To describe the number, type, and targets of international travel measures adopted during the first year of the pandemic, we constructed a dataset based on previously generated COVID-19 policy response trackers [19]. We first identified and compared leading trackers to understand their coverage of travel measures (see Sect. 1 of the Appendix). The Public Health and Social Measures (PHSM) database, which was coordinated by WHO, was selected as our primary source of data because it consolidated many trackers into a single database, collected data at the individual travel measure-level beginning in December 2019, and was the most credible database available. Despite WHO efforts to harmonize variables and data across the underlying trackers, PHSM did not apply standardized terminology to the different types and subtypes of international travel measures captured in the underlying datasets. It also lacked data on the targeted jurisdiction(s) and its travel measure-level data had not been validated (e.g., implementation date and type of measure).

Based on a critical review of terminology [16], we defined an international travel measure used in the context of COVID-19 as an action taken to control the movement of people across two or more national jurisdictions with the stated intent of preventing, controlling, or mitigating travel-related public health risks. Using the April 20, 2021, version of PHSM, we captured measures implemented by 237 countries/territories/areas (CTAs) from the start of the pandemic in late December 2019 to the end of 2020 (roughly the first year of the pandemic). A CTA was defined as any officially assigned ISO-3166 country or country subdivision code, which includes countries, overseas territories, and other special administrative regions or territories with unique political or administrative arrangements, and the ability to adopt their own travel measures. Please see Appendix Fig. 1 and Table 3 for a more detailed description of our taxonomy and how we defined the different types of measures in our dataset. Our dataset initially included the 234 CTAs identified in the PHSM database, but we then split out Hong Kong, Taiwan, and Macao from Mainland China as they have their own ISO-3166 code and have independently implemented travel measures. There are 249 officially recognized CTAs in ISO-3166, however, we further split out the Dutch islands of Bonaire, Sint Eustatius, and Saba as they were distinct in PHSM and had not implemented the same measures, for a total of 252 potential CTA . We trimmed the dataset to measures implemented from December 31, 2019, to December 31, 2020, which we refer to as the first year of the pandemic for simplicity. We focused on this period because this was when there was the most intense use of international travel measures. Plus, tracking and reporting of such measures was generally reliable during the first year. We became less confident in the quantity and quality of reporting of measures over time, likely due to reporting fatigue, and limited capacity to track the sheer volume and frequent adaptation of measures. Some of the underlying datasets that comprise the PHSM also ceased collecting data around this period [19].

A protocol was developed to code and validate the PHSM data using consistent terminology. A team of coders was trained on the protocol and each coder was assigned a set of CTAs, for which they would code all measures in PHSM. New measures were added to our dataset under one of two scenarios. First, while reading websites linked as sources to the PHSM database, coders sometimes identified previously unrecorded travel measures, which were then added. Second, if a measure in the PHSM database included more than one measure type according to our travel measure taxonomy (see Appendix Fig. 1), the original entry was split so that there was one entry in our dataset for each travel measure type. Compared to PHSM, new records accounted for 40% of our dataset.

Each entry in our dataset records the implementing CTA, measure type and subtype, one or more targeted CTA(s), and the date of adoption. Measure type refers to the broad measure type in our taxonomy, such as testing or border closure. Measure subtype provides specific details such as the timing of testing (e.g., pre-arrival) or whether a CTA closed its land, sea, air, or all borders. Appendix Tables 3 and 4 provide a complete list of measure types and subtypes. In PHSM, the status changes of measures were reported, which included the introduction of a new measure, the strengthening or extension of existing measures, the lifting or finishing of measures, and other changes (see Appendix Table 5 for more details). While each of these changes were recorded as an individual measure in our dataset, our data quality analysis determined that it was not always possible to say with certainty that measures tagged as new were in fact new or that other changes were also valid. As such, we also present what we define as the “earliest” measure, or the first time we see a country introducing a particular measure type against a particular CTA target. A version of Fig. 3, which excludes measures “extending” and “finishing” is also included in the Appendix. It is also for this reason that we present cumulative incidence figures, because if we observed a particular type of measure in the dataset, it was very likely to have occurred but we have less confidence around changes to measures or the easing or ending of measures.

If the PHSM data on a measure was unclear, for example, if the original internet link was broken or if the details of a measure could not easily be obtained, then the coder would flag the measure for additional research and discussion. Flagged measures were examined by a second coder who followed an extended coding protocol, including searching online for alternative websites to clarify the details of the measure. If the coder still could not decide on a course of action, a decision was made by the lead coder on which data to keep or by a discussion with the first author of this paper.

To identify the targets of the measures, we imported our dataset into STATA18 and then extracted the targeted country, region, city, province, or other jurisdiction that had been identified in our coding process. Most measures targeted an entire CTA, which was mapped as the target for those measures. However, if a measure targeted only a sub-jurisdiction, for example, a city (e.g., Wuhan) or a province (e.g., Hubei), then we flagged these as sub-CTA measures and then mapped it to the CTA-level target (e.g., China). Many measures targeted all CTAs or inbound travelers, in which case we flagged this as a measure that targeted “all CTAs” and mapped it to all 252 potential CTAs except the implementing CTA as targets. Similarly, some measures targeted all CTAs except a shorter list of CTAs, in which case we mapped to all CTAs except the implementing and exempted CTAs as targets.

Occasionally, a measure described a specific regional organization as its target (e.g., the European Union or the Gulf Cooperation Council), in which case we used official lists as of 2020 to map these measures to member CTAs. However, if a measure targeted a vague region, for example, “Asian countries”, we did not map these measures to individual CTAs due to the potential to misinterpret the actual target. Approximately 8.4% of the measures in our database were not mapped to a CTA. Some measures targeted specific types of travelers (e.g., non-essential workers), in which case we mapped these targets to all CTAs (unless more information was given) but flagged these as individual-based measures. Full details of the mapping, including the reasons some measures remained unmapped, can be found in Appendix Table 9.

We further categorized the implementing and targeted CTAs into the 194 WHO Member States. The remaining CTAs, collectively referred to as non-Member State CTAs, were mostly island territories but also included Lichtenstein and semi-autonomous regions, such as Hong Kong, Taiwan, Kosovo, and Palestine. We also categorized the income level of Member States using the World Bank’s income classification scheme of 2022–2023. Two Member States, Niue and Cook Islands, lacked income data, but we defined them as upper-middle-income and high-income, respectively.

In our figures, measures were grouped by week of implementation where the start of the first week of 2020 is defined as Sunday, December 29, 2019. The PHEIC was declared in week 5 and WHO described COVID-19 as a pandemic in week 11. A detailed description of the full steps taken to construct our dataset is available in the Appendix. As the project relied exclusively on secondary data in the public domain, we did not seek ethical approval for this project.