SCC Analyses

Email, digital folders, and physical folders have proven to be an inefficient long-term solution for managing multiple analysis results for many datasets. Information is too scattered and important details get lost in emails and papers that are hard to find. This page should be more efficient by 1) putting all the analyses in one place, 2) allowing for a tracking of revisions and changes, 3) creating room for copious comments and explanations with the analyses, and 4) allowing multiple authors to upload, edit, and comment.

Organization: We will only analyses here when we believe they are complete. When errors are quickly identified, we will over-write the erroneous analysis. When errors are discovered some time after the analysis was posted, we'll keep the old analysis up for reference and make appropriate comments in the page. Comments should include dates to aid in identifying the latest analysis.

Pronovost Table 3 equivalent

Download doc file of the table - PronovostTable. 2009-04-20

Download pdf of the p-values comparing each time point to baseline (paired Wilcoxon Test and two-sample Wilcoxon tests). 2009-04-20

Dataset - the streetwise infections data, adult ICU data only, with post-survey infections data filling in missing values and fixing clear errors only when the post-survey data proves sufficiently consistent with the streetwise data for that hospital.

Unit of analysis - hospital.

This table includes the median (25th percentile, 75th percentile), mean - taken as the average of hospital rates, pooled mean - taken as 1000 * the sum of infections over the sum of days, and %0 - the percentage of hospitals reporting 0 infections that quarter.

Post-Survey Q30-Q34

Analysis of Yes vs No/DKN/Missing

Table of No/Missing vs Yes with Date Missing vs Yes with Date Prior to IHI Campaign vs Yes with Date after the IHI Campaign start but before randomization vs Yes with Date after randomization - Table-Q30-Q34--Counts.pdf 2009-04-20

Analysis of the total number Yes for Q30, Q33, Q34 - Summury-Measures-Q30-Q33-Q34.pdf 2009-04-20
  • Q30 ranges from 0 - 8. It sums the 4 education questions (workshops/meetings/etc, papers/materials/etc, outreach, other) for each of BSI and VAP.
  • Q33 ranges from 0 - 6. It sums the 3 feedback questions (individual, ICU, hospital) for each of BSI and VAP.
  • Q34 ranges from 0 - 14. It sums the 7 QI interventions (revise roles, increase staffing, skill mix changes, staff driven protocols, EMR changes, training, other) for each of BSI and VAP.

Survival plots of time to implementation for Q30-Q34 (assuming missing dates occurred in 2002, i.e. well before the IHI Campaign start date).

Survival plots of time to implementation for Q30-Q34 (assuming missing dates occurred in 2008, i.e. well after the randomization date).

Table of Sensitivity Analysis for Unspecified Implementation Date Q30-Q34.

Table comparing whether Education Provided was implemented before Quality Improvement Team Q30 < Q31.

Analysis compares the proportion that implemented a quality improvement team against the proportion that provided education to ICU providers, implemented a checklist system, and provided feedback on team performance Q31 vs Q30/Q32/Q33.

Dataset - these analyses use the dataset cleaned by Sam and Rob going through the hard copies by hand. This data trumps any previous analyses, which were missing two ICUs and were imperfectly cleaning using algorithms.

Unit of analysis - ICU. ICUs treated as independent.

Topic revision: r6 - 28 Apr 2009, SamuelNwosu

This site is powered by FoswikiCopyright © 2013-2022 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Vanderbilt Biostatistics Wiki? Send feedback