Methods
Our setting was the province of Ontario, which has the largest proportion (44%) of new HIV diagnoses in Canada. Our data source was the ongoing Ontario HIV Treatment Network Cohort Study (OCS); its study design has been described previously. Briefly, the OCS source population consists of voluntary, consented participants aged 16 and older diagnosed with HIV infection receiving medical care at 10 specialty HIV clinics. Participants were interviewed annually using structured questionnaires and clinical data were abstracted from clinic records. The study protocol, research instruments and forms received ethical approval from the University of Toronto Human Subjects Review Committee and from the study sites.
Chlamydia and Gonorrhoea Testing
We obtained testing data for HIV viral load and bacteriological tests for chlamydia and gonorrhoea through record linkage with the provincial Public Health Ontario Laboratories (PHOL), the sole provider of HIV viral load tests in Ontario and the primary provider for chlamydia and gonorrhoea tests submitted by STI clinics. In other clinical settings such as HIV clinics or primary care clinics, chlamydia and gonorrhoea tests may either be submitted to the PHOL or to private laboratories. The PHOL began keeping computerised records of chlamydia and gonorrhoea tests in 2008 and testing was available by culture or nucleic acid amplification testing (NAAT). Prior to 2009, NAAT was performed using the Becton Dickinson ProbeTec assay (BD Biosciences, Sparks, Maryland, USA). Starting in July 2009, the PHOL used the Gen-Probe Aptima assay (Gen-Prose, San Diego, California, USA). NAAT testing is only performed for urine, endocervical or urethral specimens, whereas culture is offered for genital and non-genital sites (eg, rectal, pharyngeal, conjunctival swabs). Culture is the only diagnostic method that allows for antibiotic resistance testing of Neisseria gonorrhoeae.
Each clinic participating in the OCS received a questionnaire to establish to which laboratories they submitted orders for chlamydia and gonorrhoea testing. Seven of 10 clinics responded that they submit all specimens to the PHOL; two were primary care clinics and the remainder were hospital-based clinics.
Analysis
There were 5933 OCS enrolees as of December 2011. We restricted the analysis to persons under observation at any time from 2008 to 2011 (1738 removed) and to participants who attended one of the seven clinics that submitted chlamydia and gonorrhoea tests to the PHOL (1030 removed). The latter exclusion ensured that all chlamydia and gonorrhoea testing ordered by the participating HIV clinic and any STI clinic in Ontario would be observable. The final sample size for analysis was 3165 participants. We conducted all statistical analyses using SAS V.9.3 (SAS Institute, Inc., Cary, North Carolina, USA). All p values were two-sided, and statistical significance was determined using the conventional p value of <0.05.
We used descriptive statistics to characterise participants included in the analysis and compared them with participants attending the three clinics excluded from the analysis. Next, we examined the proportion of participants that underwent testing at least once at any time from 2008 to 2011 and at least once in each calendar year when they were under follow-up ('annual testing').
We defined a case as a participant with ≥1 positive test result in a given calendar year. We calculated annual positivity rates among those tested. We calculated the proportion diagnosed with chlamydia or gonorrhoea among all participants under observation in each calendar year whether or not they underwent testing in that year. This is an underestimate of true prevalence since the numerator excludes undetected (likely asymptomatic) cases among untested patients.
We used multiple logistic regression and a generalised estimating equations framework with an autoregressive correlation structure to explore potential correlates of testing and risk factors for diagnosis of chlamydia and gonorrhoea. Each person-year was modelled as a unique observation and all ORs and proportions are reported with 95% CIs. Persons with unknown or missing information for a covariate were excluded from models given the small numbers with missing data. To determine whether a previous gonorrhoea or chlamydia test increased the odds of subsequent testing, we excluded the year 2008 and participants with ≤2 years of prospective follow-up (n=341) from our testing analysis. For the diagnosis outcome, we conducted a sensitivity analysis to explore whether there were any differences when restricting to testers. For both the testing and diagnosis outcomes, we first built a multivariable model containing all considered covariates, then excluded those that were neither associated with the outcome nor considered confounders for the remaining covariates.