The prevalence of End Stage Renal Disease (ESRD) is increasing in Iran and across the world (Mousavi et al., 2014). The total number of patients with ESRD condition was 27,000 in Iran in 2015 (Mousavi et al., 2014). Haemodialysis is the main treatment for this condition. This machine-operated process removes fluid and waste from the blood and rectifies the electrolyte imbalance (Hay, 1995). Most patients have to be dialysed three times week, making it necessary for them to travel to a haemodialysis facility between 140 and 160 times per year (Stephens et al., 2013). It has already been proven that poor access to haemodialysis facilities for these patients is associated with poor health outcomes, such as high mortality and morbidity rates (Moist et al., 2008). Access to health services is defined as the ease with which the services can be used by people whenever and wherever needed (McLafferty, 2003). It has five main dimensions: accessibility, availability, accommodation, affordability and acceptability (Penchansky and Thomas, 1981). The accessibility and availability dimensions are usually related to geographical factors; hence they are spatial dimensions (Mao and Nekorchuk, 2013). Affordability, accommodation and acceptability, which are independent of geographical conditions, are the non-spatial dimensions (Guagliardo, 2004). Additionally, access can be divided into two broad categories: Potential access, which is simply defined as the presence of enabling resources, and Revealed access, which is the actual use of services (Andersen, 1995).
In recent studies, accessibility to health-care services is commonly measured by driving time between the patient's residence and the closest health-care facility (Matsumoto et al., 2013; Stephens et al., 2013; Casas et al., 2016). Revealed access differs from potential access in a number of ways. For example, haemodialysis patients may have a short driving time to access the nearest haemodialysis facility, but they may decide to go to another centre because of ethnic disparities (Saunders et al., 2014), resource availability (Just et al., 2008), flexibility in planning dialysis time, sense of security, physical space, noise, pre-dialysis education or involvement in the choice of modality (Lee et al., 2008). Therefore, if the nearest haemodialysis facility is the preferred haemodialysis facility (is assumed), the potential access should be measured (Miller et al., 2014). In addition, previously published methods in estimating the driving time between patient residence location to the haemodialysis services may not show the Actual Access Time (AAT) precisely due to other confounding factors, e.g., transportation mode (Neutens, 2015), car ownership (Lovett et al., 2002) and income (Wang and Luo, 2005). Therefore, a proper method of measuring AAT as revealed access measurement to the haemodialysis services is needed.
To compensate for this knowledge gap, this study aimed to measure an AAT index as proxy for revealed accessibility of haemodialysis patients to haemodialysis facilities as well as the association between AAT and other factors that may influence accessibility.
Materials and Methods
In this research, revealed access was measured and self-reported AAT to haemodialysis facilities was used as the revealed access measurement.
Study region and data
This cross-sectional study of haemodialysis patients was carried out in northern Khorasan Province of Iran with an estimated population of 919,000 in 2015 (Wikipedia, 2016). This paper is the continuation of a recently published paper (Kiani et al., 2017). The study area is displayed on a point density map (Figure 1). Two hundred and three patients were under haemodialysis in the area but only 168 of them met both inclusion criteria and were willing to participate in the study.
Ethics approval and consent to participate
This study was approved under the code IR.MUMS.RES. 1393.756 by the Ethical Committee of Mashhad University of Medical Sciences in Mashhad, Khorasan Province, Iran. All patients completed a consent form for participating in the study.
We included only patients meeting all the following criteria: i) the patient had agreed to participate in an interview; ii) the patient had to be dialysed at least one time a week; iii) the patient had to be at least two months under haemodialysis; iv) the patient or his/her caregiver was able to communicate with researchers.
We developed and used the census sampling method with structured questions for data collection. The validity of the questions was confirmed by two native experts (a medical information specialist and a nephrologist). According to literature review and expert opinion, the following variables were chosen for the collection of patient data: Gender (Roderick et al., 1999); Age (Rodriguez et al., 2013); Residence (Rodriguez et al., 2013); Occupation (Rodriguez et al., 2013); Education (Rodriguez et al., 2013; Calice-Silva et al., 2015); Income (Rodriguez et al., 2013); Ethnicity (Rodriguez et al., 2013; Saunders et al., 2014); Transportation mode (Murray, 2008; Prakash et al., 2010); Driving time (White et al., 2006; Matsumoto et al., 2013; Stephens et al., 2013); Driving distance (Stephens et al., 2013); Car ownership; Patient's caregiver status.
The patients were asked to give AAT and other information regarding travel to the haemodialysis facilitiy of their choice. Driving time and distance from home to the referred haemodialysis facility were calculated using Google Map for each patient. When more than one path was offered by GoogleMaps, the shortest route was chosen as it is known as potential access. For patients travelling to haemodialysis facilities with their own vehicles, fuel was included in the travel cost. The estimated fuel was 10 litres per hundred km for cars and three litres for motorcycles. Since there was a measurement bias in self-reported income and there is no validated index for Iranian household income, the Townsend deprivation index (Townsend, 1987) adjusted by local experts, was used for estimating patient income levels (Table 1).
The distribution of variables was skewed even after log transformation. Therefore, we used median and inter quartile range (IQR) instead of mean and standard deviation (SD) for descriptive analyses. Moreover, non-parametric statistical tests were used for inferential statistics.
Data for patients living in rural and urban areas without any haemodialysis facility (Group A) and those living urban areas with haemodialysis facilities (Group B) were analysed separately. The effect of all variables on AAT, as a response variable for both Groups A and B was tested. The Man Witney test (Nachar, 2008) for examining the effect of binary variables on AAT and the Kruskal-Wallis test (Elliott and Hynan, 2011) were used for determining the effect of other nominal variables on AAT. The linear, quadratic, and cubic associations were examined to determine the relationship between AAT and other numerical variables, then the best model (with the maximum of R2) was chosen.
All significant univariate analysis variables were entered into the univariate general linear model (GLM) to identify factors associated with AAT. The GLM for Group A patients was developed by including AAT as a dependent variable and two covariates: distance and driving time, while car ownership, education level, income level, sex, ethnicity, patient caregiver status and transportation mode comprised fixed factors. The univariate GLM for Group B patients was partly different and included car ownership, education level, income level, sex, ethnicity, patient caregiver status and transportation mode as fixed factors. Naturally, the AAT was also the dependent variable for Group B.
All these analyses were performed twice. First, the AAT to haemodialysis facilities, followed by the Actual Return Time (ART) from the haemodialysis facility to the patient's home. The significant threshold for all calculations was set at 0.05. We used the SPSS Version 16 for statistical analysis, ArcGIS v. 9.3, and Google Map for spatial analyses.
To visualise the difference between AAT as a revealed accessibility measurement and driving time as a potential accessibility measurement, both these values were interpolated. Spatial interpolation is the process of using points with known values to estimate values at other, unknown points. We used the Inverse Distance Weighting algorithm for interpolation (Jia et al., 2016).
The general characteristics of sample populations are shown in Tables 2 and 3. According to the former, some patients had changed their home to get better access to haemodialysis (one in Group A and 12 in Group B).
Driving time and the AAT score are summarised by a box-plot in Figure 2. This figure shows a doubling of the AAT score compared to the driving time in groups A and B. Further, it took patients in Group A three times longer than patients in group B to reach their assigned haemodialysis facilities. The Mann-Whitney U test showed that there was a statistically significant difference in the AAT score between the groups, Man-Whitney U=940.000, P=0.000, with a mean rank AAT score of 115.00 for Group A and 57.00 for Group B.
Factors influencing Actual Access Time
The GLM analysis of group A showed that driving distance and driving time as spatial factors with sex, income level, existing patient's caregiver, transportation mode, education level, ethnicity, and personal vehicle ownership as the non-spatial factors affecting the AAT (Table 4); moreover, the results of the model showed that the identified factors determined the AAT in 90% of the cases (R2=0.906). In other words, 90% of changes in the response variable (AAT) in Group A could be explained by these factors.
The GLM analysis of group B showed that the non-spatial factors were the same as those identified for Group A (Table 5); furthermore, the results of the model showed that the identified factors determined the AAT in 95% of the cases (R2=0.951). In other words, 95% of changes in the response variable (AAT) in Group B could be explained by these factors. Table 5 shows that none of the spatial factors (driving distance and driving time) were identified as factors affecting the AAT.
Comparison between Actual Access Time and Driving Time
AAT and driving time were compared to visualise the accessibility to haemodialysis facilities. Figure 3 shows the interpolated AAT as a revealed accessibility index and the interpolated driving time as a potential accessibility index to visualise the access to haemodialysis facilities at the regional level.
Cities in northern Khorasan Province are small and do not have heavy traffic. Our study showed that spatial factors are not the main determinants of health care access in urban area with having haemodialysis services. However, for the Group A patients, the spatial factors had an effect on AAT because these patients had to spend a longer time travelling to the health care facility. However, this is an contended issue, Thompson et al. (2012) for example, have shown that distance was also an important factor for urban haemodialysis patient travel, while Smith et al. (1985) report that travel time to the place of treatment is a relatively unimportant aspect of the care of haemodialysis patients in metropolitan areas. It seems that the association between spatial factors and the AAT varies in urban areas and depend on other factors such as traffic and the area of the city. Therefore, different studies may have different results and policymakers should consider these issues when planning and/or modifying haemodialysis care to unmet areas in both urban and rural areas. Some earlier studies, for example Salgado et al. (2011) and Faruque et al. (2012), worked on some models to minimise inequities in haemodialysis facilities distribution through finding new facility locations. The work in our study area presented here also emphasises the need for such studies.
This study identified that non-spatial factors affecting the AAT for patients in groups A and B are the same. The fact that females in Group A had a higher AAT score than males may be related to the fact that Iranian women are less able to commute, especially when travelling from a remote area to the city. Dialysis policymakers can plan for women who come from remote or rural areas to haemodialysis facilities by providing transportation services. Additionally, such a service can reduce the negative impact of other factors on the AAT score, and a better transportation system may eliminate the income effect, transportation mode and car ownership. It should be investigated if this approach would be a way to reduce the AAT score for haemodialysis patients and if so, whether it would be cost-efficient.
To the best our knowledge, this is the first study in Iran using the AAT for identifying accessibility to haemodialysis facilities. Moist et al. (2008), in contrast to our study, showed that employment status, family support and age correlated with self-reported one-way travel time. However, like our study, Saunders et al. (2014) showed that proximity to haemodialysis facility did not alone indicate access as other factors, e.g., ethnicity, could also affect access. Matsumoto et al. (2013) used driving time, Salgado et al. (2011) driving distance and Saunders et al. (2014) Euclidean distance (the straight-line distance between two points) for measuring accessibility to haemodialysis facilities. Our study, however, showed that each of these measures used alone would not be perfect for examining revealed access to haemodialysis facilities because there are many other non-spatial factors affecting the AAT in rural and urban areas (Tables 4 and 5). For example, patients who live near a haemodialysis facility may have a high AAT score because they do not have access to a taxi or a private car to reach the haemodialysis facility, and this would be because of a poor public transport system or financial problems. Therefore, the AAT compared to driving time is a more comprehensive measurement for health policymakers to target interventions to correctly assign people and places (Figure 3). As both Figures 2 and 3 show, the AAT (as a revealed accessibility index) and driving time (as a potential accessibility index) are different. Additionally, Figure 2 shows that the AAT index determines the outlier patients better than driving time does, so the AAT index can be used for the planning for special patients with special problems. Health policymakers could first identify factors affecting the AAT and then plan to decrease this index because increased access time to haemodialysis facilities could have clinical implications with increased mortality and decreased quality of life (Stephens et al., 2013). According to the Renal Association Clinical Practice Guideline on Hemodialysis, patients in remote areas should have access to haemodialysis facilities as well as be able to return back to their home after finishing treatment within 30 minutes (Mactier et al., 2011). In our research, 75% of patients in group A could not access haemodialysis facilities in less than 30 minutes. However, with proper planning, it could be reduced to 25% (Figure 2).
Unexpectedly, patients in group B, changed their home more often than group A to gain better access to a haemodialysis facility. It seems this is not common in rural areas due to financial problems. Indeed, financial capability is an important non-spatial factor that can affect the AAT to a haemodialysis facility. In this study, the patients with a high income level had a better AAT score. Tshamba et al. (2014) emphasised that the economic status is associated with increased risk of death among haemodialysis patients. Table 3 shows that the total average round-trip cost was 7,000 Toman (about 1.8 USD) for group A and 2,576 Toman (around 0.68 USD) for group B of patients. Although this cost is low in comparison to the full cost for the haemodialysis session (190,000 Toman or close 50 USD), this is not seen by the patients as this cost is fully subsidised by the government. Considering the comparatively low transportation cost, it could perhaps be possible to subsidise that too. This study developed a comprehensive index of revealed accessibility to haemodialysis facilities including spatial and nonspatial factors into one framework. However, the results of this study can only be generalised for regions similar to northern Khorasan. Considering chronic conditions, especially for patients with ESRD, the access time to health care facilities and return time travelling home are both important because of the many trips needed to receive the needed health service. Therefore, the ART was also examined. As there was a high correlation between AAT and ART, the policymakers in the health sector should be able to use one of them to improve accessibility to haemodialysis facilities.
The factors affecting revealed accessibility to haemodialysis facilities are different when rural and urban patients are compared. Policymakers could make proper decisions by analysing the AAT determinants as a revealed accessibility measurement; however, separate analyses are needed for rural and urban patients. Driving time in both urban and rural areas is not a good proxy for measuring access to haemodialysis facilities used alone, as there are many other reasons connected with transportation, for example, women's travel problems, ethnicity disparities, patients' education level, caregivers' status, or low income. As the AAT and ART indices correlate well, policymakers could use any one of these indices to measure revealed accessibility.