Hampson - Turnover Time

Hampson - Turnover Time

by Lindsay Hampson -
Number of replies: 2

1.    Describe the study design you will employ in order to determine if your intervention has had an effect on the outcome variable of interest.

We will measure pre-intervention mean turnover time, and use this as a baseline to compare to our post-intervention mean turnover time. It likely makes the most sense to do an interrupted time series, using each site as its own control using pre-intervention data as the baseline comparison.

To accommodate for autoregulation, we can evaluate the monthly mean turnover time over the last year to see if there is a trend – if so, we will need to incorporate autocorrelation into the effect size, recognizing that we may need a larger effect size to account for trends in pre-intervention data.

2.    Define the unit-of-analysis for your main outcome evaluation, the minimum meaningful effect size, and the sample size necessary to detect this effect size.

Unit of analysis for outcome: monthly mean turnover time between scheduled cases (in minutes). This unit of analysis provides some smoothing effect for daily and weekly variations in turnover times.

Minimum meaningful effect size and necessary sample size:

Pre-intervention turnover time mean 52.2 minutes

Parnassus 99.2 scheduled turnover times per month

Group 0: Pre-Intervention Control Group

Group 1: Post-Intervention Group

Group 0 sample size

Group 1 sample size

80% Power Effect Size

Mean Turnover Time (min)

Turnover Time Reduction (min)

1 year

1 month

1200

100

29.2%

37.1 min

15.1 min

1 month

1 month

100

100

40.0%

31.3 min

20.9 min

4 months

4 months

400

400

19.8%

41.9 min

10.3 min

1 year

4 months

1200

400

16.2%

43.7 min

8.5 min

6 months

6 months

600

600

16.2%

43.7 min

8.5 min

1 year

1 year

1200

1200

11.5%

46.2 min

6.0 min

 

We will need to do separate calculations for Mt. Zion, given that it likely has different mean pre-intervention turnover times and fewer scheduled turnover times per month (62.4 vs 99.2).

We could also consider using OR pods as clusters and/or looking at changes in outliers or other measures than just mean times. To do an analysis using OR pods as clusters, we would need data on mean turnover times for each pod as well as the denominator for number of scheduled turnover times for each specific pod.

In reply to Lindsay Hampson

Re: Hampson - Turnover Time

by JESSICA COHAN -

Hi Lindsay,

Impressive job working through the numbers.  My thoughts on the analysis would be that it would be nice to have pod-level data to feed back to the the pods (to encourage them when they are doing well and inspire them when there is room for improvement) but not to base the analysis of the program as a whole at the pod level (except where you can account for clustering).  The pod-level data will also be helpful in teaching lessons about what is working for one pod but not another.  However, I think it will be more meaningful if the final analysis focuses on the numbers at a hospital-level because ultimately the goal is to develop a protocol that can be flexible enough that it works in all pods and contributes to reduced turnover times across the board.

Great job!

 

In reply to Lindsay Hampson

Re: Hampson - Turnover Time

by Ralph Gonzales -

great job!!  there may be some effect size modifications that would result from autocorrelation techniques, but often times these serve to improve your power.  Will hope to get some clarification from Chuck on how to use the Excel sample size calculators that he posted.  

re: detectable time turnover reduction, how much time reduction would be meaningful?… if they are all meaningful, you might get away with smaller study?