Rapid Responses to:

PAPERS: G C Donaldson and W R Keatinge Excess winter mortality: influenza or cold stress? Observational study BMJ 2002; 324: 89-90 [Full text]

Rapid Responses: Submit a response to this article

Rapid Responses published:

Method of calculation of influenza attributed mortality

Douglas M Fleming, Kenneth W Cross, John M Watson, Neville Q Verlander   (6 March 2002)

Authors' reply

G C Donaldson, W R Keatinge   (3 May 2002)

Method of calculation of influenza attributed mortality 6 March 2002
Douglas M Fleming, Director Birmingham Research Unit of the RCGP, Lordswood House, 54 Lordswood Road Harborne B17 9DB, Kenneth W Cross, John M Watson, Neville Q Verlander Send response to journal: Re: Method of calculation of influenza attributed mortality	Dear Editor Re GC Donaldson, WR Keatinge Excess winter mortality: influenza or cold stress? Observational study. BMJ 2002;324:89-90 We write to challenge the method of calculation of influenza attributed mortality due to influenza in the paper published by Donaldson and Keatinge. Data were used covering the period 1970 to 1999 but the regression analysis was stated to start on 1 January 1990 and (unless a misprint) thus has only explored the relationship in the last ten years. The authors conclude by suggesting that 2.4% of winter excess mortality in south-east England is attributed to influenza and quantitatively this, if it can be extrapolated, is equivalent to a national average for England and Wales (population 53 million) of approximately 1620 deaths per year. This estimate contrasts with those obtained by other groups using different methodologies: Tillett et al estimated an average of 12000 deaths attributable to influenza between 1968/69 and 1977/781; Nicholson estimated an annual average of 13800 deaths between 1975/76 and 1989/902; and Fleming estimated an annual average of 12500 deaths between 1989/90 and 1998/993. Donaldson and Keatinge have estimated total deaths attributable to influenza from deaths certified as due to influenza. They have presumably used deaths allocated to influenza as a primary cause according to the national protocol for allocating deaths by cause. Several points are relevant. 1 The data base extended over thirty years. There has been considerable variation in the attribution of deaths from influenza by doctors over this period. In addition, primary cause mortality from respiratory disease has been affected by coding procedural changes in 1984 and 1993, such that in this intervening ten year period the numbers of deaths allocated to respiratory causes was approximately half those in the period before 1984 and the period subsequent to 19934. 2 The data are based on south east England, but this area and population size is not defined. An average of 5.1 deaths per million (estimated average over the last ten years) is equivalent to approximately 50 deaths per year distributed over 365 days in a population of 10 million. The authors indicate that 143 deaths per million were registered as due to influenza in 1976. There is an implication therefore that the acknowledged deaths in that year were thirty times those in the average of the last ten years. While there are recognized to be substantial differences between influenza activity (and related mortality) between “epidemic” years and “average” years, the difference would not be expected to be as large as this. This finding questions the credibility of the methods used to estimate the average in the last ten years. 3 Crucially, we do not accept that the number of deaths attributed specifically to influenza provides a reliable indication of the extent of occurrence of influenza related deaths. Influenza mortality needs to be examined in relation to influenza virus circulation and epidemic periods. Data are not presented in the paper on the distribution of influenza reported deaths in relation to these periods. Based on the authors’ estimate of excess winter deaths from all causes (an average of 1265 per million over the last ten years, equivalent to 67000 nationally in England and Wales), our estimate of 12500 deaths attributable to influenza is equivalent to 19% rather than 2.4%, of total excess winter deaths. Yours sincerely DM Fleming* KW Cross* J M Watson** NQ Verlander** *The Birmingham Research Unit of The Royal College of General Practitioners, Lordswood House, 54 Lordswood Road, Harborne, Birmingham B17 9DB. **PHLS Communicable Disease Surveillance Centre, 61 Colindale Avenue, London NW9 5EQ 1. Tillett HE, Smith JWG, Gooch CD. (1983) Excess deaths attributable to influenza in England and Wales: age at death and certified cause. Int J Epidemiol 1983;12:344-52. 2. Nicholson KG, Impact of influenza and respiratory syncytial virus on mortality in England and Wales from January 1975 to December 1990. Epidemiol Infect (1996), 116: 51-63. 3. Fleming DM. The contribution of influenza to combined acute respiratory infections, hospital admissions and deaths in winter. Commun Dis Public Health 2000; 3:32-8. 4. OPCS. Mortality statistics: cause 1993 (revised) and 1994, series DH2 no 21 1996. ppxxv-xxvii. HMSO, London.
Authors' reply 3 May 2002
G C Donaldson , W R Keatinge Send response to journal: Re: Authors' reply	Influenza epidemics usually occur in particularly cold weather. The key difference between our analysis and the usual assessments of mortality attributable to influenza is that ours makes allowance for the excess deaths that would have been caused, in the absence of an influenza epidemic, by low daily temperatures. Deaths from influenza were extracted before 1979 as ICD-8 codes 470- 474 and thereafter as ICD-9 code 487; allowance was made for changes in coding instructions outside 1984-92 by using the specific conversion factor 0.997 for influenza,(1) (1) not the larger adjustment for the broad category of respiratory deaths mentioned by Fleming et al. Fleming et al ask what we define as south east England. It comprises Greater London, Hertfordshire, Essex, Kent, Sussex, Hampshire, Surrey, Berkshire, Oxfordshire, Buckinghamshire, and Bedfordshire. Yes, the regression analysis started from 1970, although we focused on results since 1990. We used deaths certified as primarily due to influenza as the explanatory variable in the regression to calculate total deaths related to influenza. A theoretical alternative is to use the prevalence of influenza. Viral sampling is important, but we could find no systematic daily measurement of that in the whole population throughout the last 30 years. It would in any case be difficult to allow for different lethalities of different strains of influenza and for different sensitivities and immunities of particular age groups to them. Variation with time in the tendency to certify doubtful deaths as being due to influenza might produce some error in our analysis, but hardly one large enough to affect our conclusion. This was that deaths related to influenza over the past 10 years accounted for only a small fraction of total mortality related to cold. A check for any major error can be made by seeing whether our estimated rise in mortality related to influenza in an epidemic year corresponds with the rise in total excess winter mortality that year. Our paper shows that it did, most clearly in the major epidemic year 1976. Both mortalities increased by a similar amount that year, compared with the preceding and following years. This is consistent with deaths related to influenza being at or near the number we calculated and not at the higher figures sometimes suggested previously. It is, we think, common ground that before 1970 epidemics of influenza were more frequent and more lethal than they are now. G C Donaldson senior research associate W R Keatinge emeritus professor Medical Sciences, Queen Mary and Westfield College (University of London), London E1 4NS w.r.keatinge@qmw.ac.uk 1 Office of Population Censuses and Surveys. Mortality statistics: cause. London: OPCS, 1996: table 4. (Series DH2 No 11.)