Although equitable access to healthcare services (AHS) is a leading public health priority, the importance of AHS and the attention received from the policy managers differ from one disease to another. Access to dialysis services (ADS) is a crucial necessity for patients who have to travel to a dialysis facility three times a week (Mactier, 2007; Stephens et al., 2013). Thus, poorly developed ADS lead to poor health outcomes such as increased morbidity and mortality (Moist et al., 2008; Diamant et al., 2010; Rucker et al., 2011; Thompson et al., 2012).
A vital aspect of AHS is the ease with which they can be accessed and used when needed (McLafferty, 2003). AHS has five main dimensions: accessibility, availability, accommodation, affordability and acceptability (Levesque et al., 2013; Russell et al., 2013; Saurman, 2016). The accessibility and availability dimensions are usually related to geographical factors and are therefore labelled spatial accessibility (Mao and Nekorchuk, 2013), while non-geographic-dependent dimensions such as affordability, accommodation and acceptability are considered non-spatial (Guagliardo, 2004). Additionally, AHS can be divided into two broad categories: potential AHS and realized AHS. The former is simply defined as the presence of enabling healthcare, while the latter is the actual use of its services (Andersen, 1995).
Geographic information systems (GISs) enable researchers in the healthcare field to apply combined spatial accessibility measures for the inspection of the equitability of resource allocation. Various studies have demonstrated the application of GIS tools which are useful for calculation and visualization of accessibility scores (Guagliardo, 2004; Yang et al., 2006; Matsumoto et al., 2012; Kiani et al., 2017). Stephens et al. (2013) performed a GISbased measure of ADS which included travel impedance to dialysis facilities as an indicator. A recent study performed in Iran by Kiani et al. (2017) revealed the importance of spatial accessibility to dialysis services and showed that this variable is strongly underestimated when potential travel time is calculated (Kiani et al., 2017). They developed a comprehensive measure of revealed accessibility that includes travel time and some other spatial and also non-spatial factors into one indicator framework. However, their index does not include facility capacity. Matsumoto et al. (2012) designed an algorithm embedded in a GIS-based measure of ADS which demonstrates that patients cannot always be accepted by the nearest dialysis facility due to limited capacity. Another study performed in the US emphasized the acceptability dimension and revealed that dialysis patients may have a short travel time to one dialysis facility but might decide to go to another centre due to ethnic disparity (Saunders et al., 2014). Some studies, mainly conducted in Britain (Roderick et al., 1999; Christie et al., 2005; White et al., 2006; Judge et al., 2012), calculated deprivation as an important ADS indicator which includes some major non-spatial factors. Deprivation can also potentially affect need of dialysis services (NDS) (Roderick et al., 1999; Thomas, 2005; Yang et al., 2006; Judge et al., 2012). NDS is defined as the number of patients in an area who need dialysis (Yang et al., 2006).
The diversity of GIS-based ADS measures include various indicators with spatial as well as non-spatial confounding factors that may confuse researchers and policymakers. A mapping exercise could potentially lead to a more conclusive index producing better scores than currently used. GIS-based methods are inherently spatial; some of them, such as the two-step floating catchment area (2SFCA), demonstrated the capability of integrating both spatial and non-spatial AHS factors into one framework (Bagheri et al., 2008; McGrail and Humphreys, 2009; McGrail and Humphreys, 2015). Yang et al. (2006) provide a 2SFCA platform that integrates only dialysis patients and dialysis machines within a 30-minutes potential travel time catchment. Although this represents a certain progress, to the best of our knowledge, no study so far has made an attempt to develop a truly integrated GIS-based ADS index. In an effort to do so, we decided to focus on the gaps in the current approach to GIS-based ADS measures through a systematic review of the available literature. In addition, we aimed to develop recommendations and a list of factors affecting ADS that would improve the research on the measuring ADS based on GIS.
Materials and Methods
A systematic mapping review of the available GIS-based literature was performed. It aimed to describe the extent of the study on a particular topic and to identify knowledge gaps in the study base, where further primary and/or secondary studies are needed (Grant and Booth, 2009).
The scientific literature was explored with regard to relevant communications on ADS and the use of GIS. It was done in May 2016 and the information collected updated in October 2017. We included the following electronic databases: PubMed, Web of Science, Scopus, ScienceDirect, EMBASE and IEEE Xplore. Initially, the databases were systematically searched using a variation of access concept (access, accessibility, availability, affordability, acceptability, accommodation, utilization, deprivation, disparity and equity) in connection with spatial terms (geographic information, GIS, geomapping, location-allocation, and spatial analysis) and dialysis (dialysis, haemodialysis, renal and kidney). To combine the search terms within each category, we utilized the disjunction OR, and to combine categories we utilized the conjunction AND.
In order to identify additional relevant information related to ADS and GIS, the reference lists obtained were surveyed manually. To boost our search strategy further, we also looked for the so called gray literature, i.e. reports, standards, manuals and guidelines on the topic using general search engines such as Google. No date or study design limitation was imposed in any of the research steps described. The complete search strategy is available upon request.
After the literature search had been completed, the EndNote X5 (Thomson Reuters, New York, NY, USA) reference management software was applied to aggregate all search returns. The articles were then screened and each study assessed independently for eligibility by two of the authors in different combinations. A study was considered eligible for inclusion if it included assessment of ADS as the primary or secondary outcome, while it was excluded if i) it had not clearly reported and calculated ADS indicators; ii) it lacked a methodological description of the measurement of ADS or its indicators; or iii) it consisted of a letter to the editor, an editorial, general comments, a position paper or it was an unstructured paper.
The full text of the qualified studies was read, tagged and summarized by one author and verified by one other author. A brief flow diagram of the strategy is depicted in Figure 1.
Studies deemed eligible for review underwent data extraction. For each paper, essential data items related to ADS measurement were extracted and fitted into a form with a choice of headings, such as ADS indicator; Factor(s) affecting the indicator; Method of measurement; Primary outcome measures; Secondary outcome measures; and Study design. Additional properties, such as conclusive comments and suggested measurement intervals, were also recorded when available and applicable.
Since systematic mapping reviews mainly aim to describe the state of the art of a particular topic, it is desirable to include communication types of a range as wide as possible. Due to the high diversity of study types in our review, this necessitated an informal quality assessment that was performed by classifying the literature by type of study. We used a mixed-approach scoring system as applied by Azizi et al. (2016) under similar circumstances (Azizi et al., 2016). Quality scores were assigned by two authors and verified by a third author. The weight of the literature was assigned according to their study design by the quality scoring system. In this approach, papers such as reviews and randomized controlled trials (RCTs) achieved the highest score (score 4) while score one was the lowest score. In this approach, the gray literature was given score one as formal or expert consensus regarding quality score. The summary of the quality assessment approach is outlined in Table 1.
Out of 1119 literature items collected, 76 were deemed eligible for further full-text review. After reviewing the full-text studies for final content match, 36 were selected for the review. Further details pertaining to the included studies are shown in Figure 1.
Characteristics of included studies
Three groups of studies were identified. The first group included eight cross-sectional studies addressing ADS and treating it as a primary outcome. The focus of these communications was the design of a GIS-based model intended for gauging the degree of ADS. They mostly calculated potential travel time (Roderick et al., 1999; Christie et al., 2005; White et al., 2006; Yang et al., 2006; Judge et al., 2012; Matsumoto et al., 2012; Stephens et al., 2013) or facility capacity (White et al., 2006; Yang et al., 2006; Matsumoto et al., 2012) as the key indicators of spatial accessibility. Some papers in this group also considered deprivation (Roderick et al., 1999; Christie et al., 2005; White et al., 2006; Judge et al., 2012) as an important, non-spatial indicator of ADS. One study designed a measure to calculate actual travel time revealed significant effects on the travel time of other non-spatial factors such as sex, income level, caregivers, transportation mode, education level, ethnicity and personal vehicle ownership. It demonstrated the large gap between potential travel time and actual travel time (Kiani et al., 2017). Another study showed that ignoring facility capacity and accounting only for travel time when evaluating ADS may result in misleading conclusions (Matsumoto et al., 2012). Further details pertaining to this group are outlined in Table 2.
The second group of studies (10 peer-reviewed papers) considered gauging ADS as a secondary outcome, while their primary outcome measures focused on the association between ADS and health-related outcomes (Tonelli et al., 2007; Moist et al., 2008; Diamant et al., 2010; Rucker et al., 2011; Thompson et al., 2012; Miller et al., 2014) determining the relationship between ADS and prevalence rates of dialysis patients (Kashima et al., 2012) or designing models to locate dialysis facilities (Ayyalasomayajula et al., 2011; Salgado et al., 2011; Faruque et al., 2012). Almost all of them calculated potential travel impedance as an indicator of ADS based on GIS.
Finally, the third group of 16 studies considered factors affecting ADS. Among them, were ten peer-reviewed articles (Smith et al., 1985; Tonelli et al., 2006; Hall et al., 2008; Prakash et al., 2010; Matsumoto et al., 2013; Omrani-Khoo et al., 2013; Rodriguez et al., 2013; Plantinga et al., 2014; Saunders et al., 2014; Kiani et al., 2017), one case series (Tshamba et al., 2014), one proceeding (Richard et al., 2009), and four gray literature items (Maheswaran et al., 2003; Mactier, 2007; Levinson, 2011; Amy Martin, 2013).
Considering the contents of the total number of GIS-based studies discussed here and the literature extracted from their comprehensive reference lists, 41 factors affecting ADS were determined (Table 3). Travel impedance, especially travel time, were the main indicators of spatial accessibility, the rest were mostly non-spatial. Some studies emphasize that factors such as ethnicity or the patient’s health insurance status also affect ADS (Kashima et al., 2012; Thompson et al., 2012; Saunders et al., 2014), while other studies imply that they do not affect ADS (White et al., 2006; Matsumoto et al., 2012). The factors we found in our literature search and their frequency are given in Table 3.
Our systematic mapping review of the evidence revealed that current GIS-based measures of ADS tend to calculate potential ADS instead of a realized one (Table 2). Also, we found no study including both spatial and non-spatial dimensions of ADS into one framework that could produce a more realistic score than current attempts in this direction. However, in a recent study performed in Iran, Kiani et al. (2017) developed an integrated measure of ADS by calculating a reasonable measure of actual travel time in contrast to previous reports that mainly focus on estimated potential travel time (Christie et al., 2005; White et al., 2006; Yang et al., 2006; Judge et al., 2012; Matsumoto et al., 2012; Stephens et al., 2013) However, the results are still far from a truly realized ADS index since the key ADS indicator of facility capacity generally remains ignored.
Matsumoto et al. (2012) compared two GIS-based ADS measures, the capacity-distance model and the distance model, and found the former to be more realistic than the latter. In the capacity- distance model, which addresses both travel time and facility capacity, patients were forced to travel further due to capacity limitations at closer centres. Furthermore, an American study noted that proximity to dialysis services does not directly translate into access owing to potential racial segregation (Saunders et al., 2014). However, both these studies calculated potential travel time. Meanwhile, many British studies (Roderick et al., 1999; Christie et al., 2005; White et al., 2006; Judge et al., 2012) emphasize the importance of non-spatial factors, e.g., Kiani’s et al. study (2017), considering deprivation as one of the main ADS indicators. Indeed, controversies in this field show a gap between calculated and real ADS. Although this study did not find any research integrating spatial and non-spatial factors into a GIS-embedded model for measuring ADS, various GIS-oriented studies join some spatial ADS indicators together to measure spatial accessibility (Yang et al., 2006; Matsumoto et al., 2012). For instance, Yang et al. compared two GIS-based methods, the 2SFCA approach and the kernel density method, in a case study on renal dialysis facilities in Chicago, USA. In this study, based on the main spatial ADS indicators travel time and facility capacity, the 2SFCA method produced better accessibility ratios overall (Yang et al., 2006). Although, this work only integrated spatial ADS dimensions, it has provided a platform that has been successfully used in other contexts, especially in primary healthcare where an overall measure including both spatial and non-spatial factors of access is included (Wang and Luo, 2005; Bagheri et al., 2008; McGrail and Humphreys, 2009; McGrail and Humphreys, 2015). For example, Bagheri et al. (2008) developed an integrated access to primary healthcare (APH) index which combined spatial accessibility, calculated by 2SFCA method, and a need index into one framework, while McGrail and Humphreys (2009, 2015) improved the 2SFCA method by introducing a concept based on three key components, i.e. spatial accessibility, population health needs and mobility.
The progress covered by this review, as well as the gaps revealed, raises the hypothesis that an integrated ADS index should calculate access more realistically than current GIS-based measures. However, further research examining whether this hypothesis is correct or not is needed. This mapping review provides some evidence-based recommendations that may help researchers and policymakers perform a primary study assessing this hypothesis. Three components of the 2SFCA platform should be addressed in order to construct an integrated ADS index, i.e. spatial accessibility, mobility and NDS.
Spatial accessibility must take into account both accessibility and availability. Travel time, discussed by most current studies, should be calculated as an indicator of accessibility with a threshold of 30 minutes as haemodialysis guidelines recommended (Mactier, 2007). However, actual travel time as proposed by Kiani et al. (Kiani et al., 2017) should be used together with facility capacity, expressed as the number of dialysis machines (supply) to the number of patients (demand) in each facility (Yang et al., 2006; Matsumoto et al., 2012) that seems to be the key availability indicator. It is as important as the travel time and should be incorporated into the 2SFCA framework with an appropriate threshold to construct catchment areas more realistically. Patients need dialysis thrice a week according to current haemodialysis guidelines (Mactier, 2007), which means that each machine can serve up to four patients per week (two patients on even days and two patients on odd days) leading to a supply-to-demand ratio threshold of .. If, regardless of this threshold, all catchments are constructed with a radius of 30 minutes travel time, some of them might include patients more than their facility capacity.
Mobility is defined as the population’s ability to overcome distance (Bisht et al., 2010). Taking the relative population size of those aged either <18 years or >75 years as the measure of reduced personal mobility, McGrail et al. (McGrail and Humphreys, 2009) point out that the three indicators, i.e. households without a car, individuals with reduced mobility and public transport availability, measure different aspects of mobility and that correlations between them are small. Considering this nature of mobility, it seems useful to apply these three indicators when calculating mobility in the ADS context.
Revealed NDS (the actual demand) is most likely to be the same as the potential NDS when the current ADS is calculated (Yang et al., 2006), but future ADS calculations will differ because the number of patients then is unclear, in particular as the prevalence of end-stage renal disease (ESRD) is increasing (Chadban et al., 2003; Eggers, 2011). An easy way to estimate NDS is by multiplying the annual growth rate of demand at current demand (White et al., 2006; Yang et al., 2006). But an adjustment is needed if there is an increase in diabetes which is already anticipated (Roderick et al., 1999; Yang et al., 2006) and hypertension (Roderick et al., 1999; Yang et al., 2006), the two principal causes of ESRD. In addition, deprivation (Roderick et al., 1999; Thomas, 2005; Judge et al., 2012), age (Roderick et al., 1999; Thomas, 2005), gender (Roderick et al., 1999; Yang et al., 2006) and ethnicity (Roderick et al., 1996; Judge et al., 2012) may influence the NDS differently in the future.
To our knowledge, this is the first systematic mapping review of the available literature aimed at identifying gaps in current GISbased ADS measures and developing evidence-based recommendations. However, by limiting the search strategy by specifying it for the GIS category and developing a list of factors affecting ADS elicited, we may have lost evidence in studies exclusively focused on exploring factors affecting ADS. This would be true, even though we had a comprehensive list referencing the literature elicited from our systematic mapping review. However, this was secondary outcome measure of the study, and we suggest a systematic review with a wider scope that could list all factors affecting ADF. Although we did not systematically focus on seeking factors affecting NDS, we think that our findings related to estimating NDS are appropriate and enough. Moreover, we found that hospitalization rate (Rucker et al., 2011) and mortality rate (Rucker et al., 2011; Thompson et al., 2012) are negatively associated with ADS, facts that can be used for validation of the integrated index of ADS, a proper validation of this index remains open for future study. Finally, even though we highlighted the absence of evidence- based recommendations incorporating indicators related to the acceptability dimension in an integrated ADS index, we could not alleviate this weakness of current GIS-based measures of ADS, which needs further research.
Current GIS-based measures of ADS tend to calculate potential ADS instead of a realized one and there is a need to examine whether an integrated index of ADS can calculate a realistic score. Listed factors affecting ADS are mainly non-spatial encouraging the design of an integrated ADS index produce better ADS score than those currently advocated. The mapping review strongly suggests exploring the hypothesis that a combined index of ADS including most dimensions of ADS can be developed and produce a better ADS score than current available. The 2SFCA method may be capable of providing a platform for this aim as the study recommended and researchers and policymakers are encouraged to examine and validate this hypothesis.