Length of stay index

Length of stay index is certainly better than just average (or median or geometric mean) LOS (ALOS) if you are benchmarking against other hospitals.  If your goals are just local improvement on past local performance (and you don't expect significant changes as described below), you can stick with ALOS and have a percent decrease goal, but otherwise you certainly should opt for something that gives you length of stay/predicted length of stay.  There are a few models you can use to calculate this.  That makes it so that a hospital with higher complexity of patients is not compared to a hospital of low complexity just based on ALOS.  Higher complexity expects higher LOS.  This will of course be very dependent on your documentation and coding to appropriately capture the complexity of your patients and for coders choosing the appropriate DRGs as primary and secondary.  Often outliers are removed and you may want to argue for this if you have significant outliers (we have patients that have been admitted for years due to social issues - the day they are discharged will ruin a full year's worth of metrics if not excluded).  I would not necessarily agree to a goal of LOSi of 1 unless you know your current performance.  But technically it is an appropriate goal, same as mortality index. 

We previously compared just year over year ALOS, but then we had a large push to accept all outside transfers and to transfer low acuity patients to other hospitals.  Outside transfers are of much higher complexity, and our short stay patients are now not admitted to our hospital, so we would expect our LOS to increase.  So obviously we do not want to be compared year over year ALOS anymore.  If we kept our LOS similar with a large influx of higher complexity patients that would be a huge accomplishment, but we needed the dashboards changed as the metric was a decrease in ALOS that we were being measured on.  This may create a perverse incentivize to admit low complexity patients and not accept transfers to meet the goal.

How will they account for observation patients, or will they be included in your 'inpatient' metrics?  For many of us, our metrics look worse over the last 5-8 years on LOS, however when you really look at the numbers, the short stay patients who were previously classified and included in our inpatient data are now excluded as observation, so even though (across all patient we care for in the hospital) we have cut our LOS, since the shortest stay patients are now classified as observation it has made the LOS metric for inpatients increase.  If comparing to past performance, I would try and have apples to apples and include all unless your percent of observation patients has been consistent.  Which is unlikely.



 

Thanks for the reply!

We're looking at inpatients only for these numbers, so observation patients don't apply.

I'm guessing it's pretty obvious from my original post that this LOSi metric doesn't seem fair to me.

I have 2 main concerns. First, we're comparing all of our patients, regardless of payor status, to a benchmark based on Medicare patients. There are more barriers to discharge for uninsured patients than for patients with Medicare, so it doesn't seem reasonable to me to expect us to have the same LOS numbers for them. Also, we're comparing an arithmetic mean to a geometric mean. By definition, an arithmetic mean will always be higher than a geometric mean for a given data set because geometric means mitigate the effect of outliers on a data set.

Still curious to know if this is a generally accepted way to measure the performance of an inpatient service or a number our administration came up with on their own.

We look at (GMLOS-LOS)/patient and our goal is to work towards zero - similar to your LOSi.  That said, we look at patients discharged under 15 days, and only patients we admit and discharge. 

It is imperfect, but given more and more payment models are based on the GMLOS, it is not a bad metrics but it is a marathon not a sprint and getting to 1 in your system and staying there should not be expected overnight or even in a year.

What is good about these data, is we have started diving into them to see what service line, disposition et al, Dx seem to be having the most LOS issues so we can focus our energy on improving. 

That premise would depend on accurately documenting so that the correct CMI is captured and correct DRG is assigned, which in most cases HOSPITALIST are horrible at documenting well. Also, aLOS could be influenced by wRVU, and how patients are distributed from the Admissions, and starting census, so why should a Hospitalist harm himself by working hard to appropriately discharge patients, when it will mean getting punished financially and more work of getting new patients. 

Without removing outliers, the data becomes a measurement of the hospital’s ability to move patients through the system, not the hospitalists’.

We have contemplated measuring the median los. If it’s just over 2 days then everything is working well. Anything less than 48 hours, insurance will generally refuse inpatient status

Our hospitals use the Vizient calculators to calculate the expected LOS (eLOS) for each of our cases.  This allows us to compare our performance to other AMCs of similar size that perform similar services.  Some things to consider:
- case weight/CMI is a very crude estimate of patient complexity.  A patient on chronic O2 admitted with CAP which requires more than baseline will be in their DRG with an MCC...the highest weight for that DRG.  The same patient with ATN and severe protein malnutrition will have the exact same weight and contribute the same to the CMI.
- quality risk calculators account for things like socioeconomic status, transfer status, DNR status, physical disability...all the things that are lost in simple CMI calculation.
- assuring that your band of hospitalists understands that the "reason for admission" portion of the H&P and summary are used to determine the principal DRG is very important.  Copying and pasting the HPI into this section may do your patient and your hospital a disservice.
- Observation is OP and will not even have a DRG assigned

Assure that the method used does a better job than simple DRG/weight/CMI in determining expectations
Good luck
Cliff

Clifford Kaye MD FHM
Associate Professor
Division of Hospital Medicine
Medical Director of Utilization Management/Review & Clinical Documentation Integrity