+ Reply to Thread
Results 1 to 3 of 3

Thread: Stats for emergency care delivery

  1. #1
    Points: 4, Level: 1
    Level completed: 7%, Points required for next Level: 46

    Thanked 0 Times in 0 Posts

    Stats for emergency care delivery

    Hi all- I am hoping you can help me. I'm trying to write a paper on a new model in my ED, which reduced wait times/LOS, and improved other things (e.g., % of lab errors). One of my mentors advised me to only use descriptive stats, but I feel that doing so misses the point- that is, who really cares about changing a healthcare model unless it improves things?

    Anyway- the main things I am wondering about are the wait times and the percentage data. However, I definitely do not have enough training in statistics to be confident in my data analysis.

    For the wait times/los data, I want to compare the wait times for a five month period in 2014 (before the model change) with the same five month period in 2015 (after the model change) - both have an n of ~3500, but have different exact numbers. So, I checked for normal distribution via a Kolmogorov-Smirnoff stat which was fairly large. Then I did an independent samples t-test and used the mean change calculations, with the p value and CI- is this correct? (Meaning, did I do this correctly, and, is it the right test to be using?)

    For the percentage items, i.e., the percent of lab errors, I used this with HO = 0, Ha not equal to 0. The I used the CI and p value for this as well- does this seem like a reasonable way to determine significance of change in the percentages of errors?

    I know this is a lot of questions, but I would so, so appreciate any help anyone can offer. Thank you for considering it!

  2. #2
    TS Contributor
    Points: 12,227, Level: 72
    Level completed: 45%, Points required for next Level: 223
    rogojel's Avatar
    I work in Europe, live in Hungary
    Thanked 332 Times in 312 Posts

    Re: Stats for emergency care delivery

    I would expect time data to be quite right-skewed, so the mean might not be the right measure to test. I would go for a non-parametric test, lile Kruskal-Wallis, not because the the t-test is inappropriate but more because the median is more practical in this case.

  3. #3
    Fortran must die
    Points: 58,790, Level: 100
    Level completed: 0%, Points required for next Level: 0
    noetsi's Avatar
    Thanked 915 Times in 874 Posts

    Re: Stats for emergency care delivery

    If your instructor, I assume a professor, told you to only use descriptive statistics there probably is a really good reason I love bells and whistles as much as anyone, and sometimes that is not the way to go. One alternative, which is a design of experiment exercise rather than what I think of as true statistics is to do interrupted time series. You compare results before and after an intervention and see if the results are significantly different. It is more common to model the results immediately before and immediately after the change (the intervention) but if you think seasonality is an issue looking at the 5 same months before and after the intervention might make sense.

    Its even better is to compare the two periods and also compare wait times in another similar emergency care facility that did not see the intervention. This is to deal with history, something aside from the intervention that might cause the change. If you emergency facility changes and the other one does not, it suggests that the intervention caused the result.
    "Very few theories have been abandoned because they were found to be invalid on the basis of empirical evidence...." Spanos, 1995

+ Reply to Thread


Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts

Advertise on Talk Stats