What are the latest health research applications using R? A roundup from the 2022 NHS-R conference.
Performance reports published on a regular basis are common across the NHS. Rather than copying and pasting outputs from Excel to Powerpoint, Richard Wilson made the case for creating an interactive HTML report, using R Markdown, that is reproducible and quick to deliver. A report example is available on Github using the Summary Hospital-level Mortality Indicator (SHMI) dataset.
Improving the reporting process for the radiotherapy data has been Louise Reynolds’s (NDRS/NHS Digital) priority. Her department’s reporting process that used to be time-consuming and prone to error is now automated and produces standardised output in a few minutes. Relevant packages used for this project are R Markdown, DBI, tidyverse, data.table, and openxlsx.
Daniel Weiand (Newcastle upon Tyne Hospitals NHS Foundation Trust) wrote entire academic papers using Quarto and R Markdown. Combined with Zotero, it is possible to embed citations into your document. Quarto also has cross-referencing functionalities where it is possible to integrate ggplot with captions into the text.
Two visualisation sessions were run by Cara Thompson, a freelance data consultant. In the first one, she showed us how to build a custom ggplot theme in less than 10 minutes. She highlights three reasons why ggplot themes are useful “-1. They add text hierarchy, -2. They give space to breathe, -3. They create aesthetic consistency across the project ”.
In the second one, she shared 5 tips to create bespoke colour schemes to plot. “-1. Use colour purposefully, -2. Let others help you, -3. Apply colour using a named vector, -4. Check for accessibility, and finally -5. Make use of colour interpolation.”
Chris Beeley presented Shiny endomineR, a text mining tool for clinical text developed by the NHS-R Community. The package is not finalised yet so keep an eye out for it.
Across databases, patient information can be spelled differently and thus making data integration more difficult and time consuming. The NHSBSA Data Science team has developed the addressMatchR package to identify the best address match based on a similarity algorithm.
Ruchir Shah, a radiology registrar in Oxford, was able to automatically extract paediatric lung function data from pdf format to a usable dataframe using the R Tesseract package.
Samuel Channon-Wells (Imperial College London) used electronic prescribing records to monitor the level of antibiotic use in children. Interactive dashboards were built for Oxford University Hospitals using R and R Shiny and aimed to help infectious paediatric doctors. Implementation challenges include lack of familiarity with R, assimilating existing infrastructure, and maintenance debugging. After a year of use, the initiative allowed to identify and prevent overuse of antibiotics among children patients.
Calum Polwart (South Tees Hospital NHS Foundation Trust) developed the SACT Analyser package to clean and process systematic anti-cancer treatment data (SACT). SACT is a database gathering information on anti-cancer therapy activity from all NHS providers. The data is published on a monthly basis and is key to understanding treatment patterns and outcomes at a national level. The package is still in early development.
NHSRplotthedots is a R package built by the NHS-R community to provide tools for drawing SPC charts. It supports NHSE/I programme ‘Making Data Count’ and allows users to draw XmR charts, use change points, apply rules with summary indicators.
Chris Mainey (NHS North Central London ICB) manages to pull a crash course on linear and non-linear regression modelling in less than 15 minutes. Use the mgcv package in base R to build nonlinear regression models with smoothers.
Laura Moscoviz is a Health Care and Life Science Consultant at RwHealth in London. As part of her role, she uses data science, technology, and predictive analytics to deliver insight-driven solutions to improve quality of care and operational delivery.