Analytic approach
Across the 3 cohorts, 122 CF Programs participated in the study (80 in
Cohort 1, 30 in Cohort 2, & 12 in Cohort 3), which represented 88.4%
of grant recipients. Programs in Cohorts 1 and 2 completed multiple
yearly surveys, and 39 programs had more than one provider complete the
survey for a given year. All analyses were performed at the level of the
CF program and scores from multiple providers were aggregated within
each program using the mean implementation score and the maximum number
of years at the program across providers. Descriptive statistics were
calculated for each cohort and compared across cohorts using chi-square
tests of independence. To account for nesting of programs within years,
a random intercept linear mixed effect model was used to estimate the
total implementation score by cohort and year. To examine differences in
implementation across programs and to identify predictors of
implementation scores, we used a linear regression model with the
following predictors: type of program (pediatric, adult, both), size of
the program (small, medium, large), and years of experience working on
the CF team. Cohort was included as a covariate. To examine differences
in implementation over time, we used a random effect model to predict
both initial implementation and change in implementation across time
(i.e., interactions between predictors and year) using grant year, type
of program, program size, and MHC years of experience as predictor
variables. Random effects included the intercept and year. All analyses
were performed using the lme4 v1.1-26 , Table one v0.13.0, andstats v4.0.3 R packages.