Eur. J. Epidemiol. 0392-2990 July 1992, p. 485-490
EUROPEAN JOURNAL
Vol. 8, No. 4
OF
EPIDEMIOLOGY
FIVE YEARS OF SENTINEL SURVEILLANCE OF ACUTE RESPIRATORY INFECTIONS (1985-1990): THE BENEFITS OF AN INFLUENZA EARLY WARNING SYSTEM R. S N A C K E N .1, J. LION*, V. VAN CASTEREN*, R. CORNELIS*, F. YANE**, M. MOMBAERTS**, W. AELVOET*** and A. STROOBANT* *Institute o f Hygiene a n d Epidemiology - Section Epidemiology - Rue J. Wytsman, 14 1050 Brussels - Belgium. **Institute o f Hygiene a n d Epidemiology - Section Virology (Reference Center f o r Influenza) Rue J. Wytsman, 14 - 1050 Brussels - Belgium. ***Flemish CommuniO~ - Administration o f Health - Markiesstraat I - 1000 Brussels - Belgium.
Key words: Surveillance - Influenza - ARI For the last five years, the Brussels Institute of Hygiene and Epidemiology has been involved in the surveillance of acute respiratory infections (ARI). The four indicators used (number of encounters of ARI by GP's/100 encounters, virus isolations, absenteeism and mortality) are discussed. A regression procedure is applied to the data collected by a sentinel network of general practitioners (GP's). This procedure permits the baseline to be visualized and an epidemic threshold to be determined in order to recognize early an influenza outbreak. The traditional use of flu-like illnesses as an indicator might be improved by the addition of non-specific ARI which are more precocious, especially in children. The criteria for an accurate definition of an influenza epidemic are discussed. The same mathematical model can be used for the analysis of mortality linked with an outbreak. It shows that the last epidemic in the winter 1989-1990 was responsible for about 4900 deaths directly or indirectly related to influenza.
INTRODUCTION
In 1985, the Institute of Hygiene and Epidemiology (I.H.E.) developed a tool for the surveillance of acute respiratory infections (ARI) in Belgium, comparable with the one that exists in several other countries (25). ARI, a heterogeneous entity coveting the common cold to broncho-pneumonias and including influenza, are considered to be a real public health problem owing to their high incidence, the seriousness of their complications and their cost. The mean percentage of ARI as reason for encounter with a general practitioner (GP's) varies between 20°/0 and 1 Corresponding author.
30% (4, 11, 13). The complications concern especially children and persons of 60 years and older. Influenza leads to chest complications (acute bronchitis and pneumonia) in 9.5% of cases, and this percentage of serious complications rises to 73% in the elderly (19). The case fatality among the persons who catch ARI during an outbreak of influenza also increases with age, it varies from 1.6 deaths/100.000 in young adults to 75.9 deaths/100.000 in the elderly (2). Besides morbidity and mortality, the economic consequences should be considered, taking into account the cost of care and loss of productivity (25). The essential objectives of ARI surveillance are the early recognition of an outbreak of influenza-like illnesses and the study of its spread and impact (10, 15, 16). 485
Snacken R. et al.
Eur. J. Epidemiol.
Center isolates and types strains from throat smears performed by its network of 50 GP's from patients with flu-like illness and types influenza viruses isolated by university and hospital laboratories. The strains are sent to the reference laboratory in London for fine determination of the type and the results are transmitted to the WHO (Geneva), as is information from the other 120 National Centers. . The registration of absenteeism of less than 30 days is obtained from the National Society of Railways (SNCB-NMBS). . The monthly mortality (all causes) registered by the National Institute of Statistics.
In the case of influenza, the analysis of all ARI rather than flu-like illnesses alone, will allow earlier detection of an epidemic. Knowledge of the circulating strain is absolutely necessary for the choice of adequate vaccln and for a chemoprophylaxis policy with amantadine (21). The diffusion of weekly information among GP's is essential for the success of this program. MATERIALS AND METHODS
The surveillance of ARI was based on four types of data collected by several methods: 1. A network of 38 GP's completed a form each week for the I.H.E. with the following data: number of ARI and flu-like illnesses by age and the total number of consultations (office encounter + home visits). The GP's used the WHO definition (25): ARE common cold, rhinitis, rhino-pharyngitis, tonsilitis, sinusitis, otitis media, laryngitis, tracheitis, bronchitis, bronchiolitis, pneumonia and broncho-pneumonia. Flu-like illnesses: every illness characterized by sudden onset, fever, myalgia and respiratory symptoms. In order to avoid the problems of irregular participation'of the GP's (holiday, sick leave, etc), the selected indicator for the measure of ARI is the ARI index: number of encounters for AR//100 encounters/week. From, 1989 onwards the collection of data was interrupted for the three summer months to reduce the GP's workload. 2. The registration of the influenza viruses by the Reference Center for Influenza of the IHE. This
We used the same mathematical model for the analysis of all ARI (non-specific ARI + flu-like illnesses) and of mortality from all causes. This regression model has been developped by Serfling (22) and is applied by the CDC (18) for the estimation of excess deaths related to influenza. Castagliola et al. (1) also used it for the analysis of flu-like illnesses and the forcasting of influenza epidemics. This method allows simultaneous adjustment for seasonal variation and secular trends. The observed values can be compared for a determined period with the expected values and an epidemic threshold estimated from the variance (1.96"SD), which is calculated for the non-epidemic years. RESULTS
The evolution over time of ARI shows a classical cyclical pattern with a dip in the summer and 50
50 observed values
U) O
........ e x p e c t e d v a l u e s 40
40 =
............... e p i d e m i c t h r e s h o l d (1.96 x S.D.)
>
I H1N1 X "O
N
30 '-
H3N2
30
::3 N,,,-
,m
i-..
rr 60 years
x o
"o i
10
:\ , , . . . , . ..... ,'"
`````:`.`,``.°.~`-.`,.,`,.,``,``,.`...,``.``..`.`.:`,:-```~.`-,`,.~.``..,``,``...```..--`,````f``,``,`,``,,`
'
0
1985
1986
1987
1988
Figure 3. - Evolution of ARI index in young adults and the elderly (1985-1990).
487
1989
1990
Snacken R. et aL
Eur. J. Epidemiol.
at the beginning of the outbreak, the rise of nonspecific ARI seemed to be earlier in the year. As seen in figure 2, when flu-like illnesses really began to rise at the 48'h week (vertical dotted line), the non-specific ARI were already at their maximum. The pattern of the ARI index was not identical in several age groups: in the elderly, it was less cyclical and there was no epidemic peak in 1988-1989 (Fig. 3).
The percentage of absenteeism of less than 30 days is a classical non-specific indicator obtained by the National Society of Railways. As seen in figure 4, absenteeism and the ARI index are strictly parallel and during an epidemic absenteeism may rise to 7%. The number of observed deaths (all causes) can be compared with the expected number calculated by the same Serfling regression model as the one used for
8 ARI index
40
absenteeism
6 E = N
30
X
i
"0
'i
=m
20
m
1-
,
4 .,O ¢n
La ', ,,,
EUROPEAN JOURNAL
Vol. 8, No. 4
OF
EPIDEMIOLOGY
FIVE YEARS OF SENTINEL SURVEILLANCE OF ACUTE RESPIRATORY INFECTIONS (1985-1990): THE BENEFITS OF AN INFLUENZA EARLY WARNING SYSTEM R. S N A C K E N .1, J. LION*, V. VAN CASTEREN*, R. CORNELIS*, F. YANE**, M. MOMBAERTS**, W. AELVOET*** and A. STROOBANT* *Institute o f Hygiene a n d Epidemiology - Section Epidemiology - Rue J. Wytsman, 14 1050 Brussels - Belgium. **Institute o f Hygiene a n d Epidemiology - Section Virology (Reference Center f o r Influenza) Rue J. Wytsman, 14 - 1050 Brussels - Belgium. ***Flemish CommuniO~ - Administration o f Health - Markiesstraat I - 1000 Brussels - Belgium.
Key words: Surveillance - Influenza - ARI For the last five years, the Brussels Institute of Hygiene and Epidemiology has been involved in the surveillance of acute respiratory infections (ARI). The four indicators used (number of encounters of ARI by GP's/100 encounters, virus isolations, absenteeism and mortality) are discussed. A regression procedure is applied to the data collected by a sentinel network of general practitioners (GP's). This procedure permits the baseline to be visualized and an epidemic threshold to be determined in order to recognize early an influenza outbreak. The traditional use of flu-like illnesses as an indicator might be improved by the addition of non-specific ARI which are more precocious, especially in children. The criteria for an accurate definition of an influenza epidemic are discussed. The same mathematical model can be used for the analysis of mortality linked with an outbreak. It shows that the last epidemic in the winter 1989-1990 was responsible for about 4900 deaths directly or indirectly related to influenza.
INTRODUCTION
In 1985, the Institute of Hygiene and Epidemiology (I.H.E.) developed a tool for the surveillance of acute respiratory infections (ARI) in Belgium, comparable with the one that exists in several other countries (25). ARI, a heterogeneous entity coveting the common cold to broncho-pneumonias and including influenza, are considered to be a real public health problem owing to their high incidence, the seriousness of their complications and their cost. The mean percentage of ARI as reason for encounter with a general practitioner (GP's) varies between 20°/0 and 1 Corresponding author.
30% (4, 11, 13). The complications concern especially children and persons of 60 years and older. Influenza leads to chest complications (acute bronchitis and pneumonia) in 9.5% of cases, and this percentage of serious complications rises to 73% in the elderly (19). The case fatality among the persons who catch ARI during an outbreak of influenza also increases with age, it varies from 1.6 deaths/100.000 in young adults to 75.9 deaths/100.000 in the elderly (2). Besides morbidity and mortality, the economic consequences should be considered, taking into account the cost of care and loss of productivity (25). The essential objectives of ARI surveillance are the early recognition of an outbreak of influenza-like illnesses and the study of its spread and impact (10, 15, 16). 485
Snacken R. et al.
Eur. J. Epidemiol.
Center isolates and types strains from throat smears performed by its network of 50 GP's from patients with flu-like illness and types influenza viruses isolated by university and hospital laboratories. The strains are sent to the reference laboratory in London for fine determination of the type and the results are transmitted to the WHO (Geneva), as is information from the other 120 National Centers. . The registration of absenteeism of less than 30 days is obtained from the National Society of Railways (SNCB-NMBS). . The monthly mortality (all causes) registered by the National Institute of Statistics.
In the case of influenza, the analysis of all ARI rather than flu-like illnesses alone, will allow earlier detection of an epidemic. Knowledge of the circulating strain is absolutely necessary for the choice of adequate vaccln and for a chemoprophylaxis policy with amantadine (21). The diffusion of weekly information among GP's is essential for the success of this program. MATERIALS AND METHODS
The surveillance of ARI was based on four types of data collected by several methods: 1. A network of 38 GP's completed a form each week for the I.H.E. with the following data: number of ARI and flu-like illnesses by age and the total number of consultations (office encounter + home visits). The GP's used the WHO definition (25): ARE common cold, rhinitis, rhino-pharyngitis, tonsilitis, sinusitis, otitis media, laryngitis, tracheitis, bronchitis, bronchiolitis, pneumonia and broncho-pneumonia. Flu-like illnesses: every illness characterized by sudden onset, fever, myalgia and respiratory symptoms. In order to avoid the problems of irregular participation'of the GP's (holiday, sick leave, etc), the selected indicator for the measure of ARI is the ARI index: number of encounters for AR//100 encounters/week. From, 1989 onwards the collection of data was interrupted for the three summer months to reduce the GP's workload. 2. The registration of the influenza viruses by the Reference Center for Influenza of the IHE. This
We used the same mathematical model for the analysis of all ARI (non-specific ARI + flu-like illnesses) and of mortality from all causes. This regression model has been developped by Serfling (22) and is applied by the CDC (18) for the estimation of excess deaths related to influenza. Castagliola et al. (1) also used it for the analysis of flu-like illnesses and the forcasting of influenza epidemics. This method allows simultaneous adjustment for seasonal variation and secular trends. The observed values can be compared for a determined period with the expected values and an epidemic threshold estimated from the variance (1.96"SD), which is calculated for the non-epidemic years. RESULTS
The evolution over time of ARI shows a classical cyclical pattern with a dip in the summer and 50
50 observed values
U) O
........ e x p e c t e d v a l u e s 40
40 =
............... e p i d e m i c t h r e s h o l d (1.96 x S.D.)
>
I H1N1 X "O
N
30 '-
H3N2
30
::3 N,,,-
,m
i-..
rr 60 years
x o
"o i
10
:\ , , . . . , . ..... ,'"
`````:`.`,``.°.~`-.`,.,`,.,``,``,.`...,``.``..`.`.:`,:-```~.`-,`,.~.``..,``,``...```..--`,````f``,``,`,``,,`
'
0
1985
1986
1987
1988
Figure 3. - Evolution of ARI index in young adults and the elderly (1985-1990).
487
1989
1990
Snacken R. et aL
Eur. J. Epidemiol.
at the beginning of the outbreak, the rise of nonspecific ARI seemed to be earlier in the year. As seen in figure 2, when flu-like illnesses really began to rise at the 48'h week (vertical dotted line), the non-specific ARI were already at their maximum. The pattern of the ARI index was not identical in several age groups: in the elderly, it was less cyclical and there was no epidemic peak in 1988-1989 (Fig. 3).
The percentage of absenteeism of less than 30 days is a classical non-specific indicator obtained by the National Society of Railways. As seen in figure 4, absenteeism and the ARI index are strictly parallel and during an epidemic absenteeism may rise to 7%. The number of observed deaths (all causes) can be compared with the expected number calculated by the same Serfling regression model as the one used for
8 ARI index
40
absenteeism
6 E = N
30
X
i
"0
'i
=m
20
m
1-
,
4 .,O ¢n
La ', ,,,
