http://www.dmstat1.com/res/DecileAnalysisPrimer.html
Using this link as a reference, I think a lot of you are doing expected readmissions and lift wrong or I’m misreading your comments. Expected would be the mean of the test data times total record. Not mean of the predicted readmissions.
Lift is calculated as actual/expected.
Same here pretty much, but I’m also trying to add as much commentary as possible as to how everything in the CPB affects the scope of our analysis and what data we use.
By that, I mean like when we look at who is covered, mentioning that this means all of our experience will likely be based around those who are already covered and our analysis should focus on only the newly covered population. And that I’m using all CPT codes since none seem to apply any more or less to women than any other.
Task 4 is messing me up on the final part. I’m having a hard time figuring out a good prediction on how many of our old ladies are going to need/ participate in this service. All of the information seems conflicting or not substantially explained. The utilization percentages from the studies don’t specify the population and the comment about the “number of asymptomatic 65-69 yo being equal to symptomatic 60-64 yo” or something is so confusing. They say number so I don’t believe they mean percentage. But how do i calculate that? Find the total population of old people, use the inverse of the screening rates for 65-69, then apply that number to the 60-64 yo population and calculate that percentage and apply it to my membership?
Wait, that actually sounds like a reasonable solution.
I agree with what you are getting at, and think that is a reasonable solution. I am leaning toward using the only study that has the age range we are interested in to set the screening rate assumption, and then providing sensitivity analysis using the other studies to establish a range.
I thought about doing that too and I think that is also a reasonable solution, but im worried that study might be skewed because it is including older populations (who the task force says should get screened no matter how symptomatic).
But I think you could justify that approach due to the limited information given. I feel like they tried to make this question as vague as possible just to see how candidates handle it
What is the value/purpose of the provided screening rates? I feel like I have corrupted data in my file because what’s reported for study 3 and 4 does not make any sense when compared to one another.
My initial understanding from the CPB was that we had to identify those who are estrogen deficient and at clinical risk for osteoporosis based on low weight and estrogen levels, i.e. we need to figure out the amount of females in our range with low weight and low estrogen levels. I do not understand what the screening rates even pertain to or how that is relevant, unless these are supposed to be synonymous with those that are low weight and low estrogen levels? Said another way, you would only be screened for osteoporosis IF you were either low weight and/or low estrogen levels. I guess as I’m typing this out and thinking about it more, that could be reasonable.
If anyone can shed some light on how they’ve approached this, it would be appreciated because this is wildly unclear to me.
This question was clearly meant to be vague. The studies were probably not matching on purpose, showing that different sources can have different numbers and only collecting 1 is not due diligence.
They could vary because of the sample population, date of study, etc.
I think they want you to be atleast a little creative in the approach. I’d say I’ve thought of 3 or 4 ways to approach it.
I think having sat with this for some more hours and starting actually writing it, I think I’ve come to a position that I think is reasonable and backed up by the data that has been supplied. I know these are vague on purpose but…you never know what they are looking for. Onward to task 5…
Does anyone have any insight on the decile table for task 5?
What is the difference between Predicted Readmission and Expected readmission? What is Lift?
Does anybody know how long it is taking them to grade? I just submitted on 7/21, and just would like an idea of how long I can expect to wait.
I found the textbook really helpful
Anyone working on the module currently? I will be over the next several weeks and have been searching for discussions, much like the Google Groups for FAP. Please reply here or message me if you’re interested, thanks
Still working? I’m just starting myself
I’m having some real trouble understanding the decile table in part 2 of task 5. The example in the textbook shows expected 1’s in each quintile. But if we’ve ordered the entire dataset by predicted likelihood of readmission, shouldn’t every member in the first x deciles be a predicted readmission, then after the cutoff every member in the other deciles is a predicted not readmission? I know I must be missing something extremely simple here but it’s just not making sense to me.
Is anyone working on this like to discuss or start a googld group?