Article Text
Abstract
Objective To determine whether the 2012 Olympic and Paralympic Games were associated with a change in the number of patients attending or diagnosed with a new sexually transmitted infection (STI) at sexual health clinics in London and Weymouth.
Methods We undertook an interrupted time-series analysis of surveillance data from the Genitourinary Medicine Clinic Activity Dataset (GUMCAD) collected at 33 genitourinary medicine (GUM) clinics in London and Weymouth (where Games events were concentrated) between 2009 and 2012. Mixed-effects linear regression models of weekly attendance and diagnoses, incorporating temporal trends, bank holidays, categorical month and clinic closures, were used to test for the effect of the ‘Olympic–Paralympic’ period. We subdivided the 9-week ‘Olympic–Paralympic’ period (16 July 2012 to 17 September 2012) into five periods, including three Olympic weeks, two Paralympic weeks, pre-, post- and inter-Games weeks. We also compared characteristics of patients attending during the Olympic–Paralympic period and those attending during the same period in 2011.
Results During the 3 weeks of the Olympics, there was a significant reduction in the number of new episode attendances (2020 fewer, 5.6% reduction (95% CI −8.2 to −2.9)) and the number of patients diagnosed with an STI (267 fewer, 4.8% reduction (95% CI −8.6 to −0.9)) compared to expected. There were no important differences in the profile of patients attending during the 2012 Olympic–Paralympic period and those attending during the same period in 2011.
Conclusions We conclude that a ‘business-as-usual’ approach to managing local sexual health clinics during the 2012 Olympic and Paralympics would have been appropriate.
- PUBLIC HEALTH
- EPIDEMIOLOGY (GENERAL)
- SURVEILLANCE
- GENITOURINARY MEDICINE SERVICES
- HEALTH SERV RESEARCH
Statistics from Altmetric.com
Background
Predicting the impact on services is essential for managing mass-gathering events. In the run-up to the London 2012 Olympics and Paralympics (the ‘London Games’) held in July and August 2012, speculation about the potential for the Games to affect sexual health identified a need for action from those planning and providing sexual health services.
Over eight million tickets were available for Olympic events and two million for Paralympic events. Most events were held in London venues, although the coastal town of Weymouth provided another hub hosting the sailing events. Domestic and international visitors were anticipated to increase London's population by 500 000 during the Games.1 Sexual health service providers were informed in a briefing document that while the estimates were uncertain, an increase of 5%–10% in service activity was anticipated across the health sector.2 ,3 The same briefing informed providers that a change in the nature of clinic attendees might occur, with possible increased demand for emergency contraception and post-exposure prophylaxis, and an increased proportion of patients presenting with symptoms.
The difficulty for those planning sexual health services in preparation for the London Games was the paucity of good quality data from previous Games, and similar mass-gathering events, to inform decisions about the most likely impact, and therefore what action to take. A literature review in 2010 undertaken on the topic to inform planning for sexual health services identified only one study.4 Since this literature review, more studies investigating the impact of major sporting events on sexual health have been published.5–7 One study identified in the pre-Games literature review investigated the impact of the 2000 Sydney Olympics on attendance at one sexual health clinic, and reported that compared to the same period in 1999, attendance was higher during the Olympics; there was a change in patient profile with increases in the proportion of new patients who were non-local, had symptoms or had a sexually transmitted infection (STI) diagnosed.8
We aimed to determine whether the London Games affected attendance patterns and patient profiles at genitourinary medicine (GUM) clinics in London and Weymouth, using multi-centre time-series analysis, in order to help inform planning for future mass-gathering events.
Methods
Study population
We conducted a time-series analysis study using routine surveillance data from GUM clinics. Our study population was patients attending London and Weymouth GUM clinics for new episodes of care. We excluded follow-up attendances for the same treatment episode. Olympic–Paralympic athletes and their support teams were outside our study population as their healthcare needs were provided by special clinics within the Olympic village, which we were unable to obtain data from. We restricted our analysis to clinics in London and Weymouth as these were the areas where Olympic and Paralympic events were concentrated and where visitors were considered most likely to stay. We included data from 32 of the 33 London GUM clinics and the one Weymouth clinic. One small London clinic that had only started reporting data in 2010 was excluded.
Study period
We limited our analysis to new episode attendances between 2009 and 2012. We followed the International Organisation of Standardisation convention for formatting weeks termed ‘isoweeks’. We defined a 9-week Olympic–Paralympic period in the summer of 2012. We created a categorical variable to define the period of interest with six mutually exclusive categories: 1 week before the Olympics (week 29), weeks where Olympic events took place (weeks 30–32), inter-games period where no events took place (weeks 33–34), weeks where Paralympic events took place (week 35–36), 1 week following the Paralympics (week 37), which closely matched the actual games weeks, and weeks outside the Olympic and Paralympic period (200 weeks).
Data source
Our data source was the national Genitourinary Medicine Clinic Activity Dataset (GUMCAD), an electronic surveillance system administered by the Health Protection Agency (and Public Health England since April 2013). All GUM clinics in England, which are the main provider of NHS sexual health services, have provided mandatory quarterly extracts of their clinic activity to GUMCAD since the surveillance system was established in 2009. GUMCAD holds anonymised patient-level data on attendance date and type (new episode, follow-up attendance), services provided, diagnoses and patient characteristics.
Data analysis: descriptive
We described the profile of patients attending GUM clinics during the Olympic–Paralympic period in terms of age, gender, sexual orientation, ethnic group, whether they were born in the UK and whether they were London or Weymouth residents attending London or Weymouth clinics respectively (termed ‘local residents’). We compared this with the profile of patients attending clinics in the same 9 weeks in 2011.
Data analysis: analytical
We conducted interrupted time-series analysis using a mixed-effects regression model with two outcome variables—total weekly new episode attendance and total weekly number of new STIs diagnosed at GUM clinics in London and Weymouth 2009–2012—in order to test the hypothesis that the level and type of attendance at clinics were affected by the 2012 Olympics and Paralympics. We hypothesised that the level of clinic attendances and the number of new STI diagnoses would be increased during the Olympic–Paralympic period.
Separate regression models were built using two different outcome measures. New STI diagnoses were defined as syphilis (primary, secondary and early latent), chlamydia, gonorrhoea, genital herpes (first episode), genital warts (first episode), chancroid, lymphogranuloma venereum (LGV), donovanosis, non-specific genital infection, scabies, molluscum contagiosum or a new HIV diagnosis. Outcome variables were obtained by aggregating the dataset from individual attendances to weekly totals by clinic.
We built a mixed-effects linear regression model with GUM clinic as the random effect, allowing clinics to have their own slope and intercept. We included a non-linear temporal trend (isoweek and isoweek2) for categorical ‘month’ (13 periods of 4 weeks), bank holidays and clinic closures, and one outlier, isoweek six in 2011, as variables in the model. We included an indicator variable for isoweek six in 2011 as low-observed attendance in all 33 clinics in isoweek six of 2011 occurred as a result of exceptionally high snowfall that week.
We compared the observed attendance and patients diagnosed with new STIs for each of the time-categories in the Olympic–Paralympic period with the expected values based on the model estimates and calculated the percentage change.
All our data cleaning and analysis were conducted using STATA V.12.
Results
Descriptive
During the 9-week Olympic–Paralympic period, there were 104 703 new episode attendances at GUM clinics in London and Weymouth. Seventy-two per cent of attendees were aged 16–35 years, 52% were female, 79% were heterosexual, 58% were of white ethnicity, 54% were born in the UK and 92% were local residents (see table 1). In the 2011 reference period, there were fewer new episode attendances (n=94 383). Compared with the 2011 reference period, during the Olympic–Paralympic period, there was a 5.9% reduction in the proportion of patients known to be UK born (53.9% vs 57.1%), but a 0.5% increase in the proportion known to be local residents (92.2% vs 91.7%). No other important differences were identified between patient characteristics in the two time periods.
Analytical: new episode attendance
Between 2009 and 2012, median weekly new episode attendance across all 32 clinics was 10 072 patients, range 4160–13 129 (IQR: 9492–10 919). The median weekly new episode attendance per clinic ranged from 117 to 739 patients. Between 2009 and 2012, weekly attendance time-series showed a gradual increasing trend, with considerable variation, and regular troughs corresponding to bank holidays, especially around Christmas (figure 1).
Our model found categorical month, bank holidays, clinic closures and isoweek six were all significantly associated with weekly attendance. The temporal trend between 2009 and 2012 was significantly non-linear, (linear trend 0.44 additional attendances per week per clinic (95% CI 0.19 to 0.69), squared term 0.003 (95% CI 0.001 to 0.004)), indicating an increasing number of additional attendances each week over this period. Bank holidays resulted in significant reductions in attendance, ranging from −162.3 (95% CI −171.0 to −153.5) per clinic during weeks with 2 days of Christmas bank holidays to −59.4 (95% CI −66.3 to −52.5) in weeks with a 1-day Easter bank holiday. Significant variation was observed by month, even after adjustment for temporal trends in week and bank holidays, with the greatest effect in months 12 and 13, with 18.1 fewer attendances in each clinic in month 12 (95% CI −23.6 to −12.51) and 33.9 fewer attendances in month 13 (95% CI −39.6 to −28.1).
After adjusting for the above variables, we detected a significant reduction in attendance at clinics during the 3 weeks of the Olympic Games. Our model found that on an average, clinics saw 20 fewer patient attendances a week for the 3 weeks during the Olympic Games (95% CI −30.6 to −10.3) compared to the expected attendance (table 2). In total, this equates to 2020 (95% CI −3029.4 to −1019.7) fewer patient attendances across all 33 clinics for the 3 weeks of the Olympic Games. This represented a difference of 5.6% (95% CI −8.2% to −2.9%) between observed and expected attendance. Smaller reductions were detected during the Paralympics (−10.3 (95% CI −22.3 to 1.70)) and the inter-games period (−10.9 (95% CI −22.9 to 1.10)); however, these were of borderline significance. No significant effects on attendance were detected during the week before the Olympics or the week following the Paralympics.
Impact on new STI diagnoses
Between 2009 and 2012, the median weekly number of patients diagnosed with a new STI across all 33 clinics was 1649 (range 686–1757, IQR: 1551–1757). Across the 33 clinics, the median number of patients diagnosed with a new STI ranged from 14 to 140 per clinic. Between 2009 and 2012, the pattern of weekly number of patients diagnosed with a new STI appeared to show a slight increasing trend, less pronounced than for attendance, and again there was considerable variation between weeks and regular troughs corresponding to bank holidays, especially around Christmas (figure 1).
Similar associations were detected in the multivariable model for new STIs diagnosed as in the attendance model. A significantly increasing non-linear trend was identified between 2009 and 2012 (linear trend 0.044 (95% CI 0.015 to 0.074), squared term 0.0004 (95% CI 0.0002 to 0.0005)). Bank holidays resulted in significant reductions in patients diagnosed with an STI, ranging from −23.5 (95% CI −25.5 to −121.6) per clinic during weeks with 2 days of Christmas bank holidays to −8.93 (95% CI −10.5 to −7.41) in weeks with a 1-day Easter bank holiday. Significant variation was observed by month, after adjustment for other factors, with the greatest effect in ‘months’ 12 and 13, the last 8 weeks of the year, with 3.86 fewer patients diagnosed with an acute STI per clinic in ‘month’ 12 (95% CI −5.09 to −2.63) and 7.48 fewer patients diagnosed with an acute STI per clinic in ‘month’ 13 (95% CI −8.75 to −6.21).
After adjustment for the above variables, we detected a significant reduction in the number of patients diagnosed with a new STI at clinics during the 3 weeks of the Olympic Games (p=0.018). Our model found that three fewer patients (95% CI −5.0 to −0.5) were diagnosed with an STI per clinic for the 3 weeks during the Olympic Games compared to the expected number of patients diagnosed (table 3). In total, there were 267 (95% CI −495.0 to −49.5) fewer patients diagnosed across all clinics for the 3 weeks of the Olympic Games. This represented a difference of 4.8% (95% CI −8.6% to −0.9%) between the observed and expected number of patients diagnosed with an STI; slightly lower, but similar to the reduction in total first attendances. No other significant effects were detected for the other Olympic–Paralympic periods.
Discussion
Main findings
We found that during the 3 weeks of the Olympic Games, an estimated 2020 fewer patients attended sexual health clinics in London and Weymouth for new episodes of care than we predicted based on attendance patterns over the previous 4 years. This amounted to a significant reduction in attendance of 5.6% (95% CI −8.2% to −2.9%). The number of patients attending clinics in London and Weymouth who had a new STI diagnosis was also significantly lower than expected with 267 fewer patients diagnosed during the 3 weeks of the Olympic Games, a reduction of 4.8% (95% CI −8.6% to −0.9%). No important differences in patient characteristics were observed between the Olympic–Paralympic period and the same period in 2011, including no major change in the proportion of patients who were local residents.
Strengths and limitations
We have undertaken a detailed multi-site study using time-series techniques that enabled more sophisticated modelling, which seemed to be lacking from the evidence base on the impact of mass-gathering events on sexual health services. This study has been feasible given our access to a comprehensive electronic sexual health surveillance system, which enabled us to conduct this multi-clinic investigation and to conduct an interrupted time-series analysis. Our methodology was rigorous and appropriate, allowing us to develop a more sophisticated model of ‘expected’ values, compared to previous analyses that have used crude reference periods.
Inevitably, our study has limitations. First, it is possible that residual confounding from unmeasured variables, such as the weather or other social factors, may affect our results. Second, our dataset is limited to GUM clinics in London and Weymouth, the main source of sexual health services for local residents. We do not have data on other providers of sexual health services, including private clinics or general practice surgeries. We also do not have data from the special clinic provided for athletes and the ‘Olympics and Paralympics family’, including officials and relatives in the Olympic village. Third, our investigation of the number of patients diagnosed with an acute STI should be interpreted alongside the attendance analysis as a descriptor of the profile of patients attending clinics. This study did not seek to investigate the impact of the Olympics on transmission of STIs, recognising that varied incubation periods and delay in seeking healthcare would result in a diffuse time period of effect, difficult to distinguish from background rates; that visitors may have returned home by the time of symptom onset; and that a proportion of individuals who acquired STIs related to the Olympics would remain undiagnosed. An additional limitation is the difficulty of matching the actual dates of Olympic–Paralympics events and the surveillance weeks as the number of games activities varied across the weeks. For example, the first week of the 3-week Olympic period (isoweek 30) included the Olympics Opening Ceremony on Friday, 27 July, but the majority of games events did not begin until Monday, 30 July, which occurred in isoweek 31. However, we consider the impact of this imprecision was not too important given that we were analysing the overall impact of the games periods on attendance, rather than exploring variations in attendance by the intensity of games activities.
We did not address other relevant questions surrounding the impact of the 2012 Olympics and Paralympics on sexual health such as sexual health behaviour, in particular sexual health risk taking.
Comparison with other studies
The one study identified in the pre-games literature review was a single-clinic study, which reported an increase in the proportion of patients with an STI and a change in the profile of attendees during the 2000 Sydney Olympics compared to the same period the year before.8 More recently, a cross-sectional study comparing attendance at sexual health clinics in four cities during the 2011 Rugby World Cup in New Zealand with attendance at the same clinics in the preceding 5 years found no statistically significant increase in number of attendances or STI diagnoses.6
Implications of findings
Our findings do not support the a priori contention that the London 2012 Olympics and Paralympics would increase attendance at neighbouring sexual health clinics with the potential to overstretch local resources. This study is supported by the results of our parallel survey investigating the number of Olympic visitors attending sexual health clinics in London and Weymouth during the Olympics and Paralympics, which identified just 108 Olympic visitors among 12 347 newly registered patients surveyed between 20 August and 16 September 2012, representing less than 1% of the population surveyed.9 Our findings provide support for a ‘business-as-usual’ approach to the provision of services at sexual health clinics and do not indicate a likely need for additional staff or increased opening hours at clinics during a mass gathering such as the Olympics in a major city like London. Our findings are echoed in a recent qualitative evaluation of healthcare planning for the London Olympics, which concluded that there had been a ‘tendency to overestimate demand for care’.10
We identified a significant decrease in attendance at clinics, including a proportion of patients who were likely to have received an STI diagnosis had they attended. As we did not detect any increased clinic attendance immediately after the Games, the reduction therefore was probably not just a deferral of non-urgent consultations. This reduction is unlikely to be explained by the ‘Sherman effect’, with local residents leaving the area during the Olympic–Paralympic period as the proportion of local residents attending clinics was virtually unchanged at 92% from the 2011 reference period. The reduction may be related to potential patients going to clinics outside London or Weymouth, deferring seeking treatment for the medium- to long-term or not seeking treatment at all. It is plausible that during the Olympic Games, potential patients were either distracted from seeking healthcare in preference for watching the events or were discouraged from attending clinics due to perceptions of overcrowding and transportation difficulties. It is relevant that in preparation for the Olympics and Paralympics, residents in London especially were warned to expect overcrowding and disruption in public transport services through both official sources and the media.11
It is not clear how generalisable our findings from the London 2012 Olympics and Paralympics are for other countries hosting the Games in future or for mass gatherings generally. It is well-recognised that while there are commonalities across mass-gathering events, variation in the type of event, the profile of participants and the local context of the host city and country will have important influences on preparations for and the impacts of individual events.12 We recommend that those involved in future mass-gathering events undertake similar analyses where possible in order to develop the evidence-base.
Conclusion
In our interrupted time-series analysis, we detected a significant reduction compared with expected during the 3 weeks of the 2012 Olympics in both new episode attendance and in the number of patients receiving a new STI diagnosis at sexual health clinics in London and Weymouth. We did not find any evidence of the anticipated increase in attendance or STIs during the entire Olympic–Paralympic period. We recommend that those planning similar events undertake similar analyses to develop the evidence-base. In the meantime, we suggest sexual health services provide ‘business-as-usual’ during mass-gathering events.
Key messages
Concern about the impact of mass-gathering sporting events on demand for sexual health services created uncertainty for services preparing for the 2012 London Olympics.
We found no evidence of increased attendance or sexually transmitted infection (STI) diagnoses at sexual health clinics in London and Weymouth, the Games ‘hotspots’, during the Olympic–Paralympic period.
We found evidence of a significant reduction in attendance and STI diagnoses during the 3 weeks of the Olympics.
Our findings support a ‘business-as-usual’ approach for sexual health services in mass-gathering events.
Footnotes
Handling editor David A Lewis
Contributors VH wrote the analysis plan, cleaned and analysed the data and drafted and revised the paper. VH is the guarantor. AC provided statistical support and revised the draft paper. PC, GB, GH, HM, CM, KC and JC provided support and comments on the design and analysis and revised the draft paper. All authors have approved the final version for submission.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
Transparency declaration The lead author affirms that the manuscript is an honest, accurate and transparent account of the study being reported; that no important aspects of the study have been omitted; and that any discrepancies from the study as planned (and, if relevant, registered) have been explained.