Data quality | Introduction of JIPAD | English | JIPAD | JIPAD is Japan's largest intensive care database

Data quality

Site Visits
(1) To determine the status of each ICU/PICU and provide information that may reduce the local workload
(2) To collect defective system reports or requests directly from each ICU/PICU
(3) To become acquainted with site staff and facilitate communication
Specific activities
(1) Participation in morning rounds and conferences
(2) Auditing of the system and its operations
(3) Responses to suggestions or questions regarding JIPAD operations
(1) Visited 61 ICUs between March 2015 and December 2019
The Query System
To minimize human error during the data collection process, we developed the following procedure based on interactions between participating ICU/PICUs and administrators (e.g., analysts):
(1) The participating ICU/PICU uploads individual data items in groups of 10.
(2) The administrator verifies the data and replies with comments (e.g., input errors).
(3) The participating ICU/PICU confirms the comments and, if necessary, modifies the dataset and repeats the upload.
(4) The participating ICU/PICU then enters and uploads the next 10 individual data items while considering the previously received comments.
(5) Both the participating ICU/PICU and the administrator continue this process until the uploaded data contain no input errors. The participating ICU/PICU will then be able to upload data freely.
Random Checks
We checked 10 randomly selected the individual data items for each ICU/PICU that had completed the above-described query system procedures and responded with comments regarding typing errors.
This process has enabled us to maintain data accuracy.
ANZICS and ICNARC also inspect the inputted data. ICNARC aims to inspect all data, assuming that sufficient manpower is available.
In most ICUs/PICUs that adopted the query system, the number of rudimentary errors decreased dramatically after the data had been entered for 50–60 patients. To ensure the continued maintenance of data quality, we repeated the above query system procedure whenever the personnel responsible for data input changed.
Although the query system primarily aimed to increase data accuracy, the system identified both human errors and design errors in the CSV output format adopted by the ICU/PICU.