Identifying and characterizing hepatitis C virus hotspots in Massachusetts: a spatial epidemiological approach

Approximately 71.1 million people are estimated to be currently infected with the hepatitis C virus (HCV) worldwide [1]. The number of people infected in the U.S. alone is estimated to be more than three million, of whom 75% were born between 1945 and 1965 [2]. Age-adjusted HCV mortality rates have increased steadily in recent decades [3], and notable increases in HCV cases have been reported in many states across the U.S. [4] At least 60% of prevalent HCV infections and nearly all incident HCV infections are attributed to injection drug use (IDU) [5‐8], with an increasing burden among young people who inject drugs (PWID) [3‐6] in non-urban, white communities [4].

Recent HCV surveillance data in Massachusetts (MA) identified a new HCV epidemic pattern characterized by a bimodal age distribution, with the expected peak among 45–65 year olds and an emerging peak among 15–30 year olds [9, 10]. There was a 78% increase in HCV cases among adolescents and young adults (15–24 years) in MA between 2002 and 2009, and an increase of 137% in the 15–29-year-old population between 2002 and 2013 [9, 10]. HCV surveillance is greatly underfunded across the United States [11], and limited spatial epidemiological and geostatistical analysis of surveillance data has been performed to date.

Innovative analytical approaches can foster a better understanding of disease transmission and improved preparation for targeted responses. Studies have demonstrated that GIS-guided approaches to disease screenings have a higher yield than traditional screening methods [12‐15]. Information on where infections cluster may enhance disease prevention, treatment, and care. For example, in Connecticut, researchers used a mix of statistical and geographic information system (GIS) methods to identify unique spatial distributions and characteristics for HCV and related infections [16]. Along the U.S.-Mexico border, spatial epidemiological methods were used to identify HIV clusters, and to determine that most seroconversions were occurring in a 2.5 block radius in the red light district of Tijuana [17]. In MA, we identified HCV and HIV all-cause mortality hotspots from 2002 to 2012 [18]. However, GIS and spatial analyses to date have been relatively simplistic in their approach and new, more methodologically rigorous spatial cluster analyses are needed to pinpoint locations (i.e., “hotspots”) with highest needs for intervention. The objectives of our study were to: (1) identify HCV clusters in MA using GIS and spatial epidemiology, and (2) to characterize the HCV clusters through statistical modeling.

We defined a hotspot as a location with a statistically significant cluster of census tracts with higher counts and rates of HCV than the average count/rate for all census tracts in MA [18].

Our combined use of GIS, spatial epidemiological, and statistical modeling approaches allowed us to identify and characterize statistically significant geographic HCV hotspot clusters. We observed agreement across the different geostatistical methods we conducted, which lends support to the idea that these locations in MA present significantly greater burdens of disease. Our HCV hotspots based on HCV counts highlighted nine clusters, the largest of which included Boston, Fall River, New Bedford, Worcester, and Springfield, where the burden of disease is highest. We detected hotspots for HCV rates per 100,000 persons in Boston, New Bedford, and Springfield. Given that these hotspots are based on rates, controlling for population differences across census tracts in MA, they may represent locations with some of the greatest need for enhanced HCV prevention and treatment interventions. It is notable that each of these cities has a population with lower median age than those found across MA, and that authorized syringe exchange programs have not been available in New Bedford and Springfield, despite initiation of such programs elsewhere in MA in the 1990s, and Worcester, Lawrence and a number of additional sites in 2016.

In our multivariable analyses, we found that the percent of population that was Hispanic and the percent of households receiving food stamps were independently and positively associated with HCV hotspots, while the percent of the population that was another race/ethnicity and the percent of the population that had a high school education or higher were negatively associated with HCV hotspots. These associations help us to understand the underlying sociodemographic factors that are associated with HCV clusters. The associations we observed with racial/ethnic groups differ from recent findings that highlight an emerging HCV epidemic among non-urban, white youth and younger populations [4]. However, the associations we observed between high school education, belonging to the “other” racial/ethnic groups, and HCV hotspots echo the significant health disparities that exist among different racial/ethnic minorities, educational attainment, and income levels for HCV [28]. Since the surveillance data upon which our analyses depend cover the timeframe from 2002 to 2013, it may not be possible to discern more recent patterns in disease transmission. Confirmed and probable cases of HCV infection include individuals that have past exposure, but not current infection. Analysis of more recent surveillance data may begin to uncover different patterns, including the emerging epidemic among young, suburban, white communities.

Our findings may have implications for future disease transmission risks and for the future success of HCV treatment among PWID, for which there are many sound recommendations [29]. PWID have experienced a number of barriers along the HCV care cascade, which includes diagnosis of disease, linkage to care, treatment, retention, and cure. Targeting of future treatment or cure as prevention (CasP) approaches [30, 31] could benefit from spatial epidemiological analyses that identify and characterize HCV hotspot clusters. Antiviral treatment uptake among PWID has been limited by barriers at patient, provider, and structural levels [32]. With the advent of direct acting antiviral (DAA) therapy, many payers, including most state Medicaid programs, continue to require abstinence from illicit drug use and often restrict treatment by fibrosis stage [33]. These policies systematically exclude many PWID who have not yet suffered from decades of infection and present barriers to performing real-life CasP interventions in locations and populations with the highest risk of HCV transmission. In MA, as of August 2016, all versions of Medicaid and most private insurers approve DAAs at fibrosis stage F0 or higher and without strict sobriety criteria, facilitating PWID treatment. However, multiple barriers remain for PWID along the cascade, resulting in a very small fraction likely to reach cure [31].

There are several limitations in the current study. First, HCV surveillance systems may not capture all HCV cases. While MDPH’s MAVEN surveillance and electronic laboratory reporting system has led to improved efficiency and timeliness in capturing HCV data [19], disease surveillance data only include information on individuals who have been tested and reported to health authorities. People who are not tested or are not engaged in care are not represented in these data. Race/ethnicity data are missing or incomplete for many HCV cases, making it difficult to discern representation in the HCV surveillance data by race/ethnicity. Second, the MAVEN system captures address at time of report, typically many years after initial HCV infection, and residence rather than place of infection. As a result, spatial results may represent risk for HCV transmission 10–20 years earlier. However, in the absence of large cohort studies that follow PWID over time, starting at a young age and allowing for calculation of HCV incidence rates and a better understanding of the geographic location where transmission occurred, such case reporting systems offer the best option currently available on a statewide level. Third, our dataset includes people who are positive for anti-HCV antibody through EIA. It is possible that some HCV cases tested positive for antibody but negative for HCV ribonucleic acid, which would over-estimate HCV cases of prevalent infection. Spontaneous viral clearance occurs in approximately 25% of people initially diagnosed with HCV [34]. However, it is unlikely that there are differential spatial distributions for reported HCV cases that ultimately clear the virus compared to cases that do not. Fourth, in our statistical models, we assessed associations at the census tract level, rather than the individual level. HCV surveillance (outcomes) and ACS data (explanatory variables) for the entire state are included in our analyses, and represent the best data available on the micro (i.e., neighborhood) level, and have been used in previous small area analyses with similar outcomes [35]. Finally, differences in characteristics of HCV cases with and without addresses and for matched and unmatched addresses following geocoding can help to illustrate which groups and characteristics may be underrepresented and can inform training needs for completion of future HCV case report forms, which can further enhance the precision of HCV surveillance data as well as spatial and statistical analyses.

Combined geospatial and statistical analyses can help identify regions of risk and regions of disease, and inform public health policy decisions. Through our analytical approach, we identified and characterized clusters in MA where enhanced HCV screening, prevention, treatment, and care could help to combat the HCV epidemic.