Real-time epidemic monitoring and forecasting of H1N1-2009 using influenza-like illness from general practice and family doctor clinics in Singapore

Abstract

Background: Reporting of influenza-like illness (ILI) from general practice/family doctor (GPFD) clinics is an accurate indicator of real-time epidemic activity and requires little effort to set up, making it suitable for developing countries currently experiencing the influenza A (H1N1-2009) pandemic or preparing for subsequent epidemic waves.

Methodology/principal findings: We established a network of GPFDs in Singapore. Participating GPFDs submitted returns via facsimile or e-mail on their work days using a simple, standard data collection format, capturing: gender; year of birth; "ethnicity"; residential status; body temperature (degrees C); and treatment (antiviral or not); for all cases with a clinical diagnosis of an acute respiratory illness (ARI). The operational definition of ILI in this study was an ARI with fever of 37.8 degrees C or more. The data were processed daily by the study co-ordinator and fed into a stochastic model of disease dynamics, which was refitted daily using particle filtering, with data and forecasts uploaded to a website which could be publicly accessed. Twenty-three GPFD clinics agreed to participate. Data collection started on 2009-06-26 and lasted for the duration of the epidemic. The epidemic appeared to have peaked around 2009-08-03 and the ILI rates had returned to baseline levels by the time of writing.

Conclusions/significance: This real-time surveillance system is able to show the progress of an epidemic and indicates when the peak is reached. The resulting information can be used to form forecasts, including how soon the epidemic wave will end and when a second wave will appear if at all.

Figure 1. Influenza diagnoses in Singapore in 1957 and 2009 using alternative methods.

(A) ILI in government and city council clinics, 1957 . (B) ARI in this GPFD sentinel network, 2009. (C) ILI in this GPFD network, 2009. (D) Weekly ARI in government polyclinics, 2009 . (A–C) Both daily counts (lines) and weekly averages (shaded polygons) are presented. (D) A marked drop in baseline ARI consultations can be seen immediately before the epidemic, complicating the determination of when the epidemic started using this measure.

Figure 2. Spot map showing the locations of participating GPFD clinics in Singapore.

Most populated parts of the island were represented, the exception being the Woodlands, Sembawang and Yishun areas to the North.

Figure 3. Evaluation of forecasts.

(Left) Actual (red and orange crosses) and predicted (grey shaded area) average number of patients presenting with influenza-like illness per day at the average participating GPFD. The information used to form the forecast is indicated by the red crosses. The last day of information used in forming the forecast is indicated with a red triangle. Predictions here (and in the right-hand column) take the form of decreasing credible intervals, with the region spanned by the outermost polygons corresponding to 95% credibility. Orange crosses indicate future data not used in forming the forecasts. (Right) Predicted total number of people who (i) are currently symptomatic, or (ii) have recovered, assuming no pre-existing immunity. The last day of information used in forming the forecasts is indicated with a red triangle. The cyan cross on the bottom panel indicates the age-adjusted estimate of adult seroconversion in the community from an independent study (maximum likelihood estimate and 95% confidence interval, Mark I-Cheng Chen, personal correspondence).

Figure 4. Quantification of predictive error.

Posterior absolute deviation between predicted average ILIs per GPFD and observed average, with error averaged over the one week period following the time the forecast is made.

Figure 5. Subjective posterior distributions of parameters and Posterior mean and marginal point-wise 95% credible intervals.

The reader's posterior distributions may differ from ours (see refs –[34]). In the background for reference is the number of ILIs per GPFD per day (not to scale). The line of unity is marked on the panel for the effective reproduction number, ; the posterior crosses the line of unity around the day of the peak. Prior distributions for the parameters ( is not a parameter) are indicated on the appropriate panels, using the notation for the beta distribution and for the modified normal distribution such that if then and . The prior distributions taken for the states were , and (a Dirac delta prior), where is similar to except that its support is the integers, and its mass function at is obtained by integrating the density for from to .