Statistical analysis
In NH 4 (Table S1), blood sampling was not possible, and residents’ data
were used only to assess the link between vaccination status and
SARS-CoV-2 infection and severe Covid-19 incidence. Due to the very low
number of missing data in the 6 NHs in which blood was collected, and to
the fact that residents who did ou did not contribute to blood
collection did not differ for age, gender, and comorbidity status, no
assumptions were made for missing data et missing data were not
replaced.
Categorical variables were described with frequency and proportions for
each category. The description of continuous variables was performed
using mean and standard deviation and/or median, with interquartiles,
according to the distribution.
The statistical analysis plan was pre-specified according to previous
studies from our group. 6,7, with the following
primary, secondary, and exploratory outcomes
The primary outcome (incidence of positive RT-PCR) was estimated in the
entire cohort with 95% confidence interval (CI), and compared according
to the residents’ vaccination status (non-vaccinated, one vaccine dose,
or two vaccine doses) using the Chi-2 test. Death was not considered as
a competing risk as all deaths were related to Covid-19 during the study
period. Covid-19-related deaths were considered as incident cases.
Secondary outcomes were compared in the whole cohort according to the
vaccination status using the Fisher exact test for deaths and the
Kruskal-Wallis (KW) test for RBD-IgG levels.
Other secondary outcomes according to the vaccination status were
analyzed in RT-PCR-positive residents: Covid-19 symptom severity using
the Fisher exact test, RBD-IgG and N-Ag using the KW test. The
relationship between levels of RBD-IgG and risk of incident SARS-CoV-2
was analyzed using the Fisher exact test.
Exploratory analyzes on the relationship between RBD IgG levels and
serum neutralization in RT-PCR-positive residents of one NH were
performed using Fisher’s exact test. Wilcoxon–Mann–Whitney 2-sided
tests were used to compare N-Ag in asymptomatic, mild, and severe
Covid-19 residents.
Holm’s correction was applied for the post-hoc comparisons for each
outcome. The statistical significance threshold was set at 5%.
We displayed the positive predicted values (PPVs) and negative
predictive values (NPVs) of a positive RT-PCR testing for the accepted
thresholds of IgG-RBD values (positive threshold set at 50 AU/mL,
significant threshold set at 1,050 AU/mL, and 4,160 AU/mL, indication a
high neutralizing effect). 26,27
Analyses were performed and illustrated using SAS Enterprise Guide, v7.3
(SAS Institute Inc) and GraphPad Prism 9.1.1 (GraphPad Software, Inc.,
San Diego, CA).