Role of viral and bacterial pathogens in causing pneumonia among Western Australian children: a case–control study protocol

Bhuiyan, Mejbah Uddin; Snelling, Thomas L; West, Rachel; Lang, Jurissa; Rahman, Tasmina; Borland, Meredith L; Thornton, Ruth; Kirkham, Lea-Ann; Sikazwe, Chisha; Martin, Andrew C; Richmond, Peter C; Smith, David W; Jaffe, Adam; Blyth, Christopher C

The frequency and proportion positive for each respiratory pathogen among both cases and controls will be assessed. Where numbers permit, virus subtypes will be reported separately. These proportions will be compared across the two groups to identify if there are any specific pathogens that are detected more frequently among cases than among controls. We will report the crude odds-ratio of detection for each respiratory pathogen among CAP cases versus controls. We will also report the adjusted odds ratio for each pathogen after adjustment for demographic factors (e.g. age, sex, presence of a smoker in the household, premature birth), and presence of other pathogens using multivariable logistic regression (referred to as model 1). The adjusted odds-ratio of detection among cases versus controls will be taken to be indicative of the pathogenicity of each organism; the higher the odds-ratio the higher the indicative pathogenicity of that microorganism. Using the proportion of cases infected and the adjusted odds-ratio for each pathogen, the population level attributable-fraction of CAP will be calculated using the method of Greenland and Drescher [66] . The frequency and proportion of cases and controls having more than one respiratory virus or bacteria detected will be reported. Where numbers permit, a multivariable regression model will be developed to explore the existence of significant pair-wise interactions between respiratory pathogens and disease status. Disease severity for the cases will be assessed using Respiratory Index of Severity in Children score (RISC) based on Australian children growth standard [67] [68] [69] . Where numbers permit we will explore the relationship between detection of specific pathogens and severe disease (versus control status) using multivariable logistic regression models as described for model 1 (model 2).