Abstract:BACKGROUND: Big data analytics promise insights into healthcare processes and management, improving outcomes while reducing costs. However, data quality is a major challenge for reliable results. Business process discovery techniques and an associated data model were used to develop data management tool, ICU-DaMa, for extracting variables essential for overseeing the quality of care in the intensive care unit (ICU).OBJECTIVE: To determine the feasibility of using ICU-DaMa to automatically extract variables for the minimum dataset and ICU quality indicators from the clinical information system (CIS).METHODS: The Wilcoxon signed-rank test and Fisher's exact test were used to compare the values extracted from the CIS with ICU-DaMa for 25 variables from all patients attended in a polyvalent ICU during a two-month period against the gold standard of values manually extracted by two trained physicians. Discrepancies with the gold standard were classified into plausibility, conformance, and completeness errors.RESULTS: Data from 149 patients were included. Although there were no significant differences between the automatic method and the manual method, we detected differences in values for five variables, including one plausibility error and two conformance and completeness errors. Plausibility: 1) Sex, ICU-DaMa incorrectly classified one male patient as female (error generated by the Hospital's Admissions Department). Conformance: 2) Reason for isolation, ICU-DaMa failed to detect a human error in which a professional misclassified a patient's isolation. 3) Brain death, ICU-DaMa failed to detect another human error in which a professional likely entered two mutually exclusive values related to the death of the patient (brain death and controlled donation after circulatory death). Completeness: 4) Destination at ICU discharge, ICU-DaMa incorrectly classified two patients due to a professional failing to fill out the patient discharge form when thepatients died. 5) Length of continuous renal replacement therapy, data were missing for one patient because the CRRT device was not connected to the CIS.CONCLUSIONS: Automatic generation of minimum dataset and ICU quality indicators using ICU-DaMa is feasible. The discrepancies were identified and can be corrected by improving CIS ergonomics, training healthcare professionals in the culture of the quality of information, and using tools for detecting and correcting data errors.

ricu: R's interface to intensive care data

The eICU Collaborative Research Database, a freely available multi-center database for critical care research

Integrated Data Repository Toolkit (IDRT). A Suite of Programs to Facilitate Health Analytics on Heterogeneous Medical Data

Yet Another ICU Benchmark: A Flexible Multi-Center Framework for Clinical ML

Design and Practice of the Clinical Information System for ICU

Harnessing Big Data in Critical Care: Exploring a new European Dataset

eICU Collaborative Research Database application and data extraction process

Salzburg Intensive Care database (SICdb): a detailed exploration and comparative analysis with MIMIC-IV

rEHR: An R package for manipulating and analysing Electronic Health Record data

A Common Longitudinal Intensive Care Unit data Format (CLIF) to enable multi-institutional federated critical illness research

PIC, a Paediatric-Specific Intensive Care Database

A Methodological, Real-World Data Analysis from 86 Helios Hospitals

A Multidatabase ExTRaction PipEline (METRE) for Facile Cross Validation in Critical Care Research

The MIMIC Code Repository: enabling reproducibility in critical care research

Simpler handling of clinical concepts in R with clinconcept

National ICU Registries as Enablers of Clinical Research and Quality Improvement

"fhircrackr": An R Package Unlocking Fast Healthcare Interoperability Resources for Statistical Analysis

A scalable approach for critical care data extraction and analysis in an academic medical center

Validation of the ICU-DaMa tool for automatically extracting variables for minimum dataset and quality indicators: The importance of data quality assessment

MIMIC-IV on FHIR: converting a decade of in-patient data into an exchangeable, interoperable format

Scalable Predictive Analysis in Critically Ill Patients Using a Visual Open Data Analysis Platform