Healthcare professionals’ queries on oseltamivir and influenza in Finland 2011‐2016—Can we detect influenza epidemics with specific online searches?

Background Healthcare professionals (HCPs) search medical information during their clinical work using Internet sources. In Finland, Physician's Databases (PD) serve as an Internet medical portal aimed at HCPs. Influenza epidemics appear seasonal outbreaks causing public health concern. Oseltamivir can be used to treat influenza. Little is known about HCPs’ queries on oseltamivir and influenza from dedicated online medical portals and whether queries could be used as an additional source of information for disease surveillance when detecting influenza epidemics. Methods We compared HCPs’ queries on oseltamivir and influenza from PD to influenza diagnoses from the primary healthcare register in Finland 2011‐2016. The Moving Epidemic Method (MEM) calculated the starts of influenza epidemics. Laboratory reports of influenza A and influenza B were assessed. Paired differences compared queries, diagnoses, and laboratory reports by using starting weeks. Kendall's correlation test assessed the season‐to‐season similarity. Results We found that PD and the primary healthcare register showed visually similar patterns annually. Paired differences in the mean showed that influenza epidemics based on queries on oseltamivir started earlier than epidemics based on diagnoses by −0.80 weeks (95% CI: −1.0, 0.0) with high correlation (τ = 0.943). Queries on influenza preceded queries on oseltamivir by −0.80 weeks (95% CI: −1.2, 0.0) and diagnoses by −1.60 weeks (95% CI: −1.8, −1.0). Conclusions HCPs’ queries on oseltamivir and influenza from Internet medical databases correlated with register diagnoses of influenza. Therefore, they should be considered as a supplementary source of information for disease surveillance when detecting influenza epidemics.


| INTRODUC TI ON
Accessing the Internet enables users, including healthcare professionals (HCPs), to search medical information on the web. However, general search engines, such as Google, can lead to websites' distributing medical information of varying quality to the general public, thus making it unreliable for use in clinical work. Online professional medical databases, such as MEDLINE or PubMed, provide information not only for scientific purposes but also for clinical decision making and continuing professional development. 1 These databases are used to find information on current diseases and medications, and the information is used by HCPs in patient care. 2,3 Influenza appears worldwide as a possibly severe infectious disease with a major public health concern. 4 Influenza outbreaks follow a temporal pattern and typically occur during the cold seasons of the year. 4 People are infected mainly by two types of influenza viruses, A and B, spread via air or contaminated surfaces. Influenza epidemics may cause work absence, severe complications, hospitalizations, and even deaths, thus having a large economic burden on the community. 4 For the prevention of influenza, immunization and good personal health and hygiene habits are suggested to control infection. To treat influenza, antiviral medications can be used. Oseltamivir, an antiviral agent, is used to treat seasonal or pandemic influenza in adults and children. 5,6 Oseltamivir is a neuraminidase inhibitor, which prevents the reproduction of the influenza virus. It is available both as a tablet and in a liquid form and is recommended for people with a high risk of complications and should be taken within 48 hours of the first symptoms of influenza. 4,5 The World Health Organization (WHO) has classified oseltamivir as complementary on their list of essential medicines in a health system. 7 There are several methods for the detection of influenza epidemics. The number of influenza diagnoses has traditionally been collected during medical visits to primary healthcare professionals and also from positive findings at microbiological laboratories. 8,9 The accumulation of cases in a defined time period in a population determines the start of an influenza epidemic and indicates the intensity of an outbreak. 8 Along with epidemiological and virological data from clinical encounters and microbiological laboratories, influenza search trends from Internet search engines have also been studied to detect epidemics. [10][11][12][13] Certain Google queries coincided highly with medical visits related to influenza-like symptoms, thus making influenza activity estimation geographically possible. 13 However, this surveillance method contained several flaws in timing and location of an influenza outbreak by overestimating the intensity of an epidemic and missing the first wave of an influenza pandemic. 14 In addition, searches of the general public may be affected by issues not related to an actual epidemic, such as publicity on a given disease while searches by HCPs may be less affected. 15 As general search engines are used by both the general public and HCPs, these search data can yield heterogeneous results. Our study on HCPs' information seeking behavior from Internet medical databases showed that searches and diagnoses on Lyme borreliosis associated with each other. 16 Therefore, we concluded that Internet searches could be used as an additional source of information for disease surveillance. 16 The Moving Epidemic Method (MEM) has been developed to assess the timing of influenza epidemics and to estimate their baseline and threshold. 17 It is implemented by WHO and the European Centres for Disease Prevention and Control (ECDC) to monitor influenza circulation in European countries. [17][18][19] Historical data on influenza weekly rates are analyzed with MEM, and the method includes three stages. [17][18][19] In the first stage, the length of each influenza season with start and end points is determined forming a pre-epidemic, an epidemic, and a post-epidemic period. In the second stage, the epidemic baseline and thresholds are calculated using pre-epidemic and post-epidemic values from historical seasons. In the third stage, low, medium, and high intensity thresholds are computed. Although early detection of influenza to detect outbreaks using epidemiological and virological data has been studied before, including general search engine queries, little is known about queries on influenza and oseltamivir on Internet databases by HCPs.
When searching for medical information on influenza and oseltamivir, HCPs access dedicated medical databases on the Internet.

Duodecim Medical Publications Ltd (owned by the Finnish Medical
Society Duodecim) produces and maintains an Internet-based portal called Physician's Databases (PD). 20 It is available throughout the

| RE SULTS
Visually similar patterns were found between annual queries on oseltamivir and influenza diagnoses during 2011-2016 by season ( Figure 1, panel A and B). In addition, laboratory reports of influenza  Table 1). The seasons peak during weeks 4-8 ( Figure 3). Pre-and post-epidemic thresholds throughout the seasons were at 271 and 291 queries on oseltamivir, respectively. Influenza diagnoses calculated by MEM start during weeks 2-5 (alert weeks) and end during weeks 11-14 ( Figure 1, panel B and Table 1). The seasons peak during weeks 4-9 ( Figure 3). Pre-and post-epidemic thresholds throughout the seasons were at 146 and 162 influenza diagnoses, respectively. Table 1 and Figure 3 show  Table 2). In addition, paired differences in the mean showed statistical significance between queries on oseltamivir and laboratory  Table 2.

| D ISCUSS I ON
Our study showed similar patterns and statistically significant  Table 2). In addition, high correlations and statistically significant paired differences were found between queries on oseltamivir and laboratory reports of influenza A, queries on influenza and oseltamivir, and queries on influenza and influenza diagnoses ( platforms may provide HCPs with medical information of questionable quality. 27 Although influenza searches from Google have been assessed in terms of disease surveillance, this method included several weaknesses in timing and regional scales to predict influenza epidemics. [12][13][14] Notably, general search engines cannot characterize the users performing the searches. We have shown here that HCPs' queries on oseltamivir and influenza from the dedicated Internet portal highly coincided with diagnoses and laboratory reports of influenza A and influenza B.
This study includes certain limitations to be taken into consideration. In this work, we studied queries on oseltamivir and influenza in the whole country including no data on geographical variations. The

F I G U R E 3
The MEM-calculated epidemic weeks (red) and non-epidemic weeks (green) on queries on oseltamivir, influenza diagnoses, laboratory reports of influenza A and influenza B, and queries on influenza by season. This is the first study, to our knowledge, to demonstrate high correlation and statistical significance between queries on oseltamivir and primary healthcare influenza diagnoses by using the MEM model. We found that HCPs' information searching behavior strongly associates with epidemiological data on influenza.
Therefore, we state that HCPs' queries could be used as a supplementary source of information for disease surveillance when detecting influenza epidemics. Our study depicts a possible development for infectious disease surveillance systems. While our study utilizes a database unique for Finland, similar medical databases can be used to assess data in European countries and internationally. 9,19 The combination of Internet-based query data and other surveillance data could enhance current surveillance systems. In the future, it may be possible to create algorithms that analyze HCPs' queries in real time in order to help the detection of the beginning epidemic. Information from these different sources could be combined and delivered to primary healthcare units facing the first patients at the start of an epidemic to estimate the need for primary healthcare services and workforce during epidemics.
Further studies should focus on the applicability of these results in different pathologies and other medical databases in other countries.