Enhancing Data Linkage to Break the Chain of COVID-19 Spread: The Taiwan Experience

Abstract

Digital technology has been widely used in health care systems and disease management, as well as in controlling the spread of COVID-19. As one of the most successful countries in combating the COVID-19 pandemic, Taiwan has successfully used digital technology to strengthen its efforts in controlling the COVID-19 pandemic. Taiwan has a well-established National Health Insurance System (NHIS), which provides a great opportunity to develop a nationwide data linkage model in an agile manner. Here we provide an overview of the application of data linkage models for strategies in combating COVID-19 in Taiwan, including NHIS centralized data linkage systems and “from border to community” information-driven data linkage systems during the COVID-19 pandemic. Furthermore, we discuss the dual role of digital technologies in being an “enabler” and a “driver” in early disease prevention. Lastly, Taiwan’s experience in applying digital technology to enhance the control of COVID-19 potentially highlights lessons learned and opportunities for other countries to handle the COVID-19 situation better.

Rapid Control of a SARS-CoV-2 B.1.617.2 (Delta) Variant COVID-19 Community Outbreak: The Successful Experience in Pingtung County of Taiwan

“…Ever since the emergence of Alpha-SARS-CoV-2, the Ministry of Health and Welfare (MOHW) of Taiwan used several measures, such as imposing a 14-day quarantine requirement for all arrivals, isolated suspected cases, prohibiting the export of protective masks, avoiding unnecessary hospital visits, preparing enough negative pressure isolation wards and beds at hospitals across Taiwan, establishing an efficient virus diagnosis by using new technology and data, using proactive case detection, announcing incentive plans for frontline health care workers and business, closing entertainment venues and some educational institutions, providing transparent information, asking people to maintain social distancing and use protective masks in public areas, and performing temperature checks at every entrance of shops and buildings [19]. The Taiwanese government has also used a well-established National Health Insurance System to minimize the infringement of personal privacy and maximize the benefit of public health security, prevent panic hoarding through a mask rationing system, and provide medical information in epidemic investigation [20]; furthermore, the government set up data linkage systems and information-driven data linkage systems: these important tools help us better deal with the SARS-CoV-2 situation. The experience of SARS encourages people (especially health care workers) to proactively wear full protective equipment (including eye protection, protective suits, aprons, gloves, head and foot coverings, and respirators or masks) to protect themselves from SARS-CoV-2 and reduce the risk of SARS-CoV-2 outbreaks.…”

ISPE‐Endorsed Guidance in Using Electronic Health Records for Comparative Effectiveness Research in COVID‐19: Opportunities and Trade‐Offs

“… 13 Data linkage of EHRs with other data sources (e.g., claims data with a shorter time lag, such as local insurance plan data or state‐reported Medicaid data or leveraging novel electronic data collection methods, such as software application on smartphones, to capture data in the real‐world setting) is important to address information bias due to EHR data discontinuity, although this process is often complicated by privacy concerns (e.g., the requirement of patient identifiers for data linkage), different clinical terminologies, technical specifications, and functional capabilities of different data sources. 56 , 57 Selection bias (or collider bias): It can occur if restricting an analysis to those who had a cohort‐qualifying event, such as hospitalization with COVID‐19, had been tested for active infection or who have volunteered their participation in a prospective study (i.e., conditioning on a collider variable). 58 It can also happen with studies that included only patients without missing data when the missingness did not occur completely at random (i.e., “no missing data” effectively becomes the cohort inclusion criterion).…”

“…To address this concern, a prior study has demonstrated that the patients with high EHR‐continuity have similar comorbidity profiles compared to those with low EHR‐continuity based on claims data that are not affected by EHR discontinuity 13 . Data linkage of EHRs with other data sources (e.g., claims data with a shorter time lag, such as local insurance plan data or state‐reported Medicaid data or leveraging novel electronic data collection methods, such as software application on smartphones, to capture data in the real‐world setting) is important to address information bias due to EHR data discontinuity, although this process is often complicated by privacy concerns (e.g., the requirement of patient identifiers for data linkage), different clinical terminologies, technical specifications, and functional capabilities of different data sources 56,57 Selection bias (or collider bias): It can occur if restricting an analysis to those who had a cohort‐qualifying event, such as hospitalization with COVID‐19, had been tested for active infection or who have volunteered their participation in a prospective study (i.e., conditioning on a collider variable) 58 .…”

Setting up a three‐stage pre‐endoscopy triage during the coronavirus disease 2019 pandemic: A multicenter observational study

“… 24 , 25 Patients visiting hospitals have to display their NHI card (Figure 3 ). 17 Frontline hospital staff confirmed the medical information and TOCC history of patients by checking the NHI card and screened suspected febrile patients by conducting temperature measurements (Figure 2 ). Patients with fever or symptoms of COVID‐19 were taken care of in a separate area.…”

“…The non‐laboratory‐based methods included (1) temperature measurement; (2) TOCC history checking; and (3) clinical evaluation of COVID‐19 at the OPD by clinical physicians (Table 1 ). 17 , 18 , 19 , 20 Laboratory‐based methods included the PCR assay and rapid antigen test. 6 , 8 , 21 , 22 …”