CordDqChecker-MIM provides a set of Data Quality (DQ) tools and metrics for the special issue "Quality of Data in Health Research and Health Care” in the journal “Methods of Information in Medicine (MIM)". Theses tools were used to conduct local as well as distributed data quality assessments and reporting in CORD-MI.

Acknowledgement: This work was done within the “Collaboration on Rare Diseases” of the Medical Informatics Initiative (CORD-MI) funded by the German Federal Ministry of Education and Research (BMBF), under grant number: 01ZZ1911R, FKZ-01ZZ1911R.

• The data quality framework dqLib has been used as an R package for generating specific reports on DQ related issues and metrics.

• The following DQ indicators and parameters are configured by default reports:

  Dimension	DQ Indicator
  completeness	item completeness rate, value completeness rate, subject completeness rate, case completeness rate, orphaCoding completeness rate
  plausibility	orphaCoding plausibility rate, range plausibility rate
  uniqueness	RD case unambiguity rate, RD case dissimilarity rate
  concordance	concordance with reference values

  DQ Parameter	Description
  inpatient cases	number of inpatient cases per year
  inpatients	number of inpatient per year
  RD cases	number of RD cases per year
  Orpha cases	number of Orpha cases per year
  tracer cases	number of tracer RD cases per year
  RD cases rel. frequency	relative frequency of inpatient RD cases per year per 100.000 cases
  Orpha cases rel. frequency	relative frequency of inpatient Orpha cases per year per 100.000 cases
  tracer cases rel. frequency	relative frequency of inpatient tracer RD cases per year per 100.000 cases
  missing mandatory data items	number of missing data items per year
  missing mandatory data values	number of missing data values per year
  incomplete subjects	number of incomplete inpatient records per year
  missing orphacodes	number of missing Orphacodes per year
  outliers	number of detected outliers per year
  implausible links	number of implausible code links per year
  ambiguous RD cases	number of ambiguous RD cases per year
  duplicated RD cases	number of duplicated RD cases per year

• The following references are required to assess the quality of orphacoding and can be easily updated with new versions: (1) The standard Alpha-ID-SE terminology [1], and (2) a reference for tracer diagnoses such as the list provided in [2].

  [1] BfArM - Alpha-ID-SE [Internet]. [cited 2022 May 23]. Available from: BfArM

  [2] Tahar K, Martin T, Mou Y, et al. Distributed Data Quality Assessment Across CORD-MI Consortia. doi:10.3205/22gmds116

We applied our methodology for distributed data quality assessments on synthetic data stored in multiple FHIR servers. The developed DQ-Tools were successfully tested and validated using Personal Health Train (PHT). The aggregated DQ results are stored in ./PHT/Data/Export. To create a PHT image run the command sudo docker build -t dq-train . from the main folder. The FHIR data sets and tools used for data curation are stored in the branch methods_dataCuration.

• Here are examples of cross-site reports on data quality generated using synthetic data.
• The used references are store in the folder "./PHT/Data/refData"

To analyze your data quality locally go to folder ./Local and run cordDqChecker.R to generate data quality reports.

• The script cordDqChecker.R reads data from FHIR server or from supported file formats such as CSV and Excel. The path variable specifies which data set should be imported. For Example you can define your path as following:

  • path="http://141.5.101.1:8080/fhir/" or
  • path="./Data/medData/dqTestData.csv" or
  • path="./Data/medData/dqTestData.xlsx"

• The default values of local variables are set as follows:

  • reportYear=2020
  • max_FHIRbundles=50
  • INPATIENT_CASE_NO=997
  • path="http://141.5.101.1:8080/fhir/

Once the source data path and local variables are defined, start the script to check the quality of your data. The generated reports are saved in folder "./Local/Data/Export"

• In the folder "./Local/Data/Export" you will find some examples of data quality reports generated locally using synthetic data.
• The used references are stored in "./Local/Data/refData"

• Before starting cordDqChecker.R you need to install required libraries using the script installPackages.R stored in "./Local/R"

• To avoid local dependency issues go to folder ./Local and just run the command sudo docker-compose up to get cordDqChecker running.

• The missing item rate will be calculated based on FHIR implementation guidlines of the MII core data set. Hence, mandatory items of the basic modules Person, Case and Diagnosis are required.

• To cite cordDqChecker-MIM, please use the following BibTeX entry:

  @software{Tahar_cordDqChecker-MIM,
  author = {Tahar, Kais},title = {{cordDqChecker-MIM: Data quality tools for the special issue "Quality of Data in Health Research and Health Care” in the journal “Methods of Information in Medicine (MIM)”}},
  url = {https://github.com/KaisTahar/cordDqChecker-MIM},
  year = {2022}
  }
  
  

See also: dqLib CORD-MI