PHASE-ED: Recorded Emergency Department Stay and Temporal Prediction Across Pandemic Phases at Jordanian Public Hospitals

When a patient arrives at a hospital emergency room, the clock starts ticking. This time, known as the length of stay, is more than just a number on a spreadsheet; it is a measure of how quickly a person moves from the moment they walk through the door to the moment they leave, either to go home or to a hospital bed. For doctors and hospital managers, tracking this duration is vital. It helps them understand if their systems are working smoothly or if patients are getting stuck in bottlenecks, waiting too long for care. In recent years, the global pandemic threw these systems into chaos, changing how many people showed up, how sick they were, and how hospitals responded. Now that the immediate crisis has passed, a critical question remains: have emergency rooms truly returned to normal, or has the way we measure time changed along with the way we treat patients?

A new study from Jordan, titled PHASE-ED, set out to answer this question by looking at the records of nearly 400,000 visits to a public hospital emergency department between 2018 and 2025. The researchers wanted to see how the time patients spent in the emergency room shifted before the pandemic, during the peak of the virus, and in the years following. They also wanted to test if simple computer models could predict how long a future patient would stay, based only on basic information like their age, gender, and the time of day they arrived. The goal was not just to count minutes, but to understand if the numbers told a true story of faster care or if they were hiding changes in how the hospital recorded its data.

The study began by cleaning and organizing a massive digital log of every patient who walked into the emergency room. The researchers focused on one specific hospital facility to ensure the data was consistent, looking at visits from the start of 2018 through the end of 2025. They divided this timeline into three distinct eras: the time before the pandemic, the years when the virus was actively spreading, and the recovery period after the official health emergency ended. By comparing the middle values of these groups, they found a clear trend. Before the pandemic, the typical patient stayed for 63 minutes. During the pandemic, that time dropped slightly to 60 minutes. In the recovery period, the typical stay fell further to 51 minutes.

However, the average time tells only part of the story. The researchers looked deeper into the data to see what happened to the patients who waited the longest. Before the pandemic, the longest 10 percent of stays lasted more than five hours and fifteen minutes. During the pandemic, this high-end wait time was still nearly five hours. But in the recovery period, the longest waits dropped dramatically to just under three hours. This suggests that while the typical patient saw a modest improvement, the most significant change happened for those who were stuck in the system the longest. The number of patients waiting more than six hours fell from over 8 percent before the pandemic to just over 1 percent in the recovery phase.

Yet, the researchers were careful not to assume that these shorter times meant the hospital had suddenly become much faster at treating people. They discovered a subtle but important shift in how the hospital recorded time. In July 2024, the way the system logged the exact moment a patient left changed, becoming more precise. The researchers found that the shorter stay times had already begun to appear before this technical change happened, which suggests the trend was real and not just a glitch in the recording system. Still, they noted that the data showed a strange pattern: the number of patients with an "unknown" final status dropped sharply in the recovery period. This raised the possibility that the way staff marked patients as "discharged" or "admitted" had changed, which could make the recorded times look shorter even if the actual medical care took just as long.

To see if they could predict these times in the future, the team built several computer models. They tried using a simple historical average, which simply guessed that every new patient would stay for the same amount of time as the average of the past. They also tried more complex methods, including advanced statistical techniques and machine learning algorithms that could spot patterns in age, gender, and the time of day. The results were surprising. The complex models did not perform much better than the simple historical average. The most sophisticated model reduced the prediction error by less than two minutes compared to the simple guess. This suggests that for this specific hospital, knowing a patient's age or gender did not help much in predicting how long they would stay. The time of day and the day of the week mattered, but the basic history of the hospital's performance was a surprisingly strong predictor.

The study also looked at how well these models could estimate the range of time a patient might stay, rather than just a single number. They wanted to know if they could give a window of time, such as "most patients will stay between 20 and 100 minutes," and be confident that this window was accurate. The simple historical model was good at this overall, covering the correct time range for about 91 percent of patients. However, when the researchers looked at specific groups, the picture became uneven. For patients marked as needing "Resuscitation"—the most critical category—the model was only correct about 69 percent of the time. This means that for the sickest patients, the predictions were often too wide or missed the mark entirely, highlighting that the system struggles to forecast the needs of the most urgent cases.

The researchers concluded that while the recorded time in the emergency room has definitely gotten shorter, especially for those who waited the longest, the reasons are not fully clear. The data shows a real change in the numbers, but it cannot prove whether this is because doctors are working faster, because the types of patients arriving have changed, or because the way time is written down has shifted. The study serves as a reminder that in a complex system like a hospital, a shorter number on a screen does not automatically mean better care. It suggests that hospital managers need to look beyond the total time and examine the specific steps of the process, such as how long it takes to see a doctor or how quickly a bed is found. Until those details are understood, the shorter stays remain a promising sign that needs further investigation rather than a solved problem.