Hospital LOS model per physican's groupHi
we have like 20 physician group( clusters of physicians)in the hospital. We would like to focus which physician groups should we target in order to reduce patients LOS(length of stay). Firstly we need to identify which physician groups have higher contribution ( also to what extend) on average LOS like finding out which physician groups are more likely to have higher LOS regardless of patients insurance, severity and disposition ( final destination (like home or rehab or transferred to another hospital). we need some kind of stat of this information to present to executives.
Also want to understand if there are certain physician group who have higher LOS only when patient's insurance is Medicare.. if other insurance then they do not have higher LOS.so we can target just certain insurance company of those physician group and see why insurance is taking longer time for these particular physician group.
I have moderate understanding in stat .. wanted to ask an expert opinion.
Please advise
Thank you
Best,
S
Re: Hospital LOS model per physican's groupI'm going to make two suggestions, one for you @sweetycha and the other for the consumers of the information you are conveying.
For yourself, I suggest you take a look at JMP's Statistical Thinking for Industrial Problem Solving online, free course. Upon successful completion of the course you'll have a sound grasp of the principles of Statistical Thinking, of which @statman has touched upon as well as basic analysis methods consistent with the course's learning objectives. Which are all up all in are two fold; 1. Enable the practice of Statistical Thinking and 2. "How to in JMP.". Granted the examples are not in a clinical setting, but, having done some pro bono work in a hospital setting, all the statistical thinking concepts and practices can be extended to a hospital setting.
For your consumers; First ask yourself, across the group, 'How literate are they wrt to Statistical Thinking'? If the answer is illiterate (or weak at best), then give each person a copy of Don Wheeler's "Understanding Variation, The Key to Managing Chaos" and demand each reads from cover to cover before you meet with them. Then keep your information dump to them exclusively graphical, with easy to read and interpret charts and graphs only. P - values, parameter estimates, models, and tabular output will only serve to muddy the message. My advice wrt to your graphics is "If it does not add, it subtracts." Keep them simple and uncluttered. One message per graph. Test your audiences understanding for each.
Re: Hospital LOS model per physican's groupI must admit most folks do not think about variability or what causes it (or they may think about it, but do nothing to understand it). I can't tell you how many experiments I have analyzed (over 10 thousand) where the response variable was a mean and the purpose of the experiment was to understand variation what factors effect variation.
One way to start is not to make your response binary (opportunity index) based on an "arbitrary" specification. Use the actual days (or possibly even more discriminant, hours) to perform each "operation" (this would be an efficiency metric). Thinking "lean", have you value stream mapped the entire process? Or done a Process Map? How much time does it take to perform the surgery, prep.for surgery, waiting for operating room, sedation, getting labs, filling out paper work, waiting for insurance approval, etc.? What is the mean and standard deviation of the time of each of the activities in the process. You also have to consider efficacy metrics, were each of the steps in the process completed correctly?
Regarding your example where "they are writing a discharge order later afternoon". Why are they waiting until afternoon? Is it because they were too busy doing other aspects of their job(like taking care of their patients, the customer) You have an arbitrary specification of do it by 9:00 am. How did you arrive at this spec.? Did you use data to ensure the "current" process is capable of meeting this spec.? Is this more a goal to motivate the workforce? Certainly if you set a spec. you predicted what would you do if the spec. was not achieved. How would you improve the process to meet the spec.?
I understand that your study is iterative...I think it is best to start with the attitude, the people in the process are not intentionally doing things to negatively effect the performance of the process.
There are common causes associated with the system (and also potentially special causes). It is unfair, wrong, and counter-productive to blame the individual for the failings of the system. Sure we don’t like mistakes, complaints from customers, accidents —but if we weigh in at them without understanding, then we make things worse. W. Edwards Deming
The process (the system) is "owned" by management and in order to meet the goals and requirements of the company to satisfy the customers, they must continually improve the process(es).
Re: Hospital LOS model per physican's groupThank you for your feedback. I know this is very sensitive approach to find which group should we target based on some predictors. They must be so much underlying factors associated with having higher LOS.
I like the idea of researching a variation. I have not done before other than plotting the dependent variable box plot against group.
Here our target is 1 which means patient's LOS/ GMLOS which again in my previous example of hip surgery, if national is 3 days our is 3 days then opportunity index is 1(3/3). Anything greater than 1 is our flag if we want to make our depended variable to binary
I am interested on understanding variation and how do I start?
To answer your question, what is our plan once we identify the top physican group, we just investigate further by some underlying factors like if they are writing a discharge order later afternoon , which supposed to write in first round by 9 am so patient get discharged by noon. Hence lower LOS?. this is just an example.
Re: Hospital LOS model per physican's groupI understand this is difficult/complex. Since you will be presenting data and your interpretation to executives, I can only imagine that you want the executives to take some sort of action. Be very careful in thinking through (predicting) how will they react. I would not focus solely on averages. The delta from some "target" LOS based on what I don't know, is similar to trying to understand how to improve process (e.g., manufacturing) costs using standard costs or improving efficiency using TACT (e.g., lean applications). How was the 3.5 Days for DRG established? What is the variation expected? Reporting just a mean is not very useful if you don't understand variation!!! 3.5 with a standard deviation of 0.1 is very different than 3.5 with a standard deviation of 1.0.
I would be interested in the VARIATION of LOS and what causes this. I would want to know whether the LOS was consistent.
Again, thinking through the assignment of the issue to Physician Group seems to me like placing blame, at least you should be concerned about how executives will react to this information. Will they want to take action on the Physician Groups with higher LOS? What action would that be (e.g., fire them, re-train them, issue them a warning)? Realize there will be variation associated with Physician Groups. Is it common cause variation? If so, it is a function of the system (see Deming) and taking special cause action (examples listed above) on particular Physician Groups would be inappropriate reaction and would lead ultimately to increased variability and thus less predictability.
Re: Hospital LOS model per physican's groupThank you for your detail explanation.
To start with we have very general goal of just finding which physician groups have higher LOS so we can investigate more detail of why. Ofcourse we have so many factors other than discharge disposition and insurance like patient age, comorbidities etc. But at this moment we justw natt o find out which physician group regardless of patient's insurance and disposition are more likely to have higher LOS (opportunity days)
The LOS is been adjusted with opportunity days . If a patient is admitted for a hip surgery (DRG) and the national benchmark for this DRG is 3.5 days . If at our hospital it took 5 days for same DRG then our opportunity days is 1.5 because that patient stayed 1.5 days longer than national benchmark.There are a lot more to investigate why ….. we just wanted to start from some point by finding out certain group and target why these groups are taking longer ( higher opportunity days) regarless of patient discharge disposition and insurance. We have communicated these by taking average opportunity days and the repose we got from these groups is that we serve different patients like our patient mostly get discharged to skill nursing facility so we have to wait the skill nursing facility to accept our patient hence longer LOS.
That is why we are adjusting or patient disposition and insurance which group have higher opportunity days.
Please advise
I appreciate your guidance and support
Re: Hospital LOS model per physican's groupI apologize in advance, but I can't help myself. Being a firm believer in Deming and his teachings, it seems to me you are trying to put blame on a Physician Group(s) for Length of Stay. First we need to understand what is the distribution of LOS. Is the variation common cause or is there special (assignable) cause variation? Why would one (or more) Physician Group(s) have an impact on LOS? Is it lack of experience, training? Is it poor use of metrics and goal setting by management? Seems to me Physician Groups and how they operate is a function of the system. You want to change their behavior, change the system. In trying to understand what factors impact LOS, you should develop hypotheses.
Examples of potential hypotheses:
1. Severity of the illness increases LOS because the more severe take longer to recuperate,
2. Mis-diagnosis increases LOS because treatments are not effective so alternate treatments must be attempted,
3. Patients' compromised immune systems increase LOS because treatments are less effective,
4. Insurance policy decreases LOS because their policies reduce funding for stays that are longer than allowed by policy,
5. Type of Insurance impacts variation in LOS because the policies for each type of insurance are different.
Once you have a set of hypotheses, it then becomes the questions of how do you get data to provide insight to those hypotheses. Directed sampling (where you choose samples and subgroups of samples based on rational hypotheses) or experimentation (where you actually manipulate variables in an efficient manner).
As said in other replies, presentation to executives is challenging. Particularly when they have differing backgrounds. I prefer graphical presentation of data (rather than statistical tables) so the key points can be obvious. Most executives do not understand statistical significance. They also don't understand that statistical significance is conditional. It depends on the data (how the data was acquired) and what comparisons are being made (e.g., what terms are in the model).
Re: Hospital LOS model per physican's groupI have some additional thoughts/advice and perhaps others would like to weigh in. Given your objective to understand how different groups/payment methods affect LOS and the desire to communicate this to executives, I'd suggest starting with a purely descriptive approach using tables and graphs. It will be easier to communicate and more effective. The use of a statistical model depends on viewing the data as a random sample - my guess is it is not random, but a convenience sample (probably obtained during a particular period of time). It is more accurate to just describe the patterns in your data for that particular sample of data you have.
The desire to use a statistical model stems from 2 purposes. First, to distinguish the noise from signals - in other words, how much variability would we expect due to random sampling and how much appears to represent a true systematic effect. To the extent that the sample is not random, these inferences are not really sound. They are also more difficult to communicate to a non-statistical audience. The second reason is that when you examine a large number of factors and their impacts on LOS, it easily becomes unwieldy to use tables/graphs to explain the patterns you find (but you should still try - complex tables can be very effective, even with many variables). A statistical model becomes a concise place to put a large number of factors and talk about their separate impacts on LOS, along with measures of statistical significance/variability (but subject to how accurate it is to view your sample as reasonably random).
Re: Hospital LOS model per physican's groupThank you so much... I will do what you have suggested.
I will sure have more follow up question but as of now you have answered my initial question.
Thank you once againRe: Hospital LOS model per physican's groupThe coefficient of .24 would say that that group has .24 extra LOS compared with the average of all the groups (and is statistically significant). Each of the groups has a coefficient indicating its LOS compared with the average of all other groups - one group is held out from the report (you need to click on Expanded Estimates to see all of the groups). For an interaction term, you would put both variables in the model as factors, then highlight one of them in the factor box and the other in the potential factors (holding down the control key) and clicking the box that says "cross."
Re: Hospital LOS model per physican's groupThank you so much. for shedding light to my analysis. Far way to go but I am positive to start with something. I really appreciate your guidance and support.
In Jmp I am not sure where to add the interaction term. You brought a very good point as we are exactly testing out which group to what insurance group we need to start investigating. When we do not have interaction and we pick the group that have higher beta estimate plus p value < 0.005.
It is important to learn if there is particular group that has higher LOS when insurance is ABC or something .
Can you please help me who do I interpretate the B estimate of highlighted group on my previous post?
Re: Hospital LOS model per physican's groupYou are on the right track. I'm not sure you need to use GLM - I'd start with a simple regression. Also, I can't tell how many different financial and physician groups you have - if there are many and a number have very few observations, you might consider regrouping them into a smaller number of categories. I supposed it might also be important to investigate potential interactions between financial categories and groups - the way you have it modeled, the financial categories will have the same impact for all groups. But, the general point is that looking at the estimate and its confidence interval is the main focus of your attention. And that is what you would want to communicate - but the issue of what and how to communicate is too complex to just outline here. It depends on the audience, the results, and the use it will be put to. Simplifying to communicate is good - oversimplifying is dangerous.Re: Hospital LOS model per physican's groupThank you for your prompt reply
I have started with simple GLM model like this:
where as my dependent variable is opportunity days ( we have national benchmark LOS called GMLOS and our hospital los, so its basically LOS - GMLOS = opportunity days)
Here I just want to get the following answers:
1. If patients financial class (insurance) and disposition is constant, which group has higher OPPORTUNITY DAYS.
I think individual beta estimate on this model will give this answer. M
MY result :
Am I on right track? Please advise... Also how do I interpretate the result to executives..They are basically wanted us to tell them which group they need to focus in order to reduce the average LOS.
Re: Hospital LOS model per physican's groupYou will have to provide more detail to get a more detailed answer, and you should also review some of the documentation on modeling in general. I can provide some initial guidance. Your response variable appears to be LOS. You want to include factors in your model for payer (Medicare, other,....), physician group, and other appropriate variables you have data on (type of diagnosis would seem to be important). I would suggest starting with the Fit Model platform and just doing a linear multiple regression to see whether any or all of the variables seem to be related to LOS. There are more sophisticated approaches, but I'd start simple and work you way up.
