Skip to main content
Advertisement
Browse Subject Areas
?

Click through the PLOS taxonomy to find articles in your field.

For more information about PLOS Subject Areas, click here.

  • Loading metrics

Examining word association networks: A cross-country comparison of women’s perceptions of HPV testing and vaccination

  • Bernd C. Schmid,

    Roles Conceptualization, Funding acquisition, Methodology, Project administration, Writing – original draft, Writing – review & editing

    Affiliation Department of Gynaecological Oncology, Royal Hospital for Women, Randwick, Australia

  • Jamie Carlson ,

    Roles Formal analysis, Methodology, Resources, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

    Jamie.carlson@newcastle.edu.au

    Affiliation Newcastle Business School, Faculty of Business and Law, University of Newcastle, Australia, Newcastle, Australia

  • Günther A. Rezniczek,

    Roles Conceptualization, Data curation, Funding acquisition, Investigation, Methodology, Writing – original draft, Writing – review & editing

    Affiliation Department of Obstetrics and Gynecology, Ruhr-Universität Bochum (Marien Hospital Herne), Düngelstraße 33, Herne, Germany

  • Jessica Wyllie,

    Roles Methodology, Software, Validation, Visualization, Writing – original draft, Writing – review & editing

    Affiliation Newcastle Business School, Faculty of Business and Law, University of Newcastle, Australia, Newcastle, Australia

  • Kenneth Jaaback,

    Roles Conceptualization, Funding acquisition, Writing – review & editing

    Affiliation Department of Gynaecological Oncology, University of Newcastle, School of Medicine and Public Health, John Hunter Hospital, New Lambton Heights, Australia

  • Filip Vencovsky

    Roles Conceptualization, Methodology, Software, Validation, Visualization

    Affiliation Faculty of Informatics and Statistics, University of Economics, Prague, Czech Republic

Abstract

In this study, we examined the perceptual associations women hold with regard to cervical cancer testing and vaccination across two countries, the U.S. and Australia. In a large-scale online survey, we presented participants with ‘trigger’ words, and asked them to state sequentially other words that came to mind. We used this data to construct detailed term co-occurrence network graphs, which we analyzed using basic topological ranking techniques. The results showed that women hold divergent perceptual associations regarding trigger words relating to cervical cancer screening tools, i.e. human papillomavirus (HPV) testing and vaccination, which indicate health knowledge deficiencies with non-HPV related associations emerging from the data. This result was found to be consistent across the country groups studied. Our findings are critical in optimizing consumer education and public service announcements to minimize misperceptions relating to HPV testing and vaccination in order to maximize adoption of cervical cancer prevention tools.

Introduction

Growing evidence has demonstrated that the human papillomavirus (HPV) is the most common sexually transmitted infection (STI) responsible in a range of cervical, anogenital and oropharyngeal cancer cases. Specifically, 83% of all cervical cancer cases worldwide are attributable to the HPV infection and are therefore preventable through vaccination and screening tools [1]. In spite of the varying early cervical cancer detection programs (ECDP) that exist across the globe, cervical cancer is the fourth most widespread cancer affecting women worldwide, with an estimated 527,624 new cases and 265,672 deaths since 2012 [2, 3]. Although proven to be effective in decreasing the incidence and mortality rates of cervical cancer, cytology screening programs with a call and recall system [4] have begun to be replaced with HPV testing and vaccination as primary ECDP screening tools in several countries [5, 6].

In contrast to cervical cytology, new evidence and technology has illustrated that HPV testing provides a cost-effective and more sensitive approach in detecting high-degree lesions [7], and consequently improves early detection amongst women. However, empirical research has found that despite the benefits afforded by HPV vaccination and testing, women often hold negative psychosocial and socio-cultural associations towards cervical cancer and screening [5]. These negative associations may include beliefs that cervical cancer is HIV-related and due to poor vaginal hygiene, so than screening signifies an admission of infidelity, or that screening may affect fertility [8]. Consequently, these associations may function as a deterrent towards proactive health behaviors amongst women (i.e. HPV vaccination and testing).

Variation in cervical cancer cases has therefore been linked to the presence of adequate ECDP and other relevant resources within countries, as well as the population presence of the cervical HPV infection [8]. For example, approximately half of all OECD countries have organized screening and vaccination via population-based programs [9]. Hence, with the population of women aged 15 years and older exceeding 2.7 billion worldwide [4], the efficacy of ECDPs requires commitment from the public with response to vaccination messages and cervical cancer screening recommendations [6]. Therefore, it is critical that public attitudes and perceptions of HPV vaccination and new screening methods like HPV testing are captured and understood. Such an understanding will aid in the optimization of consumer education, public service announcements and branding strategies that help to facilitate participation in vaccination and screening by women.

The sensitive, personal and private nature surrounding public health concerns and, in this context, cervical cancer and screening, has often resulted in participants being unwilling to answer direct questions [10, 11]. Thus, research designs employed in these studies need to reflect and adjust to these complexities accordingly, in a manner sensitive to the research participants. Due to these reasons, survey research has become very popular in health research [12, 13, 14, 15]. Survey research facilitates fast and cost-effective data collection, particularly when paired with online collection methods. It also facilitates highly structured data collection, which is useful when efficiency and, intuitive and quickly actionable outcomes are the focus of the research [16, 17]. To increase the sophistication of survey research techniques and to gather the data necessary for detailed research into consumer associations, researchers have developed approaches combining survey tasks and network-based associative analyses [18, 19, 20]. Building on this work, we demonstrate the utility of novel data-driven approaches to public health research, and propose these as a means to learn more about women’s perceptions of ECDP screening tools.

Utilizing this data-driven approach, in this study we analyze the perceptual word associations women hold with regard to ECDP screening tools (i.e. cervical cancer testing and vaccination) across two countries, the United States and Australia. We examined these two countries as they use the ECDP screening tools in different ways, thus enabling us to see how informed women drawn from the general population are and to identify information gaps within each country. We undertook a semi-structured data-mining exercise, which enabled the construction of co-occurrence network graphs that were then analyzed using basic topological ranking techniques. To this end, we aimed to answer the following research questions:

  1. What word associations surrounding HPV testing do women hold and are these associations consistent across similar country groups?
  2. What word associations surrounding HPV vaccination do women hold and are these associations consistent across similar country groups?
  3. What can the types of terms produced and the connections between them tell us about the usefulness of word-association research in the public health context?

Early Cervical Cancer Detection Programs (ECDP) in Australia and United States

ECDP screening, which until recently, was based solely on cytology, in the form of the Pap smear, is currently shifting to rely on HPV testing procedures instead (accompanied by a HPV vaccination program) [5, 21]. For instance, the National Cervical Screening Program in Australia is set to introduce HPV testing in December 2017 [7], while countries such as the United States and Mexico have had HPV testing and co-testing (cytology and HPV testing) as their primary screening tools since 2008 [6,7]. Subsequently, clinical guidelines aligned to frequency and age of cervical cancer screening vary across countries and are reflective of the ECDP screening tool.

Currently in Australia, the Pap test is the primary ECDP screening tool, with clinical guidelines recommending women aged 18 to 69 undergo a Pap test every two years [7]. This varies significantly with HPV vaccine and testing, such that young women (9 to 13 years of age) have two doses of the HPV vaccine, and HPV testing is recommended every five years for women aged 25 to 74 [7, 22].

In the United States, these cervical cancer clinical guidelines differ, with women aged 21 to 65 years recommended to undergo a Pap test every three years [23]. Further, women aged 30 to 65 years seeking to extend the screening interval are able to do so through a preferred method of co-testing, which comprises both cytology and HPV testing and is performed every five years [23]. However, it is necessary to note that average at-risk women aged 25 to 65 years have the ability to use the HPV test as their primary screening tool [7]

Based on the above discussion, Australia and the United States share similar ECDPs. However, a core difference between these two countries lies in their use of HPV vaccination and testing. Therefore, we aim to generate understanding of the perceptual associations that arise from women’s thinking about ECDP screening tools (i.e. Pap test, HPV testing and HPV vaccination) across these two countries. A key aspect of our work is that co-occurrence network graphs will enable greater understanding surrounding cervical cancer screening and, ultimately, work towards the optimization of consumer education and public service announcements.

Methods

Data collection

Using the consumer database from a reputable marketing research firm (SurveyMonkey), a large-scale online survey was conducted December 18–21, 2015. Participants were randomly selected from SurveyMonkey’s U.S. and Australian databases, using this study’s pre-defined selection criteria of women aged 18 to 64 years. An email invitation was sent to potential participants outlining the purpose of the study, giving instructions to complete the survey and including the link to the online survey. Implied consent to the study was provided through participants’ registration with SurveyMonkey, as well as the anonymous completion of this study’s survey. Participants who completed the survey were compensated via non-monetary incentives including donations to their preferred charity, and were given entries into a draw to win sweepstakes [24]. Further, consistent with institutional review board policies, ethics approval was not required.

A total sample of 1473 (68%) was achieved with 704 from the U.S. and 769 from Australia. The total number of incomplete responses was 697 accounting for 32% of the sample, with 346 of these, participants from the U.S. sample and 351 participants from the Australian sample. SurveyMonkey also provided basic demographic information from participants, such as age and household income brackets, which is summarized in Table 1.

In the survey, we randomly presented participants with several trigger words to which they were asked to provide, in sequential order, the first three words (i.e. response words) that came to mind. The trigger words shown to participants comprised “cervical cancer”, “cervical cancer testing” and “cervical cancer vaccination” in succession. We divided the network analysis of participants’ response words (and subsequent presentation of results) into “vaccination” (trigger words “HPV vaccination” and “cervical cancer vaccination”) and “testing” (trigger words “HPV (human papillomavirus) test” and “pap smear”), and then further divided the responses by country groups (U.S. and Australia). For example, an Australian participant shown the trigger word ‘HPV (Human papillomavirus) test’ provided the response words of “cervix”, “cancer” and “virus”.

Analysis

We used the data from the surveys to construct detailed, weighted term co-occurrence network graphs. Co-occurrence is a fundamentally simple concept, with relevance in “hard-science” applications [25] and social science applications such as analyzing textual co-occurrence patterns [26]. In this study, data was processed (including computing co-occurrence) using KNIME [27]. Network analysis was performed using Gephi [28] and Cytoscape [29].

Co-occurrence between response words was computed by taking the n-gram (i.e. set of adjacent words) co-occurrence statistic data [30] that participants typed into separated fields in response to the trigger words. Whatever participants entered into the three separate fields provided for each trigger word was then converted into three separate nodes. Subsequently, calculation of these co-occurrences between fields was conducted. The only pre-processing applied was case conversion (i.e. conversion to lower case) as we were only interested in ranking exactly matched n-grams in this study.

We also removed n-grams related to the terms “unknown” and “n/a” provided by participants as we took these to denote a non-response. We included all other n-grams. Edges were thus created connecting n-grams provided by the same unique participant, in response to the same trigger word. The co-occurrence data for each participant was then merged into separate network graphs according to each trigger word. When the participant-level data was combined, the nodes representing identical entries (in response to trigger words) were merged. Identical co-occurrence pairings (edges) were also merged. This approach enabled the most salient n-grams (i.e. nodes) to emerge as naturally as possible.

Topology

Having generated networks from the word co-occurrence data collected from participants, we then analyzed the topological properties of the resulting networks. This involved examining the sub-structures of the networks (i.e. groupings of nodes and patterns in connections between nodes), as well as ranking nodes using some basic topological measures [31]. Specifically, we computed: degree centrality (the number of connections for each node), weighted degree (number of connections adjusted for edge weight) [32] and eigenvector centrality (weighted centrality, i.e. nodes with important connections get higher ranks) [33]. These measures were computed using the full networks (see Table 2) but, for clarity of presentation, we visualize and display rankings for nodes with degree centrality >10 only.

Visualization

Visualization was performed using Gephi [28], whereby node size corresponds with degree centrality, edge size corresponds with edge weight (i.e. the number of paired occurrences between nodes) and rank tables are ordered by degree centrality.

Results

In the following sections, the network properties and structures of each of the developed ‘trigger word’ network graphs are discussed. The results show that the HPV and Pap smear testing networks illustrated similarities with the salient n-grams (i.e. “necessary”) that arose, whilst negative n-grams (i.e. “uncomfortable”) were most apparent in the Pap smear testing networks. Furthermore, upon visual inspection of the HPV and cervical cancer vaccination networks across both country groups, the preventative and beneficial nature of the trigger word “vaccination” was exhibited through the identified n-grams.

HPV and Pap smear testing networks

Network A, seeded from the trigger word “HPV Testing” for the U.S. country group comprised, 794 nodes connected by 1974 edges with an average degree of 4.904 (Fig 1). This network graph partitions into four modules, with a modularity score of 0.334: one core community related to the n-gram “cancer”, two major communities related to the n-gram “necessary” and “prevention” and one disparate community. In line with the degree and eigenvector centrality measures specified in Table 2 for the U.S. country group, the top-ranked n-grams that are most embedded in the network are “cancer”, “necessary” and “std (sexually transmitted disease)”.

In reference to Network B, which was seeded from the trigger word “Pap smear testing”, the U.S. country group is comprised of 667 nodes connected by 1790 edges with an average degree of 5.367 (Fig 2). This network graph also partitions into four modules, with a modularity score of 0.25 and comprising one core community related to the n-gram “uncomfortable”, one major community related to the n-gram “test”, a small community related to the n-gram “annual” and a disparate community. Furthermore, as shown in Table 2, the top-ranked n-grams most embedded in this network were “uncomfortable”, “necessary” and “yearly”. Although studies have shown that HPV heightens the risk of cervical cancer in women [34], the term “std” does not feature within the top 25 nodes in the Pap smear test network graph. Overall, these findings demonstrate that HPV testing is perceived as being less invasive and more favorable than the Pap smear test amongst the U.S. female participants.

thumbnail
Fig 2. Network B.

U.S. Pap smear test network visualization.

https://doi.org/10.1371/journal.pone.0185669.g002

In comparison to the U.S. country networks, the findings from the Australian network groups demonstrate similar n-grams. Specifically, Network C, seeded from the trigger words “HPV testing”, comprised 718 nodes connected by 1703 edges with an average degree of 5.136 (Fig 3). This network graph partitioned into three modules, with a modularity score of 0.316. As specified in Table 2, the network included one core community related to the n-gram “cancer” and two major communities related to the n-grams “necessary” and “good” respectively.

thumbnail
Fig 3. Network C.

Australia HPV test network visualization.

https://doi.org/10.1371/journal.pone.0185669.g003

Network D, seeded from the trigger word “Pap smear testing” within the Australian sample, comprised 625 nodes connected by 1703 edges with an average degree of 5.45 (Fig 4). This network graph partitioned into four modules, with a modularity score of 0.242. It included one core community related to the n-gram “uncomfortable”, one major community related to the n-gram “cancer”, one small community related to the n-gram “discomfort” and finally one disparate community. Across both networks, C and D, the three most embedded n-grams within these networks comprised “cancer”, “uncomfortable” and “necessary”. However, the top-ranked n-grams of these network graphs were different, such that degree and eigenvector centrality showed that “cancer” was the top-ranked n-gram of Network C, whilst “uncomfortable” was the top-ranked n-gram in Network D. This result indicates’ that Australian women are more aware than women from the U.S. that the testing process of retrieving the small sample of cells from the surface of the cervix is actually the same for both, the HPV test and the Pap smear test.

Conversely, the n-grams from the HPV test network graphs (see Figs 1 and 3) across both country groups have demonstrated more positive associations than those of the Pap smear network graphs (see Figs 2 and 4). There is greater correlation evidenced between the n-grams “prevention”, “detection”, “screening” to the terms “cancer” and “doctor” in these graphs. Finally, the Australian network graphs (see Figs 3 and 4) highlighted that participants identified the timeframe in which testing is performed to detect cervical cell changes via the HPV test or Pap smear test with n-grams “yearly” and “annual”, which is surprising given the 2-year screening interval in Australia. This perception may heighten the perceived burden for women in undertaking this preventative behavior.

HPV and cervical cancer vaccination networks

Within the U.S. sample, Network E, seeded from the trigger words “HPV vaccination”, comprised 833 nodes connected by 1972 edges with an average degree of 4.735 (Fig 5). This network graph partitions into four communities, with a modularity score of 0.276. The communities included one core community related to the n-grams “prevention” and “shot”, one major community related to the n-gram “good”, and one small community related to the n-grams “young” and “painful”. Analysis of Network F, seeded from the trigger words “cervical cancer vaccination”, showed that the network comprised 828 nodes that were connected by 1957 edges with an average degree of 4.727 (Fig 6). Network F partitions into three modules with a modularity score of 0.226, with one core community related to the n-gram “shot”, a major community related to the n-grams “necessary” and “good”, and, finally, a small community related to the n-gram “prevention”. As shown in Table 3, the three most embedded n-grams and their ranks, across both Network E and F for the U.S. country group, were identical (“shot”, “prevention” and “good”).

thumbnail
Fig 5. Network E.

U.S. HPV vaccination network visualization.

https://doi.org/10.1371/journal.pone.0185669.g005

thumbnail
Fig 6. Network F.

U.S. cervical vaccination network visualization.

https://doi.org/10.1371/journal.pone.0185669.g006

In reference to the Australian country group, Network G, which was seeded from the trigger words “HPV vaccination”, comprised 777 nodes connected by 1923 edges with an average degree of 4.95 (Fig 7). This network graph partitions into four modules, with a modularity score of 0.235. Analysis of the network shows that there is one core community related to the n-grams “prevention” and “needle”, one major community related to the n-gram “necessary” and two disparate communities. Finally, Network H, seeded by the trigger words “cervical cancer vaccination”, comprised 732 nodes connected by 1889 edges, with an average degree of 5.161 (Fig 8). This network graph partitions into three modules with a modularity score of 2.42. The communities were dispersed into two core communities and one disparate community, such that one core community related to the n-grams “prevention” and “needle”, and the other core community related to the n-gram “necessary”.

thumbnail
Fig 7. Network G.

Australia HPV vaccination network visualization.

https://doi.org/10.1371/journal.pone.0185669.g007

thumbnail
Fig 8. Network H.

Australia cervical vaccination network visualization.

https://doi.org/10.1371/journal.pone.0185669.g008

In review of Table 3, degree and eigenvector centrality measures demonstrate that the top-ranking n-grams across the Australian networks bear similarities, regarding the most embedded n-grams. Specifically, the three most embedded n-grams for the trigger word HPV vaccination were “prevention”, “needle” and “necessary”, whilst the trigger words “cervical cancer vaccination” had the n-grams “prevention”, “necessary” and “good”. It is interesting to note that the fourth ranking n-gram for HPV and cervical cancer vaccination were “good” and “necessary” respectively.

Discussion

We structure our discussion around each of the research questions outlined at the beginning of this study. First, in spite of the differences between the ECDP across both countries, the associations regarding the trigger words HPV testing produced positive perceptual associations by female participants in the U.S. and Australian samples studied. We also found that negative connotations were raised by participants relating to the uncomfortable nature of the Pap test.

The overarching theme representing this particular form of testing was signified across both country samples by n-grams such as “uncomfortable”, “awkward” and “invasive”. This theme encapsulates the negative connotations that participants have associated with the Pap smear test, implied by the terms “discomfort”, “painful” and “awkward”. Although, the Pap smear network graphs within each country sample highlight the necessary and preventative nature of the test in identifying cervical cancer, this form of cervical cancer testing engenders negative perceptions, which may function to inhibit preventative action amongst women. Consequently, better education of health professionals is required to make the testing process and service environment less uncomfortable, which may work to increase the participation rate.

The results show that both the U.S. and Australian country samples drew links between the trigger words “HPV and cervical vaccination” and sexually transmitted infections (STIs), with the terms “sex”, “std”, “virus” and “disease” reported by the participants. Both country samples also identified a correct association between the triggers word “HPV vaccination” and the n-grams “warts”. This shows that there is knowledge in each country sample that HPV can cause genital warts and is an STI. However, both country samples identified incorrect STI associations between these trigger words (“HPV vaccination”) and the n-grams “HIV”, “herpes” and “AIDS”. Such findings might indicate that women across both country samples do not differentiate between STIs, as well as holding false assumptions about HPV testing using the same medical procedures as HIV testing (a ‘simple blood test’). Therefore, this finding illustrates potential health themes for educational public service announcements and intervention programs that encourage adoption of the HPV vaccine and preventative sexual behaviors.

Taken collectively, these findings are significant based on the explicit word choices of negative connotations toward this form of cervical cancer screening may function to inhibit women’s decision-making with regard to the adoption of preventative health behavior actions (i.e. undergoing regular Pap smear tests). Consequently, it is advisable that cervical screening education programs are designed to inform the public as to the precise details of HPV vaccination and screening schedules as well as mitigate flawed assumptions and misconceptions regarding the nature of the procedures. By addressing concerns on the part of women about timing and comfort, such action may improve HPV and Pap smear-testing goals.

Second, regarding associations surrounding HPV vaccination, we found across both country samples that women hold correct and favorable associations relating to the preventative and beneficial nature of the vaccination. For example, participants consistently referred to the n-grams “school”, “teens” and “young”. This finding signifies that participants across both country samples are aware that the vaccination is administered to pre-teen females via doctors or school immunization programs [35].

Third, in terms of broad relevance for public health research, this study reveals a number of interesting insights. In particular, the approach used in this paper allowed us to assess the diversity and variation in vocabularies used by patients to describe their perceptions and associations across two country samples. We then showed a simple method for identifying the relative importance of terms to specific trigger words.

In this study, we have demonstrated the usefulness of our approach in identifying community groups and sub-structures within networked patient associations across each country sample. Further clarity can be achieved with targeted filtering of the networks to examine network sub-structures more closely. Fig 9 below, for instance, shows two prominent communities (based on node degree centrality and edge weight) extracted from the Australian HPV vaccination network (“needle”, “prevention”, “cancer” and “good”, “easy”, “safe”, respectively) and the connections shared by the nodes.

thumbnail
Fig 9. Australia HPV vaccination network graph sub-structure.

https://doi.org/10.1371/journal.pone.0185669.g009

We have also demonstrated how easily words pertaining to specific topics, content, or sentiment types can be compared. Fig 10 below, for instance, shows only terms possibly understood as negative filtered from the U.S. Pap smear network. This allows us to quickly compare connections between prominent nodes such as “uncomfortable” and “painful” with lower-ranked nodes such as “annoying” and “embarrassing”. Such filtering could also be used to filter negative and positive associations for comparison.

Limitations, future research and concluding remarks

Our study has several limitations. First, the data set used in this study may not be representative of the female population aged 18+ in the targeted countries, as the algorithm for selecting participants is not disclosed and the pool of potential participants as a whole may be inherently biased. Second, the survey was conducted in a very narrow timeframe (4 days, just before the Christmas holidays), which may have introduced some bias but would also have excluded any bias from perception shifts that might have occurred in the targeted population over an extended time. Third, we based our analysis on exact string matches without any pre-processing such as stemming, lemmatization or fuzzy matching. Thus, the relevance of some concepts may be underestimated.

A number of research directions arise from this study. Using a more representative sample of the population, future research could consider deeper analysis focused on semantics [36] and also focus more on lower ranked nodes in similar networks that may reveal associations or attributions that are less prevalent but still, perhaps, important when considered across large enough groups. Along these lines, future research should explore differences between demographic groups (e.g. young vs. old, low vs. high income or education levels, metropolitan vs regional areas, cultural background, vaccinated vs non-vaccinated) as well as groups with different contextual perceptions and biases (e.g. those who consider HPV a sexually transmitted disease, and those who do not) across different country settings.

Future research will consider applying similar approaches to unstructured data (e.g. health discussions on social media [37]). This could be an important avenue for research, given the increasing use by patients of web-based tools to gather health information [38, 39]. The approach used in this study could also apply to other public health contexts (e.g. healthy eating, drug and alcohol use, and mental health). More specifically, this could include, for example, health issues themselves (e.g. risk perceptions around skin cancer) [40], other marketing related contexts (e.g. health claims) [41], health promotion (e.g. health websites and m-health applications) [42, 43, 44], and health service management contexts (e.g. value in health services) [45]. Finally, given the cross-sectional nature of the study, future research could collect and analyze word association data collected on a real-time or longitudinal basis such as through health apps [46,47].

Supporting information

S1 File. Survey.

The survey presented to the probands.

https://doi.org/10.1371/journal.pone.0185669.s001

(PDF)

S2 File. Survey responses.

The data file (.xlsx) with the survey responses and metadata.

https://doi.org/10.1371/journal.pone.0185669.s002

(XLSX)

Acknowledgments

The authors acknowledge Dr Ben Lucas, Maastricht University, Netherlands for the constructive comments and support in data analysis for this paper.

References

  1. 1. de Martel C, Plummer M, Vignat J, Franchesi S. Worldwide Burden of Cancer Attributable to HPV by Site, Country and HPV Type. Int J Cancer. 2017;141(4): 664–670. pmid:28369882
  2. 2. Peto J, Gilham C, Fletcher O, Matthews FE. The cervical cancer epidemic that screening has prevented in the UK. Lancet. 2004;364(9430): 249–256. pmid:15262102
  3. 3. Binka C, Nyarko SH, Doku DT. Cervical Cancer Knowledge, Perceptions and Screening Behaviour Among Female University Students in Ghana. J Cancer Edu. 2016;31(2): 322–327.
  4. 4. Ferlay J, Shin HR, Bray F, Forman D, Mathers C, Parkin DM, et al. Estimates of worldwide burden of cancer in 2008: GLOBOCAN 2008. Int J Cancer. 2010;127(12): 2893–2917. pmid:21351269
  5. 5. Bruni L, Barrionuevo-Rosas L, Albero G, Serrano B, Mena M, Gómez D, et al. ICO Information Centre on HPV and Cancer (HPV Information Centre). Human Papillomavirus and Related Diseases in the World. Summary Report 19 April 2017. Available from: http://www.hpvcentre.net/datastatistics.php. Access date: 05 May 2017.
  6. 6. León-Maldonado L, Wentzell E, Brown B, Allen-Leigh B, Torres-Ibarra L, Salmerón J, et al. Perceptions and Experiences of Human Papillomavirus (HPV) Infection and Testing among Low-Income Mexican Women. PloS One. 2016;11(5): e0153367. pmid:27149525
  7. 7. CDC, Cervical Cancer Screening Guidelines for Average-Risk Women. Available from: https://www.cdc.gov/cancer/cervical/pdf/guidelines.pdf.
  8. 8. Driscoll SD. Barriers and facilitators to cervical cancer screening in high incidence populations: A synthesis of qualitative evidence. Women & Health 2016;56(4): 448–467.
  9. 9. OECD (2013), Cancer Care: Assuring Quality to Improve Survival,Paris: OECD Publishing. 2013. https://doi.org/10.1787/9789264181052-en
  10. 10. John DR, Loken B, Kim K, Monga AB. Brand concept maps: A methodology for identifying brand association networks. J Mark Res. 2006;43(4): 549–563.
  11. 11. Lee TY, Bradlow ET. Automated marketing research using online customer reviews. J Mark Res. 2011;48(5): 881–894.
  12. 12. Culotta A, Cutler J. Mining Brand Perceptions from Twitter Social Networks. Marketing Science. 2016;35(3): 343–362.
  13. 13. Tourangeau R, Yan T. Sensitive questions in surveys. Psychol Bull. 2007;133(5): 859. pmid:17723033
  14. 14. Northouse PG, Northouse LL. Communication and cancer: issues confronting patients, health professionals, and family members. J Psychosoc Onco. 1988;5(3): 17–46.
  15. 15. Bennett C, Khangura S, Brehaut JC, Graham ID, Moher D, Potter BK, et al. Reporting guidelines for survey research: An analysis of published guidance and reporting practices. PLoS Med. 2011;8(8): e1001069.
  16. 16. Schermel A, Mendoza J, Henson S, Dukeshire S, Pasut L, Emrich TE, et al. Canadians’ perceptions of food, diet, and health–a national survey. PloS One. 2014;9(1): e86000. pmid:24465832
  17. 17. Prah P, Johnson AM, Nardone A, Clifton S, Mindell JS, Copas AJ, et al. Asking about sex in general health surveys: Comparing the methods and findings of the 2010 health survey for England with those of the third national survey of sexual attitudes and lifestyles. PloS One. 2015;10(8): e0135203. pmid:26252650
  18. 18. Narayanan A, Greco M. Patient Experience of Australian General Practices. Big Data. 2016;4(1): 31–46. pmid:27441583
  19. 19. De Deyne S, Storms G. Word associations: Network and semantic properties. Behav Res Methods. 2008;40(1): 213–231. pmid:18411545
  20. 20. De Deyne S, Navarro DJ, Storms G. Associative strength and semantic activation in the mental lexicon: evidence from continued word associations. Proceedings of the 35th Annual Meeting of the Cognitive Science Society 2013: 2142–2147. Available from: https://mindmodeling.org/cogsci2013/papers/0393/paper0393.pdf.
  21. 21. WHO, Guidelines for the Prevention and Control of Cervical Cancer. Available from: http://www.who.int/reproductivehealth/topics/cancers/hpv-vaccination/en/.
  22. 22. Van Rensbergen B, Storms G, De Deyne S. Examining assortativity in the mental lexicon: Evidence from word associations. Psychon Bull Rev. 2015;22(6): 1717–1724. pmid:25893712
  23. 23. U.S. Preventative Services, Task Force, Cervical Cancer: Screening. Available from: https://www.uspreventiveservicestaskforce.org/Page/Document/UpdateSummaryFinal/cervical-cancer-screening.
  24. 24. SurveyMonkey, Designing your Survey. Available from: https://help.surveymonkey.com/articles/en_US/kb/SurveyMonkey-Audience?bc=Buying_Responses#Panel.
  25. 25. Widder S, Besemer K, Singer GA, Ceola S, Bertuzzo E, Quince C, et al. Fluvial network organization imprints on microbial co-occurrence networks. Proc Natl Acad Sci USA. 2014;111(35): 12799–12804. pmid:25136087
  26. 26. Li H, An H, Wang Y, Huang J, Gao X. Evolutionary features of academic articles co-keyword network and keywords co-occurrence network: Based on two-mode affiliation network. Physica A. 2016;450: 657–669.
  27. 27. Berthold MR, Cebron N, Dill F, Gabriel TR, Kötte T, Meinl T, et al. KNIME: The Konstanz information miner. Berlin, Heidelberg: Springer; 2008. pp. 319–326.
  28. 28. Bastian M, Heymann S, Jacomy M. Gephi: an open source software for exploring and manipulating networks. ICWSM. 2009;8: 361–362.
  29. 29. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. 2003;13(11): 2498–2504. pmid:14597658
  30. 30. Choi D, Ko B, Hwang M, Kim P. Building knowledge domain N-gram model for mobile devices. An Int Interdisciplinary J. 2014;14(11): 3583–3590.
  31. 31. Vanni T, Mesa-Frias M, Sanchez-Garcia R, Roesler R, Schwartsmann G, Goldani MZ, et al. International scientific collaboration in HIV and HPV: a network analysis. PloS One. 2014;9(3): e93376. pmid:24682041
  32. 32. Totet M. Let’s Play Gephi: Understand Degree, Weighted Degree & Betweeness centrality. 2016. Available from: http://matthieu-totet.fr/Koumin/2013/12/16/understand-degree-weighted-degree-betweeness-centrality/.
  33. 33. Lohmann G, Margulies DS, Horstmann A, Pleger B, Lepsien J, Goldhahn D, et al. Eigenvector centrality mapping for analyzing connectivity patterns in fMRI data of the human brain. PloS One. 2010;5(4): e10232. pmid:20436911
  34. 34. American Cancer Society. What are the risk factors for cervical cancer. Available from: https://www.cancer.org/cancer/cervical-cancer/causes-risks-prevention/risk-factors.html
  35. 35. Centre for Disease, Control & Prevention. Human Papillomavirus (HPV): HPV vaccine information for young women. Available from: http://www.cdc.gov/std/hpv/stdfact-hpv-vaccine-young-women.htm. Date accessed: 05 April 2016.
  36. 36. Youn H, Sutton L, Smith E, Moore C, Wilkins JF, Maddieson I, et al. On the universal structure of human lexical semantics. Proc Natl Acad Sci USA. 2015;113(7): 1766–1771.
  37. 37. Kitchens B, Harle CA, Li S. Quality of health-related online search results. Decis Support Syst. 2014;57: 454–462.
  38. 38. Song H, Omori K, Kim J, Tenzek KE, Hawkins JM, Lin WY, et al. Trusting Social Media as a Source of Health Information: Online Surveys Comparing the United States, Korea, and Hong Kong. J Med Internet Res. 2016;18(3): e25. pmid:26976273
  39. 39. Janssen E, van Osch L, de Vries H, Lechner L Measuring risk perceptions of skin cancer: Reliability and validity of different operationalizations. Br J Health Psychol. 2011;16(1): 92–112.
  40. 40. Belei N, Geyskens K, Goukens C, Ramanathan S, Lemmink J. The best of both worlds? Effects of attribute-induced goal conflict on consumption of healthful indulgences. J Mark Res. 2012;49(6): 900–909.
  41. 41. Schneider F, van Osch L, de Vries H. Identifying factors for optimal development of health-related websites: a delphi study among experts and potential future users. J Med Internet Res. 2012;14(1): e18. pmid:22357411
  42. 42. Free C, Phillips G, Galli L, Watson L, Felix L, Edwards P, et al. The effectiveness of mobile-health technology-based health behaviour change or disease management interventions for health care consumers: a systematic review. PLoS Med. 2013;10(1): e1001362. pmid:23349621
  43. 43. McColl-Kennedy JR, Vargo SL, Dagger TS, Sweeney JC, van Kasteren Y. Health care customer value cocreation practice styles. J Serv Res. 2012;15(4): 370–389.
  44. 44. Wyllie J, Lucas B, Carlson J, Kitchens B, Kozary B, Zaki M. An Examination of Not-For-Profit Stakeholder Networks for Relationship Management: A Small-Scale Analysis on Social Media. PloS One. 2016;11(10): e0163914. pmid:27711236
  45. 45. Sweeney JC, Danaher TS, McColl-Kennedy JR. Customer effort in value cocreation activities: Improving quality of life and behavioral intentions of health care customers. J Serv Res. 2015;18(3): 318–335.
  46. 46. Spanakis G, Weiss G, Boh B, Roefs A. Network Analysis of Ecological Momentary Assessment Data for Monitoring and Understanding Eating Behavior. 2016.
  47. 47. Boh B, Lemmens LH, Jansen A, Nederkoorn C, Kerkhofs V, Spanakis G, et al. An ecological momentary intervention for weight loss and healthy eating via smartphone and Internet: Study protocol for a randomised controlled trial. Trials. 2016;17(1): 1.