Population-Based Clinical Studies Using Routinely Collected Data in Hong Kong : A Systematic Review of Established Local Practices and Counterfactual Analysis

Background: Routinely collected health data are increasingly used for clinical or epidemiological research. However, no study has systematically reviewed the temporal trends of the number of publications, databases, characteristics of study cohorts, methodology, outcomes, data governance, research ethics committee (REC) approval, and variations of practice therein in Hong Kong.Methods: PubMed was systematically searched from its inception to 28th March 2023 for studies using routinely collected healthcare data in Hong Kong. Studies using non-routinely collected data were excluded. A hypothetical situation in which a healthcare system requires ethics approval from different RECs to review identical applications is proposed. Counterfactual analysis was conducted to estimate the costs saved by not requiring repeated reviews of identical ethics applications.Results: A total of 1757 entries were retrieved. Of these, 451 met the inclusion criteria. An additional search yielded three extra studies. In total 454 studies were included for analysis. The first population-based study was published in 1989. Between 2000 and 2009, 32 studies were identified (2002: n=4; 2003: n=5; 2004: n=2; 2005: n=5; 2006: n=2; 2007: n=6; 2008: n=7; 2009: n=1). There was an increase in the number of publications from 5 to 120 between 2010 and 2022. Of the investigator-led studies using Hospital Authority (HA)’s cross-cluster data (n=393), 327 (83.2%) reported having ethics approval from a single cluster/university-based REC, whilst 50 studies (12.7%) did not report obtaining any approval from a REC. To use HADCL, approval by a single hospital-based or University-based REC is accepted. Repeated submissions of identical ethics applications to different REC are estimated to cost $4.2 million HKD yearly.Conclusions: A significant increase in the outputs using routinely collected health data from Hong Kong is observed. Most studies reported holding approval from a single cluster REC before retrieval of cross-cluster HA data. Counterfactual inference indicates significant savings upon removing the requirements of repeated reviews of identical ethics applications