Identifying the effect of patient sharing on between-hospital genetic differentiation of methicillin-resistant Staphylococcus aureus

Abstract

Background: Methicillin-resistant Staphylococcus aureus (MRSA) is one of the most common healthcare-associated pathogens. To examine the role of inter-hospital patient sharing on MRSA transmission, a previous study collected 2,214 samples from 30 hospitals in Orange County, California and showed by spa typing that genetic differentiation decreased significantly with increased patient sharing. In the current study, we focused on the 986 samples with spa type t008 from the same population.

Methods: We used genome sequencing to determine the effect of patient sharing on genetic differentiation between hospitals. Genetic differentiation was measured by between-hospital genetic diversity, F ST , and the proportion of nearly identical isolates between hospitals.

Results: Surprisingly, we found very similar genetic diversity within and between hospitals, and no significant association between patient sharing and genetic differentiation measured by F ST . However, in contrast to F ST , there was a significant association between patient sharing and the proportion of nearly identical isolates between hospitals. We propose that the proportion of nearly identical isolates is more powerful at determining transmission dynamics than traditional estimators of genetic differentiation (F ST ) when gene flow between populations is high, since it is more responsive to recent transmission events. Our hypothesis was supported by the results from coalescent simulations.

Conclusions: Our results suggested that there was a high level of gene flow between hospitals facilitated by patient sharing, and that the proportion of nearly identical isolates is more sensitive to population structure than F ST when gene flow is high.

Fig. 1

The proportion of nearly identical isolates increases with the level of patient sharing (Pearson’s correlation r between log(M) and log(I) = 0.185, Mantel test P value = 0.038; I and M are the proportion of nearly identical isolates and the level of patient sharing, respectively)

Fig. 2

Isolate pairs with smaller SNP differences were more likely to come from the same hospital or hospitals with higher level of patient sharing. a Isolate pairs with smaller SNP differences were more likely to come from the same hospital (red line) than 100 permutations of random assignment of hospitals (gray lines). b In order to obtain the effect of different levels patient sharing, we calculated normalized proportion of pairs, which is the quantity (N _ki/N _i)/(N _k/N), where N is the total number of pairs of isolates, N _k is the number of pairs of isolates from hospitals with a particular amount of patient sharing k, N _i is the number of pairs of samples with less than i SNP differences, and N _ki is the number of pairs of samples coming from hospitals with a particular amount of patient sharing k differing by less than i SNPs. Samples collected from the hospitals with higher level of patient sharing were more likely to have smaller SNP difference. Even a very low level of patient sharing (0.1-0.2 %) shows higher normalized proportion of pairs with smaller SNP differences than no patient sharing

Fig. 3

The power of π, F _ST, and the proportion of nearly identical isolates to detect the effect of patient sharing. The proportion of nearly identical isolates is more powerful than π and F _ST if the threshold for nearly identical isolates is chosen properly. F _ST is more sensitive to changes in patient sharing if patient sharing is high (Model 4). π is less powerful in all four models here