We performed a cross-sectional ecological study using the “basic health area” (ABS [acronym in Catalan]) as the geographical study unit. The ABS is the regional demarcation used by the Catalan Health Service to organize primary healthcare services. An ABS includes a reference territory of a primary health care team and its population, which comprises 5,000-25,000 people. Those ABSs with fewer than 5,000 people were merged with bordering ABSs in order to protect the statistical secret. During the study period, the limits of the ABSs were sometimes modified (basically ABS splits); this problem was solved by maintaining the original division and aggregating the data. Thus, we obtained a fixed number of 355 ABSs.
New diagnoses of HIV data
We retrieved data on new HIV diagnoses reported from 2012 to 2016 in the Catalan HIV Surveillance Registry. The Registry is a population-based information system of all persons residing in Catalonia at diagnosis of HIV infection since 2001. The Registry provides the postal code, sex, country of birth, age, and transmission mode. We calculated five-year aggregated new HIV diagnoses for each ABS. Multiple-year counts provide more stable estimates of HIV diagnoses, particularly in ABSs with small populations. New diagnosis rates per 100,000 persons were calculated by dividing the incident case count by the five-year sum of the population assigned yearly to each ABS.
The ABSs were classified into five categories (rural, semi-urban, urban county, urban metropolitan, urban Barcelona) to reflect territorial differentiation in Catalonia. These categories reflect both a rural-urban gradation and different geographical locations. The percentage of men aged 15-44 years among the assigned ABS population (year 2016) and the percentage of men from Western Europe and Latin America (year 2014) at the ABS level were calculated to account for differences in the proportion of men in the most common age range of new HIV diagnoses and the influence of the migrant population, respectively. The percentage for GBMSM as the transmission mode among new HIV diagnoses was calculated by ABS. Given the very small counts in many ABSs, the Agresti-Coull binomial proportion estimation was used to obtain less extreme values; this approach also has the advantage of imputing a value for ABS with zero incidence counts21. Relative socioeconomic disadvantage across Catalonia was evaluated using a socioeconomic deprivation index built for the assignation of budgets to the primary healthcare teams in Catalonia22. This index is a composite measure based on five indicators extracted from the national health registry of 2015, as follows: percentage of manual workers, percentage of people with a low educational level, rate of premature mortality, rate of avoidable hospitalization, percentage of population exempt from pharmaceutical copayment, and percentage of population with an annual income lower than €18,000. All these predictor variables were acquired from the set of basic health and health care indicators at the ABS level provided by the Agency of Health Quality and Evaluation of Catalonia (AQuAS) and the Information System for the Development of Research in Primary Care (SIDIAP).
Finally, to identify the gay neighborhoods, all gay locations (eg, bars, discotheques, saunas, hotels, shops, sex clubs, and travel agencies) located in Catalonia were retrieved from the Spartacus Guide and Universo Gay (2018 editions) and later georeferenced to their corresponding ABS. A total of 239 gay locations were mapped.
Spatial visualization methods
Multivariate ring maps based on Huang et al and López-De-Fede et al15,23 were constructed. A ring map is a map surrounded by a set of concentric, segmented rings. Each ring displays an additional data dimension that represents an attribute of a particular location. We built maps with a base map of Catalonia in the center showing the rates of new HIV diagnoses for each ABS surrounded by a set of histograms with the distribution of the mode of transmission of HIV, sex, and country of birth for ABSs with new HIV diagnoses rates > 10 per 100,000 population. The ring maps were developed using QGIS and Illustrator.
The univariate distribution of the new diagnoses of HIV infection, as well as its association with the predictor variables, was examined from both the statistical and geographical perspective using bivariate statistics.
A negative binomial regression model with the population at risk as an offset was adjusted to account for the overdispersion of new diagnoses of HIV infection. The exponentiated coefficients of the model are incidence rate ratios (IRR) relative to the reference category. Quantitative variables were scaled to make the IRRs comparable.
The Moran I test was used to check for spatial autocorrelation of the residuals, and a second model including an autocovariate based on the residuals of the first one was fitted. The AIC, BIC, and McFadden’s pseudo R2 were calculated to test the quality of the model. All calculations were performed using R.
The study was approved by the Ethics Committee of Hospital Germans Trias i Pujol.