Novel applications of modern epidemiological methods to the public health control of influenza

Abstract

As an acute, directly transmissible viral respiratory pathogen of epidemic potential affecting all ages, influenza represents one of the most common illnesses encountered by the human population on a daily basis and poses a formidable public health threat throughout the world. This thesis investigated a number of novel applications of modern epidemiological methods to inform the public health control of influenza, specifically in the areas of outbreak risk assessment, surveillance and outbreak detection, outbreak mitigation, and impact assessment.

The first study investigated the dynamic relationship between the patterns of viral shedding in relation to clinical symptomatology in 127 secondary cases of naturally acquired influenza infection having a clinical picture of acute respiratory infection. For influenza A, viral shedding matched closely with clinical symptoms, peaking on the first 1–2 days of clinical illness followed by gradual decrease, suggesting the usefulness of using clinical illness profiles as a proxy for infectiousness. For influenza B, viral shedding rose 2 days prior to symptom onset and persisted with a bimodal pattern for 6–7 days, suggesting potential infectiousness before symptom onset and after clinical improvement.

The second study investigated viral shedding patterns in 56 virologically-confirmed secondary cases of influenza virus infection including 31 having a pauci-symptomatic and 25 an asymptomatic clinical picture. Viral shedding was detected in patients with very few or no symptoms, with a lower amount, a shorter duration, and a more rapid decline compared with symptomatic cases. This suggested their potential for transmitting the virus to close contacts in the community setting.

In the third study, an influenza-like-illness (ILI) surveillance system utilizing smart-card based technology was developed and implemented in 107 local schools covering a total of 75052 children. The absenteeism data from the system detected an upsurge of ILI activity in the community preceding other existing surveillance systems by up to 1-2 weeks. This system filled an important surveillance gap in Hong Kong and demonstrated the feasibility of building a large-scale surveillance system riding on a routinely adopted data collection approach.

The fourth study compared the use of nose and throat swabs collected either by health care professionals or self-collected for viral detection and viral load quantification in 506 outpatients with confirmed influenza infection. Results from self-swabs were consistent with statistical predictions based on earlier and later measurements, supporting its use as a valid alternative for confirmation and quantification of influenza virus in a community setting.

The fifth study examined the clinical profile of the first 130 patients with laboratory confirmed A(H7N9) infection identified during the early phase of the novel epidemic in China in 2013. The mild symptoms in all the five cases detected through a routine sentinel surveillance system contrasted sharply with the much more severe clinical picture and outcome among the other 125 cases presented through the usual health-seeking pathway, suggesting the existence of a “clinical iceberg” phenomenon in influenza A(H7N9) infections with a substantial proportion of A(H7N9) cases with mild disease unidentified in the community.

In the sixth study, a clinical prediction rule was developed to aid the diagnosis of patients with A(H7N9) during the early first wave of the novel epidemic in China. Based on gender, simple clinical history, symptoms and investigations results, the rule risk-stratified patients first presenting to a hospital emergency setting into different risk groups of 0.3%, 2.5%, 4.3%, and 44.0% probability of having A(H7N9) influenza respectively, helping to inform clinical and public health decisions.

The seventh study examined sickness absence data among health care workers from 2004 to 2009 in Hong Kong. Comparing with the non-epidemic period, seasonal influenza epidemics and the 2009 pandemic were associated respectively with an excess in overall sickness absenteeism of 8.4% and 57.7%, and in ARI-related sickness absence of 26.5% and 90.9%, highlighting their substantial impact on productivity loss.

The eighth study employed a series of nine questionnaire surveys with HCWs and quantified that an episode of ARI associated with productivity loss increased by 130.21% and 31.32% in the form of sickness absence and on-job presenteeism respectively, and a 16.51% and 15.71% reduction in the physical and mental component of HRQOL respectively. The proportion of all three forms of productivity impact contributed by ARI kicked-up sharply during the escalating phase, maintained at the peak, and went down during the declining phase of seasonal influenza epidemics.

All in all, studies from this thesis enhanced the understanding of viral shedding in influenza infections over the whole spectrum of clinical symptomatology, explored novel approaches for enhancing influenza surveillance, used routine surveillance data for informing assessment of clinical severity profile, helped the diagnosis of a novel influenza epidemic, and quantified the productivity loss impact of influenza. Major further research directions needed include better characterization of the role of asymptomatic patients in disease transmission, further study to fill remaining gaps in the smart-card based system and the use of self-swab in the field, exploration of the timely and creative use of routine surveillance data to inform epidemic severity assessment and inform clinical and public health decisions, better understanding of the contribution of presenteeism to influenza related productivity loss, and acceptance of HCWs for various alternative policy approaches for enhancing influenza vaccination.