D. L. Cooper

Tracking the spatial diffusion of influenza and norovirus using telehealth data: A spatiotemporal analysis of syndromic data

BackgroundTelehealth systems have a large potential for informing public health authorities in an early stage of outbreaks of communicable disease. Influenza and norovirus are common viruses that cause significant respiratory and gastrointestinal disease worldwide. Data about these viruses are not routinely mapped for surveillance purposes in the UK, so the spatial diffusion of national outbreaks and epidemics is not known as such incidents occur. We aim to describe the geographical origin and diffusion of rises in fever and vomiting calls to a national telehealth system, and consider the usefulness of these findings for influenza and norovirus surveillance.MethodsData about fever calls (5- to 14-year-old age group) and vomiting calls (≥ 5-year-old age group) in school-age children, proxies for influenza and norovirus, respectively, were extracted from the NHS Direct national telehealth database for the period June 2005 to May 2006. The SaTScan space-time permutation model was used to retrospectively detect statistically significant clusters of calls on a week-by-week basis. These syndromic results were validated against existing laboratory and clinical surveillance data.ResultsWe identified two distinct periods of elevated fever calls. The first originated in the North-West of England during November 2005 and spread in a south-east direction, the second began in Central England during January 2006 and moved southwards. The timing, geographical location, and age structure of these rises in fever calls were similar to a national influenza B outbreak that occurred during winter 2005–2006. We also identified significantly elevated levels of vomiting calls in South-East England during winter 2005–2006.ConclusionSpatiotemporal analyses of telehealth data, specifically fever calls, provided a timely and unique description of the evolution of a national influenza outbreak. In a similar way the tool may be useful for tracking norovirus, although the lack of consistent comparison data makes this more difficult to assess. In interpreting these results, care must be taken to consider other infectious and non-infectious causes of fever and vomiting. The scan statistic should be considered for spatial analyses of telehealth data elsewhere and will be used to initiate prospective geographical surveillance of influenza in England.

Can syndromic surveillance data detect local outbreaks of communicable disease? A model using a historical cryptosporidiosis outbreak

A national UK surveillance system currently uses data from a health helpline (NHS Direct) in an attempt to provide early warning of a bio-terrorist attack, or an outbreak caused by a more common infection. To test this syndromic surveillance system we superimposed data from a historical outbreak of cryptosporidiosis onto a statistical model of NHS Direct call data. We modelled whether calls about diarrhoea (a proxy for cryptosporidiosis) exceeded a statistical threshold, thus alerting the surveillance team to the outbreak. On the date that the public health team were first notified of the outbreak our model predicted a 4% chance of detection when we assumed that one-twentieth of cryptosporidiosis cases telephoned the helpline. This rose to a 72% chance when we assumed nine-tenths of cases telephoned. The NHS Direct surveillance system is currently unlikely to detect an event similar to the cryptosporidiosis outbreak used here and may be most suited to detecting more widespread rises in syndromes in the community, as previously demonstrated. However, the expected rise in NHS Direct call rates, should improve early warning of outbreaks using call data.

Linking syndromic surveillance with virological self-sampling.

Calls to a UK national telephone health helpline (NHS Direct) have been used for syndromic surveillance, aiming to provide early warning of rises in community morbidity. We investigated whether self-sampling by NHS Direct callers could provide viable samples for influenza culture. We recruited 294 NHS Direct callers and sent them self-sampling kits. Callers were asked to take a swab from each nostril and post them to the laboratory. Forty-two per cent of the samples were returned, 16.2% were positive on PCR for influenza (16 influenza A(H3N2), three influenza A (H1N1), four influenza B) and eight for RSV (5.6%). The mean time between the NHS Direct call and laboratory analysis was 7.4 days. These samples provided amongst the earliest influenza reports of the season, detected multiple influenza strains, and augmented a national syndromic surveillance system. Self-sampling is a feasible method of enhancing community-based surveillance programmes for detection of influenza.