Men who have sex with men (MSM) remain the group most at risk of HIV infection in the United Kingdom. In 2004, 2185 new HIV diagnoses were reported among MSM, 30% of the total number of reported new HIV cases in the United Kingdom, a 45% increase since 2000.1 Data from the unlinked anonymous survey of genitourinary medicine (GUM) clinic attendees showed that previously undiagnosed HIV prevalence among MSM was 4.7% in London and 2.4% outside London.1 An estimated one in five of all newly diagnosed MSM were diagnosed late with serious immunosuppression.2

Detection of these undiagnosed HIV infections is now a major objective of the Department of Health. Unlinked anonymous survey data from 2004 suggest that the target uptake of HIV testing (60% by the end of 2007) cited in the National Strategy for Sexual Health and HIV3 has been exceeded; 79% among MSM. The proportion of HIV-positive MSM patients remaining undiagnosed after a clinic visit, however, remained relatively high at 44%.1

A number of clinic and regional audits in the United Kingdom, as well as studies from other countries, have determined several factors influencing HIV test offer and uptake. These include clinic test offer policy (“opt in” or “opt out”),4 5 waiting time for results,4 6 perception of risk and risk behaviour,7^–9 age10 and ethnicity.11 Evidence has been inconclusive, especially around the association of risk behaviour, such as the number of recent partners and having a current sexually transmitted infection (STI), with testing rates.7 10 12 13

The aim of this cross-sectional study was to carry out a national investigation of current testing rates to determine what proportion of MSM are currently being offered and accept an HIV test, as well as to identify what clinic and patient characteristics are associated with test offer and uptake. Understanding the factors that may contribute to the early identification of HIV infections will have both individual and public health benefits and the findings from this study will be used to recommend changes to HIV testing policy in MSM.

METHODS

Questionnaires and study forms were sent to all 238 GUM clinics in the United Kingdom, as identified by the British Association for Sexual Health and HIV (BASHH) audit committee. A consecutive sample of MSM patients attending participating clinics during the week beginning 7 February 2005 was included. Individual patient questionnaires were completed from case note review for a minimum of 10 and a maximum of 30 MSM per clinic. When fewer than 10 patients were seen during the specified week, the time period was extended retrospectively (and in some cases prospectively, n  =  76 records).

All MSM attending a GUM clinic as new patients, re-attendees (having attended the clinic previously but visiting with a “new episode”), or as follow-up patients, were included in the sample. MSM known to be HIV positive and those attending specifically for an HIV test were excluded. Data were collected on patient demographics including age and ethnicity, risk factors including unprotected anal intercourse (UAI), number of sexual partners in the past three months and time since last HIV test. Whether the patient was offered and accepted an HIV test, the HIV test result and the reasons for not offering or refusal of a test offer were recorded.

When data were missing for more than 5% of the sample for a given variable, the distribution of data was compared against those for which it was not missing in relation to the outcomes and other variables. The type of patient and throughput of clinic met these criteria; 6.4% were missing for both variables. No obvious variation was found, indicating no systematic bias.

Univariate and multivariate logistic regression analyses (Stata version 8.2; Stata Corp, College Station, Texas, USA) were carried out to examine associations between clinic characteristics, sociodemographic and behavioural variables with offer and uptake of test. The models were constructed using a backward selection technique. Variables with the weakest association with the outcome were omitted first. Criteria for inclusion were p<0.10 or odds ratio greater than 2.0 and other variables were retained if they were confounders. This criterion was applied to pairwise comparisons with the reference category as well as global p values. When two variables were found to correlate, one was omitted. Likelihood ratio tests were used to assess what model best fit the data and whether variables should be inputted as categorical or linear.

RESULTS

The sample consisted of 2162 patients seen from January 2003 to May 2005 who attended one of the participating 189 clinics; 79% of clinics in the United Kingdom. The proportion of new patients, re-attendees and follow-up patients was 50%, 38% and 12%, respectively; the median age was 32 years (range 14 to 80 years). Eighty-six per cent of the sample (1853/2162) were white; other ethnic groups comprised less than 1% of the sample each. Less than 2% of the sample had a history of intravenous drug use. Data for new and re-attending were analysed together, because by definition all were attending with a new episode. Follow-up patients (n  =  240) were excluded from the multivariate analysis as it is likely they were offered an HIV test at the first visit, as were patients for whom it was clearly stated that the test was deferred as a result of a risk within the three month window period (n  =  89). Data on clinic policy are presented in table 1 and the reasons for non-offer and uptake of HIV tests in table 2. Results of the multivariate analysis are presented in tables 3 and 4.

Clinic policy

All clinics reported offering a test to all new MSM on the first visit; however, only 61% of re-attendees are offered an HIV test irrespective of risk (table 1). In addition, only a quarter of clinics have a written policy for regular testing intervals among MSM. Nearly two-thirds of clinics (79%, 122/159) test according to an identified risk, and approximately one-third of these (36%, 44/122) also test according to defined time intervals.

Clinic practice

Offer

Overall, 86% of the sample (1862/2162), 97% of new patients and 86% of re-attending patients, were offered an HIV test. Patients attending for follow-up were least likely to be offered a test (39%), whereas new patients and those attending for a routine STI screening or contact tracing were the most likely to be offered (almost 100%).

Reasons recorded on the questionnaires for not offering an HIV test are shown in table 2, the most common being that the patient had tested negative less than three months ago.

Patients attending for an STI screening were three times more likely to be offered an HIV test in comparison with those attending with a symptom of an STI, and UAI within the past three months was associated with a high rate of offer. There was a linear correlation between increasing numbers of recent sexual partners and the likelihood of an offer. The adjusted odds ratio for opt-out policy was 1.4 (0.94 to 2.13, p = 0.1). The throughput of clinic and result of last test were dropped from the multivariate model (table 3). KC60 routine surveillance data were used to estimate the throughput of clinics and grouped by interquartile ranges. The total number of diagnoses in 2003 was used as a proxy for the magnitude of patients visiting each clinic (when data were missing for 2003, 2002 data were used).

Acceptance

Overall, 82% of those who were offered a test, 84% and 79% of new and re-attendees, respectively, accepted (70% of the entire sample).

The most common reason recorded on the questionnaire for the patient refusing an HIV test was a potential exposure within the three-month window period for seroconversion (table 2).

Reason for attendance, age, number of partners and UAI in the past three months, time since last test, and waiting time for results were all associated in the univariate analyses with the probability of accepting a test. The association with the number of partners was weakened in the multivariate analysis. There was, however, an increased likelihood of accepting a test if the last HIV test was more than three months ago and the overall association was strengthened. MSM with STI symptoms were less likely to accept a test than those attending for a screening, as were older MSM. The length of time to wait for results remained associated (table 4).

DISCUSSION

We conducted a cross-sectional study of 2162 MSM attending 189 GUM clinics in the United Kingdom. This was the first national study of this kind and the aim was to examine rates of HIV testing and factors associated with the offer and uptake of HIV testing among MSM. The data show that rates of offer and uptake are high and are similar to those cited in other research.1 14

We first examined factors associated with the offer of HIV tests to MSM; rates of offer were associated with indicators of risk exposure such as the number of recent sex partners and symptoms of an STI. Whereas an increase in the number of sexual partners was associated with an increased rate of being offered an HIV test, attending a clinic with symptoms of an STI was associated with a decreased rate of offer. Data on clinic policy show that only two-thirds of clinics offer testing irrespective of risk, which is consistent with the association with offer and partners. Conversely, the association with offer and STI symptoms may highlight a distinction made by physicians between the treatment of STI and performing routine sexual health screenings. This indicates though that MSM patients with clear risk indicators are not being tested for HIV.

Univariate analysis showed that having an opt-out policy was associated with an increase in the offer of HIV testing but the adjusted odds ratio showed a weakened effect; its inclusion did not modify the effect of other variables in the model. Data found in other studies suggested that an opt-out policy may be associated with higher rates of testing.4 6 In this study it is likely that non-differential misclassification bias may have weakened the magnitude of effect, via incorrect reporting of policy or variation within clinic practice by individual staff rather than confounding. There were some differences in the distribution of risk factors of patients by the type of clinic testing policy however these factors were adjusted for in the multivariate analysis. Rather, it is possible that this variable was not measured sufficiently for its impact on the outcome to be captured effectively.

Second, we examined factors associated with the uptake of testing by MSM who were offered an HIV test. In the multivariate analysis lower uptake of an HIV test was associated with symptomatic STI infection, increasing age, no previous HIV test and the time to wait for results. Our data concur with findings from smaller UK studies.7 9 12 13 15

Despite a clear association between time to results and uptake of test, there was a very weak association between the availability of point of care testing (POCT) and uptake in the univariate analysis. As data on POCT availability were collected at the clinic level, it is possible that some patients attending clinics where it was available were not offered this testing option. As there is reasonable evidence that the time to results was associated with the offer of a test POCT may play a role in improving uptake.

When examining both the offer and uptake of HIV testing we found that a small but important minority of MSM not testing, 14% (89/641), are having their HIV test deferred because of the window period. Other research has suggested that the probability of onward transmission is highest near the point of seroconversion when viral load is high.16 A policy of testing during the window period may increase testing rates among MSM at most risk, who might otherwise keep deferring. Patients presenting with symptoms of an STI, who by definition have had a recent potential exposure to HIV infection, could be within the window period and are at a higher risk of both acquiring and transmitting an HIV infection.17 18 Only 25% of clinics in this sample reported a system for recalling patients for testing and there is evidence that patients asked to re-attend three months after a potential exposure to HIV rarely do so.19 In addition, with recent improvements in the sensitivity of HIV tests, antibodies can often be detected within six weeks of infection,20 thus deferring testing, particularly in MSM with ongoing exposure risk, may be inappropriate.

Clinic data indicate that HIV testing policies may not always be documented, that policy may differ for new and re-attending MSM and that testing policies often combine testing according to risk as well as regular testing at intervals.

Current Department of Health guidelines suggest the routine offer of an HIV test at the first visit and then subsequently according to risk. In September 2006 the Centers for Disease Control and Prevention recommended opt-out testing as part of routine healthcare, with a view to reducing barriers to testing and identifying undiagnosed infection.21 The report highlighted the importance of testing irrespective of risk and recommended repeat screening for high-risk groups such as MSM. A move towards universal testing of all MSM attending GUM clinics with a new episode in the United Kingdom would avoid the reliance on patients disclosing sensitive information, the need for healthcare staff to make judgements on likely or actual risks and could contribute to the normalisation of HIV testing.

Limitations

The main limitations of the study involve potential information bias, specifically misclassification bias. Questionnaires were completed via case note review by GUM staff and thus relied on the completeness of records and transcribing accuracy. Many records had missing data for the time since last test; this may be indicative of the lack of information regarding testing behaviour, particularly if MSM access multiple GUM sites. As described previously, issues around defining policy, and multiple processes within clinics may have also had an impact on the strength of associations.

The final study period was two years, a change from the original protocol that specified a one week period. This was in order to ensure smaller clinics contributed a minimum of 10 MSM records and were adequately represented in the sample. This may have diluted the effects of HIV testing policy on offer as it is possible that patient data may have been collected during a period of opt-in policy, for example, when testing policy later changed to opt out and it was this latter policy captured in the study. Despite this, 76% of patient records (1633/2162) were collected within a six-month period.

Despite these limitations, there is evidence that a change in policy may serve to improve the detection rate of HIV infections.