Trends, Issues and Future Directions of Urban Health Impact Assessment Research: A Systematic Review and Bibliometric Analysis

Health impact assessment (HIA) has been regarded as an important means and tool for urban planning to promote public health and further promote the integration of health concept. This paper aimed to help scientifically to understand the current situation of urban HIA research, analyze its discipline co-occurrence, publication characteristics, partnership, influence, keyword co-occurrence, co-citation, and structural variation. Based on the ISI Web database, this paper used a bibliometric method to analyze 2215 articles related to urban HIA published from 2012 to 2021. We found that the main research directions in the field were Environmental Sciences and Public Environmental Occupational Health; China contributed most articles, the Tehran University of Medical Sciences was the most influential institution, Science of the Total Environment was the most influential journal, Yousefi M was the most influential author. The main hotspots include health risk assessment, source appointment, contamination, exposure, particulate matter, heavy metals and urban soils in 2012–2021; road dust, source apposition, polycyclic aromatic hydrocarbons, air pollution, urban topsoil and the north China plain were always hot research topics in 2012–2021, drinking water and water quality became research topics of great concern in 2017–2021. There were 25 articles with strong transformation potential during 2020–2021, but most papers carried out research on the health risk assessment of toxic elements in soil and dust. Finally, we also discussed the limitations of this paper and the direction of bibliometric analysis of urban HIA in the future.


Introduction
The European Union Environmental Impact Assessment (EIA) Directive (2011/92/EU as amended by 2014/52/EU) requires EIAs to consider the effects that a project might have on human health [1]. Over the past two decades, health impact assessment (HIA) practice has expanded across the world [2], HIA has been regarded as an important means and tool for urban planning to promote public health and further promote the integration of health concept. More and more professionals in the fields of spatial planning, transportation planning and public health all over the world are beginning to pay attention to and use HIA tools [3][4][5][6][7]. Many cities around the world set "improving public health" as an important development goal to promote the construction of healthy cities by expanding their understanding and continuously carrying out healthy city planning practice, constructing an HIA theoretical model and carrying out evaluation practice [8][9][10][11][12].
Over the past decade, scholars have increasingly focused on urban HIA research and have issued an increasing number of papers; this makes it difficult for us to grasp the

Data Acquisition
According to the data resources required by CiteSpace, we took the Web of Science Core Collection as the data collection platform. As people are paying increasing attention to the impact of contaminated sites on the environment and public health, health risk assessment methods were used to describe and quantify the health impact on neighbouring people and guide public health interventions [16]. At the same time, we found that if only "health impact assessment" was used as the search term, the total number of literature studies was less than 400 under the same search conditions. Therefore, this paper juxtaposes "health risk assessment" as a search term. Figure 1 shows the number of papers published each year from 1992 to 2021 according to the retrieval strategy used in this paper. We found that the number of papers published in the early stages was small. CiteSpace software (Version 5.8.R3, Chaomei Chen, Drexel University, Philadelphia, PA, USA) developers pointed out that the longer the literature time span, the poorer the knowledge map. Therefore, this paper focuses on the bibliometric analysis of urban health impact assessment research over the last ten years.
Finally, the search strategy of bibliometric analysis was as follows: This study regards the Web of Science Core Collection as a data-collection platform according to data resources required in CiteSpace; The bibliometric search strategy can be described as the following: TS = (urban OR city OR cities) AND TS = ("health risk assessment" OR "health impact assessment"), Publication years = (1 January 2012 to 14 November 2021), Indexes = (SCI-EXPANDED, SSCI, ESCI), Document types = "Articles". A total of 2215 publications were selected on 14 November 2021. Finally, the search strategy of bibliometric analysis was as follows: This study regards the Web of Science Core Collection as a data-collection platform according to data resources required in CiteSpace; The bibliometric search strategy can be described as the following: TS = (urban OR city OR cities) AND TS = ("health risk assessment" OR "health impact assessment"), Publication years = (1 January 2012 to 14 November 2021), Indexes = (SCI-EXPANDED, SSCI, ESCI), Document types = "Articles". A total of 2215 publications were selected on 14 November 2021.

Bibliometric Analysis Methods
Bibliometrics is an interdisciplinary study that utilizes mathematics, statistics, and bibliography to quantitatively analyze academic literature [17]. The common methods of bibliometrics include statistical analysis, citation analysis, sharing analysis, etc. In addition, mapping knowledge can show the relationship between the development process and structure of scientific knowledge, focus on the evolution process of a certain knowledge field, and help scholars understand the hot spots, frontiers, and trends of research in this field [18].
Generally, the knowledge of the development status on a research subject is carried out using reviews as systematic or integrative approaches. However, the statistical analysis of journals, authors, countries, institutions and influence can help us quickly grasp the basic information and development status of literature in a certain field [19]. We used the Online Analysis Platform for Bibliometrics (https://bibliometric.com/ (accessed on 16 November 2021)) to conduct publication year, journal, countries and influence analysis, so as to quickly obtain the information of influential institutions, authoritative publications and experts in relevant fields.
Partnership analysis mainly analyzes the relationship between countries, the relationship between institutions and the relationship between authors. We used Online Analysis Platform for Bibliometrics and VOSviewer to analyze the cooperation network among countries, institutions, and authors.
Keyword co-occurrence analysis is an effective method, which can find hot topics and develop research frontiers in specific research fields. In this paper, we made a keyword co-occurrence network map by VOSviewer and monitored hot topics in urban HIA research through keyword co-occurrence network analysis.
Co-citation analysis can map the knowledge structure of a research field, detect the trends in the research field (by the authors engaged in these topics and their interrelationships) and highlight the findings with significant impact [20]. Understanding trends and

Bibliometric Analysis Methods
Bibliometrics is an interdisciplinary study that utilizes mathematics, statistics, and bibliography to quantitatively analyze academic literature [17]. The common methods of bibliometrics include statistical analysis, citation analysis, sharing analysis, etc. In addition, mapping knowledge can show the relationship between the development process and structure of scientific knowledge, focus on the evolution process of a certain knowledge field, and help scholars understand the hot spots, frontiers, and trends of research in this field [18].
Generally, the knowledge of the development status on a research subject is carried out using reviews as systematic or integrative approaches. However, the statistical analysis of journals, authors, countries, institutions and influence can help us quickly grasp the basic information and development status of literature in a certain field [19]. We used the Online Analysis Platform for Bibliometrics (https://bibliometric.com/ (accessed on 16 November 2021)) to conduct publication year, journal, countries and influence analysis, so as to quickly obtain the information of influential institutions, authoritative publications and experts in relevant fields.
Partnership analysis mainly analyzes the relationship between countries, the relationship between institutions and the relationship between authors. We used Online Analysis Platform for Bibliometrics and VOSviewer to analyze the cooperation network among countries, institutions, and authors.
Keyword co-occurrence analysis is an effective method, which can find hot topics and develop research frontiers in specific research fields. In this paper, we made a keyword co-occurrence network map by VOSviewer and monitored hot topics in urban HIA research through keyword co-occurrence network analysis.
Co-citation analysis can map the knowledge structure of a research field, detect the trends in the research field (by the authors engaged in these topics and their interrelationships) and highlight the findings with significant impact [20]. Understanding trends and emerging topics in a research area is important for future and current researchers, policymakers, funding agencies and other stakeholders [21]. CiteSpace enables us to understand a certain field quickly and systematically. CiteSpace can label co-citation clusters and use time-sliced snapshots to form timeliness and pivotal points [14]. In this study, we used CiteSpace (Version 5.8.R3) to find major research areas in the knowledge domain and journal overlay maps. In order to better understand the development of urban HIA research in different periods, we carried out co-citation analysis in two periods: (1) Set a time span from 2012 to 2016, set "years per slice = 1", "node types = reference", "Selection Criteria: Select top = 100" and "Pruning = Pathfinder and Pruning sliced networks" in CiteSpace to analyze their intellectual structure and the dynamics of co-citation clusters; (2) Set a time span from 2017 to 2021, set "years per slice = 1", "node types = reference", "Selection Criteria: g-index, k = 50" and "Pruning = Pathfinder and Pruning sliced networks" in CiteSpace to analyze their intellectual structure and the dynamics of co-citation clusters.

Discipline Co-Occurrence Analysis
Via the "dual-map overlay" function of CiteSpace, we can construct the discipline cooccurrence network, as shown in Figure 2. The original map shows more than 10,000 journals indexed in Web of Science, which are divided into different disciplines in different colors and are located in different locations of the source (or left) region and the reference (or right) region. Among them, "source" refers to the relevant papers in a certain field for a certain period of time and "reference" refers to all references of the above papers. For example, the discipline "ECONOMICS, ECONOMIC, POLITICAL" is in lake-blue and it ranks the 10th in the source region and the 12th in the reference region [22]. Then we added a layer containing the 2215 bibliographic records on urban HIA, which became the colorful links between the source region and the reference region.
span from 2017 to 2021, set "years per slice = 1", "node types = reference", "Selection Criteria: g-index, k = 50" and "Pruning = Pathfinder and Pruning sliced networks" in CiteSpace to analyze their intellectual structure and the dynamics of co-citation clusters.

Discipline Co-Occurrence Analysis
Via the "dual-map overlay" function of CiteSpace, we can construct the discipline cooccurrence network, as shown in Figure 2. The original map shows more than 10,000 journals indexed in Web of Science, which are divided into different disciplines in different colors and are located in different locations of the source (or left) region and the reference (or right) region. Among them, "source" refers to the relevant papers in a certain field for a certain period of time and "reference" refers to all references of the above papers. For example, the discipline "ECONOMICS, ECONOMIC, POLITICAL" is in lake-blue and it ranks the 10th in the source region and the 12th in the reference region [22]. Then we added a layer containing the 2215 bibliographic records on urban HIA, which became the colorful links between the source region and the reference region.
From Figure 2, the impact of the city and health impact assessment research was relatively widespread, and the source journals of the paper were mainly located in "7. VET-ERINARY, ANIMAL, SCIENCE". As for the distribution of the reference journals, "7. VETERINARY, ANIMAL, SCIENCE" links going to "2. ENVIRONMENTAL, TOXICOL-OGY, NUTRITION", "3. EARTH, GEOLOGY, GEOPHYSICS", "10. PLANT, ECOLOGY, ZOOLOGY", "8. MOLECULAR, BIOLOGY, GENETICS" and "5. HEALTH, NURSING, MEDICINE" account for higher percentage. This just reflects that the research on urban HIA is a typical interdisciplinary research field.  From Figure 2, the impact of the city and health "7. VETERINARY, ANIMAL, SCI-ENCE". As for the distribution of the reference journals, "7. VETERINARY, ANIMAL, SCI-ENCE" links going to "2. ENVIRONMENTAL, TOXICOLOGY, NUTRITION", "3. EARTH, GEOLOGY, GEOPHYSICS", "10. PLANT, ECOLOGY, ZOOLOGY", "8. MOLECULAR, BIOLOGY, GENETICS" and "5. HEALTH, NURSING, MEDICINE" account for higher percentage. This just reflects that the research on urban HIA is a typical interdisciplinary research field.

Publication Characteristics Analysis
The top 10 countries in the number of annual publications are shown in Figure 3. Since the Online Analysis Platform for Bibliometrics can only automatically generate the top ten countries, the situation of annual publications in other countries cannot be seen in this figure. The number of articles issued in China, Iran and India has shown a rapid growth trend from 2012 to 2021.
As shown in Table 1

Publication Characteristics Analysis
The top 10 countries in the number of annual publications are shown in Figure 3. Since the Online Analysis Platform for Bibliometrics can only automatically generate the top ten countries, the situation of annual publications in other countries cannot be seen in this figure. The number of articles issued in China, Iran and India has shown a rapid growth trend from 2012 to 2021. As shown in Table 1

Partnership Analysis
The cooperative relationship between countries is shown in Figure 4. Among them, authors from CHINA, IRAN, USA, and SPAIN have more cooperation with authors from other countries.

Partnership Analysis
The cooperative relationship between countries is shown in Figure 4. Among them, authors from CHINA, IRAN, USA, and SPAIN have more cooperation with authors from other countries.
The cooperation between institutions is shown in Figure 5. Institutions include universities, research institutes and national research institutions; According to the statistics of the number of articles jointly published by institutions, the organization represented by the red node has a cooperative relationship with the node where the mouse is located. The core of institutional cooperation network mainly includes Chinese Academy of Sciences, Tehran University of Medical Sciences, Beijing Normal University, Indian Institute of Technology, etc.
The cooperation between authors is shown in Figure 6. In the main author cooperation network, the authors who play a key role include Mohammadi M J, Radfard M, Rojas-rueda D, Lu XW, Liu G, Wang Q, Li J, etc. The cooperation between institutions is shown in Figure 5. Institutions include universities, research institutes and national research institutions; According to the statistics of the number of articles jointly published by institutions, the organization represented by the red node has a cooperative relationship with the node where the mouse is located. The core of institutional cooperation network mainly includes Chinese Academy of Sciences, Tehran University of Medical Sciences, Beijing Normal University, Indian Institute of Technology, etc.   The cooperation between institutions is shown in Figure 5. Institutions include universities, research institutes and national research institutions; According to the statistics of the number of articles jointly published by institutions, the organization represented by the red node has a cooperative relationship with the node where the mouse is located. The core of institutional cooperation network mainly includes Chinese Academy of Sciences, Tehran University of Medical Sciences, Beijing Normal University, Indian Institute of Technology, etc. The cooperation between authors is shown in Figure 6. In the main author cooperation network, the authors who play a key role include Mohammadi M J, Radfard M, Rojasrueda D, Lu XW, Liu G, Wang Q, Li J, etc.

Influence Analysis
The top 10 institutions with influence are shown in Table 2. Univ Tehran Med Sci, Beijing Normal Univ and Chinese Acad Sci ranked the top three in the total number of references.

Influence Analysis
The top 10 institutions with influence are shown in Table 2. Univ Tehran Med Sci, Beijing Normal Univ and Chinese Acad Sci ranked the top three in the total number of references. The top 10 publication titles with influence are shown in Table 3. "Science of the Total Environment", "Ecotoxicology and Environmental Safety", and "Environmental Science and Pollution Research" ranked the top three in the total number of references. The top 10 authors with influence are shown in Table 4. Yousefi M, Mahvi AH and Lu XW ranked the top three in the total number of references.

Co-Citation Analysis
The co-citation map of references estimated the scientific relevance of publications [23]. Co-citation cluster analysis was used to explore the research patterns, new trends and their relationships in urban HIA research. Timeline visualization in CiteSpace described clusters through a horizontal timeline. Each cluster was displayed in left-to-right order. The largest cluster was shown at the top of the view. Large nodes or nodes with a red tree structure are particularly interesting because they are either highly referenced, have referenced bursts, or both. Below each timeline, the top three references cited in a particular year are displayed [14].

Timeline View (2012-2016)
Based on the literature records from 2012 to 2016, we generated a cited reference map with 458 nodes and 984 links ( Figure 8). The results show, the mean silhouette (S = 0.9113) and the modularity (Q = 0.7918), that the modular Q value was greater than 0.7, which indicated that it was reasonable for the network to be divided into loosely coupled clusters.

Co-Citation Analysis
The co-citation map of references estimated the scientific relevance of publications [23]. Co-citation cluster analysis was used to explore the research patterns, new trends and their relationships in urban HIA research. Timeline visualization in CiteSpace described clusters through a horizontal timeline. Each cluster was displayed in left-to-right order. The largest cluster was shown at the top of the view. Large nodes or nodes with a red tree structure are particularly interesting because they are either highly referenced, have referenced bursts, or both. Below each timeline, the top three references cited in a particular year are displayed [14].

Timeline View (2012-2016)
Based on the literature records from 2012 to 2016, we generated a cited reference map with 458 nodes and 984 links (Figure 8). The results show, the mean silhouette (S = 0.9113) and the modularity (Q = 0.7918), that the modular Q value was greater than 0.7, which indicated that it was reasonable for the network to be divided into loosely coupled clusters. The five salient clusters were sorted by size ( Table 5). The literature with high cocitation frequency became a representative publication in the cluster, which affected the annotation of the cluster and revealed the research frontier.
Cluster #0 was the largest cluster and represented "polycyclic aromatic hydrocarbons" with 64 members. In the representative publications, Yu et al. (2014) identified the potential sources of polycyclic aromatic hydrocarbons (PAHs) in 87 urban street dust samples from Tianjin as a Chinese megacity and evaluated the risk of PAHs to urban residents using the incremental lifetime cancer risk model [24].
Cluster #1 represented "street dust" with 58 members. In the representative publications, Lu et al. (2014) applied the potential ecological risk index to assess the risk of heavy metals in street dust of cities in China on urban ecosystems and applied the human exposure model to assess the risk of heavy metals to human health. The authors emphasized that further research on street dust exposure parameters and transportation factors was needed to reduce the uncertainty related to risk calculation [25].
Cluster #2 represented "potential health risk" with 45 members. In the representative publications, Fu et al. (2015) estimated the rice's potential health risk to inhabitants in Fuzhou, China by target hazard quotient (THQ), hazard index (HI) and target cancer risk (TR) [26].
Cluster #3 represented "air pollution" with 42 members. In the representative publications, Lai et al. (2013) systematically evaluated and analyzed the mortality and incidence rate of four typical air pollutants (PM10, NO2, SO2, and O3) in the Chinese population. The authors emphasized that the short-term impact of air pollution could be used for health impact assessment, but the evidence of long-term impact was still insufficient [27].
Cluster #4 represented "physical activity" with 40 members. In the representative publications, Gerike et al. (2016) investigated the determinants of active mobility and the evaluation of measures to increase it through a large-scale longitudinal survey conducted in seven Physical Activity through Sustainable Transport Approaches (PASTA) case study cities. The results would provide data on the health benefits of cycling and/or walking to the WHO's online health economic assessment tool [28]. The five salient clusters were sorted by size ( Table 5). The literature with high cocitation frequency became a representative publication in the cluster, which affected the annotation of the cluster and revealed the research frontier.
Cluster #0 was the largest cluster and represented "polycyclic aromatic hydrocarbons" with 64 members. In the representative publications, Yu et al. (2014) identified the potential sources of polycyclic aromatic hydrocarbons (PAHs) in 87 urban street dust samples from Tianjin as a Chinese megacity and evaluated the risk of PAHs to urban residents using the incremental lifetime cancer risk model [24].
Cluster #1 represented "street dust" with 58 members. In the representative publications, Lu et al. (2014) applied the potential ecological risk index to assess the risk of heavy metals in street dust of cities in China on urban ecosystems and applied the human exposure model to assess the risk of heavy metals to human health. The authors emphasized that further research on street dust exposure parameters and transportation factors was needed to reduce the uncertainty related to risk calculation [25].
Cluster #2 represented "potential health risk" with 45 members. In the representative publications, Fu et al. (2015) estimated the rice's potential health risk to inhabitants in Fuzhou, China by target hazard quotient (THQ), hazard index (HI) and target cancer risk (TR) [26].
Cluster #3 represented "air pollution" with 42 members. In the representative publications, Lai et al. (2013) systematically evaluated and analyzed the mortality and incidence rate of four typical air pollutants (PM 10 , NO 2 , SO 2, and O 3 ) in the Chinese population. The authors emphasized that the short-term impact of air pollution could be used for health impact assessment, but the evidence of long-term impact was still insufficient [27].
Cluster #4 represented "physical activity" with 40 members. In the representative publications, Gerike et al. (2016) investigated the determinants of active mobility and the evaluation of measures to increase it through a large-scale longitudinal survey conducted in seven Physical Activity through Sustainable Transport Approaches (PASTA) case study cities. The results would provide data on the health benefits of cycling and/or walking to the WHO's online health economic assessment tool [28].   Based on the literature records from 2017 to 2021, we generated a cited reference map with 888 nodes and 1470 links (Figure 9). The results show that with the mean silhouette (S = 0.9258) and the modularity (Q = 0.7937), the modular Q value was greater than 0.7, which indicated that it was reasonable for the network to be divided into loosely coupled clusters. Based on the literature records from 2017 to 2021, we generated a cited reference map with 888 nodes and 1470 links (Figure 9). The results show that with the mean silhouette (S = 0.9258) and the modularity (Q = 0.7937), the modular Q value was greater than 0.7, which indicated that it was reasonable for the network to be divided into loosely coupled clusters. The nine salient clusters were sorted by size ( Table 7). The literature with high cocitation frequency became a representative publication in the cluster, which affected the annotation of the cluster and revealed the research frontier.
Cluster #0 represented "road dust" with 121 members. In the representative publications, Moryani et al. (2020) evaluated the pollution level of heavy metals (Cu, Pb, Zn, CD, Ni, Sb, Cr) in five parts of road dust in four different regions of Karachi and Shikarpur by calculating the pollution index enrichment factor and Igeo; the health risk assessment was carried out according to the carcinogenic risk methods and hazard index [47].
Cluster #1 represented "source apportionment" with 119 members. In the representative publications, Cai et al. (2019) studied the levels, distribution, and source apportionment of metals in soils from a typical rapidly developing county, Southern China [48].
Cluster #2 represented "drinking water" with 68 members. In the representative publications, Hamed et al. (2018) assessed the nitrate concentration and also the microbial quality of bottled water in a number of brands produced in the Torbat-e Heydarieh city in 2017 [49].
Cluster #3 represented "chemical fractionation" with 63 members. In the representative publications, Sah et al. (2019) identified the migration potential of metals (As, Cd, Co, The nine salient clusters were sorted by size ( Table 7). The literature with high cocitation frequency became a representative publication in the cluster, which affected the annotation of the cluster and revealed the research frontier.
Cluster #0 represented "road dust" with 121 members. In the representative publications, Moryani et al. (2020) evaluated the pollution level of heavy metals (Cu, Pb, Zn, CD, Ni, Sb, Cr) in five parts of road dust in four different regions of Karachi and Shikarpur by calculating the pollution index enrichment factor and Igeo; the health risk assessment was carried out according to the carcinogenic risk methods and hazard index [47].
Cluster #1 represented "source apportionment" with 119 members. In the representative publications, Cai et al. (2019) studied the levels, distribution, and source apportionment of metals in soils from a typical rapidly developing county, Southern China [48].
Cluster #2 represented "drinking water" with 68 members. In the representative publications, Hamed et al. (2018) assessed the nitrate concentration and also the microbial quality of bottled water in a number of brands produced in the Torbat-e Heydarieh city in 2017 [49].
Cluster #3 represented "chemical fractionation" with 63 members. In the representative publications, Sah et al. (2019) identified the migration potential of metals (As, Cd, Co, Cr, Ni, and Pb) in urban fine particulate matter and assessed the health risks of these metals to infants, young children, children, males, and females. The research conclusions showed that urban aerosols have potential risks to humans [50].
Cluster #4 represented "volatile organic compounds" with 61 members. In the representative publications, Gu et al. (2020) analyzed the chemical characteristics and sources of volatile organic compounds (VOCs) in Tianjin, China, using 1-h resolution VOC-species data between 1 November 2018 and 15 March 2019 [51].
Cluster #5 represented "air pollution" with 52 members. In the representative publications, Lehtomaki et al. (2020) quantified the number of deaths caused by ambient air pollution in Nordic countries using selected assessment tools, identified the main differences and stressed that high spatial resolution should be used to avoid underestimating the health effects of ambient air pollution [52].
Cluster #6 represented "polycyclic aromatic hydrocarbons" with 46 members. In the representative publications, Najmeddin and Keshavarzi (2019) used toxic equivalency factors and increased lifetime cancer risk to assess the health risk of PAHs in PM 10 and road dust samples [53].
Cluster #7 represented "north China plain" with 45 members. In the representative publications, Zhang et al. (2019) evaluated the potential health risks of PAHs (including gas and particle phases) by combining methods of benzo[a]pyrene equivalent concentration and incremental lifetime cancer risk and identified the potential source regions of PM 2.5 -bound PAHs in Jinan by the Concentration Weighted Trajectory model [54].
Cluster #8 represented "respiratory diseases" with 37 members. In the representative publications, Geravandi et al. (2017) studied the relationship between the number of hospitalized respiratory diseases (including asthma attacks, acute bronchitis, and chronic obstructive pulmonary disease) caused by PM 10 and normal/dust event days in Ahvaz, Iran, from 2010 to 2012 [55]. Table 7. Summary of the largest nine clusters in 2017-2021.   The main limitation of citation-based indicators is that they may ignore newly published articles. SVA is a method to focus on the impact of newly published papers on the conceptual structure of related knowledge fields [101]. The SVA program looks for new connections that may change the global structure [102]. The purpose of applying SVA is to evaluate the potential of an article to establish abnormal or unexpected connections between different clusters. From the perspective of scientific discovery theory, many significant contributions come from the idea of crossing borders [14]. To assess the recent papers' transformative potentials, we used the SVA of CiteSpace-we used 3-year span sliding windows. Table 9 shows a list of articles with a high transformative potential based on the ∆Modularity and ∆Cluster Linkage. Table 9. Some of the articles with the strongest transformative potentials, M is ∆Modularity, C-L is for ∆Cluster Linkage, C-D is for ∆Centrality Divergence.

Trajectories of Citations across Cluster Boundaries
There were six examples of articles with high modularity change rates, as shown in Figure 10. The visualization reveals the distribution of the references cited by these articles across different clusters.
(1) Zhong et al. (2020) analyzed two particle size distributions of heavy metals in street dust from an industrial city, explored their possible sources and assessed their health risks [103]. The papers cited in this paper are mainly distributed in clusters #0, #1, #9, #10, and #12.
(2) Xia et al. (2020) explored the source of six potentially toxic elements (PTEs) in topsoil of three different land use types (residential land, industrial land, and farmland) in Tonghua City and evaluated the ecological risk and human health risk of PTEs in different types of soil [104]. The papers cited in this paper are mainly distributed in clusters #0, #1, #10, and #12.
(3) Shahab et al. (2020) evaluated the pollution level and health risk from heavy metals in road dust collected in three functional areas in the tourist city Guilin using the geoaccumulation index (Igeo), ecological risk index, spatial interpolation, and array-based risk assessment model [64]. The papers cited in this paper are mainly distributed in clusters #0, #1, #12 and #15.
(4) Ahamad et al. (2021) identified the sources of potentially toxic elements (As, Cd, Cr, Cu, Fe, Hg, Mn, Ni, Pb, Zn) of 48 samples of soil and road dust from industrial clusters in the Sonbhadra region of Uttar Pradesh (India) and assessed their human health risks [58]. The papers cited in this paper are mainly distributed in clusters #0, #1, #9 and #14.
(5) Wang et al. (2021) evaluated the pollution characteristics and spatial distribution characteristics of 12 PTEs (Mn, Ni, Cu, Zn, Hg, Cd, As, Cr, Pb, Tl, Co, and Sb) in a large copper smelter in Central China and evaluated the potential ecological and health risks of PTEs by combining with positive matrix decomposition [105]. The papers cited in this paper are mainly distributed in clusters #0, #1, #2, #6, #9, and #12.

Discussions
The limitations of this study are shown as follows: First, the data only comes from English papers in the WoS Core Collection, resulting in the exclusion of a large number of English documents in Scopus, PubMed/Medline and other databases, as well as a large number of documents in other language countries [106]. It is necessary to conduct bibliometric analysis of urban HIA related literature in other databases and urban HIA related literature in other languages [107], so as to grasp the global research progress more comprehensively on urban HIA.
Second, as urban HIA is a global topic, the situation of each country is different, so the research content may also be different. In the future, we can select country samples for comparative research, so as to understand the research progress of various countries in the field of urban HIA.
Finally, due to the limitation of the software itself, in the co-citation analysis, some important research topics in the field of urban HIA might have been omitted, for example, the urban HIA practice issues [1,2,[108][109][110] and HIA in urban planning [111][112][113]. These important research topics should be employed in a bibliometric analysis in the future.

Conclusions
This paper used the bibliometric method to analyze the trends, issues and future directions of urban HIA research. We advance the following main conclusions: First, the main research directions in the field were Environmental Sciences and Public Environmental Occupational Health.
Second, China contributed most articles; the Univ Tehran Med Sci was the most influential institution; Science of the Total Environment was the most influential journal; Yousefi M was the most influential author.
Third, the main hotspots included health risk assessment, source appointment, contamination, exposure, particulate matter, heavy metals and urban soils in 2012-2021.
Fourth, the road dust, source apposition, PAHs, air pollution, urban topsoil and north China plain were always hot research topics in 2012-2021. Drinking water and water quality became research topics of great concern in 2017-2021.
Finally, there were 25 articles with strong transformation potential during 2020-2021, but most papers carried out research on health risk assessment of toxic elements in soil and dust.