Infectious Diseases
ISSN: 2374-4235 (Print) 2374-4243 (Online) Journal homepage: http://www.tandfonline.com/loi/infd20
Spatial epidemiology of HIV-hepatitis co-infection in the State of Michigan: a cohort study Zahid Butt, Sue Grady, Melinda Wilkins, Elizabeth Hamilton, David Todem, Joseph Gardiner & Mahdi Saeed To cite this article: Zahid Butt, Sue Grady, Melinda Wilkins, Elizabeth Hamilton, David Todem, Joseph Gardiner & Mahdi Saeed (2015) Spatial epidemiology of HIV-hepatitis co-infection in the State of Michigan: a cohort study, Infectious Diseases, 47:12, 852-861 To link to this article: http://dx.doi.org/10.3109/23744235.2015.1066931
Published online: 16 Jul 2015.
Submit your article to this journal
Article views: 21
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=infd20 Download by: [University of Otago]
Date: 18 September 2015, At: 22:23
Infectious Diseases, 2015; 47: 852–861
Original article
Spatial epidemiology of HIV-hepatitis co-infection in the State of Michigan: a cohort study Zahid Butt1, Sue Grady2, Melinda Wilkins3, Elizabeth Hamilton4, David Todem1, Joseph Gardiner1 & Mahdi Saeed1,5
Downloaded by [University of Otago] at 22:23 18 September 2015
1Department
of Epidemiology & Biostatistics, 2Department of Geography, and 3Program in Public Health, Michigan State University, East Lansing, 4Body Art, Viral Hepatitis & HIV Analysis Unit, HIV/STD/VH/TB Epidemiology Section, Michigan Department of Community Health Lansing, and 5Departments of Large Animal Clinical Sciences and Epidemiology, 165 Food Safety and Toxicology Center, Michigan State University, East Lansing, MI, USA
Abstract Background: Acquired immunodeficiency syndrome (AIDS) is a continuing global public health threat affecting millions of individuals. In 2009, 33.3 million people worldwide were living with human immunodeficiency virus (HIV) infection. HIVinfected individuals are at an increased risk of acquiring hepatitis B and hepatitis C viral (HBV, HCV)infections because of shared transmission routes. The purpose of this study was to identify geographical clusters of HIV-(HBV/HCV) co-infection in the State of Michigan. Methods: Retrospective cohort data on HIV-infected individuals were matched to all hepatitis B and C cases in Michigan during the period of January 1, 2006 through December 31, 2009. A prevalence map of HIV infection was created and spatial clusters of HIV-hepatitis B or C co-infection were detected using GeoDa’s bivariate local Moran’s I and SaTScan’s discrete Poisson model. Results: A bivariate cluster of high prevalence HIV and hepatitis B or C was identified in the Detroit Metropolitan Area and surrounding counties. A Poisson cluster of HIVhepatitis B or C co-infection was identified, relative risk (RR) 1.38 (p 0.029) in the western and northwestern counties of Lower Michigan, controlling for sex, race, and AIDS status. Conclusion: This study identified elevated HIV-hepatitis B or C co-infection unexplained by sex, race or AIDS status in counties outside of the Detroit Metropolitan Area where HIV prevalence was highest in Michigan. The findings from this study may be used to target future public health policy and healthcare interventions for HIV-hepatitis co-infection in these areas.
Keywords: HIV/AIDS, hepatitis, HIV and hepatitis co-infection, cluster analysis
Introduction In 2009, 33.3 million people worldwide were living with human immunodeficiency virus (HIV) [1]. Of these, approximately 2.6 million (7.8%) individuals had newly acquired HIV infections [1]. In 2009, an estimated 1.2 million persons were living with acquired immune deficiency syndrome (AIDS) in the United States, with about 17 000 AIDS-related deaths [2]. According to the Michigan Department of Community Health (MDCH), an estimated 19 500 people were currently living with HIV/AIDS in the state, and of these, 14 895 were reported in the surveillance system through January 2011 [3]. Individuals infected with HIV are at an increased risk of developing opportunistic infections, because of
their immunodeficiency. In addition to opportunistic infections, hepatitis B virus (HBV) and hepatitis C virus (HCV) infections are common in individuals infected with HIV/AIDS because of the shared disease transmission routes. In sub-Saharan Africa, a review of studies on HIV and hepatitis co-infection reported prevalence of 15.0% for mean HBsAg and 7.0% for mean anti-HCV in HIV-positive populations [4]. In Brazil, studies from 2000 to 2007 on HIV and hepatitis co-infection reported a prevalence of 4.4% for HIV and HCV HBV co-infection [5] and 4.4% for HIV and HCV co-infection [6]. Prospective cohort studies conducted in Denmark estimated a prevalence of 6.0% for HIV and chronic HBV co-infection [7] and 16.0% for HIV and HCV co-infection [8]. Two
Correspondence: Zahid A. Butt MBBS MSc PhD, Department of Epidemiology & Biostatistics, B601 West Fee Hall, Michigan State University, East Lansing, MI 48824, USA. Tel: 1 517 353 7946. E-mail: [email protected] (Received 21 December 2014; accepted 18 June 2015) ISSN 2374-4235 print/ISSN 2374-4243 online © 2015 Informa Healthcare DOI: 10.3109/23744235.2015.1066931
Downloaded by [University of Otago] at 22:23 18 September 2015
studies conducted in the United States on HIV and hepatitis co-infection reported a prevalence of 7.6% during 1998–2001 [9] and 11.0% for HIV and chronic HBV from 1998 to 2008 [10]. In March 2014 the Centers for Disease Control and Prevention (CDC) reported that of people with HIV in the United States, approximately 10% were co-infected with HBV and 25% were co-infected with HCV [11]. The majority of HIV studies have focused on the prevalence of morbidity or premature mortality and often do not take into account the spatial variation in disease or risk factors for HIV/AIDS to identify highrisk areas for public health prevention and healthcare intervention programs. Studies of hepatitis, HIV, and other sexually transmitted diseases that have looked at spatial variation have either studied the spatial clustering of HIV infection by applying scan statistics such as the Bernoulli [12] and Poisson [13] methods or sexually transmitted infections by applying the Bernoulli [14] method in SaTScan™ [15]. Other studies have looked at the spatial clustering of gonorrhea cases using k-function techniques and SaTScan’s Poisson model [16,17] and hepatitis C using spatial filtering and SaTScan’s Poisson models [18] in urban and rural settings. Two challenges of detecting clusters for relatively rare diseases such as HIV and hepatitis are the ability to simultaneously assess their joint spatial relationships as well as the ability to measure geographic heterogeneity in large geographic areas with low population density – a problem that could lead to the misinterpretation of the underlying geographic distribution [12]. To address these problems, the bivariate local Moran’s I will estimate the strength of spatial autocorrelation between two diseases across a study area and the spatial scan statistic SaTScan™ [15] has the ability to scan over geographic areas with fixed population parameters to capture adequate cases to identify disease clusters in a variety of geographic settings, scales, and for rare events [12,19–24]. The spatial scan statistic SaTScan™ [15] also has good power to detect localized clusters and can account for multiple testing in the data [25]. The Adult and Adolescent Spectrum of Disease surveillance project at the MDCH, focusing on HIV and hepatitis co-infection, reported that of HIV-infected individuals in Michigan, 12.0% were also diagnosed with HBV infection and 20.0% were also diagnosed with HCV infection [26]. However, no study has looked at the spatial structure of hepatitis B and C co-infection among HIV-infected individuals using the bivariate local Moran’s I and spatial scan statistic. The purpose of this study was to identify geographical clusters of HIV-(HBV/HCV) co-infection in the State of Michigan. Identification of the spatial clusters of HIV and HBV or HCV co-infection in Michigan will provide valuable knowledge for public health epidemiologists, policymakers, and health managers as to where to plan
Spatial epidemiology of HIV-hepatitis co-infection 853 and implement interventions and allocate resources. To the best of our knowledge, this is the first study to identify geographic clusters of HIV-(HBV/HCV) coinfection in the state of Michigan using data from a statewide population-based surveillance system. Material and Methods Data sources For this study, HIV/AIDS-infected individuals of all ages residing in Michigan’s counties were included and matched with all HBV and HCV cases in the period of January 1, 2006 through December 31, 2009 using a common identifier (identification number). Data were retrospectively obtained for the study period from the enhanced HIV/AIDS Reporting System (eHARS) maintained by the HIV/STD/VH/TB Epidemiology Section and the Michigan Disease Surveillance System (MDSS), which is the statewide communicable disease reporting system maintained by the Surveillance and Infectious Disease Epidemiology Section of MDCH. Ethical approval was obtained from the institutional review board at Michigan State University and MDCH. An employee of the HIV/STD/VH/TB Epidemiology Section performed the record linkage to ensure confidentiality and privacy of the participants. De-identification of the study participant data was done according to Health Insurance Portability and Accountability Act guidelines on public health information and no participants were interviewed or contacted. For this study, HIV/AIDS cases were defined using the CDC 1993 revised classification system of HIV [27]. The outcome variable was ‘Co-infection’ which was recoded into ‘co-infected’ 1 and ‘not coinfected’ 0. An individual with HIV/AIDS was categorized as co-infected if he or she had been concomitantly infected with confirmed HBV or HCV (acute and chronic) based on the CDC case definition [28–32] and residing in the state of Michigan during the period 2006 through 2009. Any individuals with HIV/AIDS who had a diagnosis of acute or chronic hepatitis B or C before 2006 were not included in this study. The characteristics of HIV-infected and co-infected individuals utilized in this research included sex, race, and current HIV status (status through December 2009). The variable sex was categorized as dichotomous, male 1 and female 2; race and ethnicity were classified into white non-Hispanic 1, black nonHispanic 2, Hispanic 3, and other 4; and current HIV status was dichotomized into HIV without AIDS 1 and AIDS 2. Statistical and cluster analyses A map of HIV prevalence by county was created for the state by using information on HIV cases from
Downloaded by [University of Otago] at 22:23 18 September 2015
854 Z. Butt et al. 2006 to 2009 based upon the administrative boundaries of MDCH [3]. Descriptive analyses of the HIV and hepatitis B or C data were performed in SAS version 9.2 (SAS Institute, Inc., Cary, NC, USA). The co-infected and non-co-infected individuals were compared using the Pearson chi-squared statistic and two-sided p values were reported. Spatial cluster analyses were performed using GeoDa version 1.6.7 [33] and bivariate local Moran’s I and SaTScan version 9.014, and discrete Poisson methods [34,35] were implemented. The bivariate local Moran’s I statistic estimates the spatial autocorrelation between HIV and HBV/HCV using a k-nearest neighbor weighting scheme. The autocorrelation statistic is visualized as the slope in the scatter plot with HBV/HCV as a spatially lagged variable on the vertical axis and HIV on the horizontal axis. Counties with high HIV and high hepatitis (High-High) are located in the upper right quadrant, Low-Low counties are in the lower left quadrant, Low-High counties are in the upper left quadrant, and High-Low counties are in the lower right quadrant. These county estimates are mapped in ArcGIS version 10.2 (Environmental Systems Research Institute, Redlands, CA, USA). Scan statistics are used for the detection and evaluation of clusters by applying a scanning window across the study area (i.e. the centroids of each county across Michigan). Within the window at each location, the number of observed and expected cases of HIV-hepatitis coinfected individuals is calculated and a likelihood function is estimated [36]. The window with the maximum likelihood is the most likely cluster, that is, the cluster that is least likely to have occurred by chance. Secondary clusters are also identified and ordered by the program according to the likelihood ratio test statistic. The discrete Poisson model assumes that the number of cases in each location is Poisson distributed. The null hypothesis for the model is that the expected number of cases in each scanning window is proportional to its population size [36]. The significance of the likelihood ratio was tested using 999 Monte Carlo simulations. In this study, a circular window was used and the maximum circle size included up to 25% of the HIV-infected population being analyzed within a 160 km (100 mile) radius [36]. In all, 25% of the HIV population was scanned as this was the approximate national HCV prevalence and a 160 km radius was hypothesized to be the extent within which transmission was likely to occur. A purely spatial Poisson analysis was performed. A significant p value was considered to be 0.05. The data for both the bivariate local Moran’s I and Poisson method were aggregated at the county level using SAS version 9.2 (SAS Institute, Inc.). For the bivariate local Moran’s I analysis a file with case
counts of HIV and HBV/HCV by county was constructed. For the discrete Poisson analysis, a case (hepatitis) and a population (HIV/AIDS infection) file were created with the county as the geographic unit and covariates sex, race, and current HIV status. A coordinates file containing the latitude and longitude at the centroid of each county was used to define the locations in both analyses.
Results A total of 13 936 individuals were infected with HIV/ AIDS during the study period (Table I). The prevalence of HIV-HBV or -HCV co-infection among these individuals was 4.2% (n 578). The majority of HIV-infected individuals also had AIDS (68.0%), were males (80.6%), and were black non-Hispanic (66.0%). There was a significant difference between HIV and HBV or HCV co-infected and nonco-infected individuals in regards to the sex (p 0.007), race (p 0.0001), and current HIV status (p 0.0001) of individuals. The HIV prevalence in Michigan by county of residence at time of diagnosis is provided in Figure 1 [3]. This map shows a higher prevalence of HIV infection in counties that contain urban areas in southern Lower Michigan, particularly in Wayne County (370 per 100 000 population), which contains the city of Detroit. Seven other counties with urban areas in Michigan also had high HIV prevalence rates including Washtenaw (151), Ingham (151), Berrien (148), Oakland (134), Genesee (127), Kent (127), and Kalamazoo (116), respectively. Importantly, in rural Michigan a higher HIV prevalence was also observed in the counties of Table I. Comparative analysis of HIV and hepatitis B or C co-infected (n 578) and non-co-infected (n 13 358) individuals by sex, race, and current HIV status in Michigan, 2006–2009.
Variable Sex Female Male Race White (non-Hispanic) Black (non-Hispanic) Hispanic Otherb Current HIV status HIV (not AIDS) AIDS aChi-squared
Co-infected, n (%)
Nonco-infected, n (%)
112 (19.4) 466 (80.6)
3235 (24.2) 10123 (75.8)
0.007
161 (27.9)
4797 (35.9)
0.0001
381 (66.0)
7710 (57.7)
13 (2.3) 23 (3.9)
557 (4.2) 294 (2.2)
185 (32.0) 393 (68.0)
6040 (45.2) 7318 (54.8)
p valuea
0.0001
test. Hawaiian & Pacific Islanders, Alaskan Native and American Indian and multiracial. bOther Asian,
Downloaded by [University of Otago] at 22:23 18 September 2015
Spatial epidemiology of HIV-hepatitis co-infection 855
Figure 1. HIV prevalence and cases of hepatitis B or C by county, 2006–2009. Note: To mitigate the effect of small numbers of cases, reported HIV prevalence in multi-county health departments are listed for the health department as a whole and not the individual counties.
Luce, Alger, Mackinac, and Schoolcraft (269 per 100 000 population) in the Upper Peninsula. The global Moran’s I 0.153 demonstrated spatial clustering of HIV and HBV/HCV in Michigan (Figure 2a). Five counties were identified as High-High (Genesee, Macomb, Oakland, Wayne, and Washtenaw) in the Detroit Metropolitan Area, with two contiguous counties (St Clair and Monroe) Low-High. Ten counties were Low-Low – Alcona, Crawford, Iosco, and Oscoda in Upper Lower Michigan and Baraga, Gogebic, Houghton, Iron, Marquette, and Ontonagon in the Upper Peninsula (Figure 2b). The Poisson cluster analysis identified two spatial clusters (Figure 3). The first most likely cluster had a relative risk (RR) of 1.38 (p 0.029) after adjusting for sex, race, and current HIV status (Table II). This cluster was located in western and northwestern Upper Lower Michigan and northern interior rural counties (n 46). Within this cluster there were 706 HIV- and 135 HBV/HCV-infected individuals who matched on sex, race, and HIV status. Of the
135 HIV-HBV/HCV co-infected individuals a majority were located in Genesee (n 38, local RR 1.1), Kent (n 32, local RR 1.9), and Ingham (n 16, local RR 1.0) counties. The secondary cluster was non-significant (RR 1.17, p 0.99) after adjustment. The secondary cluster included 583 HIV and 101 HIV-HBV/HCV co-infected individuals who matched on sex, race, and HIV status within three counties: Macomb (n 25, local RR 1.1), Oakland (n 70, local RR 1.1), and St Clair (n 6, local RR 1.9), all located in the Detroit Metropolitan Area. Importantly, this secondary cluster did not include Wayne County and Detroit. The local RR’s demonstrate the geographic variability of HIV-HBV/ HCV co-infection within these two discrete Poisson spatial clusters (Figure 3). Discussion HIV, HBV, and HCV share common transmission routes, which include parenteral, perinatal, and sexual
Downloaded by [University of Otago] at 22:23 18 September 2015
856 Z. Butt et al.
Figure 2. (a) Bivariate local Moran’s I scatter plot of HIV and hepatitis B or C by county in Michigan, 2006–2009. (b) Bivariate local Moran’s I clusters of HIV and hepatitis B or C by county in Michigan 2006–2009.
transmission [37]. The elevated cluster identified in the bivariate local Moran’s I was in the Detroit Metropolitan Area and surrounding counties where HIV prevalence and the numbers of HBV/HCV cases
were high. However, the most likely Poisson cluster was in western and northwestern Lower Michigan and northern interior rural counties outside of the Detroit Metropolitan Area, demonstrating that the
Downloaded by [University of Otago] at 22:23 18 September 2015
Spatial epidemiology of HIV-hepatitis co-infection 857
Figure 3. Discrete Poisson clusters of HIV and hepatitis B or C co-infection by county in Michigan, 2006–2009.
geography of HIV-HBV/HCV co-infection is different from the geography of HIV prevalence, which could be explained through the sex, race, and HIV status adjustment, other underlying differences in the populations at risk, and/or characteristics of the environments in urban versus rural settings. Additionally, there could be some unobserved factors that might influence the transmission of HBV and HCV infections among HIV-infected individuals that are different from the transmission of HIV infection
among the general population that have not been explained in previous research. One potential explanation for these geographic differences in HIV-HBV/HCV co-infection and nonco-infection is the lack of available or accessible health services in the counties with significant clustering. In this regard, enumeration of the health facilities as a proxy for health services revealed that most of the counties in cluster 1 had a limited number of hospitals as compared with other more densely
Table II. HIV-hepatitis B or C co-infection spatial clusters with high rates identified by SaTScan discrete Poisson method, Michigan 2006–2009.a Cluster Most likely cluster Secondary cluster aControlling
HIV/AIDS population (n) 706c 583d
Observed cases (n) 135 101
for sex, race, and current HIV status. risk observed/expected. cCluster includes 46 counties. dCluster includes three counties. bRelative
Expected cases (n) 104.77 88.48
Relative riskb
Log-likelihood ratio
p value
1.38 1.17
7.51 1.05
0.029 0.99
Downloaded by [University of Otago] at 22:23 18 September 2015
858 Z. Butt et al. populated counties [38]. According to the Michigan health facility atlas [38], Leelanau, Lake, Newaygo, and Benzie counties had one hospital each. A lack of hospitals or other healthcare services could affect access to healthcare and early diagnosis and treatment of HIV and hepatitis. Another reason could be that these counties are areas where injection drug use (IDU) is common. Data from MDCH indicate that 29.0% of all injecting drug users living with HIV in Michigan reside in counties with significant coinfection clustering [25]. Additionally, Traverse City in Grand Traverse, which borders Leelanau County, has a Traverse City Michigan syringe exchange program. Syringe exchange programs in a locality have been shown to be a proxy indicator for substantial IDU in a study conducted on hepatitis C [17]. Furthermore, Allegan County has been designated as a High Intensity Drug Trafficking area (HIDTA) [39] in Michigan by the Office of the National Drug Control Policy [40]. It is therefore conceivable that IDU as well as other high-risk behaviors are more common in this county as a result of drug-related activities, which may have led to an increased RR of HIV and HBV/HCV co-infection. To explore these potential
explanations for the clustering of HIV-HBV/HCV co-infection, the demographic, current HIV status, and route(s) of transmission characteristics were examined for co-infected and non-co-infected individuals residing within and outside of clusters 1 and 2 (outside of clusters 1 and 2 herein referred to as ‘outside’ (Table III). The characteristics of coinfected and non-co-infected individuals were compared by cluster 1 vs 2, cluster 1 vs outside, and cluster 2 vs outside using a Pearson chi-squared statistic and two-sided p values (Table IV). A summary of these findings is presented. In regards to sex, over 75% of co-infected and non-co-infected individuals inside and outside of HIV-HBV/HCV clusters were males. For co-infected individuals, there were no significant differences in sex composition across clusters 1, 2, and outside. For non-co-infected individuals, however, there were significant differences in the sex composition between males and females across clusters and outside (p 0.000). These findings demonstrate that the sex composition for the co-infected was less variable than the sex composition for the non-co-infected between clusters 1 and 2 and outside. Public health
Table III. Comparative analysis of HIV and hepatitis B or C co-infected and non-co-infected individuals by demographic, HIV status, and transmission characteristics inside and outside of clusters 1 and 2, Michigan, 2006–2009. Characteristics
Co-infected (%) Cluster 1
Sex Female Male Age (years) 20 2024 2529 3039 4049 50 Race Black (non-Hispanic) White (non-Hispanic) Hispanic Othera Current HIV status HIV (not AIDS) AIDS Route of transmission Blood transfusion Heterosexual female Heterosexual male High-risk heterosexual IDU MSM MSM & IDU Other/unknownb
Cluster 2
Non-co-infected (%) Outside
Cluster 1
Cluster 2
Outside
16.9 83.1
17.6 82.4
18.1 81.9
22.4 77.6
16.5 83.5
21.0 79.0
6.1 4.7 20.3 35.8 22.3 10.8
4.0 10.1 18.2 37.4 20.2 10.1
2.4 10.5 13.2 37.1 24.5 12.4
5.6 13.1 17.5 35.5 19.8 8.5
4.7 12.0 17.2 34.9 22.2 9.0
6.1 12.4 15.5 34.4 22.2 9.3
42.6 48.0 4.7 4.7
46.1 45.1 2.0 6.9
79.3 16.8 1.3 2.6
33.1 56.5 7.7 2.7
34.9 58.9 3.2 3.0
71.8 23.5 3.2 1.5
31.1 68.9
30.4 69.6
30.8 69.2
46.8 53.2
47.7 52.3
44.0 56.0
7.4 4.7 6.1 6.8 18.9 34.5 12.2 9.5
4.9 3.9 7.8 4.9 22.5 44.1 5.9 5.9
1.6 1.6 6.7 7.5 26.4 39.6 9.6 7.0
0.7 5.9 9.0 13.3 6.5 51.8 4.4 8.5
1.0 4.1 7.5 8.3 5.3 61.2 3.5 9.1
0.4 6.0 8.7 14.0 12.5 44.8 3.8 9.8
Co-infected: cluster 1 (n 148), cluster 2 (n 102), outside (n 386). Non-co-infected: cluster 1 (n 2869), cluster 2 (n 2368), outside (n 9053). IDU, intravenous drug use; MSM, men having sex with men. aOther multiracial, Asian, Hawaiian & Pacific Islander, Alaskan Native, American Indian. bOther/unknown perinatal and unknown routes of transmission.
Spatial epidemiology of HIV-hepatitis co-infection 859
Table IV. Comparative analysis of HIV and hepatitis B or C co-infected and non-co-infected individuals by demographic, HIV status, and transmission characteristics inside and outside of clusters 1 and 2, Michigan, 2006–2009. Comparison
Co-infected
Downloaded by [University of Otago] at 22:23 18 September 2015
c2 Cluster 1 vs 2 Sex Age Race Current HIV status Route of transmission Cluster 1 vs outside Sex Age Race Current HIV status Route of transmission Cluster 2 vs outside Sex Age Race Current HIV status Route of transmission
Non-co-infected c2
df
p valuea
27.142 7.556 54.136 0.448 67.609
1 7 5 1 8
0.000 0.373 0.000 0.503 0.000
0.571 0.107 0.000 0.955 0.006
18.430 15.848 1335.762 6.652 98.680
1 8 5 1 8
0.000 0.045 0.000 0.010 0.000
0.909 0.618 0.000 0.932 0.292
97.586 11.031 1128.778 10.336 259.216
1 8 5 1 8
0.000 0.200 0.000 0.001 0.000
df
p value
0.268 3.019 3.294 0.013 6.619
1 7 5 1 7
0.605 0.883 0.655 0.908 0.470
0.321 11.804 70.149 0.003 21.555
1 7 5 1 8
0.013 5.342 49.332 0.007 9.63
1 7 5 1 8
Co-infected: cluster 1 (n 148), cluster 2 (n 102), outside (n 386). Non-co-infected: cluster 1 (n 2715), cluster 2 (n 2267), outside (n 8672). Cluster 1, n 2869; cluster 2, n 2368; outside cluster 1 and 2, n 9053. aChi-squared test.
messaging for the proportion of co-infected men and women may be consistent across the state. Messages targeted for non-co-infected men and women will need to be tailored with the greatest proportion of men in cluster 2 and the greatest proportion of women in cluster 1. In regards to age, at least onethird of co-infected and non-co-infected individuals were between the ages of 30 and 39 years. There were no significant differences in age for co-infected and non-co-infected, except for non-co-infected individuals residing in cluster 1 vs outside (p 0.045). In cluster 1 there appears to be a slightly higher proportion of non-co-infected young adults compared with outside. These younger age groups in cluster 1 and the 30–39-year age group across the state should continue to be targeted for HIV and HIV-HBV/HCV co-infection public health messaging. In regards to race, for the co-infected there were no significant differences between clusters 1 and 2; however, there were significant differences in race between cluster 1 vs outside (p 0.000) and cluster 2 vs outside (p 0.000). In cluster 1 the co-infected were racially mixed with relatively similar proportions of white (48.0%) and black (42.6%) nonHispanics. In cluster 2 the co-infected were also racially mixed (white 45.1%, black 46.1%). However, outside of these clusters a majority of coinfected were black (79.3%) and white (16.8%) non-Hispanics. Importantly within HIV-HBV/HCV
co-infection clusters, race was less of a risk factor than outside of these clusters. For non-co-infected individuals the differences in race between clusters 1 and 2 and outside were statistically significant (p 0.000). The non-co-infected individuals within clusters 1 and 2 were not racially mixed, with slightly more white (56.5%, 58.9%) than black (33.1%, 34.9%) non-Hispanic subjects. Cluster 1 also had twice as many Hispanics as cluster 2 or outside (7.7% vs 3.2% vs 3.2%). Outside, the proportion of black non-Hispanics was higher (71.8%) and the proportion of white non-Hispanics was lower (23.5%) than in clusters 1 and 2. The spatial patterns of racial differences in the co-infected outside and nonco-infected in clusters 1 and 2 and outside appear to be linked to racial residential segregation and place. In regards to current HIV status, co-infected individuals had a higher proportion of AIDS than non-co-infected individuals in cluster 1 (68.9% vs 53.2%), cluster 2 (69.6% vs 52.3%), and outside (69.2% vs 56.0%). There were no significant differences in current HIV status for co-infected individuals in clusters 1 and 2 and outside. For non-co-infected individuals there were significant differences in current HIV status between cluster 1 and outside (p 0.010) and cluster 2 and outside (p 0.001), with outside having a slightly higher prevalence of AIDS. These findings suggest that healthcare and treatment for HIV within clusters is better than outside but there is a need for improved care and treatment of HIV-hepatitis co-infection. Finally, men having sex with men (MSM) was the most important route of HIV transmission among HIV-HBV/HCV co-infected and non-co-infected individuals. There were significant differences in route of transmission for the co-infected in cluster 1 vs outside (p 0.006). Their major differences were in the proportion of MSM (34.5%, 39.6%), IDU (18.9%, 26.4%), and MSM and IDU (12.2%, 9.6%). For non-co-infected individuals there were significant differences in route of transmission between clusters 1 and 2 and outside (p 0.000), demonstrating more variability in routes of transmission for non-co-infected compared with co-infected. Importantly, the proportion of MSM was higher across clusters 1 and 2 and outside for non-co-infected but the proportion of IDU was higher across clusters 1 and 2 and outside for IDU and MSM and IDU. These findings support the potential explanations for HIV-HBV/HCV clusters, including the need for improved healthcare to slow the progression of HIV-AIDS in co-infected individuals and the need for additional emphasis on IDU in addition to MSM as routes of transmission in HIV-HBV/HCV. Limitations of this study are that the HIV surveillance system did not have information on newly
Downloaded by [University of Otago] at 22:23 18 September 2015
860 Z. Butt et al. diagnosed cases of HIV, so unfortunately incidence rates were not calculated. In addition, the data on hepatitis included both acute and chronic hepatitis cases but the number of acute cases was extremely small and therefore did not lend itself to a study of hepatitis incidence. Furthermore, the date of diagnosis of hepatitis in the surveillance system was quite incomplete and therefore a temporal analysis or spatio-temporal analysis were not possible. From a public health perspective it would have been ideal to study these incident cases of HIV and hepatitis B or C and HIV-HBV/HCV co-infection to more directly target timely interventions in affected counties. Another limitation is it was unclear if the significantly low bivariate local Moran’s I clusters are due to low HIV and low HBV/HCV (Low-Low) spatial autocorrelation or if these are counties with under-reporting and underdetection of hepatitis cases as the MDSS is based on a passive surveillance system. The HIV surveillance system in Michigan is based on data for those persons who have been confidentially reported by name. Infected individuals who have not been tested, have been tested only anonymously, or have been tested by name but not reported, are not included in the MDSS, which could also lead to under-reporting of HIV cases [25]. A final limitation is that this study was conducted at the county level and there may be significant variation in population characteristics healthcare and public health interventions. While having HIV and hepatitis data for the Zip Code would address this problem, the countylevel data are population-based including all the statewide data collected by the surveillance system on HIV-infected individuals as well as persons having HBV or HCV co-infection residing in Michigan’s counties. The number of HIV and hepatitis cases at the county level was an adequate sample by which to investigate HIV and HIV-HBV/HCV co-infection and will provide a basis for future research on a smaller scale. In conclusion, this study identified significant clusters of HIV-HBV/HCV co-infection in counties that would not be considered high risk because of low population density and low HIV prevalence. In this respect, spatial cluster analysis serves as an important tool to delineate infectious disease clusters, which could be missed by other analytic methods that do not consider their geography. The results from this study can guide policymakers and health managers to target interventions and disease control measures to reduce HIV and HBV/HCV in these counties. In addition, efficient allocation of resources to these areas can be considered as a means to obtain maximum benefit from any measures undertaken to prevent the spread of hepatitis infection among the high-risk population of HIV-infected individuals.
Future research should continue to focus on identifying risk factors that are associated with clustering of HIV-HBV/HCV coinfection in these counties as well as modifiable factors that may help to prevent these infections. Acknowledgments The authors would like to thank the staff at MDCH for their continuous support during the study. We are also thankful to Dr Kim Kirkey for providing valuable insight regarding surveillance of hepatitis infections. Declaration of interest: The authors report no conflicts of interest. This work was supported by the Dissertation Completion Fellowship awarded by the Graduate School and the Department of Epidemiology at Michigan State University. References [1] Global Report. UNAIDS report on the Global AIDS Epidemic 2010. Available from: http://www.unaids.org/en/media/unaids/ contentassets/documents/unaidspublication/2010/20101123_ globalreport_en.pdf [2] United States of America. HIV and AIDS estimates 2009. Available from: http://www.unaids.org/en/regionscountries/ countries/unitedstatesofamerica/ [3] Quarterly HIV/AIDS report, Michigan. January 2011. Available from: http://www.michigan.gov/documents/mdch/ Jan_2011_344131_7.pdf [4] Barth RE, Huijgen Q, Taljaard J, Hoepelman AI. Hepatitis B/C and HIV in sub-Saharan Africa: an association between highly prevalent infectious diseases. A systematic review and meta-analysis. Int J Infect Dis 2010;14:1024–31. [5] Portelinha Filho AM, Nascimento CU, Tannouri TN, Troiani C, Ascêncio EL, Bonfim R, et al. Seroprevalence of HBV, HCV and HIV co-infection in selected individuals from state of São Paulo, Brazil. Mem Inst Oswaldo Cruz 2009;104:960–3. [6] Victoria MB, Victoria Fda S, Torres KL, Kashima S, Covas DT, Malheiro A. Epidemiology of HIV/HCV co-infection in patients cared for at the Tropical Medicine Foundation of Amazonas. Braz J Infect Dis 2010;14: 135–40. [7] Omland LH, Weis N, Skinhøj P, Laursen A, Christensen PB, Nielsen HI, et al. Impact of hepatitis B virus co-infection on response to highly active antiretroviral treatment and outcome in HIV-infected individuals: a nationwide cohort study. HIV Med 2008;9:300–6. [8] Weis N, Lindhardt BO, Kronborg G, Hansen AB, Laursen AL, Christensen PB, et al. Impact of hepatitis C virus coinfection on response to highly active antiretroviral therapy and outcome in HIV-infected individuals: a nationwide cohort study. Clin Infect Dis 2006;42:1481–7. [9] Kellerman SE, Hanson DL, McNaghten AD, Fleming PL. Prevalence of chronic hepatitis B and incidence of acute hepatitis B infection in human immunodeficiency virusinfected subjects. J Infect Dis 2003;188:571–7. [10] Chun HM, Fieberg AM, Hullsiek KH, Lifson AR, Crum-Cianflone NF, Weintrob AC, et al. Epidemiology of
Downloaded by [University of Otago] at 22:23 18 September 2015
Hepatitis B virus infection in a US cohort of HIV-infected individuals during the past 20 years. Clin Infect Dis 2010;50:426–36. [11] Viral hepatitis populations. Available from: http://www.cdc. gov/hepatitis/Populations/hiv.htm [12] Tanser F, Bärnighausen T, Cooke GS, Newell ML. Localized spatial clustering of HIV infections in a widely disseminated rural South African epidemic. Int J Epidemiol 2009;38: 1008–16. [13] Wand H, Ramjee G. Targeting the hotspots: investigating spatial and demographic variations in HIV infection in small communities in South Africa. J Int AIDS Soc 2010;13:41. [14] Westercamp N, Moses S, Agot K, Ndinya-Achola JO, Parker C, Amolloh KO, et al. Spatial distribution and cluster analysis of sexual risk behaviors reported by young men in Kisumu, Kenya. Int J Health Geogr 2010;9:24. [15] Kulldorff M and Information Management Services, Inc. SaTScan™ v9.0: Software for the spatial , temporal and space-time scan statistics. Available from: http://www.satscan.org/, 2009. [16] Jennings JM, Curriero FC, Celentano D, Ellen JM. Geographic identification of high gonorrhea transmission areas in Baltimore, Maryland. Am J Epidemiol 2005;161:73–80. [17] Bernstein KT, Curriero FC, Jennings JM, Olthoff G, Erbelding EJ, Zenilman J. Defining core gonorrhea transmission utilizing spatial data. Am J Epidemiol 2004;160:51–8. [18] Trooskin SB, Hadler J, St Louis T, Navarro VJ. Geospatial analysis of hepatitis C in Connecticut: a novel application of a public health tool. Public Health 2005;119:1042–7. [19] Haque U, Hashizume M, Sunahara T, Hossain S, Ahmed SM, Haque R, et al. Progress and challenges to control malaria in a remote area of Chittagong hill tracts, Bangladesh. Malar J 2010;9:156. [20] Mirghani SE, Nour BY, Bushra SM, Elhassan IM, Snow RW, Noor AM. The spatial-temporal clustering of Plasmodium falciparum infection over eleven years in Gezira State, The Sudan. Malar J 2010;9:172. [21] Haque U, Huda M, Hossain A, Ahmed SM, Moniruzzaman M, Haque R. Spatial malaria epidemiology in Bangladeshi highlands. Malar J 2009;8:185. [22] Ageep TB, Cox J, Hassan MM, Knols BG, Benedict MQ, Malcolm CA, et al. Spatial and temporal distribution of the malaria mosquito Anopheles arabiensis in northern Sudan: influence of environmental factors and implications for vector control. Malar J 2009;8:123. [23] Klassen AC, Kulldorff M, Curriero F. Geographical clustering of prostate cancer grade and stage at diagnosis, before
Spatial epidemiology of HIV-hepatitis co-infection 861 and after adjustment for risk factors. Int J Health Geogr 2005;4:1. [24] Osei FB, Duker AA. Spatial dependency of V. cholera prevalence on open space refuse dumps in Kumasi, Ghana: a spatial statistical modelling. Int J Health Geogr 2008;7:62. [25] Song C, Kulldorff M. Power evaluation of disease clustering tests. Int J Health Geogr 2003;2:9. [26] 2010 Epidemiologic Profile of HIV/AIDS in Michigan. Available from: http://www.michigan.gov/documents/mdch/2010_ EPI_PROFILE_ENTIRE_335483_7.pdf [27] 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Recomm Rep 1992;4:1–19. [28] Hepatitis B, acute. Available from: http://wwwn.cdc.gov/ nndss/conditions/hepatitis-b-acute/case-definition/2000/ [29] Chronic Hepatitis B virus. Available from: http://wwwn.cdc. gov/nndss/conditions/hepatitis-b-chronic/case-definition/2012/ [30] Hepatitis C virus infection, acute. Available from: http:// wwwn.cdc.gov/nndss/conditions/hepatitis-c-acute/casedefinition/2012/ [31] Hepatitis C Virus infection, past or present. Available from: http://wwwn.cdc.gov/nndss/conditions/hepatitis-c-past-orpresent/case-definition/2012/ [32] Michigan Department of Community Health. Health care professionals’ guide to the Michigan communicable disease rules (version 2006.1). Available from: http://www.michigan. gov/documents/hlth_care_prof_guide_167371_7.pdf [33] Anselin L, Rey SJ. GeoDa 1.6.7 (March 2015). Available from: https://geodacenter.asu.edu/ [34] Kulldorff M. A spatial scan statistic. Commun Statist Theory Meth 1997;26:1481–96. [35] Kulldorff M, Nagarwalla N. Spatial disease clusters: detection and inference. Stat Med 1995;14:799–810. [36] Kulldorff M. SaTScan™ User Guide for version 9.0; 2010 pp 1–107. [37] Heymann DL, editor. Control of communicable diseases manual, 19th edn. American Public Health Association; 2008. [38] Michigan Department of Health and Human Services. 2010 Health Facilities Atlas. Available from: http://www.michigan. gov/mdch/0,1607,7-132-2945_5106-63274--,00.html [39] The High Intensity Drug Trafficking Area Program: an overview. Available from: http://www.ncjrs.gov/ondcppubs/publications/enforce/hidta2001/overview.html#4 [40] Michigan HIDTA. Available from: http://www.whitehousedrugpolicy.gov/hidta/mich.html
ISSN: 2374-4235 (Print) 2374-4243 (Online) Journal homepage: http://www.tandfonline.com/loi/infd20
Spatial epidemiology of HIV-hepatitis co-infection in the State of Michigan: a cohort study Zahid Butt, Sue Grady, Melinda Wilkins, Elizabeth Hamilton, David Todem, Joseph Gardiner & Mahdi Saeed To cite this article: Zahid Butt, Sue Grady, Melinda Wilkins, Elizabeth Hamilton, David Todem, Joseph Gardiner & Mahdi Saeed (2015) Spatial epidemiology of HIV-hepatitis co-infection in the State of Michigan: a cohort study, Infectious Diseases, 47:12, 852-861 To link to this article: http://dx.doi.org/10.3109/23744235.2015.1066931
Published online: 16 Jul 2015.
Submit your article to this journal
Article views: 21
View related articles
View Crossmark data
Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=infd20 Download by: [University of Otago]
Date: 18 September 2015, At: 22:23
Infectious Diseases, 2015; 47: 852–861
Original article
Spatial epidemiology of HIV-hepatitis co-infection in the State of Michigan: a cohort study Zahid Butt1, Sue Grady2, Melinda Wilkins3, Elizabeth Hamilton4, David Todem1, Joseph Gardiner1 & Mahdi Saeed1,5
Downloaded by [University of Otago] at 22:23 18 September 2015
1Department
of Epidemiology & Biostatistics, 2Department of Geography, and 3Program in Public Health, Michigan State University, East Lansing, 4Body Art, Viral Hepatitis & HIV Analysis Unit, HIV/STD/VH/TB Epidemiology Section, Michigan Department of Community Health Lansing, and 5Departments of Large Animal Clinical Sciences and Epidemiology, 165 Food Safety and Toxicology Center, Michigan State University, East Lansing, MI, USA
Abstract Background: Acquired immunodeficiency syndrome (AIDS) is a continuing global public health threat affecting millions of individuals. In 2009, 33.3 million people worldwide were living with human immunodeficiency virus (HIV) infection. HIVinfected individuals are at an increased risk of acquiring hepatitis B and hepatitis C viral (HBV, HCV)infections because of shared transmission routes. The purpose of this study was to identify geographical clusters of HIV-(HBV/HCV) co-infection in the State of Michigan. Methods: Retrospective cohort data on HIV-infected individuals were matched to all hepatitis B and C cases in Michigan during the period of January 1, 2006 through December 31, 2009. A prevalence map of HIV infection was created and spatial clusters of HIV-hepatitis B or C co-infection were detected using GeoDa’s bivariate local Moran’s I and SaTScan’s discrete Poisson model. Results: A bivariate cluster of high prevalence HIV and hepatitis B or C was identified in the Detroit Metropolitan Area and surrounding counties. A Poisson cluster of HIVhepatitis B or C co-infection was identified, relative risk (RR) 1.38 (p 0.029) in the western and northwestern counties of Lower Michigan, controlling for sex, race, and AIDS status. Conclusion: This study identified elevated HIV-hepatitis B or C co-infection unexplained by sex, race or AIDS status in counties outside of the Detroit Metropolitan Area where HIV prevalence was highest in Michigan. The findings from this study may be used to target future public health policy and healthcare interventions for HIV-hepatitis co-infection in these areas.
Keywords: HIV/AIDS, hepatitis, HIV and hepatitis co-infection, cluster analysis
Introduction In 2009, 33.3 million people worldwide were living with human immunodeficiency virus (HIV) [1]. Of these, approximately 2.6 million (7.8%) individuals had newly acquired HIV infections [1]. In 2009, an estimated 1.2 million persons were living with acquired immune deficiency syndrome (AIDS) in the United States, with about 17 000 AIDS-related deaths [2]. According to the Michigan Department of Community Health (MDCH), an estimated 19 500 people were currently living with HIV/AIDS in the state, and of these, 14 895 were reported in the surveillance system through January 2011 [3]. Individuals infected with HIV are at an increased risk of developing opportunistic infections, because of
their immunodeficiency. In addition to opportunistic infections, hepatitis B virus (HBV) and hepatitis C virus (HCV) infections are common in individuals infected with HIV/AIDS because of the shared disease transmission routes. In sub-Saharan Africa, a review of studies on HIV and hepatitis co-infection reported prevalence of 15.0% for mean HBsAg and 7.0% for mean anti-HCV in HIV-positive populations [4]. In Brazil, studies from 2000 to 2007 on HIV and hepatitis co-infection reported a prevalence of 4.4% for HIV and HCV HBV co-infection [5] and 4.4% for HIV and HCV co-infection [6]. Prospective cohort studies conducted in Denmark estimated a prevalence of 6.0% for HIV and chronic HBV co-infection [7] and 16.0% for HIV and HCV co-infection [8]. Two
Correspondence: Zahid A. Butt MBBS MSc PhD, Department of Epidemiology & Biostatistics, B601 West Fee Hall, Michigan State University, East Lansing, MI 48824, USA. Tel: 1 517 353 7946. E-mail: [email protected] (Received 21 December 2014; accepted 18 June 2015) ISSN 2374-4235 print/ISSN 2374-4243 online © 2015 Informa Healthcare DOI: 10.3109/23744235.2015.1066931
Downloaded by [University of Otago] at 22:23 18 September 2015
studies conducted in the United States on HIV and hepatitis co-infection reported a prevalence of 7.6% during 1998–2001 [9] and 11.0% for HIV and chronic HBV from 1998 to 2008 [10]. In March 2014 the Centers for Disease Control and Prevention (CDC) reported that of people with HIV in the United States, approximately 10% were co-infected with HBV and 25% were co-infected with HCV [11]. The majority of HIV studies have focused on the prevalence of morbidity or premature mortality and often do not take into account the spatial variation in disease or risk factors for HIV/AIDS to identify highrisk areas for public health prevention and healthcare intervention programs. Studies of hepatitis, HIV, and other sexually transmitted diseases that have looked at spatial variation have either studied the spatial clustering of HIV infection by applying scan statistics such as the Bernoulli [12] and Poisson [13] methods or sexually transmitted infections by applying the Bernoulli [14] method in SaTScan™ [15]. Other studies have looked at the spatial clustering of gonorrhea cases using k-function techniques and SaTScan’s Poisson model [16,17] and hepatitis C using spatial filtering and SaTScan’s Poisson models [18] in urban and rural settings. Two challenges of detecting clusters for relatively rare diseases such as HIV and hepatitis are the ability to simultaneously assess their joint spatial relationships as well as the ability to measure geographic heterogeneity in large geographic areas with low population density – a problem that could lead to the misinterpretation of the underlying geographic distribution [12]. To address these problems, the bivariate local Moran’s I will estimate the strength of spatial autocorrelation between two diseases across a study area and the spatial scan statistic SaTScan™ [15] has the ability to scan over geographic areas with fixed population parameters to capture adequate cases to identify disease clusters in a variety of geographic settings, scales, and for rare events [12,19–24]. The spatial scan statistic SaTScan™ [15] also has good power to detect localized clusters and can account for multiple testing in the data [25]. The Adult and Adolescent Spectrum of Disease surveillance project at the MDCH, focusing on HIV and hepatitis co-infection, reported that of HIV-infected individuals in Michigan, 12.0% were also diagnosed with HBV infection and 20.0% were also diagnosed with HCV infection [26]. However, no study has looked at the spatial structure of hepatitis B and C co-infection among HIV-infected individuals using the bivariate local Moran’s I and spatial scan statistic. The purpose of this study was to identify geographical clusters of HIV-(HBV/HCV) co-infection in the State of Michigan. Identification of the spatial clusters of HIV and HBV or HCV co-infection in Michigan will provide valuable knowledge for public health epidemiologists, policymakers, and health managers as to where to plan
Spatial epidemiology of HIV-hepatitis co-infection 853 and implement interventions and allocate resources. To the best of our knowledge, this is the first study to identify geographic clusters of HIV-(HBV/HCV) coinfection in the state of Michigan using data from a statewide population-based surveillance system. Material and Methods Data sources For this study, HIV/AIDS-infected individuals of all ages residing in Michigan’s counties were included and matched with all HBV and HCV cases in the period of January 1, 2006 through December 31, 2009 using a common identifier (identification number). Data were retrospectively obtained for the study period from the enhanced HIV/AIDS Reporting System (eHARS) maintained by the HIV/STD/VH/TB Epidemiology Section and the Michigan Disease Surveillance System (MDSS), which is the statewide communicable disease reporting system maintained by the Surveillance and Infectious Disease Epidemiology Section of MDCH. Ethical approval was obtained from the institutional review board at Michigan State University and MDCH. An employee of the HIV/STD/VH/TB Epidemiology Section performed the record linkage to ensure confidentiality and privacy of the participants. De-identification of the study participant data was done according to Health Insurance Portability and Accountability Act guidelines on public health information and no participants were interviewed or contacted. For this study, HIV/AIDS cases were defined using the CDC 1993 revised classification system of HIV [27]. The outcome variable was ‘Co-infection’ which was recoded into ‘co-infected’ 1 and ‘not coinfected’ 0. An individual with HIV/AIDS was categorized as co-infected if he or she had been concomitantly infected with confirmed HBV or HCV (acute and chronic) based on the CDC case definition [28–32] and residing in the state of Michigan during the period 2006 through 2009. Any individuals with HIV/AIDS who had a diagnosis of acute or chronic hepatitis B or C before 2006 were not included in this study. The characteristics of HIV-infected and co-infected individuals utilized in this research included sex, race, and current HIV status (status through December 2009). The variable sex was categorized as dichotomous, male 1 and female 2; race and ethnicity were classified into white non-Hispanic 1, black nonHispanic 2, Hispanic 3, and other 4; and current HIV status was dichotomized into HIV without AIDS 1 and AIDS 2. Statistical and cluster analyses A map of HIV prevalence by county was created for the state by using information on HIV cases from
Downloaded by [University of Otago] at 22:23 18 September 2015
854 Z. Butt et al. 2006 to 2009 based upon the administrative boundaries of MDCH [3]. Descriptive analyses of the HIV and hepatitis B or C data were performed in SAS version 9.2 (SAS Institute, Inc., Cary, NC, USA). The co-infected and non-co-infected individuals were compared using the Pearson chi-squared statistic and two-sided p values were reported. Spatial cluster analyses were performed using GeoDa version 1.6.7 [33] and bivariate local Moran’s I and SaTScan version 9.014, and discrete Poisson methods [34,35] were implemented. The bivariate local Moran’s I statistic estimates the spatial autocorrelation between HIV and HBV/HCV using a k-nearest neighbor weighting scheme. The autocorrelation statistic is visualized as the slope in the scatter plot with HBV/HCV as a spatially lagged variable on the vertical axis and HIV on the horizontal axis. Counties with high HIV and high hepatitis (High-High) are located in the upper right quadrant, Low-Low counties are in the lower left quadrant, Low-High counties are in the upper left quadrant, and High-Low counties are in the lower right quadrant. These county estimates are mapped in ArcGIS version 10.2 (Environmental Systems Research Institute, Redlands, CA, USA). Scan statistics are used for the detection and evaluation of clusters by applying a scanning window across the study area (i.e. the centroids of each county across Michigan). Within the window at each location, the number of observed and expected cases of HIV-hepatitis coinfected individuals is calculated and a likelihood function is estimated [36]. The window with the maximum likelihood is the most likely cluster, that is, the cluster that is least likely to have occurred by chance. Secondary clusters are also identified and ordered by the program according to the likelihood ratio test statistic. The discrete Poisson model assumes that the number of cases in each location is Poisson distributed. The null hypothesis for the model is that the expected number of cases in each scanning window is proportional to its population size [36]. The significance of the likelihood ratio was tested using 999 Monte Carlo simulations. In this study, a circular window was used and the maximum circle size included up to 25% of the HIV-infected population being analyzed within a 160 km (100 mile) radius [36]. In all, 25% of the HIV population was scanned as this was the approximate national HCV prevalence and a 160 km radius was hypothesized to be the extent within which transmission was likely to occur. A purely spatial Poisson analysis was performed. A significant p value was considered to be 0.05. The data for both the bivariate local Moran’s I and Poisson method were aggregated at the county level using SAS version 9.2 (SAS Institute, Inc.). For the bivariate local Moran’s I analysis a file with case
counts of HIV and HBV/HCV by county was constructed. For the discrete Poisson analysis, a case (hepatitis) and a population (HIV/AIDS infection) file were created with the county as the geographic unit and covariates sex, race, and current HIV status. A coordinates file containing the latitude and longitude at the centroid of each county was used to define the locations in both analyses.
Results A total of 13 936 individuals were infected with HIV/ AIDS during the study period (Table I). The prevalence of HIV-HBV or -HCV co-infection among these individuals was 4.2% (n 578). The majority of HIV-infected individuals also had AIDS (68.0%), were males (80.6%), and were black non-Hispanic (66.0%). There was a significant difference between HIV and HBV or HCV co-infected and nonco-infected individuals in regards to the sex (p 0.007), race (p 0.0001), and current HIV status (p 0.0001) of individuals. The HIV prevalence in Michigan by county of residence at time of diagnosis is provided in Figure 1 [3]. This map shows a higher prevalence of HIV infection in counties that contain urban areas in southern Lower Michigan, particularly in Wayne County (370 per 100 000 population), which contains the city of Detroit. Seven other counties with urban areas in Michigan also had high HIV prevalence rates including Washtenaw (151), Ingham (151), Berrien (148), Oakland (134), Genesee (127), Kent (127), and Kalamazoo (116), respectively. Importantly, in rural Michigan a higher HIV prevalence was also observed in the counties of Table I. Comparative analysis of HIV and hepatitis B or C co-infected (n 578) and non-co-infected (n 13 358) individuals by sex, race, and current HIV status in Michigan, 2006–2009.
Variable Sex Female Male Race White (non-Hispanic) Black (non-Hispanic) Hispanic Otherb Current HIV status HIV (not AIDS) AIDS aChi-squared
Co-infected, n (%)
Nonco-infected, n (%)
112 (19.4) 466 (80.6)
3235 (24.2) 10123 (75.8)
0.007
161 (27.9)
4797 (35.9)
0.0001
381 (66.0)
7710 (57.7)
13 (2.3) 23 (3.9)
557 (4.2) 294 (2.2)
185 (32.0) 393 (68.0)
6040 (45.2) 7318 (54.8)
p valuea
0.0001
test. Hawaiian & Pacific Islanders, Alaskan Native and American Indian and multiracial. bOther Asian,
Downloaded by [University of Otago] at 22:23 18 September 2015
Spatial epidemiology of HIV-hepatitis co-infection 855
Figure 1. HIV prevalence and cases of hepatitis B or C by county, 2006–2009. Note: To mitigate the effect of small numbers of cases, reported HIV prevalence in multi-county health departments are listed for the health department as a whole and not the individual counties.
Luce, Alger, Mackinac, and Schoolcraft (269 per 100 000 population) in the Upper Peninsula. The global Moran’s I 0.153 demonstrated spatial clustering of HIV and HBV/HCV in Michigan (Figure 2a). Five counties were identified as High-High (Genesee, Macomb, Oakland, Wayne, and Washtenaw) in the Detroit Metropolitan Area, with two contiguous counties (St Clair and Monroe) Low-High. Ten counties were Low-Low – Alcona, Crawford, Iosco, and Oscoda in Upper Lower Michigan and Baraga, Gogebic, Houghton, Iron, Marquette, and Ontonagon in the Upper Peninsula (Figure 2b). The Poisson cluster analysis identified two spatial clusters (Figure 3). The first most likely cluster had a relative risk (RR) of 1.38 (p 0.029) after adjusting for sex, race, and current HIV status (Table II). This cluster was located in western and northwestern Upper Lower Michigan and northern interior rural counties (n 46). Within this cluster there were 706 HIV- and 135 HBV/HCV-infected individuals who matched on sex, race, and HIV status. Of the
135 HIV-HBV/HCV co-infected individuals a majority were located in Genesee (n 38, local RR 1.1), Kent (n 32, local RR 1.9), and Ingham (n 16, local RR 1.0) counties. The secondary cluster was non-significant (RR 1.17, p 0.99) after adjustment. The secondary cluster included 583 HIV and 101 HIV-HBV/HCV co-infected individuals who matched on sex, race, and HIV status within three counties: Macomb (n 25, local RR 1.1), Oakland (n 70, local RR 1.1), and St Clair (n 6, local RR 1.9), all located in the Detroit Metropolitan Area. Importantly, this secondary cluster did not include Wayne County and Detroit. The local RR’s demonstrate the geographic variability of HIV-HBV/ HCV co-infection within these two discrete Poisson spatial clusters (Figure 3). Discussion HIV, HBV, and HCV share common transmission routes, which include parenteral, perinatal, and sexual
Downloaded by [University of Otago] at 22:23 18 September 2015
856 Z. Butt et al.
Figure 2. (a) Bivariate local Moran’s I scatter plot of HIV and hepatitis B or C by county in Michigan, 2006–2009. (b) Bivariate local Moran’s I clusters of HIV and hepatitis B or C by county in Michigan 2006–2009.
transmission [37]. The elevated cluster identified in the bivariate local Moran’s I was in the Detroit Metropolitan Area and surrounding counties where HIV prevalence and the numbers of HBV/HCV cases
were high. However, the most likely Poisson cluster was in western and northwestern Lower Michigan and northern interior rural counties outside of the Detroit Metropolitan Area, demonstrating that the
Downloaded by [University of Otago] at 22:23 18 September 2015
Spatial epidemiology of HIV-hepatitis co-infection 857
Figure 3. Discrete Poisson clusters of HIV and hepatitis B or C co-infection by county in Michigan, 2006–2009.
geography of HIV-HBV/HCV co-infection is different from the geography of HIV prevalence, which could be explained through the sex, race, and HIV status adjustment, other underlying differences in the populations at risk, and/or characteristics of the environments in urban versus rural settings. Additionally, there could be some unobserved factors that might influence the transmission of HBV and HCV infections among HIV-infected individuals that are different from the transmission of HIV infection
among the general population that have not been explained in previous research. One potential explanation for these geographic differences in HIV-HBV/HCV co-infection and nonco-infection is the lack of available or accessible health services in the counties with significant clustering. In this regard, enumeration of the health facilities as a proxy for health services revealed that most of the counties in cluster 1 had a limited number of hospitals as compared with other more densely
Table II. HIV-hepatitis B or C co-infection spatial clusters with high rates identified by SaTScan discrete Poisson method, Michigan 2006–2009.a Cluster Most likely cluster Secondary cluster aControlling
HIV/AIDS population (n) 706c 583d
Observed cases (n) 135 101
for sex, race, and current HIV status. risk observed/expected. cCluster includes 46 counties. dCluster includes three counties. bRelative
Expected cases (n) 104.77 88.48
Relative riskb
Log-likelihood ratio
p value
1.38 1.17
7.51 1.05
0.029 0.99
Downloaded by [University of Otago] at 22:23 18 September 2015
858 Z. Butt et al. populated counties [38]. According to the Michigan health facility atlas [38], Leelanau, Lake, Newaygo, and Benzie counties had one hospital each. A lack of hospitals or other healthcare services could affect access to healthcare and early diagnosis and treatment of HIV and hepatitis. Another reason could be that these counties are areas where injection drug use (IDU) is common. Data from MDCH indicate that 29.0% of all injecting drug users living with HIV in Michigan reside in counties with significant coinfection clustering [25]. Additionally, Traverse City in Grand Traverse, which borders Leelanau County, has a Traverse City Michigan syringe exchange program. Syringe exchange programs in a locality have been shown to be a proxy indicator for substantial IDU in a study conducted on hepatitis C [17]. Furthermore, Allegan County has been designated as a High Intensity Drug Trafficking area (HIDTA) [39] in Michigan by the Office of the National Drug Control Policy [40]. It is therefore conceivable that IDU as well as other high-risk behaviors are more common in this county as a result of drug-related activities, which may have led to an increased RR of HIV and HBV/HCV co-infection. To explore these potential
explanations for the clustering of HIV-HBV/HCV co-infection, the demographic, current HIV status, and route(s) of transmission characteristics were examined for co-infected and non-co-infected individuals residing within and outside of clusters 1 and 2 (outside of clusters 1 and 2 herein referred to as ‘outside’ (Table III). The characteristics of coinfected and non-co-infected individuals were compared by cluster 1 vs 2, cluster 1 vs outside, and cluster 2 vs outside using a Pearson chi-squared statistic and two-sided p values (Table IV). A summary of these findings is presented. In regards to sex, over 75% of co-infected and non-co-infected individuals inside and outside of HIV-HBV/HCV clusters were males. For co-infected individuals, there were no significant differences in sex composition across clusters 1, 2, and outside. For non-co-infected individuals, however, there were significant differences in the sex composition between males and females across clusters and outside (p 0.000). These findings demonstrate that the sex composition for the co-infected was less variable than the sex composition for the non-co-infected between clusters 1 and 2 and outside. Public health
Table III. Comparative analysis of HIV and hepatitis B or C co-infected and non-co-infected individuals by demographic, HIV status, and transmission characteristics inside and outside of clusters 1 and 2, Michigan, 2006–2009. Characteristics
Co-infected (%) Cluster 1
Sex Female Male Age (years) 20 2024 2529 3039 4049 50 Race Black (non-Hispanic) White (non-Hispanic) Hispanic Othera Current HIV status HIV (not AIDS) AIDS Route of transmission Blood transfusion Heterosexual female Heterosexual male High-risk heterosexual IDU MSM MSM & IDU Other/unknownb
Cluster 2
Non-co-infected (%) Outside
Cluster 1
Cluster 2
Outside
16.9 83.1
17.6 82.4
18.1 81.9
22.4 77.6
16.5 83.5
21.0 79.0
6.1 4.7 20.3 35.8 22.3 10.8
4.0 10.1 18.2 37.4 20.2 10.1
2.4 10.5 13.2 37.1 24.5 12.4
5.6 13.1 17.5 35.5 19.8 8.5
4.7 12.0 17.2 34.9 22.2 9.0
6.1 12.4 15.5 34.4 22.2 9.3
42.6 48.0 4.7 4.7
46.1 45.1 2.0 6.9
79.3 16.8 1.3 2.6
33.1 56.5 7.7 2.7
34.9 58.9 3.2 3.0
71.8 23.5 3.2 1.5
31.1 68.9
30.4 69.6
30.8 69.2
46.8 53.2
47.7 52.3
44.0 56.0
7.4 4.7 6.1 6.8 18.9 34.5 12.2 9.5
4.9 3.9 7.8 4.9 22.5 44.1 5.9 5.9
1.6 1.6 6.7 7.5 26.4 39.6 9.6 7.0
0.7 5.9 9.0 13.3 6.5 51.8 4.4 8.5
1.0 4.1 7.5 8.3 5.3 61.2 3.5 9.1
0.4 6.0 8.7 14.0 12.5 44.8 3.8 9.8
Co-infected: cluster 1 (n 148), cluster 2 (n 102), outside (n 386). Non-co-infected: cluster 1 (n 2869), cluster 2 (n 2368), outside (n 9053). IDU, intravenous drug use; MSM, men having sex with men. aOther multiracial, Asian, Hawaiian & Pacific Islander, Alaskan Native, American Indian. bOther/unknown perinatal and unknown routes of transmission.
Spatial epidemiology of HIV-hepatitis co-infection 859
Table IV. Comparative analysis of HIV and hepatitis B or C co-infected and non-co-infected individuals by demographic, HIV status, and transmission characteristics inside and outside of clusters 1 and 2, Michigan, 2006–2009. Comparison
Co-infected
Downloaded by [University of Otago] at 22:23 18 September 2015
c2 Cluster 1 vs 2 Sex Age Race Current HIV status Route of transmission Cluster 1 vs outside Sex Age Race Current HIV status Route of transmission Cluster 2 vs outside Sex Age Race Current HIV status Route of transmission
Non-co-infected c2
df
p valuea
27.142 7.556 54.136 0.448 67.609
1 7 5 1 8
0.000 0.373 0.000 0.503 0.000
0.571 0.107 0.000 0.955 0.006
18.430 15.848 1335.762 6.652 98.680
1 8 5 1 8
0.000 0.045 0.000 0.010 0.000
0.909 0.618 0.000 0.932 0.292
97.586 11.031 1128.778 10.336 259.216
1 8 5 1 8
0.000 0.200 0.000 0.001 0.000
df
p value
0.268 3.019 3.294 0.013 6.619
1 7 5 1 7
0.605 0.883 0.655 0.908 0.470
0.321 11.804 70.149 0.003 21.555
1 7 5 1 8
0.013 5.342 49.332 0.007 9.63
1 7 5 1 8
Co-infected: cluster 1 (n 148), cluster 2 (n 102), outside (n 386). Non-co-infected: cluster 1 (n 2715), cluster 2 (n 2267), outside (n 8672). Cluster 1, n 2869; cluster 2, n 2368; outside cluster 1 and 2, n 9053. aChi-squared test.
messaging for the proportion of co-infected men and women may be consistent across the state. Messages targeted for non-co-infected men and women will need to be tailored with the greatest proportion of men in cluster 2 and the greatest proportion of women in cluster 1. In regards to age, at least onethird of co-infected and non-co-infected individuals were between the ages of 30 and 39 years. There were no significant differences in age for co-infected and non-co-infected, except for non-co-infected individuals residing in cluster 1 vs outside (p 0.045). In cluster 1 there appears to be a slightly higher proportion of non-co-infected young adults compared with outside. These younger age groups in cluster 1 and the 30–39-year age group across the state should continue to be targeted for HIV and HIV-HBV/HCV co-infection public health messaging. In regards to race, for the co-infected there were no significant differences between clusters 1 and 2; however, there were significant differences in race between cluster 1 vs outside (p 0.000) and cluster 2 vs outside (p 0.000). In cluster 1 the co-infected were racially mixed with relatively similar proportions of white (48.0%) and black (42.6%) nonHispanics. In cluster 2 the co-infected were also racially mixed (white 45.1%, black 46.1%). However, outside of these clusters a majority of coinfected were black (79.3%) and white (16.8%) non-Hispanics. Importantly within HIV-HBV/HCV
co-infection clusters, race was less of a risk factor than outside of these clusters. For non-co-infected individuals the differences in race between clusters 1 and 2 and outside were statistically significant (p 0.000). The non-co-infected individuals within clusters 1 and 2 were not racially mixed, with slightly more white (56.5%, 58.9%) than black (33.1%, 34.9%) non-Hispanic subjects. Cluster 1 also had twice as many Hispanics as cluster 2 or outside (7.7% vs 3.2% vs 3.2%). Outside, the proportion of black non-Hispanics was higher (71.8%) and the proportion of white non-Hispanics was lower (23.5%) than in clusters 1 and 2. The spatial patterns of racial differences in the co-infected outside and nonco-infected in clusters 1 and 2 and outside appear to be linked to racial residential segregation and place. In regards to current HIV status, co-infected individuals had a higher proportion of AIDS than non-co-infected individuals in cluster 1 (68.9% vs 53.2%), cluster 2 (69.6% vs 52.3%), and outside (69.2% vs 56.0%). There were no significant differences in current HIV status for co-infected individuals in clusters 1 and 2 and outside. For non-co-infected individuals there were significant differences in current HIV status between cluster 1 and outside (p 0.010) and cluster 2 and outside (p 0.001), with outside having a slightly higher prevalence of AIDS. These findings suggest that healthcare and treatment for HIV within clusters is better than outside but there is a need for improved care and treatment of HIV-hepatitis co-infection. Finally, men having sex with men (MSM) was the most important route of HIV transmission among HIV-HBV/HCV co-infected and non-co-infected individuals. There were significant differences in route of transmission for the co-infected in cluster 1 vs outside (p 0.006). Their major differences were in the proportion of MSM (34.5%, 39.6%), IDU (18.9%, 26.4%), and MSM and IDU (12.2%, 9.6%). For non-co-infected individuals there were significant differences in route of transmission between clusters 1 and 2 and outside (p 0.000), demonstrating more variability in routes of transmission for non-co-infected compared with co-infected. Importantly, the proportion of MSM was higher across clusters 1 and 2 and outside for non-co-infected but the proportion of IDU was higher across clusters 1 and 2 and outside for IDU and MSM and IDU. These findings support the potential explanations for HIV-HBV/HCV clusters, including the need for improved healthcare to slow the progression of HIV-AIDS in co-infected individuals and the need for additional emphasis on IDU in addition to MSM as routes of transmission in HIV-HBV/HCV. Limitations of this study are that the HIV surveillance system did not have information on newly
Downloaded by [University of Otago] at 22:23 18 September 2015
860 Z. Butt et al. diagnosed cases of HIV, so unfortunately incidence rates were not calculated. In addition, the data on hepatitis included both acute and chronic hepatitis cases but the number of acute cases was extremely small and therefore did not lend itself to a study of hepatitis incidence. Furthermore, the date of diagnosis of hepatitis in the surveillance system was quite incomplete and therefore a temporal analysis or spatio-temporal analysis were not possible. From a public health perspective it would have been ideal to study these incident cases of HIV and hepatitis B or C and HIV-HBV/HCV co-infection to more directly target timely interventions in affected counties. Another limitation is it was unclear if the significantly low bivariate local Moran’s I clusters are due to low HIV and low HBV/HCV (Low-Low) spatial autocorrelation or if these are counties with under-reporting and underdetection of hepatitis cases as the MDSS is based on a passive surveillance system. The HIV surveillance system in Michigan is based on data for those persons who have been confidentially reported by name. Infected individuals who have not been tested, have been tested only anonymously, or have been tested by name but not reported, are not included in the MDSS, which could also lead to under-reporting of HIV cases [25]. A final limitation is that this study was conducted at the county level and there may be significant variation in population characteristics healthcare and public health interventions. While having HIV and hepatitis data for the Zip Code would address this problem, the countylevel data are population-based including all the statewide data collected by the surveillance system on HIV-infected individuals as well as persons having HBV or HCV co-infection residing in Michigan’s counties. The number of HIV and hepatitis cases at the county level was an adequate sample by which to investigate HIV and HIV-HBV/HCV co-infection and will provide a basis for future research on a smaller scale. In conclusion, this study identified significant clusters of HIV-HBV/HCV co-infection in counties that would not be considered high risk because of low population density and low HIV prevalence. In this respect, spatial cluster analysis serves as an important tool to delineate infectious disease clusters, which could be missed by other analytic methods that do not consider their geography. The results from this study can guide policymakers and health managers to target interventions and disease control measures to reduce HIV and HBV/HCV in these counties. In addition, efficient allocation of resources to these areas can be considered as a means to obtain maximum benefit from any measures undertaken to prevent the spread of hepatitis infection among the high-risk population of HIV-infected individuals.
Future research should continue to focus on identifying risk factors that are associated with clustering of HIV-HBV/HCV coinfection in these counties as well as modifiable factors that may help to prevent these infections. Acknowledgments The authors would like to thank the staff at MDCH for their continuous support during the study. We are also thankful to Dr Kim Kirkey for providing valuable insight regarding surveillance of hepatitis infections. Declaration of interest: The authors report no conflicts of interest. This work was supported by the Dissertation Completion Fellowship awarded by the Graduate School and the Department of Epidemiology at Michigan State University. References [1] Global Report. UNAIDS report on the Global AIDS Epidemic 2010. Available from: http://www.unaids.org/en/media/unaids/ contentassets/documents/unaidspublication/2010/20101123_ globalreport_en.pdf [2] United States of America. HIV and AIDS estimates 2009. Available from: http://www.unaids.org/en/regionscountries/ countries/unitedstatesofamerica/ [3] Quarterly HIV/AIDS report, Michigan. January 2011. Available from: http://www.michigan.gov/documents/mdch/ Jan_2011_344131_7.pdf [4] Barth RE, Huijgen Q, Taljaard J, Hoepelman AI. Hepatitis B/C and HIV in sub-Saharan Africa: an association between highly prevalent infectious diseases. A systematic review and meta-analysis. Int J Infect Dis 2010;14:1024–31. [5] Portelinha Filho AM, Nascimento CU, Tannouri TN, Troiani C, Ascêncio EL, Bonfim R, et al. Seroprevalence of HBV, HCV and HIV co-infection in selected individuals from state of São Paulo, Brazil. Mem Inst Oswaldo Cruz 2009;104:960–3. [6] Victoria MB, Victoria Fda S, Torres KL, Kashima S, Covas DT, Malheiro A. Epidemiology of HIV/HCV co-infection in patients cared for at the Tropical Medicine Foundation of Amazonas. Braz J Infect Dis 2010;14: 135–40. [7] Omland LH, Weis N, Skinhøj P, Laursen A, Christensen PB, Nielsen HI, et al. Impact of hepatitis B virus co-infection on response to highly active antiretroviral treatment and outcome in HIV-infected individuals: a nationwide cohort study. HIV Med 2008;9:300–6. [8] Weis N, Lindhardt BO, Kronborg G, Hansen AB, Laursen AL, Christensen PB, et al. Impact of hepatitis C virus coinfection on response to highly active antiretroviral therapy and outcome in HIV-infected individuals: a nationwide cohort study. Clin Infect Dis 2006;42:1481–7. [9] Kellerman SE, Hanson DL, McNaghten AD, Fleming PL. Prevalence of chronic hepatitis B and incidence of acute hepatitis B infection in human immunodeficiency virusinfected subjects. J Infect Dis 2003;188:571–7. [10] Chun HM, Fieberg AM, Hullsiek KH, Lifson AR, Crum-Cianflone NF, Weintrob AC, et al. Epidemiology of
Downloaded by [University of Otago] at 22:23 18 September 2015
Hepatitis B virus infection in a US cohort of HIV-infected individuals during the past 20 years. Clin Infect Dis 2010;50:426–36. [11] Viral hepatitis populations. Available from: http://www.cdc. gov/hepatitis/Populations/hiv.htm [12] Tanser F, Bärnighausen T, Cooke GS, Newell ML. Localized spatial clustering of HIV infections in a widely disseminated rural South African epidemic. Int J Epidemiol 2009;38: 1008–16. [13] Wand H, Ramjee G. Targeting the hotspots: investigating spatial and demographic variations in HIV infection in small communities in South Africa. J Int AIDS Soc 2010;13:41. [14] Westercamp N, Moses S, Agot K, Ndinya-Achola JO, Parker C, Amolloh KO, et al. Spatial distribution and cluster analysis of sexual risk behaviors reported by young men in Kisumu, Kenya. Int J Health Geogr 2010;9:24. [15] Kulldorff M and Information Management Services, Inc. SaTScan™ v9.0: Software for the spatial , temporal and space-time scan statistics. Available from: http://www.satscan.org/, 2009. [16] Jennings JM, Curriero FC, Celentano D, Ellen JM. Geographic identification of high gonorrhea transmission areas in Baltimore, Maryland. Am J Epidemiol 2005;161:73–80. [17] Bernstein KT, Curriero FC, Jennings JM, Olthoff G, Erbelding EJ, Zenilman J. Defining core gonorrhea transmission utilizing spatial data. Am J Epidemiol 2004;160:51–8. [18] Trooskin SB, Hadler J, St Louis T, Navarro VJ. Geospatial analysis of hepatitis C in Connecticut: a novel application of a public health tool. Public Health 2005;119:1042–7. [19] Haque U, Hashizume M, Sunahara T, Hossain S, Ahmed SM, Haque R, et al. Progress and challenges to control malaria in a remote area of Chittagong hill tracts, Bangladesh. Malar J 2010;9:156. [20] Mirghani SE, Nour BY, Bushra SM, Elhassan IM, Snow RW, Noor AM. The spatial-temporal clustering of Plasmodium falciparum infection over eleven years in Gezira State, The Sudan. Malar J 2010;9:172. [21] Haque U, Huda M, Hossain A, Ahmed SM, Moniruzzaman M, Haque R. Spatial malaria epidemiology in Bangladeshi highlands. Malar J 2009;8:185. [22] Ageep TB, Cox J, Hassan MM, Knols BG, Benedict MQ, Malcolm CA, et al. Spatial and temporal distribution of the malaria mosquito Anopheles arabiensis in northern Sudan: influence of environmental factors and implications for vector control. Malar J 2009;8:123. [23] Klassen AC, Kulldorff M, Curriero F. Geographical clustering of prostate cancer grade and stage at diagnosis, before
Spatial epidemiology of HIV-hepatitis co-infection 861 and after adjustment for risk factors. Int J Health Geogr 2005;4:1. [24] Osei FB, Duker AA. Spatial dependency of V. cholera prevalence on open space refuse dumps in Kumasi, Ghana: a spatial statistical modelling. Int J Health Geogr 2008;7:62. [25] Song C, Kulldorff M. Power evaluation of disease clustering tests. Int J Health Geogr 2003;2:9. [26] 2010 Epidemiologic Profile of HIV/AIDS in Michigan. Available from: http://www.michigan.gov/documents/mdch/2010_ EPI_PROFILE_ENTIRE_335483_7.pdf [27] 1993 revised classification system for HIV infection and expanded surveillance case definition for AIDS among adolescents and adults. MMWR Recomm Rep 1992;4:1–19. [28] Hepatitis B, acute. Available from: http://wwwn.cdc.gov/ nndss/conditions/hepatitis-b-acute/case-definition/2000/ [29] Chronic Hepatitis B virus. Available from: http://wwwn.cdc. gov/nndss/conditions/hepatitis-b-chronic/case-definition/2012/ [30] Hepatitis C virus infection, acute. Available from: http:// wwwn.cdc.gov/nndss/conditions/hepatitis-c-acute/casedefinition/2012/ [31] Hepatitis C Virus infection, past or present. Available from: http://wwwn.cdc.gov/nndss/conditions/hepatitis-c-past-orpresent/case-definition/2012/ [32] Michigan Department of Community Health. Health care professionals’ guide to the Michigan communicable disease rules (version 2006.1). Available from: http://www.michigan. gov/documents/hlth_care_prof_guide_167371_7.pdf [33] Anselin L, Rey SJ. GeoDa 1.6.7 (March 2015). Available from: https://geodacenter.asu.edu/ [34] Kulldorff M. A spatial scan statistic. Commun Statist Theory Meth 1997;26:1481–96. [35] Kulldorff M, Nagarwalla N. Spatial disease clusters: detection and inference. Stat Med 1995;14:799–810. [36] Kulldorff M. SaTScan™ User Guide for version 9.0; 2010 pp 1–107. [37] Heymann DL, editor. Control of communicable diseases manual, 19th edn. American Public Health Association; 2008. [38] Michigan Department of Health and Human Services. 2010 Health Facilities Atlas. Available from: http://www.michigan. gov/mdch/0,1607,7-132-2945_5106-63274--,00.html [39] The High Intensity Drug Trafficking Area Program: an overview. Available from: http://www.ncjrs.gov/ondcppubs/publications/enforce/hidta2001/overview.html#4 [40] Michigan HIDTA. Available from: http://www.whitehousedrugpolicy.gov/hidta/mich.html
