Infectious diseases (ID), a leading cause of morbidity and mortality worldwide, place a huge burden on the world, particularly in certain countries, regions, and territories, along with increasingly changing climate conditions and globalization (1, 2). It is important to understand the spatiotemporal patterns through various mining methods, and the complex process of disease transmission, its potential influences, and its driving mechanisms. This would provide insightful guidance to health authorities for making and implementing targeted interventions for infectious diseases.

Previous studies have pointed out that the spatiotemporal patterns of infectious disease prevalence are a macroscopic result of the interaction between the virus/pathogens, vectors, and hosts (3). Spatiotemporal patterns refer to the spatiotemporal distribution characteristics of geographical objects, including frequency (frequent paths, mainly for moving objects), aggregation, periodicity (periodic and seasonal characteristics of research objects), correlation (changes in research objects over time), and co-location (frequent spatial proximity between objects) (4, 5). Thus, mining infectious disease spread and transmission patterns on various spatial scales and temporal intervals would detect the diseases' epidemiological characteristics, explore their potential influences, clarify their influence mechanisms, predict epidemic risks under various conditions, based on the spread and transmission simulation, and evaluate the feasibility or effectiveness of preventive and control measures implemented by relevant health authorities (5, 6). In other words, this paradigm covers nearly every aspect of infectious diseases, except for the isolation or laboratory testing of the pathogens, through which effective interventions for various infectious diseases were achieved in the past decades (5, 6).

Due to successful performance in other research fields, a series of spatiotemporal analytical methods or models, such as kernel density analysis, spatial-time scan statistics, geo-detector, and random forest, have been appropriately employed for mining the spatial and temporal patterns of various communicable diseases (7–22). Meanwhile, different data sources and acquisitions were required for completing spatiotemporal pattern mining since the spread and transmission were commonly caused by different viruses or pathogens through specific vectors or direct human-to-human transmission. They included gene sequence data for infectious diseases from public databases, vector sentinel monitoring data, and infectious disease cases data based on streaming or reporting that covered pathogens, vectors, hosts, and environmental suitability (pathogens' survival and transmissibility, vectors' breeding, and hosts' being infected) (3). In addition, the rising utility of individual activity information, derived from social media or other mobility data, provided opportunities for clarifying the role of individuals in disease transmission and assessing the risk of being infected (23). The question remains, since the results and related conclusions were achieved on a certain fine scale (i.e., relative microscopic level), were the containment measures and interventions, based on these microscopic findings, appropriately implemented at mesoscale or macroscale? In addition, some other concerns in this field include hot diseases and their spectrum and the academic leadership of international collaborations network among countries, institutes, or scholars, which have seldom been focused on in previous literatures.

As efficient tools for reviewing literature, based on numerous publications, bibliometrics and content analysis have been widely applied in various fields to quantitatively explore the number of publications and their variations, arrange the concerned aspects and their timelines, and summarize the progress or proper prospects (24–27). These could provide useful clues and unique insights for scholars and health authorities. For example, Su et al. (28) summarized the disciplines involved, the main contributing countries, the main research themes, and the shortcomings of the existing research in the application of big data in carbon emissions and environmental management research based on bibliometric methods. Agnusdei and Coluccia (27) used bibliometrics and content analysis to summarize the trends and prospects in sustainable agri-food supply chain research, and four research themes were identified and several research gaps were discussed. Therefore, bibliometrics and content analysis were employed in this study to investigate the counts of publications regarding spatiotemporal patterns, dominant research directions of mining these patterns, the spectrum of infectious diseases concentrated upon in these studies, and the present scenario of international leadership and academic collaboration networks among countries.

This study identified the number of publications and citations, the spectrum of infectious diseases, major research directions and data-methodological-theoretical characteristics, academic collaborations, and leadership characteristics over the past 30 years in the mining of spatiotemporal patterns of infectious disease prevalence. The findings provide useful insights for advancing the detection of future spatiotemporal patterns of infectious diseases, the analysis of the influencing factors and driving mechanisms, academic cooperation, and prevention and control synergy.

The ever-increasing number of publications is likely related to a series of pandemics. The wide spread of SARS in 2003 (52) changed the field from a slow-rising phase to a fast-rising phase. Three extreme points observed in the fast-rising phase (2010, 2014, and 2017) corresponded with influenza (a pandemic transmitted from Mexico) (53), worldwide outbreaks of dengue (54), cholera epidemics of 2010 (55), unprecedented Ebola epidemics in West Africa (56), the most serious outbreak of dengue in China (57) in 2014, the cholera epidemic with thousands of infections in Yemen (58), the plague in Madagascar with more than 2,600 cases (59), and the 2017 influenza prevalence in the Northern Hemisphere (largest prevalence in the past decade) (59). The period of 2019-2021 (rapid rising phase), for the current world, was marked by the ongoing COVID-19 pandemic (60, 61), dengue fever epidemics in the subtropical and tropical regions (38), and the influenza pandemic worldwide (62). In summary, the characteristics of the infectious disease spectrum and its changes were, to some extent, strongly related to the pandemic or epidemic severity of major infectious diseases. Therefore, as the spectrum of infectious diseases illustrated (Figure 3), despite the remarkable achievements made by mankind to fight infectious diseases, such as SARS (63), cholera (64), plague (65), and avian influenza (66), the number of publications are likely to continue to rise since new infectious diseases (e.g., COVID-19) are emerging and pre-existing diseases (e.g., dengue and influenza) remain rampant.

This study uncovered the paradigmatic features of research in this field, including the logical nature of the three research directions and data-methods-theories/ideas. In spatiotemporal pattern detection (RD_1), wavelet analysis (67) was applied because it could detect periodic patterns of infectious disease epidemics and their changes over time. Spatiotemporal scanning (68) was widely employed owing to its ability to simultaneously detect specific clustering time intervals and spatial ranges. Co-location pattern mining algorithms (69) detected urban sites with a significant spatial correlation of infectious disease with case locations at the point scale. This method holds promise, however, there has been little research regarding it. RD_2 was used to identify the factors influencing the formation of the spatiotemporal patterns detected in RD_1, where the geo-detector (70) method, proposed in 2010, has become widely employed to simultaneously consider factors individually and interactively. With the improvement and perfection of machine learning and deep learning methods, methodologies, such as Bayesian models and Maxent (71), have often been applied to predict future epidemic risks of infectious diseases (RD_3) based on the influencing factors identified in RD_2. These can overcome the multi-collinearity between independent variables and prevent overfitting. In particular, they can detect the contribution of each variable while predicting the risk, and could become one of the dominant methods in this research direction. As can be seen, the three research directions were logically rigorous, which supported the framework of the research paradigm. However, the theories and methods involved in each direction are iterated and updated with time, indicating that the field deserves and requires in-depth research.

The spread of infectious diseases has no national boundaries and international cooperation is the only way and, the inevitable result of, fighting infectious diseases (25, 72). Over the last 30 years, developed countries have had a higher intensity of cross-country collaboration than developing countries, with more publications, h-index (citation), well-known relevant institutions, and scholars. These reflect stronger activity and leadership, which could be related to the prevalence of certain infectious diseases (52–56, 58, 59, 73). There has been a long-term accumulation of scientific research on the prevalence and control of infectious diseases in developed countries and a higher strategic profile (21, 62, 74–83). It was encouraging to note that China (a developing country) had a promising future in terms of international impact with a rapidly growing and leading level of intensity in international cooperation, publications, and citations, although its leadership was temporarily inferior to the USA and England. This could be related to the risk of traditional infectious diseases with long epidemiological history (84–88) and the increasing risk of importation of epidemics with frequent foreign exchanges (89) that requires China to make continued efforts in disease surveillance, data acquisition and sharing platform construction, and the development of universal methods or models (90, 91). However, China has a shortfall in terms of research areas of interest (only 13.5% of the publications focused on other countries) and research focus (smaller numbers of micro-level related research) compared to the USA and England. Thus, developing countries, such as China, need to expand the breadth and depth of their research to gain a broader, more in-depth, and stronger role in these collaborations (92, 93).

Although remarkable progress has been achieved, several points remain ill-considered and unresolved.

The spread, prevalence, and control of infectious diseases are complex ecological-geographical processes and the scale of effect (including time and space) is an important feature of this process because of the scale dependence on geographical objects (94, 95). Therefore, scale consistency is the primary issue that needs to be carefully considered when translating research findings into implementable measures. This has often been overlooked in existing research, although researchers chose the appropriate spatiotemporal scales before conducting a study. This study considered the spatial scale consistency by using a regular grid as a scale for conducting research by upscaling (fine-scale to coarse scale) and making recommendations. Regular grids can avoid the statistical and analytical issues that are likely to be caused by the research units assigned by the administrative divisions (districts, towns, streets, and villages) in relevant studies due to irregular and changeable shapes (96). In particular, these spatial grids have been gradually considered as the final units where prevention and control measures can produce practical effects in urban regions (89, 97). The appropriate grid scale can be chosen by comparing the ability (as reflected by some indicators, such as Moran's I) of a range of regular grids at different scales to characterize the spatiotemporal characteristics of infectious disease epidemics (51). However, if the resolution of the raw data is limited (e.g., at the county scale), when translating the findings into control measures at a finer scale (e.g., at the township scale), it must be carefully investigated if the findings had similar characteristics at a finer scale. It is important to note that the more the scale of the study differed from the scale of the proposed control measures, the results would be less credible.

Human mobility plays an important role in the transmission of infectious diseases (23, 98, 99), but most of these studies have focused only on infectious diseases that can be transmitted from human to human (e.g., COVID-19). Infectious illnesses include human-to-human (direct), human-vector (animal/insect)-human (indirect) types, etc. (1, 100). Human-to-human infectious diseases spread more rapidly and latently (e.g., COVID-19), in which human behavior factors have attracted more attention because of their direct impacts on the disease transmission and prevalence (23, 72, 101). In comparison, for human-vector (animal/insect)-human infectious diseases, the vector factors rather than human behavior factors were mainly concentrated in current studies although human behavior factors may pose impacts on the disease transmission and prevalence (102, 103). Actually, human often possesses higher mobility (an aspect of human behavior) than that of vectors (insects or animals), by which some infectious diseases (e.g., dengue, malaria, etc.,) transmitted by mosquitoes may be heavily affected by human behavior factors, especially when he/she has been infected or is a pathogen carrier (99, 104). The challenges of quantifying the impact of human mobility on insect- vector diseases at a fine spatial scale are related to the availability of data and the fusion of data from multiple sources (human mobility and infectious disease epidemics). With the advent of big data, data availability can be expected related to population location heat maps, bus/subway smart card swipe data, and taxi track data (105, 106). For fusing data from multiple sources, this study suggests clustering, random forest, and deep learning (e.g., convolution neural network) to convert human mobility data with fine time granularity into patterns at the same temporal scale as the incidence rate. Hence, the association between human mobility patterns (category variables) and the prevalence of infectious diseases can be detected at fine spatial scales using methods such as geo-detectors. Thus, human mobility characteristics can be further explained based on the point of interest (POI) and land use types of different mobility patterns to infer the processes and causes of infectious disease epidemics.

This study had certain limitations. The data were obtained only from the WoS core database, which resulted in some relevant articles being excluded. In the future, the authors could consider integrating data from multiple databases, such as Scopus and PubMed. In addition, we searched for publications based on the existence of specific keywords in the title, abstract, and keywords of the article and may have missed some relevant publications. Additional information (e.g., conclusions) must be combined to search the publications.

This study revealed that the epidemiological and control situation of infectious diseases over the past 30 years has led to an increase in the publication of spatiotemporal pattern mining studies and a change in the spectrum of focused diseases. Developed countries have stronger leadership than developing countries in this field. Existing research, on the whole, has followed the logic and paradigm of pattern-factors-mechanisms. The authors suggest that international cooperation must continue to advance and developing countries, such as China, must strengthen their breadth and depth of research to enhance their influence in the field. Scholars need to strengthen the exploration of scale effects and human mobility in the field. Thus, this study provides useful insights for scholars and relevant health authorities to understand the progress of existing research and the remaining gaps, and develop relevant research, prevention, and control strategies.

WL: conceptualization, methodology, visualization, validation, writing–original draft, supervision, and writing–review and editing. HR: data curation, investigation, formal analysis, writing–original draft, and writing–review and editing. All authors contributed to the article and approved the submitted version.