Diagnosis and surveillance of human influenza virus infection

Abstract

Background: Early and accurate diagnosis of influenza helps start correct treatment and

prevention strategies at individual level. Ongoing systematic collection, analysis and

dissemination of the surveillance data from aggregated diagnostic results and other early

indicators help gather the foremost disease information for all subsequent control and

mitigation strategies in the community. Disease information from surveillance results then

feed back to medical practitioners for improving diagnosis. By improving this loop of

disease information transfer in terms of accuracy and timeliness, interventions for disease

control can be applied efficiently and effectively.

Methods: Several new influenza diagnosis and surveillance methods were explored and

evaluated by comparing with laboratory reference test results. Logistic regression models

were applied to synthesize a refined clinical guideline for human influenza infections. The performance of QuickVue rapid diagnostic test was evaluated in a community setting.

Weekly positive rates from the above two diagnostic methods, together with three other

different syndromic surveillance systems, including data from school absenteeism, active

telephone survey and internet based survey were evaluated according to the US CDC

public health surveillance systems guideline in terms of their utility, correlations and

aberration detection performance. Different combinations of surveillance data streams and

aberration detection algorithms were evaluated to delineate the optimal use of multi-stream

influenza surveillance data. A framework of efficient surveillance data dissemination was

synthesized by incorporating the merits of the online national surveillance websites and the

principles of efficient data presentation and dashboard design.

Results: A refined clinical diagnostic rule for influenza infection using fever, cough runny

nose and clinic visit during high influenza activity months as predictors was scored the

highest amount all other current clinical definitions. Time series weekly positive rate from

this rule showed better correlation with reference community influenza activity than many

other current clinical influenza definitions. The QuickVue rapid diagnostic test has an

overall diagnostic sensitivity of 68% and specificity 96%, with an analytic sensitivity

threshold of 105 to106 viral copies per ml. Weekly aggregated QuickVue and school

absenteeism surveillance data was found to be highly correlated with hospital laboratory

and community sentinel surveillance data, but the telephone and internet survey was only

moderately correlated. Multiple univariate methods performed slightly better than

multivariate methods for aberration detections in general. More sophisticated outbreak

detection algorithms did not result in significant improvement of outbreak detection