Interpreting trends in consultations for influenza disease – varying ‘multipliers’Archived

Three articles in one issue of Eurosurveillance all relate to recent experience with monitoring influenza  illness in the community:  

Electronic real-time surveillance for influenza-like illness: experience from the 2009 influenza A(H1N1) pandemic in Denmark Harder KM, Andersen PH, Bæhr I et alEuro Surveillance 2011; 16(3):pii=19767

This paper  from Denmark shows an initiative to enhance surveillance for influenza-like illness (ILI) via the establishment of real-time notifications through an on-call out of hours primary care system. The authors describe the establishing of a year-round electronic clinical reporting system, in collaboration with a national medical on-call service (DMOS) recording all cinical diagnoses through an electronic health record with daily transfer of data to the national surveillance centre. There is no virological confirmation. The paper then shows the in the 2009 pandemic . Based on a weekly number of all consultations iof around 60,000 and recorded ILI activity peaked in week 46 2009 (when 9.5% of 73,723 consultations were classified as ILI). During that week, the incidence of ILI reached a maximum on the 16th November for individuals between 5 and 24 years of age, followed by peaks in children under 5 years and adults aged between 25 and 64 years; the incidence of ILI among older people (65 years or older) peaked a little later up on the 27th November. The authors argue that this novel system was useful because it was timelier than the established influenza sentinel surveillance system and allowed for a detailed situational analysis (including that of the age groups) on a daily basis.

Early spread of the 2009 infuenza A(H1N1) pandemic in the United Kingdom – use of local syndromic data, May–August 2009 Smith S, Smith GE, Olowokure B et al,Euro Surveillance 2011; 16(3):pii=19771

This paper describes  syndromic surveillance data obtained from two systems coordinatd by the UK’s Health Protection Agency (HPA).  These are the HPA/QSurveillance system which is based on diagnoses from a very large general practice (primary care) population of 23 million patients and thewhole country  (England) HPA/National Health System (NHS) Direct direct call nurse-led help-line covering the whole of England The researchers looked at reports  for monitoring the spread of influenza A(H1N1)2009 at local level during the Spring/Summer wave of the 2009 pandemic in the UK (England) especially in London and the West Midlands where transmission was most intense. During the early weeks, the diagnoses and syndromic indicators sensitive to influenza activity in both schemes remained low and the majority of cases were travel-related. Both systems then detected the first evidence of community transmission in the West Midlands region (following a school-based outbreaks) these were followed a little later by London where ultimately  rates of influenza like illness recorded by general practitioners (GP’s) were higher than in the West Midlands. The  ILI activity in these two regions peaked a week before the rest of the UK in the Spring-summer wave. Initially the UK was pursuing a policy of attempted containment and the data from  both systems were mapped at local level and used alongside laboratory data and local intelligence to assist in the identification of hotspots, to direct limited public health resources.

Two waves of pandemic influenza A(H1N1)2009 in Wales – the possible impact of media coverage on consultation rates, April – December 2009 Keramarou M, Cottrell S, Evans MR et al , Euro Surveillance 2011; 16(3):pii=19772

This paper  describes the epidemiology of the influenza A(H1N1)2009 pandemic in Wales (UK) between April and December 2009 using integrated data from a number of independent sources: primary care (general practitioner) reporting  , population virological surveillance, hospital admissions/deaths and monitoring of media enquiries . In the UK, the 2009 pandemic had a distinct two-wave pattern of general practice ILI consultations. The results show that the Spting/summer  wave peaked in late July 2009 at 100 consultations per 100,000 general practice population and attracted intensive media coverage. Hwoever the positivity rate for the A(H1N1)2009 influenza in testing of sentinel specimens did not exceed 25% and only 44 hospitalisations and one death were recorded. In contrast the the autumn/winter  wave peaked in late October 2009 and it was characterised by lower ILI consultation rates (65 consultations per 100,000 general practice population) and low profile media activity, but there was  a higher positivity rate (60%) and substantially more hospital admissions (n=379) and deaths (n=26). The authors argue that the large number of ILI-related consultations observed through the community surveillance schemes during the first wave of the 2009 pandemic was probably due to intensive media activity rather than viral circulation in the population, thus contributing to an overestimation of the true incidence of influenza.

ECDC Comment, 28th January 2011:There are a series of measures that have been used to detect changes in influenza transmission or to estimate the levels and intensity of influenza transmission. These include:

• media interest and internet enquiries about influenza;
• diary keeping and internet surveys;
• use of medications;
• community indicators and reports for example using ‘call-lines’ (with or without virological confirmation);
• consultations within primary care for influenza like illness or acute respiratory infection (with or without virological confirmation);
• influenza positivity within specimens gathered sentinel surveillance;
• laboratory reporting;
• serological surveys;
• hospital admissions with moderate or severe disease associated with influenza;
• deaths associated with influenza.

The findings in these studies demonstrate some of the strengths and weaknesses of surveillance using community enquiries and primary care consultations.  They can provide timely information and close monitoring through them can allow decision makers and national public health authorities to make evidence-based decisions. For example in detecting the acceleration of transmission. This is particularly true for the Danish  and first UK study.(1,2)  But these analysis have drawbacks, especially when it comes to estimating the level of transmission and burden of infection and disease.  The second UK study in Wales suggests  how intense media activity and public concern can influence consultations.(3)  Changes in community indicators and consultations can be influenced at least in their level if not their pattern. This is so-called ‘consultation bias’ i.e. more people seeking advice or care. There are also strong age-effects with for example parents seeking care for their children while older people tend not to seek primary care. In some countries , medical certificates are needed for those who have had to miss work while that is not the case elsewhere. In addition there also occasions when medical authorities prompt consultations to deliver countermeasures such as vaccinations or treatment.  The result of this is that the ‘mutipliers’ the number of transmissions that a single consultation represents can vary considerably over time or from place to place. This may makes interpretation of consultations and communications for influenza misleading even when they are confirmed  virologically.(3)

Direct evidence of this comes from the Fluwatch research study. This is a large community cohort which almost uniquely uses diaries accounts of illness, serial serology and PCR testing to objectively measure the incidence of new influenza infections. This study has been underway since the 2006/7 influenza seasons and is carried out in a setting where these can be compared with GP consultations for influenza like illness. Its results for the pandemic  and the preceeding seasons were presented last summer at the Options Influenza conference in Hong Kong.(4)  Essentially the Fluwatch results confirmed the  findings of Keramarou et al in Wales.(3) In the Spring/Summer wave in the UK while primary care consultation rates were high the actual incidence of influenza like illness and virologically confirmed influenza was lower than in a normal winter. That is the ‘multipiers’ were low, each GP consultation for influenza like illness represented about 11 actual cases of influenza like illness in the community and 5 cases of PCR confirmed infections. This contrasted with what Fluwatch found in a normal influenza season when the multipliers from a single influenza like illness episode seen in primary care would be between 50 and 70 and for PCR confirmed illness between 20 and 50.(4)  Clearly the levels of reported illness in the community as expressed in consultation rates have to be interpreted cautiously.  It would also seem unlikely that the Fluwatch multipliers would necessarily be transferrable from one country to another.