Estimates of Excess Medically Attended Acute Respiratory Infections in Periods of Seasonal and Pandemic Influenza in Germany from 2001 / 02 to 2010 / 11

Background: The number of patients seeking health care is a central indicator that may serve several different purposes: (1) as a proxy for the impact on the burden of the primary care system; (2) as a starting point to estimate the number of persons ill with influenza; (3) as the denominator data for the calculation of case fatality rate and the proportion hospitalized (severity indicators); (4) for economic calculations. In addition, reliable estimates of burden of disease and on the health care system are essential to communicate the impact of influenza to health care professionals, public health professionals and to the public. Methodology/Principal Findings: Using German syndromic surveillance data, we have developed a novel approach to describe the seasonal variation of medically attended acute respiratory infections (MAARI) and estimate the excess MAARI attributable to influenza. The weekly excess inside a period of influenza circulation is estimated as the difference between the actual MAARI and a MAARI-baseline, which is established using a cyclic regression model for counts. As a result, we estimated the highest ARI burden within the last 10 years for the influenza season 2004/05 with an excess of 7.5 million outpatient visits (CI95% 6.8–8.0). In contrast, the pandemic wave 2009 accounted for one third of this burden with an excess of 2.4 million (CI95% 1.9–2.8). Estimates can be produced for different age groups, different geographic regions in Germany and also in real time during the influenza waves. Citation: der Heiden Ma, Köpke K, Buda S, Buchholz U, Haas W (2013) Estimates of Excess Medically Attended Acute Respiratory Infections in Periods of Seasonal and Pandemic Influenza in Germany from 2001/02 to 2010/11. PLoS ONE 8(7): e64593. doi:10.1371/journal.pone.0064593 Editor: Benjamin J. Cowling, University of Hong Kong, Hong Kong Received December 13, 2012; Accepted April 16, 2013; Published July 16, 2013 Copyright: 2013 der Heiden et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: The Robert Koch Institute is working under the auspices of the Federal Ministry of Health. There was no extra funding. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: anderHeidenM@rki.de

3 cians. Therefore, as a rst step we convert the practice information into infor-4 mation about physicians. This is easy for practices with only one physician.

5
For practices with more than one physician we proceed in the following way: if 6 all physicians have the same specialty assume that each of these physicians have 7 served the an equal proportion of the patients for the whole practice. In case that 8 a pediatrician runs a practice together with a GP we assume that all patients 9 younger than 15 year were served by the pediatrician, while all older patient were 10 served by the GP. In case that an internist in primary care runs a practice to-11 gether with a GP we assume that all patients younger than 15 year were served 12 by the GP, while the internist and the GP serve an equal proportion of all patient 13 older than 14.  Table   33 S1.

34
We describe now how the number of MAARI was estimated for each age group We denote by P r,a,s,t the set of AGI physicians in the region r with specialty s, 40 who sent a report in the week given by t. We denote by R i,a,t the number of 41 MAARI in age group a, that are reported by physician i in the week given by t.

42
We assume that the AGI physicians reported all cases of MAARI according 43 to the case denition. Based on this assumption the {R i,a,t , i ∈ P r,a,s,t } for xed 44 region r, specialty s and calendar week given by t can be treated as indepen-45 dent identically distributed random variables. Hence, we can estimate the mean 46 number of MAARI of age group a per physician as 47 R r,a,s,t = i∈Pr,a,s,t R i,a,t |P r,a,s,t | . (2) We assume that the consultation behavior of patients of AGI physicians is 48 representative for the consultation behavior of all patients. Let n r,a,s be the total 49 number of physicians of specialty s in region r relevant for age group a. Then we 50 obtain the total number of MAARI, M r,a,s,t , as a projection of the mean number 51 of MAARI per physician to the corresponding total number of physicians (see 52   Table S1) 53 M r,a,s,t = n r,a,s * R r,a,s,t . (3) The standard error of M r,a,s,t equals the standard error of the mean (2) multiplied 54 by n r,a,s .

55
The total number of MAARI of an age group is given by the sum of MAARI 56 attended by pediatricians and those attended by GP's:

57
M r,a,t = M r,a,ped,t + M r,a,gp,t .
Since the two summands are statistically independent of each other, the resulting 58 standard error is 59 σ Mr,a,t = σ 2 Mr,a,ped,t + σ 2 Mr,a,gp,t .
Since the total number of consultations results from the estimation of a mean of 60 independent identically distributed random variables R i,a,t , i ∈ P r,a,s,t , it is asymp-61 totically normally distributed. The 95% condence interval is then approximated 62 by 63 M ± r,a,t = M r,a,t ± Φ −1 (.975) * σ Mr,a,t , where Φ denotes the cumulative distribution function of the normal distribution, 64 in particular Φ −1 (.975) ≈ 1.96.

65
Let p r,a,t denote the population of age group a in region r and time t.

66
Then the MAARI incidence is given by 67 M I r,a,t = M r,a,t p r,a .