Chatterjee, Assignment # 8, Time series

Chatterjee, Assignment # 8, Time series

by Purba Chatterjee -
Number of replies: 2

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

  • For my pilot study, I will use pre and post measures of blood pressure in GMC.
  • For the larger scale roll-out in the San Francisco Health Network (CHN), I am planning on using the stepped wedge cross-over trial design so each of the clinics within the network will get the intervention in a staggered manner.

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.

The unit-of-analysis for my main outcome evaluation is at the patient-level. The two key study outcomes are: 1) reduction in blood pressure rates in GMC; 2) evidence of treatment intensification by GMC providers. The effect size is 5 points reduction in systolic and diastolic blood pressure.

Sample Size – I am assuming there are 80 clinicians with approximately 2000 patients and 25 patients per clinician (cluster) and a within clinician variation of 0.3 so the design effect is 8.2. With an effect size of 0.3 and SD of 0.5 it says to use 45 per group so the CRT is 8.2*45 = 369 patients per group and 369/8.2 = 45 clinicians per arm.

In reply to Purba Chatterjee

Re: Chatterjee, Assignment # 8, Time series

by Christina Mangurian -

Dear Purba,

Great job!  I'm wondering rather than doing pre-post, if you might be able to do interrupted time series.  Is it possible to look at blood pressure rates and treatment intensification at 3 time points before and 3 time points after?  If so, it might be more suggestive of an impact of your intervention than just the pre-post.  Just an idea.

In reply to Purba Chatterjee

Re: Chatterjee, Assignment # 8, Time series

by Ralph Gonzales -

Nice job Purba! I think you set it up right… but a couple of comments:

1. the ICC is pretty strong (0.3), so you might also examine an ICC of 0.10.  

2. the effect size of 0.3 is pretty strong.  If you anticipate a 5% decrease in systolic blood pressure, for example, and the average SBP is 150, then this would be an effect size of 5/150 = 0.033.  Maybe a decimal point error?  You might re-check this one.  I think for a pilot test a 5% reduction is pretty aggressive… it would be reasonable to power the study for a 10% reduction to accommodate the limited sample size you will have.