Clinical and Microbiological Profile of Serratia spp. Isolates and an Exploratory Comparison of Antimicrobial Resistance by Acquisition Setting: A Two-Year Retrospective Study

Abstract

Purpose

To describe the clinical and microbiological profile of Serratia spp. isolates recovered at a tertiary-care hospital and to explore whether resistance differed by the acquisition category recorded in the source data.

Patients and Methods

We retrospectively reviewed 157 consecutive clinical isolates from 146 patients recovered from September 27, 2019, through September 27, 2021. The primary between-group analysis retained one first-listed isolate per patient (n=146); Fisher’s exact tests, odds ratios with 95% confidence intervals, and Holm adjustment across 12 antimicrobials were used. A generalized estimating equation analysis of all 157 isolates, clustered by patient, was performed as a sensitivity analysis.

Results

Serratia marcescens accounted for 128/157 isolates (81.5%), and sputum or tracheal aspirate accounted for 86/157 (54.8%). Resistance was lowest for trimethoprim/sulfamethoxazole (2/157, 1.3%), gentamicin (6/157, 3.8%), amikacin (7/157, 4.5%), and meropenem (7/157, 4.5%), and highest for imipenem (27/157, 17.2%). In the patient-level analysis, 90 isolates were community-classified and 56 were hospital-classified. Ertapenem resistance was 6/90 (6.7%) versus 11/56 (19.6%; odds ratio 3.42, 95% CI 1.19–9.86), but the comparison did not remain significant after Holm adjustment (nominal p=0.031; adjusted p=0.342). No antimicrobial showed a multiplicity-adjusted between-group difference. The clustered all-isolate analysis yielded the same overall interpretation.

Conclusion

This study provides a historical local description of Serratia isolates collected in 2019–2021. The acquisition-category comparisons were exploratory and did not identify a robust resistance difference after accounting for repeated isolates and multiple comparisons. These data should not be used alone to guide current empirical therapy.

Plain Language Summary

We reviewed Serratia bacteria found in routine patient samples at one hospital during 2019-2021. Most isolates were Serratia marcescens, and respiratory samples were the most common source. We compared antibiotic resistance between isolates that the hospital records classified as hospital acquired or community acquired. To avoid counting the same person more than once in the main comparison, we used one isolate per patient. We also repeated the analysis with all isolates while accounting for repeated samples from the same patient. After correcting for the number of antibiotics compared, we found no clear resistance difference between the two groups. The results are best viewed as a historical local record rather than a guide to current treatment.

Introduction

Serratia species are opportunistic Gram-negative members of the order Enterobacterales. S. marcescens is the species most often implicated in human disease and can be recovered from respiratory, urinary, bloodstream, wound, and ocular specimens. Clinical recovery does not by itself establish infection because colonization is common, particularly in patients with devices or prolonged healthcare exposure.^1,2

Resistance in S. marcescens reflects intrinsic and acquired mechanisms. Chromosomal AmpC beta-lactamase, efflux systems, permeability changes, acquired extended-spectrum beta-lactamases, and carbapenemases can contribute to clinically important phenotypes.^2–4 Species-level identification and resistance interpretation are therefore important, while automated identification of uncommon Serratia species and phenotypic beta-lactamase assignments may require confirmation.^2,4

Recent adult studies have shown that Serratia epidemiology and resistance vary by population, specimen type, institution, and time period. A long-term adult bloodstream-infection series reported increasing resistance over time, whereas a recent hospital survey described substantial local variation across agents.^3,4 Turkish single-center studies also demonstrate heterogeneous specimen and susceptibility patterns.^5,6 These differences support transparent local surveillance but limit transfer of historical estimates to current empirical treatment.

The distinction between community onset, community association, hospital onset, and healthcare association requires timing and prior-healthcare-exposure information. The source dataset for the present study contained a binary acquisition category but did not retain the variables needed to reconstruct those definitions. We therefore treated acquisition category as an exploratory recorded classification rather than a causal exposure. The primary aim was to describe clinical and microbiological characteristics of Serratia spp. isolates collected in 2019–2021. The secondary aim was an exploratory comparison of antimicrobial resistance by recorded acquisition category.⁷

Patients and Methods

Study Design and Setting

This retrospective, single-center study was conducted at Pamukkale University Faculty of Medicine Hospital, a tertiary-care hospital in Denizli, Türkiye. The study period extended from September 27, 2019, to September 27, 2021. The microbiology extraction included consecutive Serratia spp. isolates recovered from clinical specimens submitted during routine care. The extraction was limited to clinical cultures and did not include surveillance cultures. Isolates were analyzed as laboratory observations; infection versus colonization was not adjudicated. The analytic extract did not encode polymicrobial status, so this feature could not be evaluated.

The isolate-level variables were Serratia species, specimen type, acquisition category, and categorical antimicrobial susceptibility results. Patient-level variables were age, sex, clinical department, recorded death status, and one extracted primary comorbidity category. The comorbidity field was mutually exclusive in the source dataset and should not be interpreted as a complete multimorbidity profile. There were no missing values in the variables used for the analyses.

Acquisition Category

The hospital-acquired/community-acquired variable was abstracted from contemporaneous infection-control records, where classification had been made under the National Healthcare-Associated Infections Surveillance Guide of the Republic of Türkiye Ministry of Health.⁷ The analytic extract did not retain admission time, culture time, recent hospitalization, long-term-care residence, dialysis, or outpatient-procedure history. Consequently, the classification could not be independently reconstructed or separated into hospital-onset, community-onset, community-associated, and healthcare-associated categories. Throughout this report, it is treated as the recorded acquisition category.

Repeated Isolates and Analysis Cohorts

The 157 isolates came from 146 patients. Nine patients contributed more than one isolate: seven contributed two and two contributed three; the maximum was three isolates per patient. All repeat isolates for a patient had the same recorded acquisition category. Seven of nine repeat-patient sets had the same species, whereas all nine included more than one specimen type. Culture dates were not retained in the analytic file, so intervals between repeat cultures and infection episodes could not be reconstructed.

For the primary analysis in the revised manuscript, one isolate per patient was retained using the first-listed record in the source dataset (n=146). This reproducible rule does not necessarily identify the chronologically earliest isolate because collection dates were unavailable. An all-isolate sensitivity analysis retained all 157 observations and accounted for within-patient clustering.

Microbiological Methods

Urine specimens were inoculated onto 5% sheep blood agar and eosin methylene blue (EMB) agar (Becton Dickinson, USA) and incubated at 37°C for 18–24 h. Non-urine specimens were inoculated onto 5% sheep blood agar, EMB agar, and chocolate agar (Becton Dickinson, USA), as appropriate, and incubated for 24–48 h.

Organism identification and antimicrobial susceptibility testing were performed using the BD Phoenix 100 system (Becton Dickinson, USA). The NMIC/ID-433 panel was used for non-urine specimens and the UNMIC/ID-432 panel for urine specimens. Conventional identification, when required, included catalase, oxidase, triple sugar iron agar, urease, citrate, and motility–indole–ornithine tests.

Antimicrobial susceptibility was assessed using the Phoenix system and the Kirby–Bauer disk-diffusion method. BD BBL Sensi-Disc and Bioanalyse AST disks were used: amoxicillin/clavulanic acid (30 µg), ampicillin (10 µg), ampicillin/sulbactam (20 µg), ceftriaxone (30 µg), ceftazidime (10 µg), piperacillin/tazobactam (30/6 µg), gentamicin (10 µg), cefepime (30 µg), amikacin (30 µg), ertapenem (10 µg), imipenem (10 µg), meropenem (10 µg), ciprofloxacin (5 µg), levofloxacin (5 µg), and trimethoprim/sulfamethoxazole (25 µg). When Phoenix and disk-diffusion results disagreed, both tests were repeated from a pure culture; if uncertainty persisted, the disk-diffusion result was accepted. Results were interpreted using the EUCAST breakpoint table current in each calendar year: version 9.0 in 2019, version 10.0 in 2020, and version 11.0 in 2021. S denotes susceptible at standard exposure, I susceptible at increased exposure, and R resistant. Every isolate had a recorded category for each of the 12 agents included in the analysis. No retrospective breakpoint reinterpretation was performed because raw MIC and inhibition-zone data were unavailable.^8,9

The analytic extract retained the final categorical susceptibility result for each isolate–antimicrobial pair, but did not retain isolate-level method provenance. Accordingly, we could not reconstruct which method generated each final reported category or confirm that both Phoenix and disk diffusion were applied to every isolate–antimicrobial pair. The laboratory workflow for discordant Phoenix and disk-diffusion results was to repeat both methods from a pure culture and, if uncertainty persisted, to accept the disk-diffusion result.

Statistical Analysis

Continuous variables are summarized with mean, standard deviation, and range; categorical variables are summarized with counts and percentages. Overall resistance proportions are accompanied by Wilson 95% confidence intervals. The primary comparison used the patient-level dataset and defined resistance as R versus S or I. For each antimicrobial, Fisher’s exact test and an odds ratio for hospital-classified versus community-classified isolates with a 95% confidence interval were reported. Zero cells were handled with the Haldane-Anscombe correction for interval estimation. Holm’s procedure adjusted the 12 p values for multiple comparisons. Because the comparison was exploratory and event counts were limited, no multivariable model was fitted.

The all-isolate sensitivity analysis used a logistic generalized estimating equation with acquisition category as the predictor, patient as the clustering unit, an exchangeable working correlation structure, and robust standard errors. A second sensitivity analysis repeated the patient-level comparisons after restriction to S. marcescens. Analyses were performed using IBM SPSS Statistics version 25.0 for the original descriptive work and independently reproduced in Python 3.12 with pandas, SciPy, and statsmodels for the revision. All tests were two-sided.

Ethics Approval

The study was approved by the Non-Interventional Clinical Research Ethics Committee of Pamukkale University (meeting no. 21, November 30, 2021; approval letter dated December 3, 2021; document no. E.137014). The requirement for individual informed consent was waived because the study used retrospectively collected routine-care data. The study was conducted in accordance with the Declaration of Helsinki. Data were analyzed in de-identified form.

Results

Study Cohort

The study included 157 isolates from 146 patients. The primary patient-level cohort contained 90 community-classified and 56 hospital-classified isolates. Mean age was 64.9 years (SD 14.8; range 19–90), 108/146 patients (74.0%) were male, and 21/146 (14.4%) had death recorded. The participant and isolate flow and repeat-isolate characteristics are detailed in Supplementary Table S1. Patient characteristics and the extracted primary comorbidity category are summarized in Table 1.

Table 1.

Patient Characteristics and Single Extracted Primary Comorbidity Category (n=146)

Characteristic	Value
Age, mean (SD), years	64.9 (14.8)
Age range, years	19-90
Male sex	108 (74.0%)
Death recorded	21 (14.4%)
Malignancy	38 (26.0%)
Diabetes mellitus	23 (15.8%)
Neurologic disease	19 (13.0%)
Respiratory disease	19 (13.0%)
No comorbidity recorded	19 (13.0%)
Chronic kidney disease	12 (8.2%)
Coronary artery disease	6 (4.1%)
Benign prostatic hyperplasia	5 (3.4%)
Hypertension	5 (3.4%)

Notes: The source extraction contained one mutually exclusive primary comorbidity field per patient. It did not capture all coexisting conditions or a documented selection hierarchy; percentages therefore describe the extracted field rather than multimorbidity prevalence. Two spelling variants of malignancy in the source data were combined.

Specimen and Species Distribution

Respiratory specimens predominated: sputum and tracheal aspirates together accounted for 86/157 isolates (54.8%). S. marcescens was the predominant species (128/157, 81.5%). Blood isolates occurred only in the hospital-classified group in this dataset, whereas urine, sputum, wound, and surgical-material isolates were more frequent in the community-classified group (Supplementary Table S2). These imbalances indicate potential confounding of acquisition-category comparisons by specimen source. The complete specimen and species distributions are presented in Tables 2 and 3, respectively.

Table 2.

Distribution of Serratia Isolates by Clinical Specimen Type (n=157)

Specimen Type	n	%
Sputum	50	31.8
Tracheal aspirate	36	22.9
Urine	26	16.6
Blood	18	11.5
Wound	18	11.5
Surgical material	6	3.8
Eye swab	3	1.9

Table 3.

Species Distribution Among Serratia Isolates (n=157)

Species	n	%
Serratia marcescens	128	81.5
Serratia liquefaciens	13	8.3
Serratia plymuthica	13	8.3
Serratia rubidaea	2	1.3
Serratia odorifera	1	0.6

Antimicrobial Susceptibility Categories

Table 4 presents S, I, and R separately for all 157 isolates. Resistance was lowest for trimethoprim/sulfamethoxazole (1.3%), gentamicin (3.8%), amikacin (4.5%), and meropenem (4.5%), and highest for imipenem (17.2%). Because the estimates pool multiple Serratia species and span historical EUCAST versions, they are descriptive historical proportions rather than current treatment recommendations.

Table 4.

Antimicrobial Susceptibility Categories Among Serratia Isolates (n=157)

Antimicrobial	S, n (%)	I, n (%)	R, n (%)	R 95% CI, %
Amikacin	144 (91.7)	6 (3.8)	7 (4.5)	2.2–8.9
Cefepime	139 (88.5)	0 (0.0)	18 (11.5)	7.4–17.4
Ceftazidime	127 (80.9)	8 (5.1)	22 (14.0)	9.4–20.3
Ceftriaxone	127 (80.9)	8 (5.1)	22 (14.0)	9.4–20.3
Ciprofloxacin	146 (93.0)	0 (0.0)	11 (7.0)	4.0–12.1
Ertapenem	138 (87.9)	1 (0.6)	18 (11.5)	7.4–17.4
Gentamicin	151 (96.2)	0 (0.0)	6 (3.8)	1.8–8.1
Imipenem	114 (72.6)	16 (10.2)	27 (17.2)	12.1–23.9
Levofloxacin	146 (93.0)	0 (0.0)	11 (7.0)	4.0–12.1
Meropenem	142 (90.4)	8 (5.1)	7 (4.5)	2.2–8.9
Piperacillin/Tazobactam	147 (93.6)	1 (0.6)	9 (5.7)	3.0–10.5
Trimethoprim/sulfamethoxazole	155 (98.7)	0 (0.0)	2 (1.3)	0.4–4.5

Notes: Categories were those reported by the routine laboratory under EUCAST versions 9.0 (2019), 10.0 (2020), and 11.0 (2021). All 157 isolates had a recorded category for every listed agent. Confidence intervals are Wilson intervals. Because the cohort included multiple Serratia species, agents without a uniform species-appropriate interpretation across the pooled cohort were excluded from pooled reporting.

Abbreviations: S, susceptible at standard exposure; I, susceptible at increased exposure; R, resistant.

Exploratory Acquisition Category Comparison

In the patient-level primary analysis, the nominal comparison for ertapenem was 6/90 (6.7%) resistant community-classified isolates versus 11/56 (19.6%) hospital-classified isolates (OR 3.42, 95% CI 1.19–9.86; nominal p=0.031). This result did not remain significant after Holm adjustment (adjusted p=0.342). Piperacillin/tazobactam showed the opposite nominal pattern (8/90 vs 0/56; nominal p=0.024; adjusted p=0.284), also with a wide interval. No agent showed a multiplicity-adjusted difference (Table 5).

Table 5.

Patient-Level Antimicrobial Resistance by Recorded Acquisition Category (n=146)

Antimicrobial	Community Classified R, n/N (%)	Hospital Classified R, n/N (%)	OR (95% CI)	Nominal p	Holm Adjusted p
Amikacin	3/90 (3.3)	4/56 (7.1)	2.23 (0.48–10.36)	0.429	1.000
Cefepime	10/90 (11.1)	8/56 (14.3)	1.33 (0.49–3.61)	0.610	1.000
Ceftazidime	12/90 (13.3)	9/56 (16.1)	1.24 (0.49–3.18)	0.637	1.000
Ceftriaxone	12/90 (13.3)	10/56 (17.9)	1.41 (0.57–3.53)	0.483	1.000
Ciprofloxacin	6/90 (6.7)	5/56 (8.9)	1.37 (0.40–4.73)	0.749	1.000
Ertapenem	6/90 (6.7)	11/56 (19.6)	3.42 (1.19–9.86)	0.031	0.342
Gentamicin	2/90 (2.2)	4/56 (7.1)	3.38 (0.60–19.12)	0.204	1.000
Imipenem	12/90 (13.3)	11/56 (19.6)	1.59 (0.65–3.89)	0.354	1.000
Levofloxacin	6/90 (6.7)	5/56 (8.9)	1.37 (0.40–4.73)	0.749	1.000
Meropenem	3/90 (3.3)	4/56 (7.1)	2.23 (0.48–10.36)	0.429	1.000
Piperacillin/tazobactam	8/90 (8.9)	0/56 (0.0)	0.09 (0.005–1.52)	0.024	0.284
Trimethoprim/sulfamethoxazole	1/90 (1.1)	0/56 (0.0)	0.53 (0.02–13.19)	1.000	1.000

Notes: One first-listed isolate per patient was retained. Resistance was R versus S/I. Odds ratios compare hospital-classified with community-classified isolates. Fisher’s exact test was used for every row; Holm’s procedure adjusted the 12 p values. The Haldane-Anscombe correction was used for odds-ratio interval estimation when any cell was zero. No adjusted p value was below 0.05.

The all-isolate clustered sensitivity analysis produced a similar ertapenem estimate (OR 3.22, 95% CI 1.12–9.29; nominal p=0.030), which also did not survive Holm adjustment (Supplementary Table S3). The S. marcescens-restricted patient-level analysis likewise produced no multiplicity-adjusted difference (Supplementary Table S4). These analyses support a descriptive, exploratory interpretation rather than a claim of a setting-specific resistance association.

Discussion

This historical single-center series was dominated by S. marcescens and respiratory specimens. Most isolates were categorized as S at standard exposure for the tested agents, although imipenem, third-generation cephalosporins, cefepime, and ertapenem showed higher R proportions than the remaining reported agents. After one isolate per patient was retained and multiplicity was addressed, no acquisition-category comparison remained statistically significant. Clustered and S. marcescens-restricted sensitivity analyses did not change that conclusion.

The predominance of S. marcescens is consistent with its recognized role as the most frequent human Serratia pathogen.^1,2 Recent adult data provide more relevant context than neonatal outbreak reports. Pérez-Viso et al described 141 bloodstream-infection episodes over 15 years and documented temporal changes in resistance, underscoring the importance of study period and population.³ Cosimato et al similarly reported marked agent-level variation in a recent institutional series.⁴ Turkish studies have also shown variability in specimen distributions and susceptibility, supporting the value of local description while cautioning against broad generalization.^5,6

S. marcescens resistance biology complicates simple phenotype labels. Chromosomal AmpC, altered permeability, efflux, acquired ESBLs, and carbapenemases may produce overlapping susceptibility patterns.^2,10–12 The 2024 IDSA guidance treats S. marcescens as less likely than some Enterobacterales to develop clinically significant AmpC overexpression but still emphasizes susceptibility-guided treatment.¹³

Earlier population-based surveillance from Canada and a hospital series from Crete similarly showed that Serratia species distribution, specimen mix, and susceptibility estimates vary by population and study period.^14,15 A US bacteremia cohort additionally identified injection drug use as a major population-specific exposure, illustrating how case mix may shape observed clinical patterns.¹⁶ These findings further support cautious cross-study comparison.

The nominal ertapenem difference is hypothesis-generating only. The estimate was imprecise, did not survive Holm correction, and may reflect differences in ICU exposure, specimen mix, species composition, prior antibiotic exposure, or unmeasured healthcare contact. Prior antibiotic use was not available, so greater healthcare antimicrobial exposure cannot be tested. The absence of blood isolates in the community-classified group and the concentration of tracheal aspirates in the hospital-classified group particularly limit direct interpretation. The full patient-level clinical department distribution is provided in Supplementary Table S5.

This study has several limitations. It is retrospective and single-center, uses data from 2019–2021, and includes isolates rather than adjudicated infections. The acquisition variable could not be reconstructed from timing and prior-contact variables. Culture dates were unavailable, so the first-listed record was used for patient-level deduplication and repeat intervals could not be calculated. Raw MICs and inhibition-zone diameters were not retained in the analytic extract, and the available records did not specify the inoculum standard or quality-control strain. Uncommon species were not confirmed by MALDI-TOF MS or sequencing. The retrospective records did not support reliable characterization of a specific beta-lactamase phenotype or mechanism. The primary comorbidity field did not capture multimorbidity. Pooling species and specimen types introduces heterogeneity, and event counts did not support stable multivariable adjustment. These limitations restrict causal and clinical inference. Isolate-level AST method provenance and formal acceptance criteria were not retained in the analytic extract.

The study therefore provides a historical single-center description and an exploratory comparison that accounts for repeated observations and multiple testing. It should not be used as evidence of current local susceptibility or as a substitute for current antibiograms, patient-specific testing, or contemporary treatment guidance.

Conclusion

S. marcescens predominated among Serratia isolates collected at this tertiary-care center during 2019–2021, and respiratory specimens were the most frequent source. After patient-level deduplication and multiplicity adjustment, no antimicrobial showed a robust resistance difference by recorded acquisition category. The dataset is best interpreted as a historical descriptive baseline; current clinical decisions require contemporary local susceptibility data and isolate-specific results.

Funding Statement

This research received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

Declaration of Generative AI

OpenAI ChatGPT was used to support English-language editing and reproducible computational cross-checking of the revised analyses. The authors verified the source data, statistical outputs, interpretation, references, and final text and take full responsibility for the content.

Abbreviations

AMR, antimicrobial resistance; CI, confidence interval; EUCAST, European Committee on Antimicrobial Susceptibility Testing; GEE, generalized estimating equation; ICU, intensive care unit; OR, odds ratio; PCR, polymerase chain reaction; R, resistant; S, susceptible at standard exposure; I, susceptible at increased exposure.

Data Sharing Statement

A de-identified isolate-level dataset and analysis syntax supporting the revised analyses are available from the corresponding author on reasonable request, subject to institutional and ethical requirements.

Author Contributions

All authors made a significant contribution to the work reported, whether that is in the conception, study design, execution, acquisition of data, analysis and interpretation, or in all these areas; took part in drafting, revising or critically reviewing the article; gave final approval of the version to be published; have agreed on the journal to which the article has been submitted; and agree to be accountable for all aspects of the work.

Disclosure

The authors report no conflicts of interest in this work.