Background
The case for causality with respect to poor health outcomes is multifactorial. Increasing global mortality as a result of chronic disease continues to challenge health care provision financially [
1‐
3]. Additional contributors include increasing modifiable negative public health behaviours, most notably: smoking, poor diet and lack of physical activity [
2,
4]. Despite the growing prevalence of these modifiable risk-factors for long-term conditions, we are seeing an aging population. This is due to improvements in socio-economic status (SES), reductions in birth mortality and a decreasing prevalence of communicable disease [
5]. However, these recent advancements in individual status and wellbeing are matched by cumulative morbidity rates and a seemingly unachievable demand for healthcare resources, further fuelling the growing burden on national care provision [
6,
7]. Even though individuals may have benefitted from changes in SES, this is not a nationwide phenomenon. The literature suggests those with a lower SES are increasingly likely to be exposed to more unhealthy behaviours and outcomes compared to their more affluent counterparts [
8].
From a socio-economic perspective, measures such as the Townsend Score [
9] or Index of Multiple Deprivation [
10] (IMD) are examples of tools that attempt to correlate SES to overall wellbeing through a number of factors including employment, education, health deprivation and disability. Such tools frequently utilise infographics or images to better convey information given the public’s familiarity with this form of data visualisation. The aim of such an approach is to enhance understanding, engagement, and interest in the topic. Evidence of this application already exist within academia [
11‐
13] and research [
14,
15], but in an era of globalisation, multi-media and increasing access to technology, the role of visualisation continues to expand. Internationally, there is a wealth of literature discussing the use of visualisation tools in relation to public health data. A systematic review produced in 2014 [
16] identified various themes that impact implementation of visualisation tools. A total of 88 documents across 5 bibliographic databases met the inclusion criteria with a focus on epidemiological data of infectious disease. A key observation from the review was that the potential to provide an abundance of relevant information was offset by poor interpretation of data. Lack of follow up in terms of tool usability plans and data dissemination were reported as contributing factors to a decreased uptake in this approach to data analysis. Furthermore, service-users were found to have different needs, lack of support and a general misunderstanding of how visualisation tools can be utilised [
16].
A more positive demonstration of Geographical Information System (GIS) implementation in public health was undertaken by St James’s Hospital in Dublin, assessing population Vitamin D levels in the local catchment area [
17]. The study visually demonstrated the relationship between seasonal changes, gender, and age in the variability of Vitamin D concentrations. This approach supported the concept of service provision based on identifying high risk areas for example, lower Vitamin D levels in men vs women living in the same area. From this observation, male patients could be the target group for service provision in specified localities. Similarly, Curtis et al. [
18] produced evidence depicting regional trends in diabetes prevalence contrasted with available regional resources. This resulted in data analysis that helped categorise “high risk – low resource (HRLR)” populations. The capacity of visualisation to highlight HRLR areas or populations has value, providing an additional layer of evidence for evaluating and identifying correlations between current health inequalities and available services [
19,
20].
A systematic review by Luan and Law [
21] highlights the utilisation of web-based GIS public health surveillance. The authors describe variations in functionality across several tools such as Google Maps, OpenCalais and ArcGIS to name but a few. Elements perceived as integral for overall functionality include interactivity and usability, irrespective of an individual’s technical background or experience with GIS. Furthermore, the authors emphasise that translation of raw health data via GIS should produce easily interpretable results which can be effectively communicated. Other notable examples include Mapbox, Open Layers and GIS Cloud, however the choice is often determined by user preference and desired functionality.
As part of a collaborative funded project with Public Health Croydon (PHC) and Croydon University Hospital (CUH), the researchers were tasked to create an accessible mapping model to demonstrate variations in socio-economic deprivation across the Borough of Croydon. Upon completion, PHC and CUH assigned the research team with local authority and hospital data to evaluate two case studies based on local needs using the prototype mapping model. Further work was conducted to assess the usability of the model with users outside of the public health domain, which centred on identifying the perceptions of a range of stakeholders (SHs) in the use of interactive visualisation models of public health data. In addition, the study explored the potential uses and value of data visualisation techniques, barriers and challenges to data use and access as well as currently employed methods.
Discussion
As part of a collaborative initiative with PHC and CUH, this study has developed an interactive GIS methodology based on local needs that can deliver visualised public health data while identifying its usability and applications across different SH groups. This study also explored the influence of socio-demographics and smart technology use on the acceptability of the mapping model.
The literature suggests that the average SUS score for a system is 68 [
29]. However, beyond a score of 68 exists ‘above average’ percentile rankings for systems on graded criteria. The overall SUS score for the model in this study was 83.17. This score is indicative of a ‘B’ grade in terms of usability, falling within the top 10% of scores (> 80.3%). The literature suggests that a ranking of > 80.3% not only defines a system as ‘good’ in terms of usability, but also demonstrates an increased likelihood of users recommending the technological system to a friend [
29].
The results are consistent with other methods that assess acceptability such as the Technology Acceptance Model (TAM) [
34]. This method identifies factors such as PEOU and perceived usefulness PU as the most important in determining the likelihood of a new technology being adopted. This study demonstrated that participants identified numerous applications for the mapping model. This result validated the PU aspect of the model assessment in accordance with TAM [
34]. PEOU as a concept was also established through the quantitative SUS measure, with participants strongly agreeing the model was easy to use. This finding was also apparent through reported participant confidence with the model. The mapping model was able to provide a succinct summary of various public health data sets through visualisation of both a 12-month observational study (readmission diagnosis) and retrospective data analysis (paediatric vaccination rates), with participants reporting ease of use as the most positive SUS outcome. These results support the concept of mapping as a tool for information provision by improving understanding and satisfaction in concurrence with the literature [
35].
The case studies were specifically designed and analysed in response to local needs identified by leads across PHC and CUH with further testing conducted to examine the usability and acceptability of the model with SHs not working within public health. With respect to overall distribution of MMR vaccinations for children aged two & five, there is a relatively even spread of recorded vaccinations across the Borough, with greater reporting in more densely populated areas as would be expected. However, when compared to the demographically marked births, there is an apparent disparity with those recorded as White European being registered in more affluent areas versus those registered as either Black African or Asian births in less affluent areas. With the apparent population diversity and greater exposure to deprivation for non-White European persons, it was expected that overall vaccination rate by ward would closely mirror the reports from the IMD for Croydon. Surprisingly, some of the best vaccination reports are from the most deprived wards in the North and East of the Borough, whereas those with the lowest deprivation reported some of the worst vaccination uptake for children aged 2 and 5. Case study 3 also shows the distribution of GP surgeries, with a greater number concentrated in the north of the Borough, however this mainly accounts for the greater population density. A 2016 US study [
36] found a significant relationship between paediatric MMR vaccination and affluence, with those from a more affluent background demonstrating lower vaccination uptake. The authors note the trend to decline vaccination based on personal belief is fairly novel. However, despite the unexpected result produced by the mapping model, no full conclusion can be made between IMD, vaccination uptake and ethnicity within the scope of this study. It may be that those living in the UK in more affluent areas may prefer to vaccinate their children privately, and/or outside of the Borough. Nevertheless, these results were of great interest to study participants who found they could deduce these types of correlations using the interactive elements of the model.
Manovich and McInerny et al [
37,
38] discuss the positive impact of interactivity on interpreting information displayed by visual data models. This evidence supports the themes identified throughout the study, with the SHs suggesting the most important contribution to usability of the mapping model being the interactive component. The “overlaying” function also supplemented SH’s ability to interpret data displayed by the mapping model. Tippett [
39] highlights the value of this visual and constructivist approach e.g., manipulating data layers on the map, as a process through which data comprehension is enhanced. As a result, improvements to usability and the adoption of technology are also observed, as demonstrated by this study through PEOU as a factor of the TAM. Furthermore, the study provides evidence that the interactive functions of the map were well integrated through analysis of the SUS scores.
Interestingly, despite an overall mean SUS score of 83.17, some disparity was noted between participants with a range of 52.5. The lowest SUS score provided was 47.5, with the next closest score being 70. An examination of causal relationships between the result and the demographics for the participant identified one potential implicating factor, the participant’s age. The lowest scoring participant (SUS Score = 47.5) fell into the 60–69 age category. The literature indicates a lower level of acceptability and adoption of smart technologies above the age of 60, and hence this may have been a factor impacting the usability of the mapping model for the participant [
40]. However, in contrast the eldest participant in the 70–79 age category provided a SUS score of 87.5. Additionally, this was also the only participant in the study to not own any smart technology. Although this finding relates to a small sample (
n = 2), it demonstrates that age as a demographic is not always an exclusive indicator of technology acceptance. Chung et al. [
40] support this conclusion in a cross-sectional population TAM study of 248 participants which found no statistical significance between age and PEOU. It may be the case that the participant who provided a score of 47.5 also felt they lacked some understanding or confidence in answering the SUS statements, however this was not concordant with the attitudes of other SHs in answering the SUS survey. Furthermore, no correlation between frequency of smart technology use or demographics and SUS scores was seen in this study, a finding in agreement with the current literature [
40].
The study aimed to determine the desired purpose for the mapping model based on SH views. As previously discussed, numerous applications were proposed by the SHs during the interviews. Overall, the mapping model was identified as a powerful tool in enabling the visualisation of public health data, as demonstrated by the case studies. SHs described the value of the mapping model as a tool for generating discussion with respect to outcomes and population needs. SHs suggested the purpose of the model could be as a tool for planning the provision of health services or identifying already locally available services. This suggestion mirrors previous NHS strategy, referred to as Sustainability and Transformation Plans (STP), which aimed to highlight the needs of the local population to better streamline service delivery. Rummery [
41] emphasises the need for a methodology linking data and outcome; this study has provided evidence that SHs recognise the map as an appropriate model for this function. SHs recognise the map as a tool that enables users to identify correlations in data, specifically generating conversation regarding visual trends. This functionality to visualise trends in data is another approach to quantifying the impact of factors on public health similarly to IMD, as well as other examples of GIS technology use discussed in the literature [
16‐
18,
42]. Marmot et al [
43] emphasise the implications of growing health inequality in the UK. Therefore, tackling the impact of SES disparity on health will be both a challenge and a priority in the future of public health service provision.
One correlation of interest to participants, deduced from case study 2, was the abundance of civil service organisations dispersed across Croydon. A range of support organisations were mapped including those targeting alcohol and substance misuse, disability and specialist epilepsy care for young people. Case study 2 demonstrated how GP surgeries and pharmacies are greatly outnumbered by civil service organisations with 819 recorded at the time of this study. In the advent of social prescribing, the researchers question whether such civil organisations would be suited to support patients post-discharge to alleviate the pressures on primary care. Evidence suggests that SES has a significant impact on readmission, hence integrating health and social care when discharge planning may prove to be beneficial [
44,
45]. However, numerous barriers including a lack of infrastructure, poor implementation and a lack of funding persist [
46]. Developing methods, such as the mapping model, may serve to alleviate some of these pressure through the utilisation, streamlining and distribution of data among SHs when service planning, such as in case study 2.
This study has demonstrated the usability of an accessible and low-resource intensive mapping model as a method of data visualisation. More work is needed to determine the impact of such a model in areas such as commissioning and social prescribing. Limitations for the use of data visualisation can be categorised into three main areas including poor interpretation of data, lack of understanding for the role of this methodology as well as accessibility, as described by Caroll et al [
16]. This study attempted to offset this by including a diverse study population with five distinct SH groups, and an interactive mapping model element to support data interpretation across the three case studies. Additional focus groups to examine user-experience may have benefitted the analysis of the model’s usability and acceptability, hence this was a limitation for the study. This study had a small recruitment sample, therefore quantitative results should be treated cautiously with further work required, however content analysis identified saturation of themes throughout the interviews.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.