Imagine a world where a simple infection could become a deadly threat, claiming over a million lives annually due to bacteria that laugh in the face of our strongest antibiotics. That's the stark reality of antibiotic resistance, and it's hitting global health harder than ever. But here's where it gets controversial: is simply cutting back on antibiotic use enough to turn the tide, or are there deeper factors at play that we're overlooking? Dive in with us as we explore this pressing issue, uncovering thresholds that could redefine how we fight back against resistant bacteria—and ponder the debates that rage around them.
Introduction
Antibiotic resistance stands as one of the most formidable challenges facing global public health today. According to recent estimates, it directly contributes to approximately 1.27 million deaths worldwide each year from infections caused by resistant bacteria. Gram-negative bacteria (GNB), a group of microbes known for their complex outer membranes that make them harder to kill, are increasingly dominating hospital-acquired infections. Among these, carbapenem-resistant Gram-negative bacteria (CRGN) present a particularly stubborn barrier to effective antibiotic treatment. Data from China's Antimicrobial Surveillance Network (CHINET) for 2022 reveal that isolation rates of clinical GNB are a staggering 2.51 times higher than those of Gram-positive bacteria. The most prevalent GNB species identified include Escherichia coli (E. coli), Klebsiella pneumoniae (K. pneumoniae), Pseudomonas aeruginosa (P. aeruginosa), and Acinetobacter baumannii (A. baumannii), mirroring trends seen in prior years. Furthermore, resistance to carbapenems— a class of powerful antibiotics used as last-resort options—among these four species rose steadily from 2005 to 2018 but began declining from 2019 to 2021.
Carbapenem antibiotics play a crucial role in combating severe infections caused by GNB. CRGN encompasses strains like carbapenem-resistant Escherichia coli (CREC), carbapenem-resistant Acinetobacter baumannii (CRAB), carbapenem-resistant Klebsiella pneumoniae (CRKP), and carbapenem-resistant Pseudomonas aeruginosa (CRPA), all of which are frequently encountered in hospital settings. These resistant bugs not only increase mortality rates but also amplify healthcare costs and the overall burden of antimicrobial resistance. For instance, resistance in GNB often stems from plasmids—mobile genetic elements carrying multiple resistance genes—that force clinicians to resort to even more potent or toxic antibiotics, heightening risks to patient safety and driving up expenses. Studies in the European Union and European Economic Area highlight CRKP and CREC as pathogens linked to the highest fatality rates among infected individuals. This resistance complicates the development of effective treatment plans significantly. And this is the part most people miss: while controlling carbapenem use might seem straightforward, there's ongoing debate about whether it alone can curb CRGN detection rates. Some research points to a clear link between carbapenem consumption and CRGN prevalence, with overuse potentially fueling resistance. Yet, other studies challenge this, suggesting the relationship isn't always straightforward. Although antimicrobial stewardship programs—strategies to promote responsible antibiotic use—can slow resistance, over-restrictive policies might backfire. Past investigations have proposed thresholds for antibiotic consumption, defining the maximum safe levels while exploring the interplay between usage and resistance. Advanced modeling techniques, like nonlinear time series analysis, such as spline regression, vector autoregressive models (VAR), and generalized additive models (GAM), offer better insights into these dynamics than simpler linear approaches.
Building on this, our study delves into the connection between CRGN detection rates and the antimicrobial use density (AUD) of six antibiotic classes in a Chinese tertiary hospital from 2015 to 2023. Through nonlinear time-series analysis, we pinpointed critical thresholds for carbapenems, fluoroquinolones, third-generation cephalosporins, aminoglycosides, penicillins, and glycopeptides, paving the way for smarter anti-infective strategies in clinical settings.
Methods
Data Source
Our data on antimicrobial consumption and carbapenem resistance rates, collected quarterly, comes from the Second Qilu Hospital of Shandong University. This 2,431-bed tertiary care facility in northern China serves as a university-affiliated comprehensive hospital with educational roles. To ensure accuracy, we applied strict inclusion criteria: patients with hospital-acquired infections, qualified pathogen samples, and complete clinical records. Exclusion factors included infections or suspected infections present before admission, hospital stays shorter than 48 hours, contaminated or unqualified samples, and repeated isolates from the same patient site.
Bacterial Isolates and Susceptibility Testing
Susceptibility testing adhered to guidelines from the National Health Commission of the People's Republic of China for clinical microbiology labs. For quality assurance, we used standard strains like E. coli ATCC®25922, K. pneumoniae ATCC®700603, and P. aeruginosa ATCC®27853. Meropenem served as the representative carbapenem for testing. We employed the MicroScan WalkAwayplus system, interpreting results per Clinical and Laboratory Standards Institute (CLSI) standards.
Frequency and Intensity of Antimicrobial Drug Use
Antimicrobials were categorized using the Anatomical Therapeutic Chemical (ATC) system, focusing on six commonly prescribed classes: penicillins (including penicillin G, amoxicillin, ampicillin, and piperacillin/tazobactam), aminoglycosides (amikacin and etilmicin), fluoroquinolones (moxifloxacin and levofloxacin), carbapenems (imipenem and meropenem), glycopeptides (vancomycin and telicoplanin), and third-generation cephalosporins (ceftriaxone, cefotaxime, cefixime, cefoperazone/sulbactam, and ceftazidime). Usage frequency was gauged by quarterly drug consumption in grams divided by the defined daily dose (DDD), as recommended by the World Health Organization (WHO). AUD was then calculated as DDDs multiplied by 100, divided by patient-days, expressed in DDDs per 100 patient-days. Patient-days here represent the total inpatient days when CRGN was identified.
Statistical Analysis
Analysis was conducted using R version 4.3.2. Regression discontinuity designs (RDDs), a powerful method for assessing causal impacts of interventions without randomized trials, helped us evaluate AUD's effect on CRGN resistance rates. We modeled a cubic regression of the intervention variable (AUD) and outcome variable (CRGN detection rates) using global parameter estimation regression discontinuity (RD). The model formula is:
Yi = β + β1 * (time - cutpoint)^k + β2 * treatment * (time - cutpoint)^k + Ɛi
Here, Yi is the overall CRGN detection rates, treatment is the combined AUD of six antibiotic classes, β denotes the treatment effect at the cutoff, cutpoint is the time breakpoint, β1 and β2 are coefficients for the kth term of the time difference and its interaction with treatment, and Ɛi is the residual error.
For deeper insights, we used GAMs to examine nonlinear associations between antibiotic consumption and CRGN rates, identifying thresholds. The GAM equations are:
Y = g(μ) + fi(Xi) + s(time)
GAMs excel at uncovering predictor relationships without assuming linear patterns, using iterative fitting. Adjusted R² measures correlation strength (closer to 1 means stronger link), lag coefficients account for time delays based on Akaike Information Criterion (AIC, where lower values indicate better fit), and a time smoothing term reduces noise. X1 to Xn are AUDs for different antibiotics, fi(Xi) is a nonparametric function, and g(μ) links to the outcome. This approach captures nonlinear ties between AUD, resistance, and time. We focused on models with P < 0.05 and adjusted R² > 0.3 as significant. To find thresholds, we maximized gradient change rates in regression models.
Results
Trends in AUD and Detection Rates of CRGN
Over the study period, hospitalized patients infected with CREC, CRAB, CRKP, and CRPA numbered 142, 5,299, 1,534, and 1,921, respectively. Cases with simultaneous CRAB and CRKP infections totaled 3, CRAB and CRPA 2, and CRKP and CRPA 5. Figure 1 illustrates AUD trends for all six antibiotic classes from 2015 to 2023. Consumption remained stable from Q1 2015 to Q4 2019, dipped from Q4 2019 to Q1 2021, then gradually rose from Q1 2021 to Q3 2023. Figure 2 shows CRGN detection rates: CREC stayed steady from Q1 2015 to Q3 2023, while CRAB consistently exceeded 50%, peaking above 90% from Q4 2018 to Q2 2019 before declining. CRKP and CRPA rates climbed from Q1 2015 to Q2 2019, then fell through Q4 2019 to Q3 2023.
[Figure 1: Antibiotic Consumption Trends 2015–2023. Measured in DDDs/100 patient-days across classes. Abbreviations: DDD, defined daily dose.]
[Figure 2: CRGN Detection Rate Trends 2015–2023. Abbreviations: CRGN, carbapenem-resistant Gram-negative bacteria; CREC, carbapenem-resistant Escherichia coli; CRKP, carbapenem-resistant Klebsiella pneumoniae; CRAB, carbapenem-resistant Acinetobacter baumannii; CRPA, carbapenem-resistant Pseudomonas aeruginosa.]
Regression Discontinuity Designs
Using Q1 2020 as the COVID-19 breakpoint, we analyzed discontinuities via locally weighted smoothed fits. Figures 3a and 3b reveal breaks in AUD and CRGN rates at this point. A linear RD model showed a significant AUD link to CRGN rates (β = 0.499, P = 0.039), with a lower AIC indicating good fit (Table 1).
[Figure 3: Smoothed Fits for AUD and CRGN Rates vs. Time. (a) AUD over time; (b) CRGN rates over time. Abbreviations: AUD, antimicrobial use density; CRGN, carbapenem-resistant Gram-negative bacteria.]
[Table 1: RD Results for CRGN Detection Rate Changes.]
Multifactor Time Series Analysis Based on GAM Models
Among the six classes, CRAB rates correlated significantly with carbapenems (lag = 1, P = 0.016, adjusted R² = 0.521), fluoroquinolones (lag = 1, P = 0.001, adjusted R² = 0.347), and glycopeptides (lag = 1, P = 0.001, adjusted R² = 0.643) (Table 2). Deeper analysis showed CRAB rates dropping when carbapenem, aminoglycoside, and glycopeptide use fell below thresholds of 5.82, 0.06, and 0.77 DDDs/100 patient-days (Figure 4).
CRKP rates linked strongly to carbapenems (lag = 1, P = 0.001, adjusted R² = 0.808), aminoglycosides (lag = 3, P = 0.006, adjusted R² = 0.787), and glycopeptides (lag = 1, P = 0.005, adjusted R² = 0.797). Resistance declined below thresholds of 5.82, 7.36, and 0.96 DDDs/100 patient-days for these classes.
[Table 2: GAM Analysis of CRGN Rates and Antibiotic Use, 2015–2023.]
[Figure 4: Thresholds via Gradient Change in GAMs. (a) Aminoglycosides and CRKP; (b) Carbapenems and CRKP; (c) Carbapenems and CRAB; (d) Fluoroquinolones and CRAB; (e) Glycopeptides and CRAB; (f) Glycopeptides and CRKP. Abbreviations: CRGN, carbapenem-resistant Gram-negative bacteria; CRKP, carbapenem-resistant Klebsiella pneumoniae; CRAB, carbapenem-resistant Acinetobacter baumannii; GAM, generalized additive model.]
Discussion
Our research uncovered nonlinear links between CRGN detection and six antibiotic classes from 2015 to 2023. Fluoroquinolones and third-generation cephalosporins dominated usage, aligning with prior studies but exceeding European levels. Total consumption plummeted in Q1 2021, coinciding with COVID-19 management in Jinan, China, before rising post-pandemic. CRGN rates for all four species declined from 2019 to 2022, likely due to altered hospital prescribing, fewer inpatients, and national antibiotic rationalization policies. COVID-19 had no direct data impact, as infected patients weren't treated here. Enhanced stewardship, lab-clinical collaboration, and awareness curbed resistance.
CRGN rates correlated significantly with carbapenems, fluoroquinolones, aminoglycosides, and glycopeptides (P < 0.05), with thresholds at 5.82, 0.06, 7.09, and 0.77 DDDs/100 patient-days. These vary by region due to differences in healthcare, usage, and resistance patterns, suggesting localized thresholds are key. High AUD increases bacterial exposure, potentially selecting mutants producing AmpC enzymes. For example, β-lactam exposure can foster ceftazidime resistance in K. pneumoniae. These benchmarks guide stewardship, balancing resistance delay with overuse avoidance. When carbapenem AUD hit 5.82 DDDs/100 patient-days, triggering multidisciplinary reviews lowered it and CRKP rates. CRAB rates, highest among CRGN and rising in China, responded to quarterly threshold monitoring, indirectly reducing carbapenem use. This highlights factors beyond consumption, urging precise carbapenem criteria. More research on quinolones and cephalosporins' impacts is needed.
But here's where it gets controversial: some argue that strict thresholds stifle innovation or ignore patient needs, potentially leading to undertreatment in complex cases. Is this a slippery slope toward rationing life-saving drugs? Others counter that without them, resistance spirals out of control. What do you think—should hospitals set hard limits, or prioritize individualized care?
Limitations include single-center data, unaddressed confounders like stay length or illness severity, and retrospective design lacking individual-level details. Resistance mechanisms and variables like patient demographics or infection control also play roles. Broader, multicenter studies are essential.
Conclusions
In essence, AUD ties directly to CRGN detection. Thresholds for carbapenems, aminoglycosides, fluoroquinolones, and glycopeptides stand at 5.82, 0.06, 7.09, and 0.77 DDDs/100 patient-days. Regular monitoring offers benchmarks to curb overuse, aligning with WHO principles. At threshold approaches, prioritize optimized plans or alternatives. Diversified antibiotic use and AUD reduction can minimize resistance gene selection. This framework supports targeted therapies and resistance control.
What are your thoughts on balancing antibiotic thresholds with patient care? Do they empower doctors or create unnecessary hurdles? Share in the comments—let's debate this critical topic!
Data Sharing Statement
The datasets utilized and analyzed in this study can be obtained from the corresponding author upon reasonable request.
Ethics Approval and Patient Consent Statement
This research received approval from the Institutional Review Board of The Second Qilu Hospital of Shandong University.
Funding
No funding supported this work.
Disclosure
The authors declare no conflicts of interest.
References
Tacconelli E, Sifakis F, Harbarth S, et al. Surveillance for control of antimicrobial resistance. Lancet Infect Dis. 2018;18(3):e99–e106. doi:10.1016/S1473-3099(17)30485-1
Xiao YH, Nishijima T. Status and challenges of global antimicrobial resistance control: a dialogue between Professors Yonghong Xiao and Takeshi Nishijima. hLife. 2024;2(12):47–49. doi:10.1016/j.hlife.2023.11.004
Lin TL, Chang PH, Chen IL, et al. Risk factors and mortality associated with multi-drug-resistant Gram-negative bacterial infection in adult patients following abdominal surgery. J Hosp Infect. 2022;119:22–32. doi:10.1016/j.jhin.2021.09.021
Thai VC, Stubbs KA, Sarkar-Tyson M, et al. Phosphoethanolamine transferases as drug discovery targets for therapeutic treatment of multi-drug-resistant pathogenic gram-negative bacteria. Antibiotics. 2023;12(9):1382–1389. doi:10.3390/antibiotics12091382
Zhang X, Tan L, Ouyang P, et al. Analysis of distribution and antibiotic resistance of Gram-positive bacteria isolated from a tertiary-care hospital in southern China: an 8-year retrospective study. Front Microbiol. 2023;14:1220363. doi:10.3389/fmicb.2023.1220363
Zhen S, Wang H, Feng S. Update of clinical application in ceftazidime-avibactam for multidrug-resistant Gram-negative bacteria infections. Infection. 2022;50(6):1409–1423. doi:10.1007/s15010-022-01876-x
Sganga G. Burden of antibiotic resistant gram negative bacterial infections: evidence and limits. J Med Microbiol Diagn. 2014;3(1):1050–1051. doi:10.4172/2161-0703.1000132
Fitzpatrick MA. Real-world antibiotic needs for resistant gram-negative infections. Lancet Infect Dis. 2020;20(10):1108–1109. doi:10.1016/S1473-3099(20)30178-X
Zhao H, Liu M, Wu J, et al. Mortality and genetic diversity of antibiotic-resistant bacteria associated with bloodstream infections: a systemic review and genomic analysis. BMC Infect Dis. 2024;24(1):1385. doi:10.1186/s12879-024-10274-7
Cassini A, Högberg LD, Plachouras D, et al. Attributable deaths and disability-adjusted life-years caused by infections with antibiotic-resistant bacteria in the EU and the European economic area in 2015: a population-level modelling analysis. Lancet Infect Dis. 2019;19(1):56–66. doi:10.1016/S1473-3099(18)30605-4
Tam VH, Hirsch EB, Lasco TM, et al. Correlation of hospital carbapenem consumption and resistance trends in selected gram-negative bacteria. Ann Pharmacother. 2012;46(7–8):1120–1122. doi:10.1345/aph.1R021
Hu Y, Ping Y, Li L, et al. A retrospective study of risk factors for carbapenem-resistant Klebsiella pneumoniae acquisition among ICU patients. J Infect Dev Ctries. 2016;10(3):208–213. doi:10.3855/jidc.6697
Li J, Li Y, Song N, Chen Y. Risk factors for carbapenem-resistant Klebsiella pneumoniae infection: a meta-analysis. J Glob Antimicrob Resist. 2020;21:306–313. doi:10.1016/j.jgar.2019.09.006
Ho CM, Ho MW, Liu YC, et al. Correlation between carbapenem consumption and resistance to carbapenems among Enterobacteriaceae isolates collected from patients with intra-abdominal infections at five medical centers in Taiwan, 2006-2010. Int J Antimicrob Agents. 2012;40:S24–S28. doi:10.1016/S0924-8579(12)70006-7
Jiao Y, Qin Y, Liu J, et al. Risk factors for carbapenem-resistant Klebsiella pneumoniae infection/colonization and predictors of mortality: a retrospective study. Pathog Glob Health. 2015;109(2):68–74. doi:10.1179/2047773215Y.0000000004
Guo W, He Q, Wang Z, et al. Influence of antimicrobial consumption on gram-negative bacteria in inpatients receiving antimicrobial resistance therapy from 2008-2013 at a tertiary hospital in Shanghai, China. Am J Infect Control. 2015;43(4):358–364. doi:10.1016/j.ajic.2014.12.010
van Leth F, Schultsz C. Unbiased antimicrobial resistance prevalence estimates through population-based surveillance. Clin Microbiol Infect. 2023;29(4):429–433. doi:10.1016/j.cmi.2022.05.006
Lopez-Lozano JM, Lawes T, Nebot C, et al. A nonlinear time-series analysis approach to identify thresholds in associations between population antibiotic use and rates of resistance. Nat Microbiol. 2019;4(7):1160–1172. doi:10.1038/s41564-019-0410-0
Aldeyab MA, Bond SE, Gould I, et al. Identification of antibiotic consumption targets for the management of Clostridioides difficile infection in hospitals- a threshold logistic modelling approach. Expert Rev Anti Infect Ther. 2023;21(10):1125–1134. doi:10.1080/14787210.2023.2263642
Bondi A, Cutoni A, Peradotto M, et al. Performance evaluation of bd phoenix and microscan walkaway plus for determination of fosfomycin susceptibility in enterobacterales. Antibiotics. 2023;12(7):1106–1113. doi:10.3390/antibiotics12071106
Chen L, Lu YW, Xu J, et al. Prediction of drug’s anatomical therapeutic chemical (ATC) code by constructing biological profiles of ATC codes. BMC Bioinf. 2025;26(1):86–108. doi:10.1186/s12859-025-06102-7
Fukushima Y, Fukushima F, Kamiya K, et al. Relation between the antimicrobial susceptibility of clinical isolates of pseudomonas aeruginosa from respiratory specimens and Antimicrobial Use Density (AUD) from 2005 through 2008. Internal Med. 2010;49(14):1333–1340. doi:10.2169/internalmedicine.49.3360
Walkey AJ, Drainoni ML, Cordella N, et al. Advancing quality improvement with regression discontinuity designs. Ann Am Thorac Soc. 2018;15(5):523–529. doi:10.1513/AnnalsATS.201712-942IP
Tanskanen J, Taipale S, Anttila T. Revealing hidden curvilinear relations between work engagement and its predictors: demonstrating the added value of Generalized Additive Model (GAM). J Happiness Stud. 2016;17:367–387. doi:10.1007/s10902-014-9599-z
Chen Y, Ebenstein A, Greenstone M, et al. Evidence on the impact of sustained exposure to air pollution on life expectancy from China’s Huai River policy. Proc Natl Acad Sci U S A. 2013;110(32):12936–12941. doi:10.1073/pnas.1300018110
Chen S, Li Z, Shi J, et al. A nonlinear time-series analysis to identify the thresholds in relationships between antimicrobial consumption and resistance in a chinese tertiary hospital. Infect Dis Ther. 2022;11(3):1019–1032. doi:10.1007/s40121-022-00608-w
Hayajneh WA, Al-Azzam S, Yusef D, et al. Identification of thresholds in relationships between specific antibiotic use and carbapenem-resistant Acinetobacter baumannii (CRAb) incidence rates in hospitalized patients in Jordan. J Antimicrob Chemother. 2021;76(2):524–530. doi:10.1093/jac/dkaa463
Ryu S, Klein EY, Chun BC. Temporal association between antibiotic use and resistance in Klebsiella pneumoniae at a tertiary care hospital. Antimicrob Resist Infect Control. 2018;7:83–89. doi:10.1186/s13756-018-0373-6
Jacoby TS, Kuchenbecker RS, Dos Santos RP, et al. Impact of hospital-wide infection rate, invasive procedures use and antimicrobial consumption on bacterial resistance inside an intensive care unit. J Hosp Infect. 2010;75(1):23–27. doi:10.1016/j.jhin.2009.11.021
Hu F, Zhu D, Wang F, et al. Current Status and Trends of Antibacterial Resistance in China. Clin Infect Dis. 2018;67(suppl_2):S128–S134. doi:10.1093/cid/ciy657
Hu F, Guo Y, Yang Y, et al. Resistance reported from China antimicrobial surveillance network (CHINET) in 2018. Eur J Clin Microbiol Infect Dis. 2019;38(12):2275–2281. doi:10.1007/s10096-019-03673-1
