The rise in Klebsiella pneumoniae (KPn) antimicrobial resistance (AMR) has contributed to high mortality rates from KPn-invasive disease (KPn-ID), including bloodstream infections (BSIs).¹ The highest modelled rates of AMR burden in 2019 were in sub-Saharan Africa, with an estimated all-age regional death rate attributable to AMR at 27.3 deaths per 100 000,² including 50 000 deaths attributable to KPn.³

Klebsiella pneumoniae is ubiquitous in the environment and can be present as a commensal in the gastrointestinal tract and nasopharynx of humans.⁴ Gastrointestinal colonisation represents a major reservoir for KPn^5,6 and is a risk factor for KPn-ID,⁷ with four-fold higher odds of KPn-ID in colonised compared with uncolonised hospitalised individuals.⁸ The higher risk of KPn-ID associated with colonisation is particularly notable in intensive care units (ICUs), where individuals with KPn colonisation have been shown to have a 6.9 times higher odds of KPn-ID compared with those not colonised with KPn.⁹ The pathogenesis of progression from asymptomatic KPn gastrointestinal colonisation to invasive disease is not fully understood,^5,6,8 although bacterial density of colonising strains and compromised host immunity have been postulated as risk factors.^5,10

There is a paucity of data on the prevalence of colonisation with antimicrobial-susceptible or ‘wild-type’ KPn in hospitalised adults in Africa. A systematic review and meta-analysis of gut mucosal colonisation with extended-spectrum beta-lactamase (ESBL) producing Enterobacterales in sub-Saharan Africa (of which 60.0% of participants were adults) reported a pooled community colonisation prevalence of 18.0% (95% confidence interval [CI]: 12.0% – 28.0%).¹¹ Furthermore, the prevalence of colonisation at hospital admission was 32.0% (95% CI: 24.0% – 41.0%) compared with 55.0% (95% CI: 49.0% – 60.0%) in individuals hospitalised for more than 24 h.¹¹ A systematic review of carbapenem-resistant Enterobacterales (CRE) infections and colonisation in mainly African hospital settings (92.9%), identified Klebsiella species as the most prevalent CRE pathogen in most invasive and colonising isolates (72.2%; n = 11315/15666).¹² Furthermore, previous antibiotic use (three out of three studies) and prior hospitalisation (two out of three studies) were the most common risk factors for colonisation with CRE.¹² Another systematic review and meta-analysis reported a pooled global prevalence of carbapenem-resistant KPn (CRKp) colonisation in hospital and community settings of 5.4% (95% CI: 3.7% – 7.4%), with a regional African prevalence of 14.3% (95% CI: 9.48% – 19.97%) based on data from two African studies.¹³

We investigated the risk factors for KPn gastrointestinal colonisation amongst hospitalised adults without KPn-ID in Johannesburg, South Africa. Furthermore, we compared clinical outcomes in colonised versus uncolonised participants. The findings on the ‘clinical and microbiological epidemiology of KPn-ID’ of the participants with KPn-ID have been previously reported.¹⁴

Six hundred and sixty participants without KPn-ID were enrolled. Nine individuals subsequently developed KPn-ID, including three with and six without KPn rectal colonisation at the time of enrolment as controls; Figure 1. All nine were subsequently re-enrolled as cases in the main cohort investigating KPn-ID,¹⁴ and were excluded from further analysis of colonisation risk factors and outcomes. The KPn-ID episode occurred at 25 days, 44 days and 134 days after having been enrolled as controls in those who were colonised by KPn; and at 9 days, 15 days, 20 days, 21 days, 68 days and 87 days after enrolment as controls in the six who were not colonised.

There were no differences in the demographic characteristics of gender, age and race between participants with and without KPn colonisation (Table 1). Rates of colonisation varied depending on the hospital study site (Table 1). Among the cohort of hospitalised adults without KPn-ID (n = 651), the rate of KPn gastrointestinal colonisation was 34.4% (95% CI: 30.4% – 37.6%; n = 224/651) overall, including 38.1% (95% CI: 33.1% – 42.9%; n = 143/375) at CHBAH, 29.3% (95% CI: 23.3% – 34.7%; n = 72/246) at CMJAH, and 30.0% (95% CI: 13.6% – 46.4%; n = 9/30) at HJH (Table 2). There was a lower odd of being colonised with KPn in participants enrolled at CMJAH (29.3%, n = 72/246) compared with those enrolled at CHBAH (38.1%, n = 143/375; aOR 0.68; 95% CI: 0.47–0.98) (Table 2).

There were no significant differences in the colonised versus uncolonised participants with regard to area of admission at the time of swab collection, previous hospitalisation 14 days prior to the current admission, receipt of antimicrobials prior to or during admission, the presence of comorbidities, and the CCI (Table 1).

Colonised and uncolonised participants had the same median CCI (4, IQR 1–7). There were no significant differences between the colonised and uncolonised people living with human immunodeficiency virus, even after stratification by cluster of differentiation 4 (CD4) count, viral load and exposure to antiretroviral therapy; Table 1. Although HIV, diabetes, metastatic cancer, congestive cardiac failure (CCF), connective tissue disease and chronic obstructive pulmonary disease (COPD) were more prevalent in colonised participants, the associations were not statistically significant (Table 1 and Table 2). Similarly, moderate-to-severe chronic kidney disease (CKD) and non-metastatic cancer were more frequent in uncolonised participants, but the associations were insignificant (Table 1 and Table 2). There were no significant differences between the colonised and uncolonised participants regarding exposure to recent surgery or interventions, urethral catheters, or severity of illness measured with qSOFA scores; Table 1 and Table 2. Length of stay (in days) between admission and enrolment (rectal swabbing) was longer in colonised versus uncolonised participants (median = 10, IQR 7–16 vs median = 8, IQR 6–17; p = 0.002); Table 1. Similarly, a length of hospital stay of more than 7 days prior to enrolment (rectal swabbing) was more likely in colonised vs uncolonised participants (80.8%; n = 181/224 vs 70.0%; n = 299/427; p = 0.003).

On multivariate regression analysis, risk factors for KPn colonisation were peptic ulcer disease (2.2%; n = 5/224 vs 0.5%; n = 2/427; aOR 7.75; 95% CI: 1.34–44.78), a CVC (3.6%; n = 8/224 vs 1.4%; n = 6/427; aOR 3.07; 95% CI: 1.00–9.40), a pHAI (4.0%; n = 9/224 vs 1.4%; n = 6/427; aOR 3.13; 95% CI: 1.06–9.28), and length of stay more than 7 days prior to enrolment (80.8%; n = 181/224 vs 70.0%; n = 299/427; aOR 1.84; 95% CI: 1.23–2.75); Table 2.

There was no difference in in-hospital case fatality risk (CFR) between participants with (3.1%; n = 7/224) and without (3.0%; n = 13/427, p = 0.955) KPn colonisation (Table 1). Of the participants, 96.9% (n = 631/651) were discharged alive, and a successful 90-day post-discharge follow-up was achieved in 66.4% (n = 419/631) (Figure 1). Participants colonised with KPn had a higher 90-day CFR compared to the uncolonised (6.5%; n = 14/217 vs 3.1%; n = 13/414, p = 0.051) and were more likely to be readmitted at 90-day follow-up (5.1%; n = 11/217 vs 4.3%; n = 18/414, p = 0.045) (Table 3). Total length of hospitalisation in days was significantly longer in colonised (median 17; IQR 11–28) compared to uncolonised (median 15; IQR 9–24) participants, p = 0.025; Table 3.

This multicentre observational study adds important evidence on risk factors associated with KPn colonisation in South Africa, a setting where the colonisation rate appears higher than in most high-income countries. We demonstrated a KPn colonisation rate amongst hospitalised adults without KPn-ID that ranged from 29.3% to 38.1% across three hospitals, substantially higher than that reported in European⁷ or Australian⁹ cohorts, suggesting a greater baseline risk in our setting.

There is a paucity of data on drug-susceptible KPn colonisation in Africa, as most high-quality studies from the region focus on AMR-related KPn colonisation and KPn-ID.^11,18,19 The observed KPn colonisation rate in our study (34.4%) is markedly higher than rates reported in hospitalised Australian (6.0% – 19.0%)⁹ and healthy Norwegian (16.3%)⁷ adults, but lower than the 62.0% prevalence in healthy Chinese adults.²⁰ A large cross-sectional study by Huynh et al.²¹ investigated KPn gastrointestinal colonisation in asymptomatic pregnant women from low-income communities in Madagascar (64.7%), Cambodia (66.4%), and Senegal (40.2%) reported an overall colonisation rate of 55.9% (n = 489/874).

Gastrointestinal colonisation rates also differ according to different KPn AMR profiles. A 2019 South African point-prevalence study investigating CRE gastrointestinal colonisation in hospitalised adults reported a prevalence of 0.2% (n = 1/439) of CRKp.²² Another South African study (2021) investigating CRKp colonisation via serially collected rectal swabs (days 1, 3, 7 and weekly until transfer, discharge or death) in adults admitted to ICU reported a cumulative prevalence of 16.1% (n = 14/97 swabs from n = 5/31 patients).²³ A study from Bloemfontein, South Africa, investigating the prevalence of, and gastrointestinal colonisation with, multidrug-resistant (MDR) organisms amongst adults with CKD undergoing outpatient haemodialysis and peritoneal dialysis reported a prevalence of MDR KPn of 21.1% (n = 15/71).²⁴ This variability in colonisation rates across different geographical regions and different institutions within the same city as demonstrated in our study, highlights the complex interaction of host, pathogen ecology, and healthcare system factors, with infection-prevention practices likely contributing to the differences.^7,19,25

Risk factors for KPn gastrointestinal colonisation have been reported primarily in high-income settings and include ICU admission and exposure to antimicrobial therapy.^7,9,10 While there was an increased risk for KPn colonisation associated with CVCs, we did not demonstrate an increased risk with ICU admission, surgery or urethral catheters, although this may be because of the relatively small number of critically ill participants enrolled from the ICU. In keeping with published literature, we found a length of hospital stay of more than 6 days prior to enrolment (rectal swabbing) and a pHAI to be independently associated with KPn colonisation,¹⁰ suggesting that hospitalisation itself drives colonisation pressure in our hospitals. In terms of comorbidities, we found peptic ulcer disease to be a significant independent risk factor for colonisation. In a community-based cross-sectional study of 2975 Norwegian adults, Raffelsberger et al.⁷ reported that the use of proton pump inhibitors (PPIs) (aOR 1.62; 95% CI: 1.18–2.22) and non-steroidal anti-inflammatory drugs (NSAIDs) within the past 6 months (aOR 1.38; 95% CI: 1.04–1.84) were independent risk factors for KPn gastrointestinal colonisation. Both PPIs and NSAIDs are associated with peptic ulcer disease treatment and aetiology, respectively, which suggests a common pathophysiological state involving gut microbiota changes that favours KPn colonisation.⁷ Huynh et al.²¹ demonstrated a higher risk of KPn colonisation following the use of antibiotics during pregnancy (in Madagascar), dry fish consumption (in Cambodia) and contact with chickens (in Senegal). These findings suggest that, in addition to the traditional risk factors associated with colonisation in hospitalised adults, food and animal exposure are likely underestimated risk factors for KPn colonisation in community settings, highlighting the need for a one-health approach.

There is scarce regional data on the clinical outcomes of KPn colonised individuals. Although gastrointestinal colonisation is widely recognised as a prerequisite for invasive disease,^{7,8,9,10,26,27} in our study, only three out of nine participants who went on to develop KPn-ID were colonised at enrolment. This finding is not unexpected as KPn gastrointestinal colonisation is a dynamic process,^10,28 and our study participants were swabbed at only one time point. A recent systematic review and meta-analysis investigating the ‘incidence and risk factors for subsequent infections amongst rectal carriers with CRKp’ reported that multi-site colonisation (OR 6.24; 95% CI: 2.38–16.33) was an independent risk factor for subsequent CRKp disease.²⁹ Although we did not find differences in in-hospital CFR amongst colonised versus uncolonised participants, we did note a trend towards poorer clinical outcomes in the colonised group with higher 90-day CFR and readmission rates.

Limitations of our study include that participants were not randomly selected but matched to invasive cases based on age group, duration of hospitalisation prior to enrolment and study site, leading to inadvertent bias in participant selection and estimation of colonisation rate. Additionally, gastrointestinal colonisation is dynamic⁹ and the fact that participants in our study were only swabbed at enrolment and from a single site (rectal swab) likely underestimated the rate of KPn gastrointestinal colonisation in our setting, and also raises the question of whether the colonising KPn isolates were acquired in hospital or represent the patient’s own gut microbiota. Also, phenotypic AST was not performed on rectal swab KPn isolates. Furthermore, approximately 33.6% of the discharged participants were not traceable for the 90-day follow-up, which may lead to an under-estimate of mortality in the study.

Our study has several important implications. The high burden of KPn colonisation in South African hospitals suggests that prevention strategies must extend beyond antimicrobial stewardship to include stricter infection-prevention practices. To accurately determine KPn gastrointestinal colonisation rates in South Africa, prospective surveillance studies should be undertaken in both community and hospital settings, with a one-health approach to investigating risk factors for KPn colonisation. Given the dynamic nature of KPn gastrointestinal colonisation, longitudinal surveillance in both community and hospitalised cohorts is essential to understand transmission pathways and to design interventions that interrupt progression from colonisation to disease.