Duration of Carbapenemase-Producing Enterobacteriaceae Carriage in Hospital Patients

To determine the duration of carbapenemase-producing Enterobacteriaceae (CPE) carriage, we studied 21 CPE carriers for »1 year. Mean carriage duration was 86 days; probability of decolonization in 1 year was 98.5%, suggesting that CPE-carriers’ status can be reviewed yearly. Prolonged carriage was associated with use of antimicrobial drugs.

The estimated mean duration of CPE carriage was 86 (95% credible interval [CrI] 60-122) days. The probability of decolonization in 1 year was 98.5% (95% CrI 95.0%-99.8%), assuming a constant decolonization rate within the time interval. The longest observed carriage duration was 387 days. We performed a sensitivity analysis that included 16 participants who became decolonized during follow-up (i.e., the last sample collected was negative for CPE). This analysis gave a mean carriage time of 77 (95% CrI 53-108) days and a 98.8% (95% CrI 96.5%-99.9%) probability of decolonization within 1 year.
As time-fixed covariates, we analyzed age, co-colonization with other multidrug-resistant organisms, presence of a urinary catheter, antimicrobial drug use during follow-up, Charlson Comorbidity Index score, and readmission; as a time-varying covariate, we used the Shannon Diversity Index score to explore the covariates' association with decolonization (Appendix Figure 3). The only factor associated with prolonged CPE carriage was antimicrobial drug use during the follow-up period (hazard ratio 0.48, 95% CrI 0.20-0.93). The rate of decolonization for CP-EC was lower than that for CP-KP (0.018 [95% CrI 0.007-0.031] per day vs. 0.030 [95% CrI 0.016-0.049] per day) (Appendix Table 2, Figure 4).

Conclusions
CPE infections are typically preceded by asymptomatic carriage, especially in vulnerable patients such as those who are immunocompromised and critically ill (5). To prevent transmission, active surveillance to identify CPE carriers is essential but may be associated with a high cost:benefit ratio if implemented without knowledge of the natural history of CPE carriage.
Previously reported CPE carriage durations vary widely; median durations range from 43 to 387 days (5)(6)(7). These variations probably result from differences in follow-up schedules, microbiological and molecular methods used to identify CPE, and criteria to define clearance. Studies that reported longer carriage duration tended to adopt an opportunistic sampling strategy and considered both clinical and fecal samples to determine carriage (7). Opportunistic sampling may lead to selection bias because patients with more healthcare contacts would have more samples collected. Infrequent and inconsistent sampling is more likely to misclassify recolonization from a new transmission event as continuous colonization, resulting in perceived longer duration of carriage.
Of note, the participants carried more species of CPE than types of carbapenemase genes. Although this observation may be the result of new acquisition events, it is more parsimoniously explained by active interspecies horizontal gene transfer, especially †Gastrointestinal surgeries (n = 7), skin and soft tissue surgeries (n = 4), neurosurgery (n = 2), removal of Tenckhoff catheter (n = 1), and urologic procedure (n = 1).
in a low-transmission setting such as Singapore. Differential rates of clearance of CP-KP and CP-EC can be related to colonizing affinity of the species and fitness cost of carbapenemase genes, which vary widely among different species (8). Further studies incorporating between-host and within-host transmission dynamics of resistance may shed light on the roles of bacterial clones and plasmids in spreading and maintaining resistance.
Our study has limitations. First, because the participants were screened after hospital admission, the time of initial colonization could not be confidently determined. However, our multistate Markov models assume a constant rate of decolonization. Our sensitivity analysis used the latest CPE-negative swab samples, so carriage duration estimates were similar to those calculated without the last known CPE-negative swab samples, suggesting that our modeling assumptions were reasonable. Second, our sample size was small and drawn from a single population, and the extent to which our findings can be extrapolated to other populations is uncertain. However, our study was rigorously conducted in terms of frequency of fecal sample collection, duration of follow-up, and number of participants in a nonoutbreak setting. The use of multistate models has been shown to preserve power with modest sample size given more frequent follow-ups (9).
Our systematic sampling and robust methods for identifying CPE enabled us to closely follow participants' carriage status. Using combined detection methods of culture on carbapenem-resistance selective media, antibiotic susceptibility testing, and PCR, we found that 4 (19.0%) patients had intervening negative samples, 1-3 weeks apart. This finding suggests that a patient should have >2 negative samples 4-6 weeks apart before considering CPE to be eliminated. The finding that the probability of decolonization is 98.5% in 1 year suggests that a policy of reviewing CPE carrier status yearly may be appropriate for this population. Further health economics analysis is needed to make institution-specific recommendations for rescreening frequency. Given the finding that antimicrobial drug use was the most important factor associated with prolonged CPE carriage, use of antimicrobial drugs in these patients should be avoided if not clinically indicated.