IF YOUR HOSPITAL is like most others, it’s measuring patient census at midnight, the traditional marker used to gauge how many patients are hospitalized.
Why has midnight census become such a universal metric? “It’s probably a measure of convenience,” says Logan Pierce, MD, a hospitalist with the University of California, San Francisco. “You want to pick a point in time that can apply to all the different services in the hospital, and midnight sounds as good as any.”
Midnight census may also be, Dr. Pierce adds, “the easiest to report on, and it takes a little more work to report at a different time.” It turns out that he put in that additional work when building a dashboard for UCSF’s hospital medicine service.
“You could target shift resources to when discharges are happening, which isn’t necessarily the normal 9-to-5 workday.”
~ Logan Pierce, MD
University of California, San Francisco
Service leaders asked Dr. Pierce—who is also assistant director of Data Core, a UCSF data acquisition group that analyzes data out of the center’s Epic EHR with a very clinical focus—to track hospitalist census at 7 a.m., not at midnight, a figure that UCSF was already recording.
Once that dashboard was up and running, “we found a consistent delta of about five or six patients” between the midnight census and that at 7 a.m., he explains. “It didn’t take people long to start asking, ‘Why do we keep seeing this difference?’ ”
That internal question led him and his research colleagues to begin tracking hourly census throughout the day, using retrospective EHR data to look back years. With Dr. Pierce as lead author, they wrote up their findings in a study on hourly census variations that was posted online by the Journal of Hospital Medicine in November 2023.
While previous studies have looked at how patient volumes surge and wane over the course of a year, this research is—surprisingly—the first to track average census variations during the day. The authors’ conclusion: “Measuring ADC at midnight, as traditionally done, may underestimate workload and therefore contribute to staffing shortages and physician burnout.”
Quantifying census changes
According to Dr. Pierce, most clinicians have a solid intuitive feel for when their daily patient load gets heavier and when it eases up. In fact, when the study was first published, other doctors wondered if diurnal patterns of highs and lows that are so obvious were even worth studying.
But Dr. Pierce notes that he wanted to quantify how much hospitalist census changes over the course of the day. “We know intuitively when it will be busier, but I couldn’t have said how much busier by how many patients,” he says. And while clinicians may grasp daily census ebbs and surges, that’s not true for administrators.
Once he started tracking his group’s 7 a.m. census, for instance, Dr. Pierce had to spend hours explaining those findings to mostly non-clinicians at UCSF who were surprised at how different—and how predictably different—the 7 a.m. census was from that at midnight.
“They just assumed that census evened out every day,” he points out. “I said, ‘No, there are definitely patterns,’ and then I put together this analysis. Once they saw this very predictable kind of sine wave, everybody said, ‘Oh yeah, that makes sense.’ ”
Dr. Pierce was himself surprised by how set the hourly patterns were. “I did not expect such a nice, neat curve that just didn’t change no matter what year or month you ran it,” says Dr. Pierce. While he and his research colleagues tracked hospital medicine’s hourly census over three years, the JHM study reports data from only 2022.
Implications of hourly variations
At UCSF, the study found, hospitalists’ average daily census peaked at 11 a.m. and was the lowest at 7 p.m., with an average delta between those two times of 9.1 patients. (The study recommends that institutions track census between 7 a.m. and 11 a.m. to find the time when it’s highest.) Meanwhile, admission rates surged in the afternoon and evening, peaking at 9 p.m., while discharge rates were highest during daytime hours and peaked at 4 p.m.
Tracking such variations is key, says Dr. Pierce, to make the case to administrators for either more staffing or for changing shift times to better match staffing to actual workload. After all, funding for many hospitalist programs, including his own, is based on average daily census.
“Our funding comes from our health system, not directly from billing for the care we provide, and I think that’s relatively common for hospital medicine groups,” he says. “Measuring your average daily census is very important because there are hard cutoffs for when you might open a new team.”
Based on the results he and his team found, UCSF approved funding for an additional dedicated admitter during the afternoon. (While UCSF always maintains dedicated hospitalist admitters, only one or two may be on during slower portions of the day, ramping up to three or four during the busiest.) That additional admitter now works an eight-hour shift from 2 p.m. to 10 p.m., allowing the other day hospitalists to concentrate on discharges.
While that’s the one change UCSF has made, Dr. Pierce points out that other groups might use such an analysis to beef up or change their jeopardy systems and potentially increase their number of doctors on-call.
Another possible change to consider: linking census variations to the shifts of case managers and social workers to better facilitate discharges.
“If discharges don’t really start happening until 11 a.m. and peak at 4 p.m., it doesn’t make sense to have case managers in-house from 8 a.m. until 3 p.m.,” Dr. Pierce points out. “You could target shift resources to when discharges are happening, which isn’t necessarily the normal 9-to-5 workday.”
Weekday and seasonal changes
The research also revealed seasonal variations that will come as no surprise to hospitalists, falling to a low during late spring and early summer and spiking in early winter.
As for census variations over the days of the week, “the biggest difference was a delta of six patients between Saturday and Tuesday,” says Dr. Pierce. “That’s probably driven by a falloff in the number of surgeries and by patients not being sent to the hospital from PCPs.”
He also notes that surgeons have historically had a much better handle on their own census variations to provide tighter OR management, and they have long relied on dynamic staffing. He also points out that surgical schedules are more predictable than hospitalist workloads.
Dr. Pierce recommends that other institutions, as well as specific service lines within them, run the same analysis to see how their census changes over the course of a day.
“If I was in another faculty—cardiology, say, or emergency medicine—I would probably do this same study,” he says. “Census peaks and troughs for other services during the day are probably different than ours.”
Phyllis Maguire is Executive Editor of Today’s Hospitalist.