Showing papers by "Derrick W. Crook published in 2014"

Prediction of Staphylococcus aureus Antimicrobial Resistance by Whole-Genome Sequencing

Abstract: Whole-genome sequencing (WGS) could potentially provide a single platform for extracting all the information required to predict an organism's phenotype. However, its ability to provide accurate predictions has not yet been demonstrated in large independent studies of specific organisms. In this study, we aimed to develop a genotypic prediction method for antimicrobial susceptibilities. The whole genomes of 501 unrelated Staphylococcus aureus isolates were sequenced, and the assembled genomes were interrogated using BLASTn for a panel of known resistance determinants (chromosomal mutations and genes carried on plasmids). Results were compared with phenotypic susceptibility testing for 12 commonly used antimicrobial agents (penicillin, methicillin, erythromycin, clindamycin, tetracycline, ciprofloxacin, vancomycin, trimethoprim, gentamicin, fusidic acid, rifampin, and mupirocin) performed by the routine clinical laboratory. We investigated discrepancies by repeat susceptibility testing and manual inspection of the sequences and used this information to optimize the resistance determinant panel and BLASTn algorithm. We then tested performance of the optimized tool in an independent validation set of 491 unrelated isolates, with phenotypic results obtained in duplicate by automated broth dilution (BD Phoenix) and disc diffusion. In the validation set, the overall sensitivity and specificity of the genomic prediction method were 0.97 (95% confidence interval [95% CI], 0.95 to 0.98) and 0.99 (95% CI, 0.99 to 1), respectively, compared to standard susceptibility testing methods. The very major error rate was 0.5%, and the major error rate was 0.7%. WGS was as sensitive and specific as routine antimicrobial susceptibility testing methods. WGS is a promising alternative to culture methods for resistance prediction in S. aureus and ultimately other major bacterial pathogens.

Evolutionary History of the Clostridium difficile Pathogenicity Locus

Abstract: The symptoms of Clostridium difficile infection are caused by toxins expressed from its 19 kb pathogenicity locus (PaLoc). Stable integration of the PaLoc is suggested by its single chromosomal location and the clade specificity of its different genetic variants. However, the PaLoc is variably present, even among closely related strains, and thus resembles a mobile genetic element. Our aim was to explain these apparently conflicting observations by reconstructing the evolutionary history of the PaLoc. Phylogenetic analyses and annotation of the regions spanning the PaLoc were performed using C. difficile population-representative genomes chosen from a collection of 1,693 toxigenic (PaLoc present) and nontoxigenic (PaLoc absent) isolates. Comparison of the core genome and PaLoc phylogenies demonstrated an eventful evolutionary history, with distinct PaLoc variants acquired clade specifically after divergence. In particular, our data suggest a relatively recent PaLoc acquisition in clade 4. Exchanges and losses of the PaLoc DNA have also occurred, via long homologous recombination events involving flanking chromosomal sequences. The most recent loss event occurred ∼30 years ago within a clade 1 genotype. The genetic organization of the clade 3 PaLoc was unique in containing a stably integrated novel transposon (designated Tn6218), variants of which were found at multiple chromosomal locations. Tn6218 elements were Tn916-related but nonconjugative and occasionally contained genes conferring resistance to clinically relevant antibiotics. The evolutionary histories of two contrasting but clinically important genetic elements were thus characterized: the PaLoc, mobilized rarely via homologous recombination, and Tn6218, mobilized frequently through transposition.

Whole-Genome Sequencing Shows That Patient-to-Patient Transmission Rarely Accounts for Acquisition of Staphylococcus aureus in an Intensive Care Unit

Abstract: BACKGROUND: Strategies to prevent Staphylococcus aureus infection in hospitals focus on patient-to-patient transmission. We used whole-genome sequencing to investigate the role of colonized patients as the source of new S. aureus acquisitions, and the reliability of identifying patient-to-patient transmission using the conventional approach of spa typing and overlapping patient stay. METHODS: Over 14 months, all unselected patients admitted to an adult intensive care unit (ICU) were serially screened for S. aureus. All available isolates (n = 275) were spa typed and underwent whole-genome sequencing to investigate their relatedness at high resolution. RESULTS: Staphylococcus aureus was carried by 185 of 1109 patients sampled within 24 hours of ICU admission (16.7%); 59 (5.3%) patients carried methicillin-resistant S. aureus (MRSA). Forty-four S. aureus (22 MRSA) acquisitions while on ICU were detected. Isolates were available for genetic analysis from 37 acquisitions. Whole-genome sequencing indicated that 7 of these 37 (18.9%) were transmissions from other colonized patients. Conventional methods (spa typing combined with overlapping patient stay) falsely identified 3 patient-to-patient transmissions (all MRSA) and failed to detect 2 acquisitions and 4 transmissions (2 MRSA). CONCLUSIONS: Only a minority of S. aureus acquisitions can be explained by patient-to-patient transmission. Whole-genome sequencing provides the resolution to disprove transmission events indicated by conventional methods and also to reveal otherwise unsuspected transmission events. Whole-genome sequencing should replace conventional methods for detection of nosocomial S. aureus transmission.

Genome Sequencing of an Extended Series of NDM-Producing Klebsiella pneumoniae Isolates from Neonatal Infections in a Nepali Hospital Characterizes the Extent of Community- versus Hospital-Associated Transmission in an Endemic Setting

Abstract: NDM-producing Klebsiella pneumoniae strains represent major clinical and infection control challenges, particularly in resource-limited settings with high rates of antimicrobial resistance. Determining whether transmission occurs at a gene, plasmid, or bacterial strain level and within hospital and/or the community has implications for monitoring and controlling spread. Whole-genome sequencing (WGS) is the highest-resolution typing method available for transmission epidemiology. We sequenced carbapenem-resistant K. pneumoniae isolates from 26 individuals involved in several infection case clusters in a Nepali neonatal unit and 68 other clinical Gram-negative isolates from a similar time frame, using Illumina and PacBio technologies. Within-outbreak chromosomal and closed-plasmid structures were generated and used as data set-specific references. Three temporally separated case clusters were caused by a single NDM K. pneumoniae strain with a conserved set of four plasmids, one being a 304,526-bp plasmid carrying bla(NDM-1). The plasmids contained a large number of antimicrobial/heavy metal resistance and plasmid maintenance genes, which may have explained their persistence. No obvious environmental/human reservoir was found. There was no evidence of transmission of outbreak plasmids to other Gram-negative clinical isolates, although bla(NDM) variants were present in other isolates in different genetic contexts. WGS can effectively define complex antimicrobial resistance epidemiology. Wider sampling frames are required to contextualize outbreaks. Infection control may be effective in terminating outbreaks caused by particular strains, even in areas with widespread resistance, although this study could not demonstrate evidence supporting specific interventions. Larger, detailed studies are needed to characterize resistance genes, vectors, and host strains involved in disease, to enable effective intervention.

Mobile elements drive recombination hotspots in the core genome of Staphylococcus aureus

Abstract: Horizontal gene transfer is an important driver of bacterial evolution, but genetic exchange in the core genome of clonal species, including the major pathogen Staphylococcus aureus, is incompletely understood. Here we reveal widespread homologous recombination in S. aureus at the species level, in contrast to its near-complete absence between closely related strains. We discover a patchwork of hotspots and coldspots at fine scales falling against a backdrop of broad-scale trends in rate variation. Over megabases, homoplasy rates fluctuate 1.9-fold, peaking towards the origin-of-replication. Over kilobases, we find core recombination hotspots of up to 2.5-fold enrichment situated near fault lines in the genome associated with mobile elements. The strongest hotspots include regions flanking conjugative transposon ICE6013, the staphylococcal cassette chromosome (SCC) and genomic island νSaα. Mobile element-driven core genome transfer represents an opportunity for adaptation and challenges our understanding of the recombination landscape in predominantly clonal pathogens, with important implications for genotype–phenotype mapping.

Reduction of Invasive Pneumococcal Disease 3 Years After the Introduction of the 13-Valent Conjugate Vaccine in the Oxfordshire Region of England

Abstract: © The Author 2014. Background: The 7-valent pneumococcal conjugate (PCV7) vaccine's impact on invasive pneumococcal disease (IPD) is well described, but few reports exist on the additional impact of the 13-valent vaccine (PCV13). Methods: We calculated the IPD incidence across all ages in a surveillance project following implementation of PCV7 (in September 2006) and PCV13 (in April 2010) in children aged < 2 years (11 hospitals; 4935 cases). Results: The overall incidence decreased from 10 cases/100 000 persons per year in 1996-1997 to 8 cases/100 000 persons per year in 2007-2008 and 7 cases/100 000 in 2012-2013. Declines were greater in children aged < 2 years (from 37 cases/100 000 in 1996-1997 to 29 and 14 cases/100 000 in 2007-2008 and 2012-2013, respectively). The incidence of IPD due to PCV7 serotypes decreased in all ages after PCV7 introduction (P < .001), whereas the incidence of IPD due to the additional 6 serotypes in PCV13 and to nonvaccine types (NVTs) increased in children aged ≥2 years (P < .001 for both comparisons). The incidence of IPD due to the 6 additional serotypes in PCV13 declined significantly after PCV13 introduction in all ages (P ≤.01), and the incidence of IPD due to NVTs declined significantly in children aged ≥2 years (P =.003). In 2011-2013, the overall incidences of IPD due to PCV7 serotypes, the 6 additional serotypes in PCV13, and NVTs were 0.3, 2.8, and 4.4 cases/100 000; the incidences among children aged < 2 years were 0.9, 2.4, and 10.8 cases/100 000, respectively. Conclusions. The annual incidence of IPD due to vaccine serotypes (1-3 cases/100 000) among children aged < 2 years and nontarget groups demonstrates the success of PCV7 and PCV13. A substantially higher incidence of IPD due to NVTs indicates the importance of ongoing surveillance and extension of vaccine polyvalency.

Prevalence of Staphylococcus aureus protein A (spa) mutants in the community and hospitals in Oxfordshire

Abstract: Staphylococcal protein A (spa) is an important virulence factor which enables Staphylococcus aureus to evade host immune responses. Genotypes known as “spa-types”, based on highly variable Xr region sequences of the spa-gene, are frequently used to classify strains. A weakness of current spa-typing primers is that rearrangements in the IgG-binding region of the gene cause 1-2% of strains to be designated as “non-typeable”. We developed an improved primer which enabled sequencing of all strains, containing any type of genetic rearrangement, in a large study among community carriers and hospital inpatients in Oxfordshire, UK (6110 isolates). We identified eight novel spa-gene variants, plus one previously described. Three of these rearrangements would be designated “non-typeable” using current spa-typing methods; they occurred in 1.8% (72/3905) asymptomatically carried and 0.6% (14/2205) inpatient S. aureus strains. Some individuals were simultaneously colonized by both formerly non-typeable and typeable strains; previously such patients would have been identified as carrying only currently typeable strains, underestimating mixed carriage prevalence and diversity. Formerly non-typeable strains were found in more spa-types associated with multilocus sequence type ST398 (35%), common among livestock, compared to other groups with any non-typeable strains (1-4%), suggesting particular spa-types may have been under-represented in previous human studies. This improved method allows us to spa-type previously non-typeable strains with rearrangements in the spa-gene and to resolve cases of mixed colonization with deletions in one or more strains, thus accounting for hidden diversity of S. aureus in both community and hospital environments.

Whole-Genome Sequencing Demonstrates That Fidaxomicin Is Superior to Vancomycin for Preventing Reinfection and Relapse of Infection With Clostridium difficile

Abstract: Whole-genome sequencing was used to determine whether the reductions in recurrence of Clostridium difficile infection observed with fidaxomicin in pivotal phase 3 trials occurred by preventing relapse of the same infection, by preventing reinfection with a new strain, or by preventing both outcomes. Paired isolates of C. difficile were available from 93 of 199 participants with recurrences (28 were treated with fidaxomicin, and 65 were treated with vancomycin). Given C. difficile evolutionary rates, paired samples ≤2 single-nucleotide variants (SNVs) apart were considered relapses, paired samples >10 SNVs apart were considered reinfection, and those 3-10 SNVs apart (or without whole-genome sequences) were considered indeterminate in a competing risks survival analysis. Fidaxomicin reduced the risk of both relapse (competing risks hazard ratio [HR], 0.40 [95% confidence interval {CI}, .25-.66]; P = .0003) and reinfection (competing risks HR, 0.33 [95% CI, 0.11-1.01]; P = .05).

Multiple-strain colonization in nasal carriers of Staphylococcus aureus

Abstract: Staphylococcus aureus is a commensal that can also cause invasive infection. Reports suggest that nasal cocolonization occurs rarely, but the resources required to sequence multiple colonies have precluded its large-scale investigation. A staged protocol was developed to maximize detection of mixed-spa-type colonization while minimizing laboratory resources using 3,197 S. aureus-positive samples from a longitudinal study of healthy individuals in Oxfordshire, United Kingdom. Initial typing of pooled material from each sample identified a single unambiguous strain in 89.6% of samples. Twelve single-colony isolates were typed from samples producing ambiguous initial results. All samples could be resolved into one or more spa types using the protocol. Cocolonization point prevalence was 3.4 to 5.8% over 24 months of follow-up in 360 recruitment-positives. However, 18% were cocolonized at least once, most only transiently. Cocolonizing spa types were completely unrelated in 56% of samples. Of 272 recruitment-positives returning ≥12 swabs, 166 (61%) carried S. aureus continuously but only 106 (39%) carried the same single spa type without any cocolonization; 31 (11%) switched spa type and 29 (11%) had transient cocarriage. S. aureus colonization is dynamic even in long-term carriers. New unrelated cocolonizing strains could increase invasive disease risk, and ongoing within-host evolution could increase invasive potential, possibilities that future studies should explore.

Machine learning for the prediction of antibacterial susceptibility in Mycobacterium tuberculosis

Abstract: The prevalence of antibiotic resistance in pathogens is far outpacing our ability to develop new antibiotics. This necessitates the development of diagnostic tests that can determine bacterial susceptibility. For Mycobacterium tuberculosis (MTB), this is particularly urgent given that current methods for testing susceptibility take up to two months. The decreasing cost and time required for whole genome sequencing (WGS) offers the possibility of using genome-wide mutational patterns in bacterial DNA to determine drug susceptibility. However, the computational framework for taking advantage of this data has not yet been developed. This paper describes a machine-learning approach for predicting bacterial susceptibility from genomic data. The presence or absence of over 500 single nucleotide polymorphisms (SNPs) found in a dataset of 652 bacterial isolates from the Oxford University Hospitals NHS Trust and elsewhere in the UK were used as features for a number of classification algorithms. Susceptibility and resistance were defined based upon phenotypic growth patterns, and the results from the proposed machine learning method were compared to predictions based upon the presence of a set of known resistance-conferring mutations. Misclassified isolates were also examined for commonalities, revealing eleven potentially new resistance-conferring mutations. The prediction of drug susceptibility using the proposed approach was very promising. Classification accuracy of 93% was obtained for predicting resistance to isoniazid, a key first-line antibiotic drug for MTB. The proposed method was capable of particularly high sensitivity, ranging between 95-100% across the four drugs examined. There is great potential to further develop this framework to find new resistance-conferring mutations.

The complexity and diversity of the Pathogenicity locus in Clostridium difficile clade 5

Abstract: The symptoms of Clostridium difficile infection are caused by two closely related toxins, TcdA and TcdB, which are encoded by the 19.6 kb Pathogenicity Locus (PaLoc). The PaLoc is variably present among strains, and in this respect it resembles a mobile genetic element. The C. difficile population structure consists mainly of five phylogenetic clades designated 1-5. Certain genotypes of clade 5 are associated with recently emergent highly pathogenic strains causing human disease and animal infections. The aim of this study was to explore the evolutionary history of the PaLoc in C. difficile clade 5. Phylogenetic analyses and annotation of clade 5 PaLoc variants and adjoining genomic regions were undertaken using a representative collection of toxigenic and nontoxigenic strains. Comparison of the core genome and PaLoc phylogenies obtained for clade 5 and representatives of the other clades identified two distinct PaLoc acquisition events, one involving a toxin A(+)B(+) PaLoc variant and the other an A(-)B(+) variant. Although the exact mechanism of each PaLoc acquisition is unclear, evidence of possible homologous recombination with other clades and between clade 5 lineages was found within the PaLoc and adjacent regions. The generation of nontoxigenic variants by PaLoc loss via homologous recombination with PaLoc-negative members of other clades was suggested by analysis of cdu2, although none is likely to have occurred recently. A variant of the putative holin gene present in the clade 5 A(-)B(+) PaLoc was likely acquired via allelic exchange with an unknown element. Fine-scale phylogenetic analysis of C. difficile clade 5 revealed the extent of its genetic diversity, consistent with ancient evolutionary origins and a complex evolutionary history for the PaLoc.

The complexity and diversity of the Pathogenicity locus in Clostridium

Abstract: The symptoms of Clostridium difficile infection are caused by two closely related toxins, TcdA and TcdB, which are encoded by the 19.6 kb Pathogenicity Locus (PaLoc). The PaLoc is variably present among strains, and in this respect it resembles a mobile genetic element. The C. difficile population structure consists mainly of five phylogenetic clades designated 1–5. Certain genotypes of clade 5 are associated with recently emergent highly pathogenic strains causing human disease and animal infections. The aim of this study was to explore the evolutionary history of the PaLoc in C. difficile clade 5. Phylogenetic analyses and annotation of clade 5 PaLoc variants and adjoining genomic regions were undertaken using a representative collection of toxigenic and nontoxigenic strains. Comparison of the core genome and PaLoc phylogenies obtained for clade 5 and representatives of the other clades identified two distinct PaLoc acquisition events, one involving a toxin A+B+ PaLoc variant and the other an A−B+ variant. Although the exact mechanism of each PaLoc acquisition is unclear, evidence of possible homologous recombination with other clades and between clade 5 lineages was found within the PaLoc and adjacent regions. The generation of nontoxigenic variants by PaLoc loss via homologous recombination with PaLoc-negative members of other clades was suggested by analysis of cdu2, although none is likely to have occurred recently. A variant of the putative holin gene present in the clade 5 A−B+ PaLoc was likely acquired via allelic exchange with an unknown element. Fine-scale phylogenetic analysis of C. difficile clade 5 revealed the extent of its genetic diversity, consistent with ancient evolutionary origins and a complex evolutionary history for the PaLoc.

Can rapid integrated polymerase chain reaction-based diagnostics for gastrointestinal pathogens improve routine hospital infection control practice? A diagnostic study

Abstract: Background Every year approximately 5000–9000 patients are admitted to a hospital with diarrhoea, which in up to 90% of cases has a non-infectious cause. As a result, single rooms are ‘blocked’ by patients with non-infectious diarrhoea, while patients with infectious diarrhoea are still in open bays because of a lack of free side rooms. A rapid test for differentiating infectious from non-infectious diarrhoea could be very beneficial for patients. Objective To evaluate MassCode multiplex polymerase chain reaction (PCR) for the simultaneous diagnosis of multiple enteropathogens directly from stool, in terms of sensitivity/specificity to detect four common important enteropathogens: Clostridium difficile, Campylobacter spp., Salmonella spp. and norovirus. Design A retrospective study of fixed numbers of samples positive for C. difficile (n = 200), Campylobacter spp. (n = 200), Salmonella spp. (n = 100) and norovirus (n = 200) plus samples negative for all these pathogens (n = 300). Samples were sourced from NHS microbiology laboratories in Oxford and Leeds where initial diagnostic testing was performed according to Public Health England methodology. Researchers carrying out MassCode assays were blind to this information. A questionnaire survey, examining current practice for infection control teams and microbiology laboratories managing infectious diarrhoea, was also carried out. Setting MassCode assays were carried out at Oxford University Hospitals NHS Trust. Further multiplex assays, carried out using Luminex, were run on the same set of samples at Leeds Teaching Hospitals NHS Trust. The questionnaire was completed by various NHS trusts. Main outcome measures Sensitivity and specificity to detect C. difficile, Campylobacter spp., Salmonella spp., and norovirus. Results Nucleic acids were extracted from 948 clinical samples using an optimised protocol (200 Campylobacter spp., 199 C. difficile, 60 S. enterica, 199 norovirus and 295 negative samples; some samples contained more than one pathogen). Using the MassCode assay, sensitivities for each organism compared with standard microbiological testing ranged from 43% to 94% and specificities from 95% to 98%, with particularly poor performance for S. enterica. Relatively large numbers of unexpected positives not confirmed with quantitative PCR were also observed, particularly for S. enterica, Giardia lamblia and Cryptosporidium spp. As the results indicated that S. enterica detection might provide generic challenges to other multiplex assays for gastrointestinal pathogens, the Luminex xTag® gastrointestinal assay was also run blinded on the same extracts (937/948 remaining) and on re-extracted samples (839/948 with sufficient material). For Campylobacter spp., C. difficile and norovirus, high sensitivities (> 92%) and specificities (> 96%) were observed. For S. enterica, on the original MassCode/Oxford extracts, Luminex sensitivity compared with standard microbiological testing was 84% [95% confidence interval (CI) 73% to 93%], but this dropped to 46% on a fresh extract, very similar to MassCode, with a corresponding increase in specificity from 92% to 99%. Overall agreement on the per-sample diagnosis compared with combined microbiology plus PCR for the main four/all pathogens was 85.6%/64.7%, 87.0%/82.9% and 89.8%/86.8% for the MassCode assay, Luminex assay/MassCode extract and Luminex assay/fresh extract, respectively. Luminex assay results from fresh extracts implied that 5% of samples did not represent infectious diarrhoea, even though enteropathogens were genuinely present. Managing infectious diarrhoea was a significant burden for infection control teams (taking 21% of their time) and better diagnostics were identified as having major potential benefits for patients. Conclusions Overall, the Luminex xTag gastrointestinal panel showed similar or superior sensitivity and specificity to the MassCode assay. However, on fresh extracts, this test had low sensitivity to detect a key enteric pathogen, S. enterica; making it an unrealistic option for most microbiology laboratories. Extraction efficiency appears to be a major obstacle for nucleic acid-based tests for this organism, and possibly the whole Enterobacteriaceae family. To improve workflows in service microbiology laboratories, to reduce workload for infection control practitioners, and to improve outcomes for NHS patients, further research on deoxyribonucleic acid-based multiplex gastrointestinal diagnostics is urgently needed. Funding The Health Technology Assessment programme of the National Institute for Health Research.

The hospital microbiome project: meeting report for the UK science and innovation network UK-USA workshop ‘beating the superbugs: hospital microbiome studies for tackling antimicrobial resistance’, October 14th 2013

Abstract: The UK Science and Innovation Network UK-USA workshop ‘Beating the Superbugs: Hospital Microbiome Studies for tackling Antimicrobial Resistance’ was held on October 14th 2013 at the UK Department of Health, London. The workshop was designed to promote US-UK collaboration on hospital microbiome studies to add a new facet to our collective understanding of antimicrobial resistance. The assembled researchers debated the importance of the hospital microbial community in transmission of disease and as a reservoir for antimicrobial resistance genes, and discussed methodologies, hypotheses, and priorities. A number of complementary approaches were explored, although the importance of the built environment microbiome in disease transmission was not universally accepted. Current whole genome epidemiological methods are being pioneered in the UK and the benefits of moving to community analysis are not necessarily obvious to the pioneers; however, rapid progress in other areas of microbiology suggest to some researchers that hospital microbiome studies will be exceptionally fruitful even in the short term. Collaborative studies will recombine different strengths to tackle the international problems of antimicrobial resistance and hospital and healthcare associated infections.

Multidrug-resistant Escherichia coli soft tissue infection investigated with bacterial whole genome sequencing.

Abstract: A 45-year-old man with dilated cardiomyopathy presented with acute leg pain and erythema suggestive of necrotising fasciitis. Initial surgical exploration revealed no necrosis and treatment for a soft tissue infection was started. Blood and tissue cultures unexpectedly grew a Gram-negative bacillus, subsequently identified by an automated broth microdilution phenotyping system as an extended-spectrum β-lactamase producing Escherichia coli. The patient was treated with a 3-week course of antibiotics (ertapenem followed by ciprofloxacin) and debridement for small areas of necrosis, followed by skin grafting. The presence of E. coli triggered investigation of both host and pathogen. The patient was found to have previously undiagnosed liver disease, a risk factor for E. coli soft tissue infection. Whole genome sequencing of isolates from all specimens confirmed they were clonal, of sequence type ST131 and associated with a likely plasmid-associated AmpC (CMY-2), several other resistance genes and a number of virulence factors.

Reply to Mills and Linkin.

Abstract: TO THE EDITOR—In their letter [1], Mills and Linkin question the conclusion of our recent study that patient-to-patient transmission of Staphylococcus aureus rarely occurs on an intensive care unit [2]. Over 14 months, we detected only 7 patient-to-patient transmissions using whole-genome sequencing, a finding that strongly supports our conclusions. Furthermore, patient-to-patient transmission accounted for only a minority of the observed acquisitions. It is important to emphasize that we conducted our study in the context of enhanced infection prevention measures including scrupulous hand hygiene, environmental decontamination, and washing of patients with chlorhexidine. Our results suggest that these measures are highly effective in preventing transmission. Mills and Linkin base their criticism on calculation of confidence intervals (CIs) around proportions; in fact our observed 18.9% (7/37) of S. aureus acquisitions that could plausibly represent patient-to-patient transmissions, based on whole-genome sequencing, has an exact binomial 95% CI [3] of 8.0%–35.1% (this method better allows for the small number of acquisitions than the Wald/ normal approximation used in [1]). Mills and Linkin pay particular attention to methicillin-resistant S. aureus (MRSA) acquisitions (actually 5/15 [33%; 95% CI, 11.8%–61.6%] patient-to-patient transmissions from Figure 2 in our article; patient 35 acquired 2 MRSA strains of differing spa types). Approaches that focus on effects within small subgroups that have not been predefined are widely recognized as flawed [4, 5].We did not predefine subgroups as we knew the number of S. aureus acquisitions to be limited before we performed sequencing to identify transmissions. The 95% CI within any subgroup will widen substantially because of the even smaller size, making inference on the basis of 95% CI alone problematic. If, as recommended [4, 5], we consider heterogeneity across subgroups (albeit in exploratory/post-hoc analyses), comparing the 33% of MRSA acquisitions that are plausible transmissions with the corresponding 9% (2/22; 95% CI, 1.1%–29.2%) of methicillin-sensitive S. aureus (MSSA) acquisitions, we find weak evidence for such an effect (two-sided P = .10 by Fisher’s exact test). In our study context of enhanced infection prevention practice, larger numbers would be needed to confirm whether or not MRSA could be responsible for greater patient-to-patient transmission than MSSA, and to estimate the proportion of acquisitions that could be due to patient-to-patient transmission more precisely.