STI epidemic re-emergence, socio-epidemiological clusters characterisation and HIV coinfection in Catalonia, Spain, during 2017–2019: a retrospective population-based cohort study

Study design and data source

We conducted a retrospective population-based cohort analysis of all confirmed cases of the notifiable STIs, syphilis, gonorrhoea, chlamydia and LGV, in Catalonia between 1 January 2017 and 31 December 2019. Data were obtained from the Catalan HIV/STI Registry,¹⁶ which uses information from the Epidemiological Repository of Catalonia (REC, in Catalan), an electronic database used by the Epidemiological Surveillance Network of Catalonia (XVEC, in Catalan). REC collects information from two sources: (1) the microbiological notification system (SNM, in Catalan) of confirmed cases from microbiological laboratories; and (2) the mandatory disease notification system (MDO, in Catalan) based on physician reporting of clinically suspected/probable and laboratory-confirmed cases as per established case definitions. Information collected through an epidemiological questionnaire that records clinical, epidemiological and behavioural variables are included along with the mandatory notification in REC (online supplemental table S1). Case definitions for surveillance reporting are standardised according to the European Union definitions established by the European Centre for Disease Prevention and Control (online supplemental table S2).^{17 18}

Analysis variables

We extracted data around epidemiological, sociodemographic and clinical variables as detailed in online supplemental table S3. All individuals who had experienced at least one STI episode during the study period were linked, through the Spanish healthcare system personal identification code (CIP), to the Catalan HIV/STI Registry to identify HIV coinfections either before or after the recorded STI episode. In addition to the CIP, the Catalan HIV/STI Registry surveillance team performs checks of duplicates at least twice annually using a unique STI episode number (assigned to each notification and disease), name and date of birth. For our analysis, a deduplicated, HIV/STI-linked and anonymised version was provided.

A Basic Health Area (BHA; Àrea Bàsica de Salut (ABS), in Catalan) is a territorial unit of coverage served by a primary healthcare team. Each BHA typically serves a population of approximately 5000–25 000 people. The socioeconomic level of the BHAs was classified according to a deprivation index (calculated by the Agency of Health Quality and Assessment of Catalonia) which was attributed to each individual according to their residential address. The deprivation index is a composite measure based on indicators such as proportion of residents with low educational level, proportion of manual workers, proportion of residents with an annual income below a specified amount, and rate of premature mortality. Deprivation indices were categorised in quintiles, with the first quintile being the least deprived.¹⁹

Clinical variables that were extracted included reinfections, multiple STI episodes and coinfection with HIV. Reinfection was defined as an episode of the same STI detected after a defined period, which differed for each STI, following the previously recorded infection in the same individual during the study period. Multiple STI episodes were defined as total number of episodes of any STI reported for the individual during the study period (online supplemental table S3). As information regarding treatment response was not available, episodes occurring outside of the specific time frames for each STI were assumed not to be a persistent infection resulting from treatment failure.

K-means clustering of STIs

We implemented k-means to define STI clusters by sociodemographic characteristics. Specifically, the k-means clustering methodology is an unsupervised machine learning approach that seeks to group heterogeneous units (in our case BHAs) into clusters based on similarities in characteristics (variable values and categorical distribution among the BHA).¹⁵ A clustering algorithm is a procedure for grouping a series of vectors according to a specific criterion, which could be distance or similarity. Proximity is defined in terms of a distance function and uses a k-means clustering method based on Euclidean distance to quantify similarities or differences between observations. This method is sensitive to outliers and requires both internal and external validation processes using different combinations of variables in the algorithms. The internal clustering validation considered an average Euclidean distance for each cluster and similarities between cases according to the correlation matrix of distances to determine the optimum number of clusters for which intracluster variation is minimum. Our external validation process was based on a description of the possible socio-epidemiological variables to determine the most appropriate clustering algorithm for our data set. Based on these validations, the researchers ended up with an exact number of clusters formed by defined variables which were included in the algorithm. In our case, the following variables were chosen to identify and build the final three socio-epidemiological clusters of STI by BHA: incidence rate by each STI, percentage of women, percentage of people with HIV coinfection, median age among all STI cases in each BHA and deprivation index of each BHA.

Statistical analyses

We performed a descriptive analysis to summarise epidemiological, clinical, sociodemographic and geographical variables for the total confirmed STI cases and by STI clusters. Continuous variables were summarised as median and interquartile range (IQR), while categorical variables were reported as absolute frequencies and percentages. Annual incidence rates of STIs are described per 100 000 population for the total confirmed STI cases and by STI cluster, and calculated based on census information from the Statistical Institute of Catalonia (IDESCAT) (online supplemental table S4). Incidence trends were analysed using the χ² test for linear trend. For identifying and building clusters in k-means clustering, STI incidence rates were described per 1000 population due to the small population size per BHA.

We assessed risk factors associated with HIV coinfection among individuals diagnosed with STIs using multivariable logistic regression models to estimate odds ratios (ORs) and 95% confidence intervals (CIs). Individuals with more than one STI episode were counted once (first episode), and successive episodes in the same individual were grouped in a variable that considers the number of episodes, and included in the models. Sexual preference, country of birth and education level were excluded from the models because more than 50% of values were missing. We used backward stepwise elimination regression to include all analysed variables that showed statistical significance (p<0.05) by the Wald test in the final multivariable logistic regression model. All analyses were performed using R Statistical Software V.3.6.1).

Ethics approval statement

Data from the Catalan HIV/STI Registry, REC and all aggregated variables used in the study were handled according to international recommendations, the Helsinki Declaration revised by the World Medical Organization in Fortaleza in 2013 and Spanish Law 3/2018 on Data protection and Public Health 33/2011. Patient information was anonymised and deidentified prior to analysis and therefore no informed consent was required.

Patient and public involvement

Patients were not directly involved in this study; only data from the nationally notifiable disease surveillance system were used.

STI epidemic and trends

Between 2017 and 2019, a total of 42 283 cases of STIs were reported among 34 600 individuals in Catalonia (table 1). Throughout the study period, half of all reported STIs were chlamydia (50.1%) and almost a third were gonorrhoea (31.6%). Reinfections accounted for 10.8% of all reported cases. Among the subjects affected by STIs, the events of gonorrhoea had the highest reinfection rate (15.7%), while chlamydia had the lowest occurrence of reinfection (6.7%).

The number of STI cases doubled from 9687 in 2017 to 18 872 in 2019 (table 2). The incidence rate of STIs increased by 91.3% from 128.2 cases per 100 000 population in 2017 to 248.9 cases per 100 000 population in 2019. The annual incidence rate of STIs for the period 2017–2019 was 185.5 per 100 000 population. Incidence rates increased significantly (p<0.001) from 2017 to 2019 for all STI types except for syphilis cases which remained stable over the 3 years, with the highest increase in number of cases seen in chlamydia (188.8%) followed by gonorrhoea (63.8%) and LGV (56.1%). In 2017, chlamydia and gonorrhoea represented 36.8% and 36.0% of all reported STIs, respectively, but by 2019 chlamydia accounted for 55.1% of all cases. Gonorrhoea showed the second greatest increase from 2017 to 2019, with 47.5% occurring in individuals under 30 years of age throughout the study period. This increase in the number of confirmed STI cases from 2017 to 2019 was remarkably higher in women (132% vs 75% in men) and individuals below the age of 30 years (125% vs 68% in those ≥30 years). Indeed, women under 30 years of age presented the highest decrease in both number of cases, with an increase of 155.8% versus 93.6% in men below 30 years and in incidence rates with an increase of 154.1% (from 193.6 to 491.9 per 100 000 population) versus 93.6% in men under 30 years (from 202.9 to 384.0 per 100 000 population).

The vast majority of reported cases occurred in men for all STI types except chlamydia, of which 61.9% occurred in women (table 1). Among all STI cases, 78.6% were reported in individuals below 40 years of age. Chlamydia was reported most frequently among individuals below 30 years of age (66.1%), while syphilis occurred most in those above 30 years of age (77.1%). Among the 15 023 (35.5%) reported STI cases for which information regarding sexual preference was available, half (54.5%) were reported in women who have sex with men (WSM), 21.8% in MSM and 21.0% in MSW (table 1).

When examining the distribution of STI cases according to deprivation index, the highest proportion of cases was seen in less deprived areas, with 24.3% of all cases reported in the first quintile. In more deprived areas (fifth quintile), chlamydia (56%) and gonorrhoea (29%) occurred more frequently than syphilis (14%) and LGV (1%). Data around country of birth and education level were limited due to high rates of missing data (56.6% for country of birth and 76.5% for education level). Nevertheless, we report that among cases with available information, 72.4% were observed among individuals born in Spain and 85.0% in those with secondary or higher education (online supplemental table S5).

The incidence rate of STI cases was disproportionately higher in Barcelona (83.3%) compared with the other six regions combined (table 1). Barcelona reported the highest incidence rate of STIs, while Alt Pirineu i Aran recorded the lowest consistently throughout the study period (table 2). In 2019, the incidence rate of STIs was 307.8 cases per 100 000 population in Barcelona and 45.7 cases per 100 000 population in Alt Pirineu i Aran. Nevertheless, incidence rates of STIs increased significantly from 2017 to 2019 in all regions, regardless of STI type. Similarly, the large majority of STI cases occurred in urban BHAs (70.9%) throughout the study period.

Factors associated with HIV coinfection among individuals with STIs

In total, 6% of STI episodes affected HIV-positive individuals with a higher proportion of HIV coinfection observed with cases of syphilis and LGV (13% and 25%, respectively) and the lowest with cases of chlamydia (2%) (table 1). Factors associated with HIV coinfection among individuals with STIs in the multivariable analyses are shown in table 3. The likelihood of HIV coinfection was greater among males (adjusted OR (aOR) 23.69; 95% CI 16.67 to 35.13 compared with females) and in urban BHAs (aOR 1.32; 95% CI 1.04 to 1.69 compared with rural BHAs). All age groups from 20 years and above and having multiple STI episodes were also associated with greater odds of HIV coinfection. BHA deprivation indices beyond the first quintile were associated with lower likelihood of HIV coinfection among individuals with STIs.

Identification and characterisation of the socio-epidemiological clusters of STIs

Of the 373 Catalan BHAs, five (Garraf rural, Polinyà-Sentmenat, Ribes-Olivella, Roquetes-Canyelles and Viladecans 3) were excluded from the K-means clustering analysis because their delimitations and populations changed during the study period. In these five BHAs, 679 episodes were reported during the 3 years of the study period. This fact and having 5773 episodes with no information available about BHA of residence reduced the sample size for the cluster analysis from 42 283 to 35 831 STI cases. Of the 368 BHAs included in the analysis, we identified three distinct clusters (table 4). Among the included BHAs, the incidence rate of STIs in 2017–2019 was 160.6 per 100 000 population per year.

Of the three clusters, the socio-epidemiological characteristics of STI-infected individuals in cluster A most closely resembled that of the total cases reported in the Catalan surveillance system that were included in the cluster analysis. Among the 109 BHAs in cluster A, median age was 31 years compared with 29 years among all reported STI cases, median deprivation index was 31.9 versus 39.8, the proportion of men was 67.4% versus 58.1% and HIV coinfection rate was 8.8% versus 6.1% (table 4).

Cluster B consisted of the largest number of BHAs (251) and had the highest deprivation index (44.9) of all three clusters and compared with the total. The incidence rate of STIs was lower in cluster B compared with that of the total reported cases included in the cluster analysis (136.3 vs 160.6 per 100 000 population), but represented the majority of all reported STI cases (55.7%). STI-infected individuals in cluster B were the youngest among all groups (26 years) and were predominantly women (53.0%), whereas men represented the majority in all other groups. Compared with the total, cluster B consisted of more rural BHAs (15.8% vs 11.0%) and had a higher proportion of heterosexual men and women (approximately 12% higher) and chlamydia cases (61.7% vs 55.0%). Rates of multiple STI episodes and HIV coinfection in cluster B were the lowest of all three clusters and compared with the total (table 4).

Cluster C consisted of only eight BHAs and had the lowest deprivation index (25.6) among all groups (table 4). The incidence rate of STIs in cluster C was the highest among all groups (721.0 per 100 000 population), with all cases reported in urban BHAs. STI-infected individuals in cluster C were the oldest among all groups (34 years) and had the highest proportion of MSM. Similar to other clusters and the total reported cases, chlamydia remained the most common STI type; however, cluster C was characterised by higher rates of gonorrhoea (33.2%), syphilis (23.6%), LGV (5.4%), multiple STI episodes (24.0%) and HIV coinfection (15.7%).

Almost 60% of STI cases in cluster B occurred in BHAs in the three lowest quintiles of STI incidence rates, while more than 60% in cluster A occurred in areas of high STI incidence rates (fourth and fifth quintiles). All 4359 STI cases in cluster C were reported in BHAs in the highest quintile of STI incidence rate. This correlated well with the fact the number of STI cases per BHA was higher in clusters A and C (105.8 and 544.9 cases per BHA, respectively) than in the total (97.4 cases per BHA), which indicates higher proportion of high incidence rates (table 4 and figure 1).

Patient consent for publication

Not required.