Using health-seeking pattern to estimate disease burden from sentinel surveillance

Abstract

The general health-seeking behavior has been well described in different populations. However, how different symptoms have driven health-seeking behavior was less explored. Compared with syndromic surveillance, symptom-specific behavior may more realistically reflect responses from the public. Moreover, the existing sources of syndromic surveillance data are affected by the choice of health service and health seeking behavior and hence may not be representative of actual disease burden. This study described health-seeking behavior of Hong Kong general population due to different symptoms, at different times of the year. By understanding health-seeking behavior at different times of the year, we could estimate the disease burden in the population, and population subgroup from multiple surveillance data. We will use influenza-like illness (ILI) as an example. We conducted 4 rounds of longitudinal randomized telephone surveys over 1.5 year, to describe symptom-specific health-seeking behavior at times with different level of disease activity. We collected information if healthcare service was being sought for and types of medical consultation after onset of different symptoms. The information was further utilized to estimate the ILI burden in specific age groups. Fever is the symptom most likely to prompt people to seek medical attention. We further focused on ILI, defined as fever ≥37.8℃ plus either cough or sore throat. There were 111 episodes of ILI in the study period. Of which, 85.6%, 6.3% and 18.9% visits GP, GOPC and CMP respectively (some have sought for multiple health service types), while 9% did not seek for any medical attention. Based on the relation in the Methods, we estimated the weekly age-specific ILI burden over time from GP, GOPC and CMP surveillance. The estimated ILI burden for the population aged ≤15 and ≥55 years captured the peak in February-March in 2015 from surveillance of ILI institutional outbreaks (e.g. schools, elderly care centers). However, no such increased ILI burden was observed in the 16-54 years age group in the period. Syndromic surveillance data from different sources (e.g. medical consultation or google flu trends) were affected by different health seeking or reporting behavior. By understanding and quantifying these behaviors at different times, it is possible to estimate disease burden in the population.