Wednesday, August 12, 2015
Present: Marianne Burke, Nancy Gell, Kairn Kelley, Ayodelle LeBruin, Ben Littenberg, Connie van Eeghen
Start Up: Family drama is everywhere.
1. Discussion: Nancy Gell: Cancer Survivorship Exercise Adherence
a. Nancy provided preliminary findings from analysis of “steps to wellness” data base
b. The underlying issue: how would you intervene during Oncology Rehab to improve later adherence?
i. What are the characteristics of the program that result in adherence 3 months after discharge?
1. Age: not a predictor of interest; may be a confounder. May indicate a need for a different intervention.
2. Attendance: a variable that measures interaction of both patient and program
3. Program: doesn’t vary
c. Research question: what are the characteristics of patients that predict long term exercise
d. Analytical plan:
i. Use data at the start of STW as potential predictors
ii. Consider data at the end of STW as potential predictors as well
iii. Outcome data are limited to the 79 patients that completed the 3 months post data collection process
1. Missing ~100 people’s data; that’s a lot to assign a “zero” score to
a. The missing data are not missing randomly
b. They did not finish the program (Baseline 288 – 162 = 126)
i. Many are likely to be non-exercisers
ii. Did not complete yet? These were taken out.
c. They finished the program but could not be found (154 – 79 = 85)
i. Did they die?
2. 54% of the 79 were exercisers: 43; 43/288 = 15% long term exercisers (Meets the PA recommendation); or 1 in 7 patients is a success
3. If the program costs $100/person, then a success costs $700
iv. But we do know that whether the 288 met PA recommendations; for those that don’t have data afterwards, they are assigned their baseline status
v. With the 10:1 rule, there is room for about 5 predictors
1. 1 variable for every 10 positive patient outcomes
vi. Build a descriptive model and find the most strongly associated variables
1. Age, sex, timing BMI, CA type (Table 1 characteristics), also including fatigue, PHQ9
a. Eliminate those for which no action is possible from a programmatic perspective
b. Outcomes: meets PA recommendations or GLTEK
2. Identify most likely predictors
3. Sub-analyses: sub groups
4. Univariate analysis (loss of interactions)
e. Come back with next step!
a. August 14: Kairn: Update on manuscript draft
b. August 21: Marianne: Update
c. August 28: Nancy: Preliminary findings from analysis of “steps to wellness” data base
d. Future: Connie’s Pfizer application and reviewer comments
Posted by Connie at 8/12/2015 03:12:00 PM