Clinical decision support systems should be an unmixed blessing for busy healthcare workers. Who wouldn’t want a sharp-eyed sidekick on hand to spot patient safety issues or opportunities to apply clinical best practices? Instead, many clinicians have encountered systems that behave more like a clueless micromanager, popping in dozens of times a day to ask irrelevant questions and slow down urgent tasks.
Trust plummets. Burnout rises. Alert fatigue sets in.
Where does alert fatigue in healthcare come from?
“Implementing the first EMR-based clinical decision support system years ago, I've experienced the complexities of managing alert fatigue,” recalled Dr. Steven Lane, now Chief Medical Officer of Health Gorilla, in a recent webinar on optimizing treatment decisions with Synapse Medicine CEO Dr. Clément Goehrs, Pharmacy HIT Collaborative Executive Director Shelly Spiro and clinical pathologist Dr. Hung Luu.
Alert fatigue happens when health IT systems flood clinicians with so many warnings that they stop paying attention. The problem has been with us for decades, and it isn’t unique to healthcare. Before alert fatigue, alarm fatigue has long plagued high-stakes settings from hospitals to nuclear power plants. Wherever they work, people tend to become desensitized to an everyday onslaught of beeps, buzzes and flashing lights. Meanwhile, truly critical warnings disappear in the noise.
The dawn of the electronic medical record (EMR) brought alerts that were quieter, but often no less distracting—and sometimes even less helpful or relevant. One 2013 study found that on average, prescribers accepted only one in every thousand prescribing alerts they received.
It’s dangerous when clicking “dismiss” becomes the default. A 2018 study of medication-related alerts judged that about 40% of alert overrides were clinically inappropriate. In the worst cases, real patient harm brings alert fatigue into national headlines.
“As someone surrounded by healthcare professionals, I understand the common disregard for constant alerts on clinical conditions,” said Dr. Goehrs during the webinar. “This alert fatigue is a genuine issue.” The solution is complex, but he thinks about it through three simple lenses: quality, quantity and timing.
1. To solve alert fatigue, start with data quality
“Quality involves ensuring the information is up-to-date and from trusted sources,” said Dr. Goehrs. That can involve a tremendous amount of work behind the scenes. For instance, relevant medication alerts must start with a high-quality drug database—which he estimates requires about 10,000 updates every year. Especially for critical alerts like drug interactions, decisionmakers should look for data sources that are clinically sound and comprehensive, aggregating and synthesizing information from multiple official sources.
The data underlying clinical alerts must also be fine-grained enough to avoid false alarms. Drug interactions are a useful example here, too. “In the context of drug interactions and classification, there's an ongoing effort to refine our approach,” noted Shelly Spiro. “The United States Pharmacopeia, in its second five-year cycle, acknowledges the flaws in the current evidence-based medicine regarding medications. A key issue is the inaccuracy in drug classifications, which forms the basis for drug-drug interaction alerts. To address this, we've developed value sets focusing on individual medications rather than broad drug classes.”
More relevant alerts start here: with data that’s trusted, comprehensive and precise. A review of clinical validation studies of medication decision support systems found that in general, “clinical relevance would improve with higher data quality, by including more patient characteristics in the [clinical decision support systems] and by the selection of high impact or prevalent problems.”
That brings us to the second step in fixing alert fatigue: heightening impact by controlling quantity.
2. For efficient healthcare alerts, control the quantity
Research has shown that in general, alert acceptance falls as alert frequency rises. But the relationship isn’t always straightforward. One study found that it wasn’t necessarily the total number of alerts, but the number of repetitive alerts that caused override rates to rise.
It’s not always possible to reduce the amount of information that’s needed to make the best decision. But it’s usually possible to make that information far easier to take in. Fixing alert quantity can also be about “user experience and innovative UI designs to display drug interactions visually instead of in list form, making it easier for pharmacists and physicians to absorb the information quickly,” said Dr. Goehrs.
Combining related alerts into one simple module—instead of presenting them across many screens and clicks—also gives clinical decision support systems an opportunity to shine the brightest light on the biggest risks. During the prescribing process, this could look like:
- Sorting potential adverse drug events by frequency to display the most common adverse effects first
- Using color and order to visualize drug interactions by severity level
- Consolidating all alerts during a prescription into one summary of alerts
Considering quantity through the lens of user experience is crucial. When researchers at one health system applied principles of user-centered design, their alerts became significantly more effective. While the generic alerts never prompted prescribers to change their orders, 14% of the improved alerts did.
That’s a big increase, but there’s plenty of room to grow.
3. For the best user experience, fine-tune alert timing
“Timing is crucial—providing information when it's most needed,” said Dr. Goehrs. For example, prescribing support tools might take into account the provider’s typical prescribing habits, and prioritize alerts for drug indications or side effects that are new to them.
It’s also important that alerts fit into the larger sequence of clinical events. Each alert should be action-oriented. Synapse Medicine’s drug interaction alerts are designed to supply just enough context to aid an efficient but well-informed decision: the drugs involved, the mechanism behind the interaction and the recommended course of action.
Research backs up the power of adapting alerts to clinicians’ everyday routines. A study of clinical reminders in family medicine practices found success with design features such as syncing alerts to appointment times and letting clinicians choose when to attend to alerts. With these enhancements, “there was no decline in response to reminders over time, no decline in odds of response as number of reminders increased within a visit, and no decline with increasing patient complexity.”
Accordingly, alerts should be customizable to each clinical setting. For example, Synapse Medicine’s drug interaction alerts flag the most severe interactions by default to avoid over-alerting. Organizations then have the option to display less severe levels—but it’s important that this be a deliberate choice.
“Optimizing these systems involves a delicate balance between customization and standardization across different healthcare sectors,” pointed out Dr. Lane during the webinar. That balance is worth striving for. When we achieve it, we deliver clinical decision support that truly does support clinicians in delivering safer care.
To learn more about how Synapse Medicine and others across the industry are working to deliver the right amount of high-quality data at the right time, watch the on-demand webinar: “Not Too Much, Not Too Little: Wrangling The Right Amount of Data to Optimize Treatment Decisions.”
