Length of Stay: Understanding its Shortcomings

Length of stay does not correlate directly with costs.

I am sure many of you have heard me rant about observation rates before. I’m sure at least half of you have been told that your observation rate is too high and that you need to fix it. “You’re above the benchmark rate” is what you probably heard. But of course, there is no benchmark rate, nor is there a standard way to measure a hospital’s observation rate. I have even written about it here on RACmonitor enews, where I discussed “Hirsch’s Law.” If you are unfamiliar with that law, you may want to read the article to ensure you are not a lawbreaker.

I have decided to add a new item to my rant list, which I think is even more misunderstood, and that is length of stay. “Our length of stay is too long, you need to fix it,” is something often heard. “Dr. Smith’s length of stay is too long; he needs to address it.” “Why is that pneumonia patient still here?” “The Medicare geometric mean length of stay (GMLOS) for their assigned DRG is four days, and they are still here on the fifth.” I bet every one of you have heard some version of those remarks. So, let’s break it down.

First, what are we measuring? What constitutes a patient’s length of stay? From where is your hospital data calculator pulling its data? Is it the date the patient registered in the ED? The date the inpatient admission order is written? Some manually calculated number that is entered when the claim is coded? Is the way you calculate the length of stay the same as that of the hospital to which you are comparing your data?

Length of stay is generally calculated in days based on the patient’s location and status at midnight, so what is the length of stay for the patient who comes to the ED and checks in late on Monday night, gets an order for observation on Tuesday, is admitted as inpatient on Wednesday, and then goes home on Friday? Is it four days, since the patient was physically in the hospital for four midnights, or is it three days, since that first midnight was in the ED and doesn’t count? Or is it two days, since that’s the number of inpatient days? If length of stay is calculated based on inpatient days and I was a doctor who wanted to make my length of stay look really good, I’d put every patient in observation for a couple of days before agreeing to write an admission order. That doctor’s number would look great, but at what cost?

And what do we use as a comparison? We often use the Medicare published geometric mean length of stay (GMLOS) for every DRG as the standard. But how does the Centers for Medicare & Medicaid Services (CMS) calculate that? They must calculate it from claims data, as they do when calculating payment rates. An inpatient claim has data fields for the date of the start of care, the admission date, and the date of the end of care. But the start of care will also include any services provided in the three-day payment window, so if that date is used, it may artificially add days. If the CMS calculation uses the inpatient admission date, then it will not account for any hospital days spent when the patient was an outpatient receiving observation services, and using the GMLOS will result in setting unrealistic goals. Under that paradigm, an admission with one day of observation and two days of inpatient services would underestimate the true length of stay by a third, driving doctors to hit an artificially low target.

I also dislike length of stay because it does not correlate directly with costs. Is a doctor with a 3.8-day average length of stay and average cost of $10,000 per admission better than the doctor with a LOS of 4.3 days and cost of $8,000 per admission? But does any hospital know the true costs of care for a single patient? Most calculations use charges and then adjust based on the hospital’s Medicare-designated charge-to-cost ratio, and that is a very imprecise way to measure costs.

Length of stay also often does not take into account payor source. If a contract with a payor pays a daily rate and that rate is profitable for the hospital, does it make good business sense to drive physicians to discharge patients earlier when additional days would pass scrutiny as medically necessary and be paid?  

What adjustment to the length of stay is made for patient acuity? It only takes a few complex surgical patients being assigned to a hospitalist for their LOS number to be skewed. Hospitals will often adjust for the case mix index (CMI), but is that a valid measure? The CMI is based on the assigned DRG, but the DRG assignment is based solely on diagnoses that are classified in the DRG weighting system as comorbidities, complications, major comorbidities, or major complications, specific to the primary diagnosis. There are many other factors that affect patient acuity and cost, including all the social determinants of health, none of which are considered in the DRG weighting.

Many hospitals assign a working DRG to each admission and prompt the doctor that discharge is expected as the GMLOS approaches. But GMLOS is derived from hundreds of thousands of admissions, and as the title indicates, represents a mean length of stay. That means that by definition, many patients will have shorter stays and many will have longer stays. If a patient’s stay exceeds the GMLOS, one of two things should happen: either the doctor should be asked to document any existing comorbidities or complications that would result in the admission being assigned to a higher-weight DRG with a longer GMLOS, or the doctor should be asked to ensure that documentation every day supports why the patient continues to need care in the hospital. What should not happen is that the doctor be asked to discharge the patient, the doctor be labeled as an outlier, or even be threatened with disciplinary action.

There is also the unsolved problem of patient attribution when calculating length of stay. If one hospitalist handles an admission on a Sunday, another assumes care on Monday, and then a third hospitalist discharges the patient on Friday, to whom do you assign the admission? Should hospitalists and medical doctors be held accountable for the length of stay of surgical patients who are admitted to the medical service by the surgeon? Specialists have a crucial part in determining length of stay, yet how do you track that and hold them accountable?

In many hospitals there are a full range of services available Monday to Friday, but limited availability of services on weekends. If a patient’s stay is extended because the hospital does not have a PICC team on weekends, why should the physician’s LOS statistics be affected by an administrative decision to not offer those services? There are also many patients whose length of stay is extended because of placement or insurance issues. A patient requiring guardianship may spend an extra month in the hospital until that process is completed; that admission will skew a physician’s data for months to come. It is no longer uncommon for an insurance company to delay approval for transfer to a post-acute care facility for several days, after patient stability, to minimize their costs; that is out of the control of the physician, yet their data will be affected.

Like the observation rate, I would love to see length of stay as a metric go away. But I suspect it will be with us for a long time to come. Understanding its shortcomings –and as I have shown, there are many – is hopefully a good first step.

Program Note: Register to listen to Dr. Ronald Hirsch every Monday on Monitor Mondays at 10-10:30 a.m. ET.