HIV risk and the association with accessibility coverage to medical facilities and socioeconomic status among heroin users in Kaohsiung, Taiwan (2011‐2015): A GIS approach

Abstract

Illegal drug use has caused considerable health and social costs. Access to medical treatment has been one of the most cost‐effective interventions to reduce problematic drug use. Accessibility to treatment, which can be analyzed by the spatial approach such as geographic information system (GIS), has thus been utilized as a measure of the proportion of population that reaches appropriate health services. In this study, the association of accessibility coverage to medical facilities and socioeconomic status with human immunodeficiency virus (HIV) risk for heroin users was evaluated by GIS analysis in Kaohsiung, Taiwan. Data of 7890 heroin users were collected from 2011 to 2015 and categorized into five risk groups according to their income and distance to treatment settings. The results of this GIS‐based analysis show that the areas with over 50% accessibility coverage rate had less amounts of HIV cases with statistical significance. Inconvenient access to medical facilities could result in poor treatment outcomes such as higher HIV incidences. Therefore, in order to effectively reduce HIV incidences among the heroin users, the accessibility coverage and locations of medical treatment facilities should meet the needs of drug users. It is also advised that the formulation of policies associated with illegal drug use problems should be evidence‐based and geographical indicators could serve for this purpose from either prevention or intervention perspective.

1. INTRODUCTION

Illegal drug use has caused considerable health and social costs. Among all illegal drug use‐related health burdens, transmission of HIV by injecting drug use has been a global threat for more than 30 years. There were approximately 33.3 and 36.7 million people living with HIV globally in 2010 and 2015, respectively, of whom 4.7 and 5.1 million people in 2010 and 2015, respectively in Asia and the Pacific, and about 300 000 heroin users were newly infected in 2015 in Asia and the Pacific.1, 2 In Asia, injection drug use was supposed as a major risk factor of HIV transmission that resulted in the escalating situation of HIV infection.3

In Taiwan, heroin and methamphetamine have been the predominant illicit drugs since 1990s, with an increasing trend of illegal ketamine use.2, 4 Since 2005, due to the escalating HIV incidences among the injection drug users, the national harm reduction program (HRP) was implemented.5 The measures of HRP mainly included methadone maintenance treatment program (MMTP) and needle exchange program (NEP).6, 7 We have previously shown that education and HIV testing were essential for the effective HIV control among injecting heroin users.5 However, the factors contributing to such outcomes other than education remain to be scrutinized, especially at the local levels.8 The accessibility in the geographical dimensions is usually analyzed by the spatial approach such as geographic information system (GIS). In a previous research, addiction treatment was regarded as one of the most cost‐effective interventions to reduce problematic substance use,9 and treatment completion in therapeutic communities was the most important factor in reducing illegal drug use.10 They also found one variable correlate to treatment retention was the distance traveled to the medical facilities.10

In Taiwan, although the nationwide HRP has been shown to curb the spread of HIV incidences, it is not known if accessibility coverage of medical facilities would affect the outcomes of drug users who seek treatment, especially those with higher risks of contracting HIV. With such a GIS approach performed in Kaohsiung city, we hope that the formulation of local drug use‐associated policy could be more solid and evidence‐based.

2. METHODS

2.1. Data Source and Collection

In Taiwan, Kaohsiung city is by area the largest municipality at 2951.85 km², and the second most populous (by city proper) with a population of approximately 2.77 million. Kaohsiung city is the political, economic, transportation, manufacturing, refining, shipbuilding, and industrial center of southern Taiwan. In this study, Kaohsiung city was chosen because its drug use‐related data systems have been well established and it ranked the third highest cumulative HIV infection cases among 22 cities/counties.

There are 38 districts in Kaohsiung city, and the geographical mapping layers information, including country village boundary maps, was obtained from Taiwan Geospatial One‐Stop Portal, Information Center, Ministry of the Interior (TGOS, http://tgos.nat.gov.tw/tgos/web/tgos_home.aspx). Illegal drug‐related data from 2011 to 2015 including (a) heroin users and medical facilities' addresses, (b) drug types, (c) criminal patterns (d) arrested data, and (e) HIV cases were obtained from incumbent agencies of Kaohsiung city government such as bureaus of health, police, education, and social welfare. The demographic data were retrieved from Department of Statistics, Ministry of the Interior, and Civil Affairs Bureau, Kaohsiung city government. The data of age and low‐income households were retrieved from Department of Budget, Accounting and Statistics, Kaohsiung city government.

2.2. Data management

In order to have more accurate evaluation, preprocessing was performed before analysis:

1. The first step was to exclude the unclear, missing, or incomplete residence addresses such as address information with no house number from original data. If the data with the street name and house number, but no district name, they were checked and corrected manually. The heroin users' data (n = 8639) from 2011 to 2015 were collected from various departments of Kaohsiung city government and the data of resident addresses not located in Kaohsiung city or incomplete, and criminal patterns without taking heroin (n = 749) were also excluded. As a result, a total of 7890 heroin users’ data were included for analysis.

2. The second step was to transform address of residence and medical facilities (including methadone treatment facilities and NEP spots) into longitude and latitude coordinates. The address data were uploaded to the TGOS website, and then transformed into X, Y projected coordinates. Finally, we input these coordinates information via GIS to plot on the map layers

2.3. Geographic Information System

Geographic accessibility emphasizes the degree of isolation in geography of medical services providers and demand clients.11 According to the World Health Organization (WHO), accessibility can be defined as “a measure of the proportion of the population that reaches appropriate health services” (WHO, Regional Office for Europe 1998). There have been many reports analyzed the geographic accessibility of medical services in different regions.8, 12, 13 It was found that the distance was an obvious impediment to access to health services in regions with fewer populations.14 However, access to medical or social services was a multidimensional issue between patients and the service delivery system.15 The locations and distributions of treatment facilities in geography could affect individuals’ willingness to participate in treatment programs and their retention rate.16 A 2004 study identified four factors that could influence individuals' willingness to retain in treatment, namely, availability of drugs, neighborhood disadvantage, availability of community resources, and the influence of a particular location on individual's mental state or behavior.17, 18

GIS is a system utilized to capture, display, and analyze geographical data on the map. In this study, we adopted ArcGIS (version 10.4; Esri's Network Analyst, USA) as the tool for geographical analysis. GIS could use geocode to reveal more specific and detailed geography information, and help to verify the geographical assumptions about the allocation of medical services. Via GIS analysis, not only the distribution of treatment facilities could be illustrated but also the geographical accessibility for heroin users in need.19

Coverage assessment was identified as a helpful indicator to evaluate goals of health services by WHO, including availability coverage, accessibility coverage, acceptability coverage, contact coverage, and effective coverage.20 In 2008, a similar study used availability coverage and accessibility coverage to examine the benefit of methadone treatment services in Hong Kong.21 In this study, we adopted accessibility coverage to measure how approachable services for heroin users.

Accessibility coverage is illustrated as the percentage of heroin users covered within accessible regions in each district among all heroin users in that district as the formula below. Due to the amount of treatment facilities in each district varied fairly, the amount of heroin users covered needed to be adjusted and the average amount of heroin users covered by one treatment facility by district was calculated.21

$$\frac{\text{Amount of individuals covered within}\mathrm{a}\text{accessible region in}\mathrm{a}\text{district}}{\text{Total amount of individuals in that corresponding district}}\times 100\%$$

2.4. Ethics approval and consent to participate

The Institutional Review Board of Kaohsiung Medical University Hospital approved this study (IRB No: KUMHIRB‐SV[II]‐20150067, November 24, 2015). According to the approved document, the written consent of the participants was waived for the purpose of public policy evaluation by the government or commissioned organization carrying out official duties under the condition that the collected data have been managed so as not able to identify any individual, which conforms to the Personal Data Protection Law of 30 December 2015, Taiwan.

3. RESULTS

In order to learn the influence of environment and socioeconomic status, the population in 15 to 64 years, the amount of arrests related to drug use, and the percentage of low income were selected as indicators to illustrate the risk level in each district. The spatial analysis was thus carried out and the results are shown by district reclassification and shortest distance and accessibility coverage.

3.1. Spatial analysis

3.1.1. District reclassification

The risk and socioeconomic factors were taken into account by reclassifying the districts of Kaohsiung city in this study. According to the United Nations Office on Drugs and Crime (UNODC) report, the age of predominant population with drug problems was 15 to 64 years old. Based on the expected treatment outcome scale (ETOS),22 the factors, including proportion (percentage) of low‐income households and amount of arrests, were also evaluated. The average value of each indicator from 2011 to 2015 was calculated and then scored by quintile. In our study, we derived the data via ArcGIS. The average value was reclassified by a scoring system from 1 to 5 based on the quintile where 5 stood for maximum vulnerability.22 The population in 15 to 64 years old was the major drug user group, and then each area was reclassified with the score from 1 to 5 based on the average value of 15 to 64 years old people in a decreasing order. The score of each indicator was summarized in Table 1.

Table 1.

Risk factor score

Indicator	Quintile range	Score
Population in 15‐64 years	1387.00‐10 812.60	1
	10 812.61‐22 597.60	2
	22 597.61‐34 454.80	3
	34 454.81‐99 828.60	4
	99 828.61‐268 363.20	5
The amount of arrested	2.36‐36.34	1
	36.35‐50.19	2
	50.20‐60.87	3
	60.88‐81.63	4
	81.64‐424.76	5
The percentage of low income	0.79‐1.48	1
	1.49‐1.92	2
	1.93‐2.48	3
	2.49‐4.69	4
	4.70‐11.82	5

Table 2.

Accessibility coverage rate in each area

Area	The amount of HIV cases	The amount of MMTP	MMTP accessibility coverage rate	The amount of NEP	NEP accessibility coverage rate
Risk 1	229	0	0.000	6	72.152
Risk 2	1112	4	47.668	17	119.948
Risk 3	2557	6	41.970	29	107.933
Risk 4	1457	3	29.938	24	241.495
Risk 5	937	5	75.822	17	175.704

Thirty‐eight districts in Kaohsiung city were then reclassified into five areas based on risk factor scores (Table 1) after summing up each indicator as shown in Figure 1. Each area had various characteristic of demographic, social, and economic conditions. The risk 1 area stood an area with lower risk related to drug use, and the risk 5 area could have higher risk level.

Figure 1.

Reclassified areas by risk score

3.1.2. The shortest distance and accessibility coverage

Locations of heroin users' residence (N = 8267), treatment facilities (N = 18), and NEP spots (N = 93) were geocoded via TGOS and mapped by ArcGIS. Individuals living closer to the facility were considered to have greater accessibility to medical treatment than ones were farther from it. Therefore, the shortest distance between heroin users and medical facilities was calculated to measure whether the locations of medical facilities were barriers for access treatment in our previous study.23 We found that the drug users were distributed mainly in the metropolitan regions, and the minimum distances between medical sources and heroin users were around 1.15 km in Kaohsiung city.

Furthermore, based on the previous research,16 accessibility can be calculated by travel distance or cost time between individuals' location and medical facilities or Euclidean distance.21 We assumed that an individual's maximum walking time was 30 minutes, and then adopt 1.5 km as radius to create buffers for each treatment facilities to determine accessible region in our study.

3.2. Heroin users and medical facilities

A total of 8267 reported subjects with heroin use were included in our study, and the male‐to‐female ratio was around 6.4:1. There were 18 methadone treatment facilities and 93 needle and syringe exchange sites in Kaohsiung city. As illustrated in Figure 2, it was found that the allocation of medical treatment facilities was uneven especially in some areas there were even no MMTP. However, we tried to calculate the distances between heroin users and medical facilities to assess how long clients would take to access the treatment. The distance (Mean ± SD) between clients and methadone treatment facilities was 3.671 ± 3.731 km, and 1.319 ± 1.219 km between needle/syringe exchange sites. If we considered methadone treatment facilities and needle/syringe exchange sites simultaneously, the distance between clients and medical facilities was 1.251 ± 1.2 km. Although the results showed access to medical services for heroin users was convenient in Kaohsiung city, the accessibility coverage of MMTP seemed insufficient in some areas.

Figure 2.

The distribution of treatment facilities

3.3. Accessibility coverage and HIV

After the 1.5 km radius buffers were created for methadone treatment facilities and needle/syringe exchange sites, respectively (Figure 3A,B) to determine accessible region, heroin users geographically covered by medical treatment facilities were calculated in each district according to the formulation defined in this study. Because in some districts there may be more than one facility, and we assume one user accessed one facility, therefore the amount of heroin users covered in each district was adjusted, and then we explored the accessibility coverage rate in each district. In Kaohsiung city, the average coverage rate was around 51.26%. The result indicated the approachable medical services for heroin users.

Figure 3.

The buffer area of MMTP (3A) and NEP (3B) and the distribution of heroin users

It is known that injecting drug users had higher risks with transmissible diseases such as HIV, other blood‐borne or sexual transmitted infections. Due to the elevating rate of HIV problem, the relationship between the accessibility coverage rate and the amount of HIV cases caused by drug injection was further explored (N = 6292). Based on the concept of repeated measures analysis realizing self‐differences in different times, we tried to analyze the relationship between the amount of HIV cases and the accessibility coverage rate of medical facilities. As shown in Table 3, a significant difference (P = 0.0143) was found. The amount of HIV cases in the districts with less than 50% accessibility coverage rate was much higher than those in the districts with over 50% accessibility coverage rate.

Table 3.

The relationship between the amount of HIV cases and the accessibility coverage rate of medical facilities

		Amounts of HIV
	Year	2011		2012		2013		2014		2015
Accessibility coverage rate	na	Mean	(SD)	Mean	(SD)	Mean	(SD)	Mean	(SD)	Mean	(SD)	P valueb
<50%	22	46.64	(42.45)	48.64	(48.05)	49.36	(50.24)	50.55	(50.90)	51.50	(50.41)	0.0143
≥50%	13	13.54	(19.30)	13.46	(19.06)	12.77	(17.49)	13.23	(18.40)	13.46	(18.52)

^aRepeated measures analysis.

^bThe amount of the districts.

4. DISCUSSION AND CONCLUSION

To our knowledge, there are few illegal drug use‐related studies with the application of GIS in Taiwan, especially the medical accessibility coverage rate. The illegal drug use‐related data from 2011 to 2015 were collected from Kaohsiung city government and analyzed via GIS‐based analysis to evaluate the accessibility coverage of medical facilities for heroin users. Improvement of accessibility coverage rate of medical facilities was assumed critical to reduce the amount of HIV cases among heroin users.

4.1. Indicators for drug use prevention and intervention

Illegal drug use is a complicated and multifaceted issue, and many factors would affect the effectiveness in either prevention or treatment. According to the ETOS,22 many variables, such as age, physical/social environment (neighborhood), amount of arrests in the past year, and family support etc., may present the characteristics and life experiences of individuals to contribute to drug use. There may be still other factors in relation to various geography, culture, and customs in different regions. In this study, we only took three variables into account, that is, age, arrests, and percent of low‐income population, for the sake of data availability. Some physical and social characteristics of living environment and neighborhoods had been identified as the strong predictors of behavior in substance use and mental health outcomes.24, 25 However, the environmental and social context of illicit drug use is a complex situation that requires multilevel perspectives.26 Living in a disorganized neighborhood with high rates of poverty and single parent families were related to higher rates of failure recovery.27, 28 The study conducted in Vancouver found that poorest neighborhood, extreme poverty, high crime, homelessness, and poor housing were associated with high rates of alcohol and drug abuse. Researchers also found that living in unstable housing is often a predictor for syringe sharing.29, 30 In this study, therefore, the association of accessibility coverage to medical facilities and socioeconomic status with human immunodeficiency virus (HIV) risk was evaluated by collecting data of 7890 heroin users who were categorized into five risk groups according to their income and distance to treatment settings. In the future, it may be necessary to set up a series of comprehensive indicators not only to have more detailed information of drug users but also to assess the effectiveness and efficiency of prevention policies and intervention treatment.

4.2. Application of GIS on illegal drug use problems

From our GIS‐based analysis maps, it can be found that the high‐degree of spatial aggregation phenomenon for heroin users in the metropolitan and urban areas, especially in the west‐south regions of Kaohsiung city due to its high population density. However, it was also found that heroin users clustered in some remote or marginal regions. It may be due to their geographic locations such as proximity to harbors for easy access to drugs or simply clandestine enough to avoid from being arrested. In this study, we further demonstrated the amount and distribution of medical facilities for heroin users. If GIS could be utilized to integrate relevant data, including drug users and HIV individuals' locations, precise arrested addresses, the distribution of medical facilities, the flow direction of drugs, the current prevalence and the future trend of drugs, and the allocation of medical sources, the epidemic of HIV could be precisely identified and suitably controlled.

Literatures have shown that the clients who traveled less than one mile attending to treatment program would be more likely to complete the program than those who travel longer distance.10 Similarly, a previous study in 1995 also indicated that alcoholics were less likely to participate in aftercare treatment if they lived further away from the treatment program.31 Among HIV‐positive persons in the US, when transportation was provided, there would be higher participation rate in substance use and mental health treatment programs.32 It was found in Philadelphia that the higher density of self‐help groups for drug users or alcoholism, the lower likelihood of patients with comorbid mental health and substance use disorders under continuing treatment.33 In 2010, Wong et al.34 collected data via questionnaires in specific areas of Hong Kong to explore the relationship between the distances from heroin users' residents to methadone treatment clinics based on geographic analysis. They illustrated the clinics should be situated in the center of the city for convenient purposes, and revealed the location of clinics was crucial for guaranteeing effective maintenance. Inconvenient access to medical facilities would also influence adolescents' drug use behaviors to have higher probability to initiate drug use.9 Therefore, in order to have maximum effectiveness, the locations of medical treatment facilities should meet with the needs of drug users.10, 15

4.3. Data sources, format, and limitations

In Taiwan, there have been many illegal drug use‐related data systems to monitor the trend and situation of illegal drug use, but the data of these systems have not yet been integrated. In this study, all data were obtained from the existent data systems. These illegal drug use‐related data were collected and integrated for comprehensive analysis. Unfortunately, some of the data formats were diversified and incompatible for integration among different data systems. In addition, some data of heroin users' addresses were incomplete or missing. The inaccuracy of spatial data could affect the results in geographical analysis. For the cause, it was imperative to re‐confirm the different data definition, double check the data integrity and correct the errors or supplement missing content. With such time‐consuming efforts, the format of data associated with drug use in each monitor system could be harmonized.

Another limitation was the restriction imposed by the Act of Personal Data Protection. It was therefore difficult, if not impossible, to obtain the treatment data of individual drug user. But thanks to previous researches, we could understand the key impact factors contributing to the effectiveness of medical treatment, which involved at least compliance of complete treatment regimens by the clients, and the distances traveled to the treatment facilities.24 Accordingly, we explored the relationship between the amount of HIV cases and the accessibility coverage rate of medical facilities in each area as assessment indicator.

Despite these potential limitations, our study would still illustrate the spatial distribution of heroin users and medical facilities via GIS‐based approach. Our results indicated the areas with over 50% accessibility coverage rate had fewer HIV cases with statistical significance, implying that improvement of accessibility coverage rate of medical facilities was critical to reduce the amount of HIV cases among heroin users. Inconvenient access to medical facilities could result in poor treatment outcomes such as higher HIV incidences. Therefore, in order to effectively reduce HIV incidences among the heroin users, the accessibility coverage and locations of medical treatment facilities should meet the needs of drug users.

GIS has been applied for drug policy research in recent years. From either prevention or intervention perspective, the formulation of drug policies could be more solid and evidence‐based with geographical indicators.

CONFLICTS OF INTEREST

The authors have no conflicts of interest to report.

Feng L‐Y, Lan Y‐C, Huang J‐J, Li J‐H. HIV risk and the association with accessibility coverage to medical facilities and socioeconomic status among heroin users in Kaohsiung, Taiwan (2011‐2015): A GIS approach. Kaohsiung J Med Sci. 2019;35:56–62. 10.1002/kjm2.12009