27 May 2008

Short-staffed

A commenter the other day made this comment which kind of stung:

It's not your fault you guys are under-staffed and only have about two minutes per patient to ask them questions.

Well, it kind of is my fault. You see, I do the staffing for our ER. I hire the docs, and set the schedule, and figure out how many hours of staffing we will have every day. So, if we were under-staffed, hypothetically, it would be my fault.

Which got me to thinking, Are we under-staffing the ER? Not the first time I have contemplated the issue, and once again I concluded that no, we are not under-staffed. The way we staff is to volume, which is highly predictable.

So we generally know how many patients are coming in, and when they are coming. I staff to about two patients per hour per doctor, which is somewhat more generous than other ERs do -- I know of some where three to four per hour is common; I don't think that's safe or sustainable. ACEP recommends about 2.25 patients per hour, though that number will vary dramatically depending on the acuity case mix, whether it's a teaching facility, and other factors. The graph above tells me how many new patients are presenting to triage every hour of the day; my job is to make sure that there are sufficient physician resources there at that time to actually care for those patients.

So why are we feeling like we are chronically short-staffed?

I think the big reason is demand-capacity mismatch. The capacity is more or less fixed in advance, and the demand fluctuates dramatically, which leads to chronic over- and under-staffed periods in the department.

The above graph is an average, and I staff to the average. From 10am-5pm I have 8 patients per hour presenting, so I have four doctors working. Sounds great, right? Except that real life isn't that predictable. If we have a busier than usual day, or if they come earlier than anticipated, then we are behind the eight-ball. I don't know the standard deviation, but half the days are going to be busier than average. Some of the variation is predictable: day-of-the-week is reliable (Sat, Sun, and Mon are the busiest days). Most of the variation is not, though. The month-to-month variation is highly unpredictable. Sometimes the flu season hits in December, sometimes in March; sometimes there's no noticeable spike in volume, and sometimes, as this year, there are 15% more patients than expected. The year-over-year growth is a huge factor, and is inherently unknowable.

Bearing in mind that I create the schedule two or three months in advance (yes, I could do it with shorter lead times, but that's really disruptive to people's lives), and that I am chasing a moving target, and that the fluctuations are so unpredictable, I think we do a good job. That does not, however, spare me from complaints from staff and patients when I happen to be there on a really busy day. Everybody wants more bodies. Of course, were I do respond by blindly increasing hours every time someone cornered me and complained that "we need more," then the proportion of days when we are so overstaffed that docs are sitting around drinking coffee and not making any money would increase, and salaries would go down. That is absolutely sure to generate a lot of complaints from the docs, and my experience is that when the money goes away, so do my employees, to greener pastures.

Finally, I might add to the anonymous commenter, that it's not just under-staffing that cannibalizes the time we get to spend with patients. At the optimal rate of two patients per hour, that's 30 minutes of my time per patient. During that time, in addition to the "time at the bedside," I'm going to do a lot of things -- review old records, create the new record, order tests and meds, track down the results of the tests, communicate with the nurses, communicate with consultants, communicate with family members or other historians, all while juggling the constant stream of interruptions that comes with taking care of eight to twelve patients at once. Given the complexity of modern medicine, the tremendous volume of data we create and digest, and the inherent inefficiency in the workflow, it's no surprise that of that 30 minutes we spend an average of less than ten minutes at the bedside.

10 comments:

  1. Dude, no mention of the economics? Health care is a business the last time I checked. How much are your hands tied by the economic performance of your ER when it comes to staffing?

    ReplyDelete
  2. Robert Reich is 4'10", Shaquille O'Neal is 7'1". Therefore, on average, they're 5' 11.5".

    See the problem? Averages don't tell the whole story. Average patient volume is just that - a theoretical, meaningless construct, as the patients don't show up according to the arithmetic mean.

    What you need is a more advanced model, where you can set a confidence interval (ie, 85% of patients will have a physician available N minutes after triage) and manage to that. The catch is, going in, you're going to need detailed time-of-arrival data on a comparatively large set of patients. The upside is, these models are largely automated - they'll give you a starting place and let you smooth the schedule for the realities of people not coming in fractional units, for example. Look for solutions that use monte carlo analysis - they're the most likely to give reasonable answers.

    Another way to do this would be to look at the standard deviation in the arrival data and perhaps manage to 1SD above the arithmetic mean.

    E

    ReplyDelete
  3. I make the schedule for our ER group. I can easily schedule more physicians and "overstaff it" until the hourly drops and my doctors go elsewhere.

    My scheduling has little to do with the waiting room because so much of it is out of my control and scheduling more doctors wouldn't solve the problem:

    1. nursing shortage
    2. nursing "call offs". A sense of professionalism has largely left the nursing profession.
    3. In California legislated "nursing ratios"
    4. ER boarding of admitted patients. A hospital failure issue, not physician staffing issue
    5. Lack of specialty back up.
    6. Hospitalists wanting a 3 day hospital work up completed in the ER

    ReplyDelete
  4. Didn't mean to sting you, shadowfax. I was trying to be understanding and nice.

    However, it's easier to be understanding when I understand the whole problem, which I clearly do not. Thanks for taking the time to give out more information. I come to blogs like this to learn stuff.

    Patients perceive that there is a staffing issue in ERs the same way they perceive it on a busy night in a restaurant. When you can't flag anyone down to assist you with your pain, or you can't find anyone down to refill your iced tea, you perceive that there are simply not enough people around to help at that particular time. You don't necessarily see the times when the ER or the restaurant are so quiet that half the staff are taking a nap in the back room.

    ReplyDelete
  5. Hey Shadowfax,

    You don't necessarily need models (cost about $3500) you can do most of this stuff in Excel (looks like you already are) and real data is always more reliable. I do it for our clinic and have done it for on-site emerg tents to staff to 95% CI. Excel can be connected directly to your database.

    Assuming that you've already got lots of admin support to analyze the numbers there may not be much more you can do with staffing assuming variation in demand is very high (equal to the average?). If the patient demand is too variable is there any way to mitigate it? Can you use same-day booking to mitigate some of the 4pm-10pm crowd? Is there work left to be done to improve efficiency in the ER (patient flow charts, stuff like that)? In general flow rate is 75% of the wait times and excess variation is 25%. If you want some help with the Excel modelling let me know I love that stuff and would be happy to show what we use in our clinic. Ian.
    www.waittimes.blogspot.com

    ReplyDelete
  6. forgot email
    ian underscore furst at yahoo dot com

    ReplyDelete
  7. You are right, you are the problem. You are staffing to averages, not to robust demand, as someone else has already pointed out. To blame other problems as one commenter did is making useless excuses. I consult with EDs and model them using sophisticated statistical software. I have not meet an ED yet that didn't staff to census not demand. Therefore they are adding physician hours AFTER the patients are there. By then it is too late. Also they, like you, staff to average patient load and then when if swings high, fight the fires. Why isn't it done right? Another commenter hit the nail on the head, hospitals see EDs as cost centers not as the all important front door to their hospitals where approximately 40-50% of their admits are generated. A local hospital here fired all their ED docs and put in a contracted service to cut cost. How did the group do it? They cut physician hours in the ED. Now the EDs LOS is an average 8-9 hours up from 4-5 hours. The public is screaming, and not a damned thing is being done about it. I have decided to patronize the ED in another area hospital although it is inconvenient to get there because their LOS is 5 hours. Who says patients don't make choices when it comes to the ED?

    ReplyDelete
  8. 1) Most of the bottlenecks in patient flow aren't due to inadequate physician staffing.

    2) ER docs don't like to take call, so "flexible" staffing solutions aren't very popular. One of the most satisfying perks of our job is the fact that when we're off, we're off.

    ReplyDelete
  9. Shadowfax - I did some more detailed research (we had a funding announcement on ER overcrowding up here so I needed to do it for a post today anyway). Check out this study by Michael Schull at
    http://www.ices.on.ca/file/Emergency_department_services_in_Ontario.pdf
    and look him up on pubmed. He found some good guidelines to the amount of variation in ER in both urban and rural settings. For instance over the year, the peak is 50% above mean and the trough 25% below. Hope that helps. some other links are at my post too. Ian.

    ReplyDelete