem; selection of an inappropriate system or method; poor consideration of and integration with clinical workflow; and inadequate testing and user training." [1]

Algorithms: What Are They Good For? Absolutely Anything; And That Is The Problem.

According to the article, algorithms can be created for anything; however, when it comes to CDSS specificity of a design determines the quality of CDSS. In order to make ACDSS efficient, special attention must be paid on: problem identification, specific approach to the identified problem, and a corresponding CDSS solution that integrates in to a given workflow. When approaching ACDSS as a solution to a given clinical problem, there are generally three option to chose from: Knowledge base, Expert system, and Predictive algorithm. All these three have their own strength and weakness; therefore, it is up to the stakeholder’s to make a sound decision in choosing one for the problem at hand.

Evaluation and Testing My ACDSS: Is My Algorithm Ugly?

Power, Accuracy and Applicability. According to Eberhardt, Bilchick and Stojadinovic (2012), usually, ACDSS are validated based on power and accuracy. These two factors indicate if findings are statistically significant and if algorithms are accurate. Aside from that, the article reported applicability evaluates the robustness of an algorithm of CDSS when it is applied to a different population. A minimal variance in performance between populations indicates robustness of an algorithm.

Be careful What You Wish for: Building An ACDSS

The article warns, special attention should be paid during development or selection phase of ACDSS. For decision at this level tremendously affect the overall out come. Besides, including the right stakeholders to participate in the development phase will lead to a successful implementation outcome. In general, development of ACDSS can be broken-down into four parts: problem identification, system development, system evaluation and system implementation. Finally, it is worth noting that developing and maintaining ACDSS is an iterative process that repeats the four steps in a cyclic manner.


This article gives a nutshell summary of what CDSS/ACDSS is. It also describes the main components of the system as well as what steps need to be taken during development and implementation of a system. I would say it is a good summary for CDSS.


  1. 1.0 1.1