pandemic in Denmark

To enhance surveillance for influenza-like illness (ILI)in Denmark, a year-round electronic reporting system was established in collaboration with the Danish medical on-call service (DMOS). In order to achieve real-time surveillance of ILI, a checkbox for ILI was inserted in the electronic health record and a system for daily transfer of data to the national surveillance centre was implemented. The weekly number of all consultations in DMOS was around 60,000, and activity of ILI peaked in week 46 of 2009 when 9.5% of 73,723 consultations were classified as ILI. The incidence of ILI reached a maximum on 16 November 2009 for individuals between five and 24 years of age, followed by peaks in children under five years, adults aged between 25 and 64 years and on 27 November in senior citizens(65 years old or older). In addition to the established influenza surveillance system, this novel system was useful because it was timelier than the sentinel surveillance system and allowed for a detailed situational analysis including subgroup analysis on a daily basis.


Introduction
In most industrialised countries, surveillance for influenza-like illness (ILI) is carried out by networks of sentinel general practitioners or clinics.Data from sentinel surveillance, in combination with virological data, constitute the basis for influenza surveillance, and has for many years proven to be of value [1].However, the sentinel surveillance systems have limitations.In most countries, participation in the system is voluntary and it requires time and commitment for a general practitioner to report on a regular basis.Due to a limited number of active sentinel practitioners, analysis of trends and differences by subgroups such as age or geography may also be imprecise.Furthermore, reporting from sentinel practitioners is often done on a weekly basis and only during the influenza season.Finally, the Danish sentinel system, as organised at the present, has delays due to mail delivery from the sentinel practices to the surveillance institute and other practicalities [2,3].
To enhance influenza surveillance, a year-round simple electronic reporting system was established in Denmark in collaboration with the Danish medical oncall service (DMOS).Nearly real-time surveillance of ILI was achieved by a simple checkbox for ILI inserted in the electronic health record.This system was first established in 2006 and covered the entire country in 2008.This paper describes the DMOS surveillance system and reports data from the influenza A(H1N1)2009 pandemic from May 2009 to January 2010 where this surveillance system allowed a risk assessment of ILI trends on a daily basis.

Methods
DMOS is a national public medical service replacing the function of the general practitioners after opening hours.On weekdays, this service is open for attendance from 4 pm to 8 am, and during weekends and national holidays on a 24-hours basis.The service is staffed by physicians, mainly general practitioners.DMOS can only be contacted by telephone.The duty officer will either give advice on the phone, make an appointment for a consultation (at the nearest public clinic staffed by DMOS or a home visit, depending on the circumstances), or refer for admission to hospital.
All contacts are registered in a single national computer system.In the electronic health record, demographic data are registered in a structured format, but the medical history, diagnosis and actions taken are recorded in a free text format.In agreement with the on-call physicians and the Danish Medical Association, the computer system was in 2006 modified when a checkbox for ILI was added in the userinterface of the data system.It has a 'mouse-over' function presenting the ILI definition.When the ILI checkbox is marked, the following text with the ILI definition is automatically entered in the unstructured text field: 'Influenza-like illness (ILI): sudden onset of fever, muscle pain, headache and respiratory symptoms'.The cursor is placed after this text, and the physician may enter additional clinical information.With this simple improvement it became possible to obtain structured data on ILI without interfering with the routines of the physicians.In our definition of ILI all three symptoms must be present in order to increase the specificity of the diagnosis.
On a real-time basis, data are transferred to a common external server.On working days, a surveillance data extract is transferred daily to the national public health institute for infectious diseases (Statens Serum Institut).Data are available before 1 pm.The file uploaded on Monday includes activities from Friday, 4 pm to Monday, 8 am.
The data file contains the following information on each contact: time of contact, ILI (yes/no), age in   months, sex, residence of patient (postal code), geographical region of the reporting DMOS physician, type of contact: call, followed by consultation, doctor's visit to the home of the patient, or hospital admission.When a patient contacts the on-call service more than once during one working period, only one record is generated and the information on action taken is the last action taken (e.g.visit to a clinic or admission to hospital).No personal information on individuals is transferred through this system.
At Statens Serum Institut, data are stored in a SQL database and analysed to obtain the incidence rate of ILI and the proportion of patients with ILI of all patients managed (consultation percentage).The results are analysed by age group and geographical region.During the peak influenza period, a seven-day moving

Results
The median weekly number of contacts to the DMOS was 60,029 corresponding to In children aged between five and 14 years, the incidence increased from 0.9 per 100,000 population (n=6) on 17 October to a peak of 57 per 100,000 population (n=387) on 16 November 2009.On the same day, there was a peak in the incidence of cases among individuals aged between 15 and 24 years (18 per 100,000 population, n=396).The incidence in children under five years of age peaked on 20 November (68 per 100,000 population, n=222), in adults aged between 25 and 64 years on 24 November (5 per 100,000 population, n= 68), and persons aged 65 years or more on 27 November (2 per 100,000 population, n=17).
In order to examine referral rates, the data were analysed according to three time periods determined according to influenza transmission: seasonal influenza ( ).
Referral rates were highest for seasonal influenza (47%), whereas only 26% and 28% were referred for consultation during the two pandemic waves.Patients were younger in the autumn wave of the pandemic than in the seasonal influenza period: median age (interquartile range) was 27 years (11 to 41 years) in the seasonal influenza period, 27 years (15 to 40 years) in the summer peak and 15 (6 to 32 years) in the autumn peak.We therefore adjusted for age by Poisson regression and time period remained independently associated with referral rate (Table ).
Figure 3 shows overall incidence of ILI in the sentinel practices (adjusted for number of reporting sentinel practices), incidence of ILI in DMOS as well as the number of laboratory-confirmed cases of influenza A(H1N1)2009 reported to the Department of Virology, Statens Serum Institut.
The incidence of ILI was higher in the sentinel system than in the DMOS.In both systems, marked increases in incidence were observed in week 45 and the peak appeared a week earlier in the DMOS compared with the sentinel surveillance.Thus, the peak incidence in DMOS was in week 46 of 2009 with 128 cases per 100,000 population whereas the peak incidence in the sentinel system was 432 cases per 100,000 population in week 47.The latter estimate was based on 1,864 reports from 288 practices extrapolated to the total of 3,655 general practitioners in Denmark.For comparison, the incidence of laboratory-confirmed cases of influenza A(H1N1)2009 peaked in week 46 with 1,472 cases (27 cases per 100,000 population).

Discussion
During the 2009 pandemic, the DMOS provided valuable real-time and detailed information on ILI-incidence in different age groups and geographical areas.The surveillance data were updated each week.However daily updates were used during the autumn wave of the pandemic, as illustrated in Figure 2.This enabled us to provide timely data to policy makers and health authorities.In particular, they were able to get an overview of the influenza activity during the previous day whereas the sentinel system had more than a week delay.To our knowledge, this is the first year-round, real-time electronic syndromic influenza surveillance system with national coverage that is based on reports provided by physicians.The surveillance system had several advantages among which the automatic data transfer and the daily reporting were the most important.The fact that it was added to an existing administrative system, made it simple to establish and maintain and can therefore be considered as an efficient approach to syndromic surveillance.
Other systems for influenza surveillance, including traditional surveillance for consultation of general practitioners for ILI or acute respiratory infections within their working hours, ambulance dispatches [5,6] and hospital admissions [7,8], may in emergencies or in times of lack of resources become 'saturated'.It is obvious that such systems have limited capacity (for instance, the number of ambulance dispatches will be limited by the number of ambulances and ambulance drivers, and people will find alternative ways to get to hospital during crisis).General practitioners often have a very busy schedule of planned visits and may only have a small number of slots open for acute illnesses.By contrast, the public on-call service is more flexible.There are by definition no planned visits and capacity may be increased by calling in standby medical doctors and adding more telephone lines.This may be one of the reasons that the signal from the on-call service came earlier than in the sentinel surveillance (Figure 1).However, it is also possible that there are differences in the characteristics of the patients (including age) who use the two systems and that this contributes to a later peak in the sentinel system.Importantly, we were able to demonstrate that the peak in the virological surveillance corresponded well with the peak in the DMOS system.
Another possible useful source for influenza surveillance are web queries [9,10].Web queries have the advantage of being cost-effective and timely and may serve as an early indication of unusual activity.However, since they are based on lay reporting, data are more subjective than the present system which has both the advantage of being very timely and automated while still based on evaluation by medical staff.An interesting development of influenza surveillance is Gripenet and related surveillance schemes consisting of cohorts of volunteers reporting ILI cases on a regular basis on the Internet [11].Gripenet is a fast and flexible monitoring system whose uniformity allows for direct comparison of ILI rates between countries and is useful for assessing the burden of illness.However, it requires more commitment from administrative staff and participants than does DMOS system and cases are not evaluated by medical staff.
Nevertheless, the DMOS system has its limitations.As opposed to the sentinel system, there are no virological data from the on-call physicians.Therefore, it cannot replace the sentinel system.Furthermore, sentinel doctors are committed to influenza surveillance, whereas the on-call service is staffed by a larger group of physicians with different knowledge and attitude towards influenza surveillance.Although the novel system was promoted in the regions that administer the DMOS, we have no formal evaluation of its use and the completeness of reporting.
The emergence of influenza A(H1N1)2009 outside the normal 2009/10 influenza season, the high morbidity, the high burden of illness in children and young adults, and the occurrence of several waves are all characteristics of a pandemic [12].The system described here was sufficiently sensitive to be able to detect different peaks for different age groups, and we hope that such detailed data will be of value to obtain more detailed knowledge on the pandemic.As shown in the Table, patients with pandemic influenza were less frequently referred to consultation or admitted to hospital than patients with seasonal influenza in the 2008/09 season.This confirms that in most patients, the clinical presentation in the 2009 pandemic was mild [13][14][15], but may also reflect that the public may have been concerned with the situation and that the threshold for contacting the healthcare system was lower than in periods with seasonal influenza, with the on-call physicians being the most accessible professionals.From July 2009, the Danish National Board of Health advised the public to use the telephone for getting in contact with the healthcare system and to restrict physical consultations in order to limit the spread of influenza A(H1N1)2009.A relatively low referral rate may reflect that this advice was often followed [16].
In conclusion, we established a simple, yet comprehensive and timely, system that allowed us to follow the incidence and consultation percentage of ILI during the autumn of 2009 when pandemic influenza peaked in Denmark.The system allowed for a detailed situational analysis and was useful for the health authorities' response to the pandemic, including risk communication.We propose that other countries explore the possibility of establishing such a system which may also be of relevance for other public health threats.

Figure 1
Figure 1Contacts to the on-call medical service and influenza-like illness cases, per week,Denmark, 2008Denmark,  -2010

Figure 3
Figure 3Weekly incidence of influenza-like illness cases, Denmark, 2009-2010 was presented daily on the website of Statens Serum Institut.Furthermore, a weekly report based on data aggregated over a full week were presented along with data from sentinel surveillance and virological data from the weekly influenza bulletin published every Wednesday on the Statens Serum Institut website.Because the system was recently implemented, we have not yet established a historical baseline and epidemic thresholds for these outcome measures.
a 8 December 2008 to 15 March 2009.b13 July to 11 October 2009.c 12 October 2009 to 18 April 2010.dAdjusted for age by Poisson regression analysis.Source: Danish medical on-call service.averageinfluenza, the referral rates were adjusted for age by Poisson regression (age in five-year groups as categorical variables).We used the GENMOD procedure of the SAS statistical software (SAS institute, Cary, NC, United States of America).