November 12, 2021
Like Clockwork: Long Wait Times in the Emergency Department
When you think about an emergency department (ED), an image might come to mind of a horrible bus crash, where, without warning the ED is inundated by a dozen people who were injured in the tragedy. These events do happen. But they are (fortunately) rare. More typically, an ED operates with a regularity that might surprise much of the world. Wait times, at least on average, rise and fall each day like the sun.
Never have an emergency on a Monday evening
If you’re planning to have an emergency, a contradiction in terms I suppose, avoid Monday evening. And start early. Mornings have the lowest patient volume and shortest wait times. While your mom always said nothing good ever happens after 2 am, it’s more like “nothing good ever happens 6pm to midnight”. It’s the evenings that get us into trouble.
Part of this phenomenon is human nature but doesn’t necessarily involve taking part in risky physical activities. It’s more behavioral and nuanced than that. People are more likely to let health issues slide on the weekends (it’s “my time”) and take time off work on Monday to address. Staffing levels at nursing homes may be lower on the weekends, putting pressure onto the ED on Mondays.
Same with evenings. When talking to any individual patient, no one seems particularly culpable. Yet, there must be an aggregate decision along the lines of “I can’t take a loved one to the hospital because I have to work’”. Health issues tend to build over time and are addressed as it’s more convenient for patients and their caregivers.
These behaviors result in longer waiting times for people with acute emergencies. This pattern is regular. We can measure it across hospitals geographically, and by hospital type (urban, suburban, rural). ED wait times are a sinusoid, a wave (relatively) of quick morning waits to I’ve-been-here-6-hours in the evening, day after day after day. Human nature and their patterns of activity, it seems, are almost as sure as the sun.
Predicting wait times
While the average wait time across hospital emergency rooms is quite regular, the wait time for an individual is highly variable. Elderly and might have had a stroke? You won’t (or shouldn’t) wait at all. In your 20s and think you might have fractured your ankle but aren’t bleeding out of your eyes? You’re going to wait. Emergency medicine uses a 5-point scale called the Emergency Severity Index (ESI). 1 & 2 are imminent or possible threats to life or limb. 4 & 5 are more urgent care levels. 70% of people get placed in the middle - ESI 3, the “I’m not sure what’s wrong yet” category.
Predicting wait times, then, requires predicting ESI, which in turn requires predicting the number of resources (lab tests, X-rays, etc.) a patient might need. That’s complex! At Vital, we’re using hundreds of variables, including vital signs, age, chief complaint, and other patient-specific variables, along with incoming volumes, rates of discharge, numbers of free beds, and other aggregate statistics. Of these, one might think “number of beds” was static. It’s not. When busy, EDs open “hall spots” or double up on beds per room. Other times, there may be 60 beds in the ED, but nursing staff for only 40. This makes determining how long you will wait on any given night an incredibly hard problem.
After two years, dozens of AI models, and a new form of “adaptive AI” that changes based on the data available for a patient and for a hospital (some patients have a lot of data, others very little; same with hospitals), Vital can now predict wait time ranges with 90%+ accuracy. And not just total length of stay: we predict time to triage, time to bed, time to see a doctor, and soon time for each and every lab and imaging test.
If you do have an emergency, even if it’s a Monday evening, Vital can tell you how long you’ll wait and what you can do to speed your visit along (hint: pee in a cup as soon as possible, and carry around a list of your medications and allergies). Good luck out there.