I am on a committee to provide recommendations to our Dean on how to compensate physicians for teaching. This turns out to be a lot more complicated than it might...
Hospital’s priorities are usually not aligned with how we pay hospitalists. In fact, the two are often in direct conflict with each other. In my last post, I argued that the RVU is not the best measure of productivity for a hospitalist. In this post, I have some ideas of how hospitals can align hospital priorities with the hospitalist’s income.
The first thing we need to do is to get away from the model of a rigid census cap/expectation per hospitalist. In a previous post, I discussed why the work required to take care of 15 patients at one hospital does not equal the amount of work required to take care of 15 patients at another hospital. In fact, a census of 15 patients on one floor of any given hospital is not the same as 15 patients on another floor. Quite simply, this is because the amount of physician work necessary to take care of one patient is not the same as the amount of work necessary to take care of another patient. One way of determining the proper census per hospitalist is to do a CMI-adjusted census (CMI = case mix index). The idea is that the higher the CMI, the sicker the patient and presumably, the more time required by the hospitalist to care for that patient. Let’s look at the CMI of 3 hypothetical hospital services:
Service 1: CMI = 1.30. This service admits general medicine patients but also admits to the ICU.
Service 2: CMI = 1.10. This is service admits non-ICU general medicine patients.
Service 3: CMI = 1.00. This service mainly covers lower acuity medicine patients, generally with single-issue medical problems and about half of patients being observation status patients. They have a short length of stay.
Let’s start with an assumption of 20 patient encounters per hospitalist and then divide the census by the CMI. So, for service #3, we would have 20 ÷ 1.00, which would be 20 patient encounters per hospitalist per day. On the other hand, for service #2, we would have 20 ÷ 1.10 = 18 patient encounters. Service #1 would be 20 ÷ 1.30 = 15 patient encounters. Notice that I used hospital encounters in this analysis and not daily census. Because of the differences in length of stay (and therefore differences in patient turnover) for each of the 3 services, the daily census could be the same for each of the services (eg, 13). Moreover, if you have night coverage hospitalists who are doing admissions to these services at night, the service census at the midnight census tally might be 15 for each of the services. Surgical patients inherently have a higher case mix index because of the surgical procedure so you cannot apply the same analysis for staffing surgical patients as you would with medical patients.
CMI-adjustment does several things to align the hospital and the hospitalists:
- It rewards the hospitalist to compulsively document in the chart all of the mundane co-morbidities that affect the CMI score but really don’t affect how the patient gets managed. So, for example, if a patient has a sodium level of 144 (normal 133-143) on admission, the hospitalist is going to ignore it since it is not clinically significant – adding “hypernatremia” to their admission note is extra work and why bother typing in the extra line of text if it is clinically irrelevant? However, since by adding the word “hypernatremia” to their note, the CMI goes up slightly and so the hospitalist is granted a slightly lower census target.
- The hospital’s financial margin improves because the higher the CMI, the more the hospital gets paid for that patient admission.
- The hospital’s length of stay index improves because the index is determined by the actual length of stay adjusted for the CMI.
- The hospital’s mortality index improves because the actual mortality rate is adjusted for the CMI to give the publicly reported mortality index.
Outcomes-Based Bonus Plan
Historically, bonus plans were based on productivity. At the end of the day, the productivity that really matters is total cash collections. However, we all know that when performing the same service, you get paid more for a commercially-insured patient than you do for a Medicare patient. You get paid even less for a Medicaid patient and you get paid practically nothing for most uninsured patients. So, the RVU has evolved to be a better measure of physician work effort than cash collections in order to remove the disincentive of taking care of the uninsured and Medicaid patients in the hospital since the hospital has to have someone take care of these patients.
In medicine, we often define true value in the service that we provide by the equation: value = quality ÷ cost. In other words, you can increase your value by increasing your quality or by decreasing your cost. So, what the hospital really wants is for the hospitalist to improve value of healthcare, by either improving quality (particularly in those publicly-reported quality measures on the Medicare Hospital Compare Website) or by improving the hospital’s financial margin. The financial margin in turn, can be improved by either increasing the revenue per patient-day in the hospital or by decreasing the cost per DRG. Therefore, bonuses should be based on some combination of:
- Query responsiveness. Hospitals have coding staff that comb inpatient charts looking for those co-morbidities that add up to a higher case-mix index for any given patient. The problem is that even if those co-morbidities appear in the lab results (for example, hypernatremia in the previous discussion) or appear in a non-physician’s note (for example, the dietician who mentions “protein calorie malnutrition” in his/her note), it only counts toward the CMI if a physician (or nurse practitioner or physician assistant) puts it in their note. So, hospitals have evolved a query system where co-morbidites identified by coders are reported to the hospitalist as a query and then the hospitalist decides whether or not it is valid and then addends their note accordingly. This is extra work for the hospitalist and so if they are not incentivized to answer the queries, they are going to ignore the coders and then the CMI ends up being lower.
- Patient discharge time. The earlier you get patients out of the hospital, the earlier in the day that bed can be filled by the next patient. However, you don’t need to get all of the patients discharged early in the day – your housekeeping staff can’t clean all of those rooms at the same time. The strategy is to get some of the patients out by 11:00 AM, some more out by 1:00 PM, etc. so that you have a steady flow of discharges throughout the day in order to accommodate the steady stream of patients waiting to be admitted into those beds. So, pick some numbers that work best for your hospital, for example, 20% of discharges by 11:00 AM and 40% of discharges by 1:00 PM.
- Mortality index. Because the mortality rates are one of the publicly reported items by Medicare, the hospital wants patients to die anywhere but in the hospital. For those patients who are anticipated to die, transferring a patient to a hospice facility to die is ideal. The danger of using mortality index for hospitalist bonuses is that sometimes, it can work against you from a hospital expense standpoint. For those patients who are clearly going to die in the ICU, the hospitalist could be incentivized to try to keep that patient alive a little longer in order to buff them up just enough to survive the transport to inpatient hospice or to have them die on another hospitalist’s shift so that the death doesn’t count against them. In this situation, earlier withdrawal of life support would have resulted in the hospital not having the expense of those extra days treating the patient in the ICU and the patient (and family) would have been spared making an inevitable unpleasant and uncomfortable death last longer.
- 30-day readmission rate. Hospitals get penalized by Medicare if this is too high. The hospital wants all of its rooms to be full, but to be full of patients who were not there in the past month. Hospitalists can often reduce the readmission rate by putting more time and effort into the discharge process (see post on The Most Dangerous Procedure In Medicine).
- Lower length of stay. This is a tricky one. If you discharge a patient prematurely, that patient is more likely to be readmitted within 30 days and is more likely to be dissatisfied if they perceive that they were thrown out of the hospital too early. So, length of stay should never be the sole metric for a bonus plan and should only be used when coupled with hospital readmission rates and with patient satisfaction. Also, length of stay lends itself to gaming the system since it is based on the midnight census. So, a patient admitted to the hospital at 11:30 PM already has a 1-day length of stay a half hour later at midnight. In order to improve his length of stay, the hospitalist will procrastinate putting the admission orders in for anyone showing up in the ER in the evening. If that order is placed at 12:01 AM, you just knocked a day off of that patient’s length of stay.
- Patient satisfaction. For inpatients, this is measured by the “HCAHPS” survey questions that are reported on the Medicare Hospital Compare Website. Some of these questions are specific to physician practice and can be used in a hospitalist bonus plan; other questions pertain to the patient’s overall perception of the hospital which measures the physician’s performance as a member of a larger team of providers in the hospital.
You can’t do away with RVUs completely, otherwise, the hospitalist would either not bother to submit their charges for patient encounters or they would bill everyone as a level 1 visit, thus reducing the necessity of all of the painful documentation required to bill higher levels of service. So, there has to be some why to hold the hospitalist accountable for turning in their bills and to insure that they are actually billing for the level of service that they are providing. Most electronic billing programs will allow you to see what the distribution of level 1, 2, and 3 CPT codes for any given physician. This distribution can be compared to internal benchmarks of all of the other hospitalist or to external benchmarks, such as the Vizient benchmark data for academic medical centers.
The Bottom Line
Ultimately, the strategy is to align the hospitalist’s reward system with the financial margin of the hospital. To do this, you need to think beyond hospitalist census caps and RVUs.
March 7, 2017