kumc-bmi/bc_qa

GPC Breast Cancer Cohort Characterization; formerly https://bitbucket.org/gpcnetwork/bc_qa/

GPC Breast Cancer Data Quality Reporting

based on earlier work by Jianghua He, with contributions from Bradley McDowell and Theresa Shireman, under the direction of Elizabeth Chrischilles

by Dan Connolly, with Russ Waitman, Tamara McMahon, and Vince Leonardo
Medical Informatics Division, Univeristy of Kansas Medical Center

Copyright (c) 2015 Univeristy of Kansas Medical Center
Share and Enjoy according to the terms of the MIT Open Source License.

Background

On 23 Dec 2014, GPC honest brokers were requested to run a breast cancer cohort query and submit results (see bc_qa2.Rmd for details). All participating sites have now done so, and we are evaluating the results (227) using automated reports built with R Markdown.

An initial QA report was sent to each site 23 Feb 2014.

Site Usage

In future iterations, sites are encouraged to run this report on their own before sumitting:

1. Get the bc_qa code
   • Visit bc_qa downloads and get a zip file, or
   • clone the https://bitbucket.org/gpcnetwork/bc_qa repository
2. Build Query Terms and Exclusion Criteria article
   1. In RStudio, install.packages(pkgs=c('RSQLite', 'ggplot2', 'reshape', 'xtable'))
   2. Open bc_qa2.Rmd and Knit HTML
      • output: bc_terms_results.RData
3. Build QA for SITE article
   1. Use DataBuilder or equivalent to generate sqlite file.
   2. Copy dataset-example.R to dataset.R and edit filename etc.
   3. Knit bc_excl.Rmd

Central Usage

As new submissions come in, members of the breast cancer research team can reproduce the analysis of data from all sites:

1. Fetch all the data files.
   • Knit bc_fetch.Rmd to build bc_fetch_results.Rmd
2. Build Query Terms and Exclusion Criteria article as above
3. Build any QA for SITE articles you like, using dataset.R as below.
4. Build Data by Site presentation
   • Open bc_qa_p1.Rpres in R Studio and use the presentation tab.
5. To mail results to all sites
   1. comment out SITE <- ... in dataset.R and run report-all-sites.R.
   2. Move the report-SITE.html files to data-files.
   3. Use report_mail.py to mail the reports.

load("bc_fetch_results.RData")

SITE <- 'KUMC'  # Salt to taste

(function (s) {
  list(
    conn=fetch$site.data(s),
    about=subset(fetch$dataset, site == s)
  )
})(SITE)