A New Model for Ranking Schools of Public Health: The Public Health Academic Ranking

Objectives: As there is no ranking designed for schools of Public Health, the aim of this project was to create one. Methods: To design the Public Health Academic Ranking (PHAR), we used the InCites Benchmarking and Analytics™ software and the Web Of Science™ Core Collection database. We collected bibliometric data on 26 schools of Public Health from each continent, between August and September 2022. We included 11 research indicators/scores, covering four criteria (productivity, quality, accessibility for readers, international collaboration), for the period 2017–2021. For the Swiss School of Public Health (SSPH+), a network gathering faculties across different universities, a specific methodology was used, with member-specific research queries. Results: The five top schools of the PHAR were: London School of Hygiene and Tropical Medicine, Public Health Foundation of India, Harvard T.H. Chan School of Public Health, SSPH+, Johns Hopkins Bloomberg School of Public Health. Conclusion: The PHAR allows worldwide bibliometric ordering of schools of Public Health. As this is a pilot project, the results must be taken with caution. This article aims to critically discuss its methodology and future improvements.


INTRODUCTION
International university ranking systems were originally created to compare the performance of universities, but most have been criticized for not focusing enough on what is relevant to societies.Many rankings have been introduced since the creation of the Shanghai Ranking in 2003 [1], based on indicators from various fields such as research, education or teaching.They have several inherent well-described shortcomings [2,3], and have often been interested in ranking universities as an entity, without adapting their indicators to the different disciplines, including Public Health [4].Numerous shortcomings of current rankings have been highlighted in the literature, such as the choice, weighting and lack of inclusion of certain indicators, the issue of reputation surveys, the lack of inclusion of minorities and low-income countries, the bias in favor of English-based universities and the lack of transparency in methodology [3,5,6].As stated by Vernon et al. (2018), who conducted a systematic review on ranking systems: "There is a need for a credible quality improvement movement in research that develops new measures, and is useful for institutions to evaluate and improve performance and societal value" [7].
According to the CDC Foundation, "Public health professionals try to prevent problems from happening or recurring through implementing educational programs, recommending policies, administering services and conducting research (. ..)" [8].Although this definition is not exhaustive and Public Health includes several other areas (such as surveillance and outbreak investigation), it highlights the broad aspects covered by Public Health, the specificity of such a discipline, the societal impact it aims to achieve, as well as the particularity of research in this field (including community-based and observational studies).Furthermore, in the wake of the SARS-CoV-2 pandemic, it is widely recognized that more robust public health efforts are needed [9].Strong Public Health requires not only excellent educational programs, but also a state-of-the-art research.As stated by Odone et al. (2017): "gaps exist between current public health needs and the extent to which public health workers are trained" [10].In this context, a valid ranking is required to help all stakeholders improving this discipline, as a majority of university leaders use rankings to monitor the performance of their own institutions and to highlight their strengths in order to attract researchers and students from around the world [11].
Therefore, given the importance of Public Health, the common use of rankings in the academic world, and the demand for a valid ranking system, we designed and pilot tested a ranking for schools of Public Health.As an initial step, we did not seek to evaluate the fields of education and teaching, but developed a ranking based purely on bibliometric indicators evaluating the research field of Public Health.These indicators are only one means of evaluating a university and it is important to emphasize that the overall evaluation of a school must include many other parameters, which will be developed in later phases of this project.

Sources and Database
Although this definition cannot claim to be exhaustive regarding the broad scope of Public Health, "core disciplines include epidemiology, environmental health sciences, health policy and management, biostatistics, and social and behavioral aspects of health" [12].In this context, we used the InCites Benchmarking and Analytics ™ software (from Clarivate ™ -named "software" in this paper) to collect data on schools of Public Health.The software is based on the publications metadata of the Web Of Science ™ Core Collection database (herein named "database") [13], that covers the main disciplines of Public Health mentioned above and many others [14].This software only lists a small number of schools of Public Health.Other schools not listed in the software were selected from the database.

Temporal Criteria for Indicators and Study Period
In order to obtain a meaningful ranking that is not subject to year-toyear variation, we used data from 2017 to 2021.We decided to select a 5 years time period as some of our criteria (particularly those concerned with citation numbers) require time to become relevant.The raw data needed for the construction of the ranking were collected between 1 August and 30 September 2022.

Indicators and Scores Constituting the Ranking
This ranking is based on 11 bibliometric indicators and scores, from four different criteria (productivity, quality, accessibility for readers, international collaboration), which have been chosen because they are relevant for the discipline of Public Health.The term "indicators" refers to the raw data that have been collected, whereas the term "scores" refers to the normalized values of the raw data.While it seems obvious to include "productivity" and "quality" indicators/scores, Public Health being of high policy relevance, it also seemed essential to include indicators/scores related to the "accessibility for readers."Finally, in areas such as pandemic management or migration policy for instance, Public Health is a discipline with a strong international focus, which is why the "international collaboration" indicator was significant.The indicators are described in Table 1.

Formulas Used in the Ranking
The formulas used for the ranking consisted of normalizing the indicators according to a "base 100" process (i.e., attribution (or assessment) of a score 100 to the school with the highest value of the indicator and assessment of the percentage of this score to the other schools, depending on the indicator values they had).The overall scores were then computed by summing up scores for the indicators in two different ways.The first formula consists of summing up all scores without any weighting.The second formula consists of the sum of six scores, with a weighting giving the same weight to each criterion.The formulas are detailed in Table 2.
The weighting given to each criterion of the two formulas is presented in Table 3. Formula 1 gives a significant weight to the "quality" criterion, while Formula 2 gives an equal weighting to all four criteria.Since all four criteria are of equal importance in Public Health, we selected Formula 2 to create the PHAR.
The raw data and the formulas for the ranking were downloaded and executed in M.S. Excel (version 2016).

Study Population: Schools of Public Health Included in the Ranking
We collected data from 26 schools of Public Health worldwide.In this pilot project, our sub-sample of selected schools was based on the following criteria: All the 12 schools of Public Health that were listed with specific dedicated affiliation in the software were selected.This search was conducted on the basis of keywords (in English and French) related to the discipline of Public Health.These keywords were: "Public Health," "Santé Publique," "Global Health," "Tropical Medicine" (these last two keywords gave no results when translated in French).We added the Swiss School of Public Health (SSPH+), with the specific approach described below.We expanded the scope of the ranking, by adding 13 schools of Public Health from different geographic locations, not listed in the software, with the aim of including at least one school from each continent.It should be noted that none of the schools initially included in our sample were subsequently excluded from the study.
Since this is a pilot project, the choice of schools was necessarily limited, but will be completed in the future stages of the project.The list of schools included in the ranking is detailed in Table 4.The use of this method allowed the collection of data for all indicators, apart from the indicators 10 and 11.For those two indicators, we did an estimation, by using the "% International Collaboration" indicator that is attributed to the affiliated University listed in the software (we used the "% International Collaboration" attributed to Columbia University and applied it to the Mailman School of Public Health).All research queries are available in the Supplementary Appendix S1.
The SSPH+ is a unique case as it is, to our knowledge, the only "virtual campus" worldwide.Some of the other ranked schools gather different universities, but have a formal affiliation, such as the Colorado School of Public Health, so the situation is different.Given that none of the Swiss universities run schools of Public Health, the SSPH+ was created in 2005, building a virtual national faculty including all public health oriented scientists who belong to any faculty of one of the 12 universities of the SSPH+ (at the time of the ranking's creation, SSPH+ included 12 universities).Given that the SSPH+ does not employ its virtual faculty members, most scientists do not use the SSPH+ as an affiliation.Thus, in contrast to all other schools of Public Health, the output of the SSPH+ faculty cannot be identified with the term "SSPH+" Instead, the record of SSPH+ corresponds to the scientific output of the SSPH+ faculty members and their research groups.Thus, a different methodology was needed to derive the ranking of the SSPH+, explained by the use of the database and the research queries.We also asked several representatives of included universities to check the research queries used in the database, to avoid errors inherent to the research methodology.Furthermore, in order to avoid mistakenly counting duplicates in the data collected in the database (although this should rarely be the case having used the conjunction word "OR" to link the different queries for the universities within the SSPH+), we manually sampled the data (i.e., we selected the ten first articles published in the database from each of the 12 universities of the SSPH+) and checked the percentage of duplicates between publications from each SSPH+ university.As this percentage turns out to be minimal (1.67%), we did not make any correction to the collected data.Finally, we conducted a preliminary validation for one university using a subset of bibliometric outputs selected by the formula.

RESULTS
The results of the PHAR are detailed in Table 5.The range of scores awarded to the included schools extends from 128 (26thranked school) to 288 (1st-ranked school).For comparison purposes, we use the "score/unit" to compare scores between Formulas 1 and 2. The ranking according to Formula 1, the changes of rank between Formulas 1 and 2, and the detailed scores of each school are available in the Supplementary Appendices S2-S4.
The The number of ranks that each school loses or gains by moving from Formula 1 to 2 is detailed in the Supplementary Appendix S3.The extremes of the ranking are the least subject to change (top five and bottom five ranks), while ranks six to 22 are the most subject to instability.The school with the greatest change in rank is the Mahidol Oxford Tropical Medicine Research Unit (MORU, +12 ranks), while six schools show only one change of ranks (positive or negative).The change of rank for the MORU can be understood as follows: it scores very well in the criteria "accessibility for readers" and "international collaboration" (weighting improving respectively of 7% between Formulas 1 and 2) whereas it scores less well for the "quality" criterion (weighting decreasing of 30% between Formulas 1 and 2).

DISCUSSION
In January 2023, Harvard Medical School withdrew from the U.S. News and World Report Ranking despite being ranked at the top of it, arguing that "Rankings cannot meaningfully reflect the high aspirations for educational excellence (. ..)" [15].This movement of withdrawal highlights the shortcomings of existing rankings, as well as the need to review their methodology to create valid and useful rankings, to meet the demand of many stakeholders who use them as a decision-making tool [16,17].In addition, several initiatives have emerged, including the Declaration on Research Assessment (DORA) a decade ago, and recently the Coalition for Advancing Research Assessment (COARA), highlighting the need for change in research assessment and the move towards giving tribute to the impact of university-based research [18,19].As it is well known that universities rely in part on rankings to obtain funding, it seems clear that rankings can potentially have considerable implications for public health policies.A wellranked school will find it easier to obtain funding, which in turn will generate a greater potential pool of Public Health experts at national level, and thus a definite impact on national public health policies.It seems therefore illusory to simply withdraw from a ranking; on the contrary, we need to collaborate and develop a valid, transparent and robust ranking that will enable us to promote this discipline, fundamentally important for our societies.
To our knowledge, the PHAR is the first international bibliometric university ranking system designed for academic Public Health.This pilot project is based on 11 bibliometric indicators and scores and ranks 26 schools of Public Health worldwide.Two different formulas were used for this pilot project, but only Formula 2 was finally selected for the PHAR, because of the same weight that this formula attributes to each criterion.
There are two main highlights in this ranking.First, although two of the top five schools were from the United States, the top ten schools included institutions from four continents.The presence of schools from India, Thailand, South Africa and Singapore is notable and may reflect the increasing importance of Public Health research in non-high income countries.The criteria of international collaboration may have allowed the identification of these schools as international leaders.Second, we can highlight two unexpected appearances in the top five ranks: the SSPH+ and the PHFI.The special case of the SSPH+ is detailed below.Concerning the PHFI, although this institute also gathers several schools, as it benefits from an affiliation listed in the software (implying that the authors sign with this affiliation), it was decided not to apply the same methodology as for the SSPH+.In fact, the PHFI corresponds to the situation of the Colorado School of Public Health.In spite of this difference, we notice that the PHFI ranks high.It is interesting to note that this school ranks exceptionally well not because of the "productivity" criterion (as might have been felt in view of the network bringing together numerous schools in the same country) but thanks to the "quality" criterion, particularly in view of the scores 3, 5 and 7, which are all at the maximum.As far as the SSPH+ is concerned, its high ranking is explained by the number of publications (underlining the importance of considering this network of highly productive schools) and also by the number of "open access" publications.As already mentioned, the PHAR helps to highlight these types of particular schools, which are often underestimated by traditional rankings.

Strengths
Compared to the 2021 Global Ranking of Academic Subjects for Public Health [20], the top five ranks of the PHAR also contain the top three schools of the previously mentioned ranking.However, the SSPH+ is not classified in the Global Ranking of Academic Subjects for Public Health.Indeed, the specific issue about the SSPH+ is that this virtual school is underestimated internationally, because it is not taken in account in the existing rankings.Despite inherent structural differences of such "virtual faculty," SSPH+ acts in many regards like a "real" school of Public Health, which is demonstrated by the extensive scientific collaboration in research (e.g., the "Corona Immunitas" study including more than 40 national studies on the immunity against SARS-CoV-2 in Switzerland [21]), the actions to improve Public Health education (e.g., the Inter-university Graduate Campus for PhD students [22]), or the strong professional network developed between these universities [23][24][25].Thus, the SSPH+ provides a national network unifying this multidisciplinary field across its current universities.Therefore, one of the main strength of the PHAR is to offer visibility to this type of unique and decentralized school, which was previously unaccounted by the traditional rankings.We perceive some strong arguments in favor of the PHAR.First, the PHAR provides full transparency about the methodology used, allowing a reproducibility by any external stakeholders interested to replicate the ranking, as recommended by the Berlin Principles [26], and unlike some of the existing rankings whose methodology is sometimes described as a "black box approach" [27].
Second, the PHAR is an objective and robust ranking, based solely on bibliometric criteria.Although bibliometric indicators are subject to criticism, they have the advantage of being easily comparable and understandable, and to have a certain degree of objectivity [28][29][30].
Third, the PHAR is constructed on a well-recognized and reliable source of data [14].Indeed, the software and database that were used in this ranking have a long experience in the field of scientometrics.
Finally, the PHAR aims to fill a gap in the ranking domain.As described by a previous work, to date, there does not exist any specific ranking designed for schools of Public Health or for "virtual schools" [4], therefore this project is innovative.

Limitations
The PHAR, while innovative, does have its limitations.
First, this ranking is based purely on bibliometric criteria, focusing on the field of research.This choice was made in view of the importance of research in Public Health, as highlighted by TABLE 5 | The Public Health Academic Ranking (i.e., according to Formula 2).To obtain the score/unit, Formula 2 score was divided by 4 (Formula 2 includes 6 scores with a respective weighting of 1, 3 × 1/3, 1 and 1) (Geneva, Switzerland, 2022).

Rank
School name/Country Formula 2 score Score/unit McLaren et al. ( 2019) [31].Nevertheless, we plan to include other indicators in later stages of the project (based for instance on the societal impact of schools), as proposed by Holmes (2021) regarding "third missions" indicators [6].
Second, the choice of the indicators can always be viewed as arbitrary, as highlighted regarding existing ranking shortcomings [2,3].Although these indicators already exist since longtime in the bibliometric field, we can still question their relevance: for instance, is publishing in Open Access journals a critical element of making papers "policy-relevant"?These questions require further research.Furthermore, the date of establishment of the ranked schools should always be taken into account, as the results of recently established schools may not be as pertinent as those of older schools.
A third limitation is that the queries used for the 14 schools for which data came from the database could include a potential source of error, and the definition of the term "school" may vary from a country to another, so these points could represent a bias.Nevertheless, we minimized this risk by asking some of the representatives of the ranked schools to do an external control of the research queries.In addition, although the use of these research queries may represent a source of error, we believe that this methodology is also one of the strengths of the ranking, as it allows the inclusion of schools usually overlooked by traditional rankings.
A fourth limitation of this pilot project is the need to restrict it to a manageable and rather small selection of schools of Public Health from all continents.Thus many schools are left out, including prestigious ones.Once algorithms of the PHAR are fully settled, the assessment may be easily extended.It is therefore important to emphasize that the accrediting bodies for schools of Public Health (e.g., APHEA in Europe), which contribute significantly to the evaluation (and accreditation) of certain schools, should not at present take the provisional ranking into account, as this could compromise the evaluation of the schools not included.
A fifth limitation of this ranking is that its scores do not take into account the size of the school.Thus, it is true that a large school will be better off in the number of publications by example than a smaller school.This certainly represents a future point of improvement.
Finally, the choice of the database may represent a bias, because Web of Science focuses mainly on publications written in English, and this represents a commonly criticized shortcoming regarding rankings [32].Nevertheless, this database has the advantage of compiling numerous bibliometric indicators that allow for easy analysis of publications.

Future Improvements
Future improvements include the integration of other evaluation domains (e.g., education, health economics), of others bibliometric indicators as well as non-bibliometric indicators, and of a non-English language database.Furthermore, this pilot project is limited in the number of included schools, but the aim is to progressively rank more schools worldwide and to classify schools according to the type of structure they present.As pointed out by de Leeuw (1995), there are at least eight different types of school of Public Health in Europe, and it would be interesting to be able to adapt the ranking to each of these types in order to bring greater refinement to the future ranking [33].Some indicators specific to Public Health should also be developed, to make the ranking more specific to this discipline and to develop an evaluation of the impact of a school, possibly based on a participative and/or qualitative process.In conjunction with the previous step, a content analysis could also be designed to profile the different strengths of the schools.
Regarding the special case of the SSPH+, and although to our knowledge there is no official threshold for sensitivity and specificity in the bibliometric field, a sensitivity and specificity assessment was done on the articles selected by the research queries for the year 2021 of one member university.For the sensitivity analyses, five SSPH+ faculty members were chosen to compare their publication record, as listed on the institutional database, with the publications chosen by the algorithm.The queries had a sensitivity of 87% (13% of the publications actually made by those five SSPH+ affiliated members were missing).With a total of 33% articles included from authors of this university not formally affiliated with SSPH+, the specificity was 67%.In order to evaluate the impact of this specificity analysis, analyses for the SSPH+ ranking were recalculated by applying a "worst case" correction factor of 33% to the three indicators relating to publication volumes (indicators 1, 6, and 8).With this correction factor, the SSPH+'s rank has moved from the fourth to the sixth rank, highlighting that the SSPH+'s rank did not change significantly despite this correction (Supplementary Appendix S5).Moreover, algorithms with a higher sensitivity have the opposite effect on the ranking.Nevertheless, this point raises the question of a future methodological improvement regarding the research queries for SSPH+, which may need to be based on the names of the faculty members rather than the institutions and research fields alone.
Finally, an improvement could be made regarding the "international collaboration" indicator and the fact that 14 schools of the PHAR use an estimate of this indicator based on the university to which they are affiliated.Regarding this point, a second sensitivity analysis was conducted, excluding this indicator in order to analyze the change of ranks without including it.In this analysis, even if the ranks are changing for some schools, four of the top five ranks are still occupied by the same schools, highlighting a certain stability even when this indicator was excluded (Supplementary Appendix S6).

Conclusion
The PHAR is the first international university bibliometric ranking designed for schools of Public Health, with a pilot project aiming to evaluate and focus on the research field of this discipline.As a result of initiatives such as COARA and DORA, research evaluation is being challenged and requires a change in the usual ways of thinking about this issue, integrating the impact of schools of Public Health on societies.However, and to provide an initial ranking of schools of Public Health, the main purpose of this article is to present and discuss the methodology used to create a valid ranking system, to underline its actual limitations, and to open the discussion for future improvements, as highlighted by Wilbers and Brankovic (2021) who stated "rankings are here to change and their status challenged" [34].
top five schools of the PHAR are: London School of Hygiene and Tropical Medicine (LSHTM), Public Health Foundation of India (PHFI), Harvard T.H. Chan School of Public Health, SSPH+, Johns Hopkins Bloomberg School of Public Health.The top five ranks are occupied by the same schools using Formulas 1 or 2, highlighting a stability between the two rankings even if the rank within the top five places changes.

TABLE 1 |
List of the 11 indicators used in the ranking, distributed in four criteria (Geneva, Switzerland, 2022).

TABLE 2 |
FormulasResearch Queries Used in the Web of Science™ Core Collection Database For the 14 schools that were not registered in the software, data were collected from the database.To find publications belonging to a certain school in the database, we developed a research query for each school.As an example, we take the Mailman School of Public Health, which is part of the used for the ranking, based on indicators normalized in "base 100" (i.e., "score").The score number corresponds to the number attributed to each indicator in Table1(Geneva, Switzerland, 2022).Overall score N °Formula (using indicators in "base 100," i.e., "score")Overallscore 1 Formula 1 = score 1 + score 2 + score 3 + score 4 + score 5 + score 6 + score 7 + score 8 + score 9 + score 10 + score 11 Overall score 2 Formula 2 = score 1 × 1 + score 3 x (1/3) + score 5 x(1/3) + score 7 x (1/3) + score 9 × 1 + score 11 × 1 TABLE3| Criterion weighting (in %) for Formula 1 and Formula 2 (Geneva, Switzerland, 2022).

TABLE 4 |
List of the 26 schools of Public Health included in the ranking (according to countries, classified in alphabetical order) (Geneva, Switzerland, 2022).