Data Quality Improvement and Internal Data Audit of the Chinese Neonatal Network Data Collection System

Jianhua Sun, Yun Cao, Mingyan Hei, Huiqing Sun, Laishuan Wang, Wei Zhou, Xiafang Chen, Siyuan Jiang, Huayan Zhang, Xiaolu Ma, Hui Wu, Xiaoying Li, Yuan Shi, Xinyue Gu, Yanchen Wang, Tongling Yang, Yulan Lu, Wenhao Zhou, Chao Chen, Shoo K. Lee, Lizhong Du, The Chinese Neonatal Network, Shoo K. Lee, Chao Chen, Lizhong Du, Wenhao Zhou, Yun Cao, Falin Xu, Xiuying Tian, Huayan Zhang, Yong Ji, Zhankui Li, Jingyun Shi, Xindong Xue, Chuanzhong Yang, Dongmei Chen, Sannan Wang, Ling Liu, Xirong Gao, Hui Wu, Changyi Yang, Shuping Han, Ruobing Shan, Hong Jiang, Gang Qiu, Qiufen Wei, Rui Cheng, Wenqing Kang, Mingxia Li, Yiheng Dai, Lili Wang, Jiangqin Liu, Zhenlang Lin, Yuan Shi, Xiuyong Cheng, Jiahua Pan, Qin Zhang, Xing Feng, Qin Zhou, Long Li, Pingyang Chen, Xiaoying Li, Ling Yang, Deyi Zhuang, Yongjun Zhang, Jianhua Sun, Jinxing Feng, Li Li, Xinzhu Lin, Yinping Qiu, Kun Liang, Li Ma, Liping Chen, Liyan Zhang, Hongxia Song, Zhaoqing Yin, Mingyan Hei, Huiwen Huang, Jie Yang, Dong Li, Guofang Ding, Jimei Wang, Qianshen Zhang, Xiaolu Ma

Background: The Chinese Neonatal Network (CHNN) is a nationwide neonatal network that aims to improve clinical neonatal care quality and short- and long-term health outcomes of infants. This study aims to assess the quality of the Chinese Neonatal Network database by conducting an internal audit of data extraction. Methods: A data audit was performed by independently replicating the data collection and entry process in all 58 tertiary neonatal intensive care units (NICU) participating in the CHNN. Eighty-eight data elements selected for re-abstraction were classified into three categories (critical, important, less important), and agreement rates for original and re-abstracted data were predefined. Three to five records were randomly selected at each site for re-abstraction, including one short- (0-7 days), two medium- (8-28 days), and two long-stay (more than 28 days) cases. Agreement rates for each data item were calculated for individual NICUs and across the network, respectively. Results: A total of 283 cases and 24,904 data fields were re-abstracted. The agreement rates for original and re-abstracted data elements were 96.1% overall, and 97.2, 94.3, and 96.6% for critical, important, and less important data elements, respectively. Individual site variation for discrepancies ranged between 0.0 and 18.4% for all collected data elements. Conclusion: The completeness, precision, and quality of data in the CHNN database are high, providing assurance for multipurpose use, including health service evaluation, quality improvement, clinical trials, and other research.