We conducted a cross-sectional study using open and publicly available data sources. Specifically, the data on the COVID-19 diagnostics used, including the polymerase chain reaction (PCR) and antigen tests, were mainly sourced from the Foundation for Innovative New Diagnostics (FIND). Considering the accessibility of data, we analyzed the usage data from March 1, 2020 to October 31, 2022. Countries were categorized into four regions based on their sociodemographic index (SDI) and six WHO regions according to geographic contiguity. SDI is a composite indicator that reflects a country’s socio-demographic level. More information about the distribution of SDI is provided in Additional file 1. Data concerning factors associated with the utilization of COVID-19 diagnostic tools encompassed five sets of country-level indicators, including the severity of COVID-19, socioeconomic status, health status, medical service capacity, and rigidity of response. Evidence of COVID-19's dynamic severity and rigidity of response was primarily collected from “Our World in Data” with recent diagnostics data from different governments. Data on socioeconomic status, health status, and medical service capacity were obtained from the Global Health Data Exchange query tool with relevant data since 2019. Detailed descriptions of data collection methods and data sources are provided in Additional file 2.

This study was conducted from November 1, 2022 to February 28, 2023. We used standardized country names to link the multilevel datasets and excluded countries with missing values on key variables. Ultimately, 161 countries and territories were included in our study. This study used publicly available data and was deemed exempt from guidelines for human research from the Institutional Review Board of Peking University.

The testing rate used in our study was determined by calculating the number of tests procured per 1000 people for each country. We defined the total testing rate as the cumulative number of tests procured per 1000 people for each country, which encompassed the total nationwide usage per 1000 people from March 1, 2020, to October 31, 2022, including both PCR and antigen tests. The study period was chosen for several reasons, including the high infection rates of the Omicron variant, variations in the frequency of testing data reporting among many countries, and the different types of tests used, which resulted in varying reporting methods between countries, potentially affecting the accuracy of testing data. In addition, to investigate the changes and time trends in testing rates, we also analyzed the total testing rate for the last twelve months (from November 1, 2021, to October 31, 2022) and the monthly testing rate (cumulative number of COVID-19 tests per 1,000 people every month). The monthly testing rate refers to the cumulative number of tests per 1,000 people for each month.

Five sets of factors associated with the usage of COVID-19 diagnostics were analyzed. The severity of COVID-19 was measured by the number of deaths and cases per 100,000 people, both in total and on a monthly basis according to the outcome variables. Socioeconomic status was assessed by Gross Domestic Product (GDP) per capita and the proportion of people aged ≥ 70. The health status was characterized by the prevalence of cardiovascular diseases, chronic respiratory diseases, diabetes diseases and neoplasms per 100,000 people [16]. The medical service capacity was quantified by using the health workforce density of per 100,000 people. The rigidity of response was measured by using the stringency index, a composite measure encompassing nine response metrics: school closures, workplace closures, cancellation of public events, restrictions on public gatherings, closures of public transport, stay-at-home requirements, public information campaigns, restrictions on internal movements, and international travel controls [17]. To better illustrate the result comparison, we computed standardized Z scores by subtracting the mean of the independent variables from each variable value and then dividing it by the standard deviation (SD).

Descriptive statistical analyses were conducted to describe the usage of COVID-19 diagnostics across the world. A negative binomial regression model was used to investigate into five sets of factors associated with the total usage and a generalized linear mixed model was used to investigate the impact of these same factors on monthly usage, with all influencing factors simultaneously included in the model. To assess the relative importance of the five sets of factors associated with the total testing rate or testing rate per month, we conducted a dominance analysis for decomposition. The dominance statistics were used as an index of effect size [18]. The Blinder–Oaxaca decomposition technique for nonlinear models was employed to decompose the differences in the usage between low-income countries (categorized as low or lower-middle countries by SDI) and high-income countries (categorized as high or upper-middle countries by SDI), thus elucidating the determinants of these disparities [19].

All associations were presented as incidence rate ratio (IRR) or coefficients with corresponding 95% confidence intervals (Cis). A two-sided p-value < 0.05 was considered statistically significant. Stata version 16.0 for Mac (Stata Corp, College Station, TX, USA) and R Studio Version 1.2.5042 (The R Project for Statistical Computing, Vienna, Austria) were used for the statistical analyses.