Table 1.
Location and Characteristics of Participating Centers.
Table 2.
Data Provided by Collaborators for Evaluation of Epidemiology of Bacteremia due to Gram Negative Bacteria.
Figure 1.
Locations of participating sites.
Figure 2.
Fraction of bloodstream infections due to Gram-negative bacteria by site (lowest to highest).
Note that X-axis is presented on a natural log scale. The area of rectangles is inverse to the variance of log(odds) of bloodstream infection due to Gram-negative bacteria; horizontal lines represent 95% confidence intervals.
Table 3.
Spearman Correlation Coefficients and P-Values Comparing Culture Volumes and Metrics for Estimation of Gram NB Risk.
Figure 3.
Fitted waveforms for bloodstream infections due to Gram-negative bacteria from Poisson models adjusted for culture frequency for data available at 13 sites.
Arithmetic means for northern hemisphere (solid curve) and southern hemisphere (dashed curve) are represented by black curves. Curves for individual sites from northern hemisphere (solid curves) and southern hemisphere (dashed curves) are presented in gray.
Figure 4.
Relationship between peak month of occurrence of bloodstream infection due to Gram-negative bacteria (Y-axis) and latitude (X-axis) (circles).
It can be seen that while there is a strong linear overall relationship between month of peak occurrence and latitude, this largely reflects the inversion of summer and winter months in the northern and southern hemispheres (fitted solid black line). Within the northern hemisphere (fitted dashed line) and the southern hemisphere (fitted gray line) there is no relationship between distance from the equator and month of peak incidence.
Figure 5.
Mean monthly fraction of bloodstream infections due to Gram-negative bacteria (Y-axis) plotted against latitude by study site (black circles).
Predicted mean monthly fractions of bloodstream infections due to Gram-negative bacteria, based on a meta-regression model that incorporated latitude-squared, percent of GDP spent on healthcare, and mean annual temperature, are plotted as gray circles. It can be seen that the 3-coefficient model resulted in excellent prediction of Gram-negative bacterial bloodstream infection fraction, and in some cases predictions are sufficiently precise that labels are superimposed.
Figure 6.
Non-linear effects of latitude on risk of bloodstream infection due to Gram-negative bacteria, among bacteremic individuals, according to distance in degrees from equator.
The figure is based on a coefficient for latitude-squared of -0.0003, as presented in Table 4. The log-odds of bloodstream infections due to Gram-negative bacteria (logit, blue curve) is calculated as −0.0003 x latitude2, while the odds ratio for GNB (relative to the odds at the equator) is this quantity exponentiated (red curve). The change in odds ratio per 10 degree increment is non-constant by latitude, and is presented as the green curve.
Table 4.
Univariable and Multivariable Meta-Regression Models Predicting Log (Odds) of Bloodstream Infection due to Gram Negative Bacteria.
Table 5.
Univariable Models of Factors Influencing Culture-Adjusted Incidence of Bloodstream Infection Due to Gram-Negative Bacteria.