Unfortunately, alerts are a too common sound in our hospital’s landscape, too many alerts generate a form of desensitization. When specifically looking at Clinical decision supports systems (CDSS), these desensitizations have a direct correlation to negative outcomes for patients and an increase in mortality. So how do we address this? Working with an expert system. An experts system or best-of-breed CDSS should provide alerts when they are potentially actionable. Delivering alerts is something we take very seriously, we always conduct an in-depth analysis of an algorithm before integrating it. The more we know and the deeper the knowledge base, the less false alerts being produced.
When we speak to alert fatigue and how they evolve, we need to look more closely on how the rule sets within the system were built and by whom. Within an expert system, typically a rigorous process is undertaken to ensure the highest quality assurance is achieved; each rule-set should follow a strict method of design, analysis, testing and validation against a retrospective set of clinical data. It is throughout this process, the “fine-tuning” or rule specificity is being continually evaluated until a certain threshold is achieved and can then be deployed. The expertise behind the work typically consists of a multidisciplinary team of infectious disease physicians and pharmacists, epidemiologists, software engineers and artificial intelligence experts.
Besides rule specificity, our expert systems will have most evaluation rules predefined that make for a quick turnaround and deployment seeing improved patient outcomes and cost savings within the first 3 months. This is possible because we already built an extensive rule set. Yes, customization is always required to match local clinical practices but needs to fall under the same processes as mentioned above (design, analysis, testing, validation) to avoid contradictory rules and triggering false alerts. So be warned, an offering that provides “open” customization may sound appealing but will potentially create an environment of contradictory rules and false alerts, while consuming a lot your hospital’s time and resources. Typically, in this scenario, measurable impacts on quality clinical practices and patient outcomes aren’t coming as fast if they ever come.