Hospital length of stay (LOS) is one of the most important metrics we use to judge hospital efficiency and to predict whether the hospital is making money or losing money on different diagnoses. LOS is measured in days with each day defined as whether a patient is considered admitted to the hospital at midnight. This is the so-called midnight census. I believe that the midnight census is no longer a valid measurement for the calculation of the duration of hospitalization.
Never admit a patient between 10 PM and midnight
If your hospital judges or bonuses hospitalists based on length of stay, then those hospitalists know to avoid writing admission orders in the two hours before midnight. The simple reason is that when the midnight hour strikes, that patient is already considered to have been in the hospital for one day when using the midnight census of admitted patients to measure length of stay.
Consider two patients, patient A and patient B who both arrive in the emergency department with pneumonia on a Tuesday evening. The ER physician determines that both patients need to be admitted to the hospital and the on-duty hospitalist is called to the ER to write admission orders. Patient A has an admission order placed at 11:59 PM and patient B has an admission order placed at 12:01 AM, two minutes later. Both patients improve with medical treatment and are ready to be discharged on Friday. Patient A is discharged at 8 AM Friday morning and patient B is discharged at 4 PM Friday afternoon. By using the midnight census to measure duration of hospitalization, patient A has a length of stay of 3 days and patient B has a length of stay of 2 days. However, patient A was actually hospitalized for 56 hours and patient B was actually hospitalized for 64 hours. Using the midnight census measurement, patient B’s hospitalization was 33% shorter than patient A’s but based on hours in the hospital, patient B’s hospital stay was 14% longer than patient A’s.
Hospitalists are aware of this and if they are judged by the number of midnights their patients are in the hospital, they will delay writing an admission order until after midnight whenever feasible in order to improve their LOS numbers.
“I’m getting my discharge orders written earlier in the day, so why isn’t my length of stay improving?”
Hospital administrators want to have patients discharged as early in the day as possible so that rooms can be cleaned and ready for the next bolus of hospital admissions. By using the midnight census, a patient’s length of stay will be the same whether that patient is discharged at 7:00 AM or 5:00 PM. Thus initiatives to get patients discharged earlier in the day will not affect the length of stay as measured by the midnight census.
In order to measure hospital efficiency, the hospital must measure both the length of stay and the time of day of discharge. However, the time of day of discharge is also fraught with flaws. For example, if a hospital bonuses its hospitalists on earlier discharge orders, the hospitalists may hold off on discharging a patient who is ready for discharge in the late afternoon and instead discharge them early the following morning so that their numbers look good. Additionally, depending on when a patient was admitted to the hospital, a patient discharged in the late afternoon may actually have a shorter duration of stay (in hours) than a patient discharged early in the morning. In that case, you don’t want to penalize the hospitalist for getting the patient out of the hospital faster, simply because that patient was discharged in the afternoon.
So, why use the midnight census to measure length of stay?
Hospitals have used the midnight census for decades. In the pre-computerization era, it was the most easy and reliable way to know how many patients were in the hospital – unit clerks or nursing supervisors would write down the number of admitted patients on each nursing unit at midnight and then report that to the hospital administration the following morning. That was also an era when hospitals typically ran at a lower capacity with the result that there were always empty beds to admit patients to and consequently, there was not pressure to get patients discharged as early in the day as possible.
In the pre-computerization era, it was difficult to track the time of day that a patient was discharged since it required someone to manually go through each patient’s paper chart to collect the time of day of that patient’s admission and discharge; many doctors did not enter the time of day that they hand wrote their orders and many nurses did not enter the time of day that they took those orders off of the patients’ charts. Electronic medical records have changed all of that and now the exact time an admission or discharge order is placed and acted on can be measured with a keystroke. Yet, the midnight census remains as a hold-over from the pre-computer era.
In addition, before the institution of diagnosis-related groups (DRGs) by Medicare in 1983, it really did not matter how long a patient was in the hospital since the hospital was usually paid by number of days that a patient was in the hospital. As a result, the longer the length of stay, the more the hospital got paid. With DRGs, hospitals got paid based on a patient’s diagnosis and not based on the length of stay. Therefore, hospitals became motivated to shorten the length of stay in order to reduce their expenses for each patient. Once again, the midnight census remains a hold-over from the pre-DRG era.
The institution of DRGs was also a turning point for the time of day that patients were hospitalized. Prior to DRGs, most hospital admissions were elective admissions and those patients often had pre-planned testing and treatments and were usually admitted to the hospital in the late morning or early afternoon. Nighttime emergency admissions through the ER were less frequent. With daytime elective admissions predominating, the midnight census was a reasonably good measure of length of stay. DRGs brought an end to most elective medical admissions with a shift to the overwhelming majority now being admitted through the emergency department with the peak in ER admissions typically in the late afternoon or early evening. With that shift, the midnight census became a less accurate metric for measuring actual length of stay.
Length of stay should be measured in hours and not in days
The midnight census is a satisfactory measure in patients with a very long length of stay – if a patient is in the hospital for 50 days, then whether that is actually 49 days or 51 days has little impact on hospital efficiency. But as the hospital length of stay becomes shorter, the midnight census becomes a less accurate measurement. Given the flaws of using the midnight census to measure length of stay, I believe that we should move to measuring LOS by the hour. Our electronic medical records makes hourly measurement quite easy.
However, there are two types of hours in the hospital – daytime hours and nighttime hours. During the daytime, hospitalists do daily patient rounds, diagnostic tests are performed, surgeries occur, and consultants evaluate patients. During the nighttime, patients receive medications but the other daytime activities do not take place. In other words, more of the stuff that needs to happen in order to evaluate and treat the patient happens during the daytime hours. For this reason, a patient will spend fewer total hours in the hospital if admitted early in the daytime than if admitted early in the nighttime. Therefore, to accurately assess hospital efficiency, length of stay should be measured in both total hours of hospitalization and daytime hours of hospitalization.
The advantages of using total and daytime hours of hospitalization, rather than the midnight census, to measure length of stay include:
- A more accurate measure of duration of patient hospitalization, especially for shorter duration hospital admissions
- A more accurate measure of duration of observation stays which are inherently ultra short-duration stays
- Elimination of the measurement bias that occurs with nighttime admissions as opposed to daytime admissions
- Better representation of the effect of early-in-the-day discharge initiatives on length of stay
- Better identification of individual hospitalists or hospitalist groups that could benefit by patient throughput efficiency training
The biggest barrier is the length of stay index
Hospitals benchmark their length of stay to other hospitals using the length of stay index. If a hospital’s length of stay for a given DRG diagnosis is 4 days and the average of hospitals across the country for that diagnosis is also 4 days, then that hospital’s length of stay index is 1.0 and the hospital has an average length of stay for that diagnosis. If the length of stay index is 1.2, then the hospital requires more inpatient days for that diagnosis and likely has greater expenses per admission. However, if the length of stay index is 0.9, then the hospital is able to treat that diagnosis with fewer inpatient days and likely has lower expenses per admission.
Hospital length of stay benchmarks use the midnight census for length of stay calculation and as long as benchmarks continue doing so, any given hospital will need to continue to measure and report midnight census-based length of stay measurements to determine how that hospital is performing compared to other hospitals.
Nationwide change to an hour-based length of stay measurement (and thus length of stay index measurement) will not happen quickly – the midnight census measurement is just too entrenched in administrative practice and data reporting. However, a hospital that internally uses an hourly measure of length of stay will have a more accurate measurement of its own efficiency and that data can be used gain a competitive advantage.
It is time to move past the midnight census.
August 20, 2021