Data Collection and Analysis
Data for outcomes and factors associated with severe RSV infection were extracted from the EMR and entered into standardized electronic case report forms maintained on a secure server.
Chi-squared test of proportions was utilized to determine differences in the distribution of severe illness by month and by surveillance year. Odds ratios (OR) and 95% confidence intervals (CI) were used to estimate the odds of severe RSV outcomes by demographic characteristics, comorbid conditions, and living situation. We built a multivariate logistic regression model using bidirectional elimination to evaluate characteristics associated with severe RSV outcomes. Variables included in the model were those hypothesized to be associated with severe illness, i.e., older age, obesity, and heart, lung, and neurologic comorbidities and any factors significantly associated with severe RSV infection in the bivariate analysis. The final model was selected by minimizing the Akaike information criterion (AIC) and maximizing the coefficient of determination (r2).
We explored whether severe RSV outcomes impacted the discharge level of care needed, using living situation as a surrogate for the level of care. To do so, we compared the pre-admission and discharge living situation for each surviving patient. Changes in the patient’s living situation were categorized as increased (e.g., living independently before hospitalization and discharged to a nursing home), decreased (e.g., living in the community with family before hospitalization and discharged to living independently), or unchanged level of care. Odds ratios and 95% confidence intervals were used to estimate the odds of change in living situation among those with and without severe RSV outcomes.
Fisher’s Exact test was utilized, when appropriate. A p-value of <0.05 was considered statistically significant. R-Studio (https://rstudio.com, Version 1.2.5001) packages and procedures were used for data analysis.