Abstract
Background
Influenza causes a large number of hospitalisations and deaths each year. This thesis is the first study to use modelling in a single country to estimate the health burden of influenza across demographic characteristics (age, sex, and ethnicity) and socioeconomic status; the proportion of influenza deaths occurring inside and outside hospital; the proportions with respiratory causes compared with cardiovascular disease and other medical illnesses; and to validate modelled estimates using observations of influenza incidence and distribution.
Aims
This thesis aims to
1. Estimate the contribution of seasonal influenza to hospitalisations in New Zealand and its sociodemographic distribution.
2. Estimate the contribution of seasonal influenza to mortality in New Zealand and its sociodemographic distribution.
3. Validate modelled estimates of influenza incidence and distribution using observed influenza hospitalisation data.
4. Investigate how and where influenza kills people by major health settings and illness categories.
Methods
The main analyses used a total of over 200 quasi-Poisson and negative binomial regression models with weekly counts of hospitalisations, deaths or isolates of influenza A, B and respiratory syncytial virus. It focused on the period 1994-2008, except for the validation Chapter, in which the study period was extended to 2015 to match the observational data. The virus’ contribution to hospitalisations and deaths coded as pneumonia and influenza (P&I), respiratory, circulatory, medical illness, and all causes were modelled.
Results
The contribution of influenza to total hospitalisations and mortality was about 9 and 23 times, respectively, larger than indicated by routine coded data. Respiratory illness was the major contributor (77%) to hospitalisations attributed to influenza whereas circulatory illness made a negligible contribution. By contrast, influenza mortality included a large (37%) contribution from circulatory illness.
The elderly (80 years of age and older) had the highest influenza-attributable hospitalisation rate (327.8 per 100,000) and mortality (214.0 per 100,000). Infants also had high rates of influenza hospitalisation (245.5 per 100,000). Influenza hospitalisation and mortality also varied markedly by ethnicity and socioeconomic status.
Direct measurement and modelling produced similar rates of influenza-associated respiratory hospitalisation across sex, ethnicity and deprivation. However, modelling found the highest rates of influenza hospitalisation in the elderly (538.2 per 100,000 in those aged 80 years plus) whereas directly measured rates were highest in children 1 to 4 years of age (262.9 per 100,000).
Overall 58.1% of influenza-associated deaths occurred in hospital. The majority of these influenza-associated deaths (66.4%) were associated with respiratory illness. By contrast, circulatory illness predominated in those dying in other places (57%). Using modelled mortality (numerator) and modelled hospitalisations (denominator) data, the case fatality risk (CFR) of influenza in hospital was 12.0%. CFR varied markedly by age, ranging from 3.8% for those 20-64 years of age to 41.3% for those aged 80 years and above.
Conclusions
These results provide strong evidence for applying modelling techniques to estimate the health burden of influenza. The marked inequalities of influenza deaths and hospitalisations by sociodemographic characteristics, illness categories and health setting support prioritised interventions (notably vaccination) for these vulnerable groups and further research to identify ways of reducing these inequalities.