RWE Reproduction Project: Antibiotic Resistance Surveillance in Southeast France.

RWE
Real-World Data
Antimicrobial Resistance
Epidemiology
Trend Analysis
Reproduction of the published RWE analysis: ‘Major Discrepancy Between Factual Antibiotic Resistance and Consumption in South of France’ (Scientific Reports, 2020), using simulated/aggregated data.
Author

Ousmane Diallo, MPH-PhD

Published

November 16, 2025

Important

Note: All plots labeled “Original published figure” are taken from Diallo et al., Scientific Reports (2020).

All analytical steps in this portfolio project (data structure design, data cleaning, harmonization of AST results, resistance calculations, DTR classification, and statistical workflow) are reproduced using fully simulated datasets to ensure GDPR and data protection compliance.

No original isolate-level or patient-level data were reused.

Overview

This portfolio project reconstructs the analytical workflow and Real-World Evidence (RWE) methodology from the peer-reviewed publication:

Diallo OO et al. Major discrepancy between factual antibiotic resistance and consumption in South of France. Scientific Reports 2020;10:18262.

The goal is not to reproduce the original results, but to demonstrate the full RWE analytic pipeline using synthetic data modeled after the real-world microbiology surveillance network used in the publication.

This approach allows:

  • faithful reconstruction of the study design,
  • demonstration of large-scale RWE data engineering skills,
  • reproduction of the DTR classification approach,
  • and illustration of AMR trend analysis techniques—

while remaining fully compliant with GDPR and ethical requirements.

Original figures from the publication are included for scientific context and comparison.

Background

Antimicrobial resistance (AMR) is often quantified using predictive mathematical models, which may not reflect real-world patterns.

The Scientific Reports study analyzed 539,037 bacterial isolates collected across 267 laboratories to quantify:

  • Real-world resistance trends over time

  • Discrepancies between resistance and antibiotic consumption

  • Prevalence of Difficult-to-Treat (DTR) phenotypes (Kadri et al.)

  • Temporal evolution of key pathogens (E. coli, K. pneumoniae, S. aureus, A. baumannii…)

This portfolio project reproduces the analytic strategy, statistical logic, and visual narratives using simulated data.

Objectives of This RWE Reproduction

  1. Reconstruct the study design used in the original 2020 publication.

  2. Create a clean analytic workflow using modern RWE standards.

  3. Simulate datasets matching the structure and distribution of the original study.

  4. Reproduce key indicators:

    • Resistance to key antibiotics

    • Temporal trends (2014–2018)

    • DTR classification

  5. Produce RWE-style visualizations (forest plots, timeline trends).

  6. Summarize clinical and public health implications.

This exercise demonstrates RWE reproducibility, analytical rigor, and epidemiological reasoning, essential in the pharmaceutical industry.

Real-World Data Structure (Simulated Version)

The original study used routine-care laboratory data from:

  • BALYSES (real-time bacterial monitoring system)

  • MARSS (Marseille AMR surveillance)

  • PACA SurvE (regional epidemiology system)

To maintain ethical compliance, this portfolio project uses:

  • simulated datasets

  • distributions calibrated on published tables

  • aggregate-level counts consistent with Scientific Reports 2020

The following data elements are reproduced:

  • Sample type (urine, blood culture, deep, respiratory)

  • Bacterial species (15 most frequent clinical pathogens)

  • Antibiotic susceptibility test (AST) outcomes

  • Key antibiotics for each species

  • Year of isolation

  • DTR phenotype classification

Methods (Reconstructed)

Study Design

  • Retrospective, observational, non-interventional RWD study (2014–2018).

  • No linked clinical data (mortality, comorbidity).

  • Unit of analysis: isolate, not patient.

Inclusion Criteria

  • Isolates from the 15 most frequent pathogens.

  • At least one key antibiotic tested per species.

Resistance Definition

  • EUCAST breakpoints

  • Intermediate considered resistant (per original study)

DTR Classification

Reproduced using the Kadri et al. algorithm:

  • GNB: resistant to all β-lactams, carbapenems, and fluoroquinolones

  • Gram-positive: resistant to (methicillin + gentamicin + vancomycin) or equivalent triads

Statistical Modeling

  • Mann–Whitney U

  • χ² and Fisher exact tests

  • Kendall correlation for trends

  • p < 0.05 considered significant

This statistical logic is faithfully reconstructed.

Key RWE Insights (From Simulated Reproduction)

Although data are simulated, trends reproduce the structure of published findings:

2. Significant reductions in resistance

  • Amikacin resistance decreased in E. coli, K. pneumoniae, P. mirabilis, K. oxytoca.

  • Imipenem resistance declined in A. baumannii.

3. Significant increases in resistance

  • Ceftriaxone resistance in:

    • E. cloacae

    • K. aerogenes

    • K. oxytoca

Species distribution (simulated RWD)

Figure 2: Most frequently isolated bacterial soecies

4. Resistance profiles across pathogens

This heatmap summarizes the resistance phenotypes by pathogen and antibiotic class, highlighting strong resistance clusters (e.g., E. faecium, A. baumannii) and preserved susceptibilities in Gram-positive organisms.

Fig 3: Heatmap resistance profiles

5. Prevalence of DTR Phenotypes

  • 0.3% of isolates

  • Mostly Gram-negative

  • Lower than rates reported in U.S. hospital networks

  • No upward trend observed

Table 5a. Prevalence of Difficult-to-Treat Resistance (DTR) phenotypes in the PACA region (January 2014 – February 2019)

Bacterial species DTR strains (n / total) Percentage (%)
Escherichia coli 85 / 246,353 0.03
Klebsiella pneumoniae 372 / 49,733 0.74
Klebsiella oxytoca 11 / 7,887 0.13
Proteus mirabilis 1 / 18,064 0.006
Serratia marcescens 3 / 4,437 0.07
Morganella morganii 5 / 4,935 0.10
Enterobacter cloacae 33 / 14,620 0.20
Enterobacter aerogenes 6 / 6,982 0.09
Pseudomonas aeruginosa 902 / 34,966 2.60
Acinetobacter baumannii 175 / 1,098 15.90
Enterococcus faecalis 3 / 36,857 0.008
Enterococcus faecium 5 / 4,871 0.10
Staphylococcus aureus 0 / 65,023 0.00
Staphylococcus epidermidis 3 / 28,527 0.01
Streptococcus agalactiae 0 / 14,684 0.00
Total 1,604 / 539,037 0.30

Table 5b. Comparison of DTR prevalence between PACA region and 173 U.S. hospitals (Kadri et al., 2018)

Bacterial species DTR strains (n / total) Percentage (%) U.S. hospitals (Kadri 2018)
Escherichia coli 85 / 246,353 0.03 0.04
Klebsiella pneumoniae 372 / 49,733 0.74 1.70
Enterobacter cloacae 33 / 14,620 0.20 0.60
Enterobacter aerogenes 6 / 6,982 0.09 0.60
Pseudomonas aeruginosa 902 / 34,966 2.60 2.30
Acinetobacter baumannii 175 / 1,098 15.90 18.30
Total 1,573 / 353,752 0.44 1.01 (471/46,521)

Clinical & Public Health Interpretation

This project supports the central conclusion of the 2020 Scientific Reports article:

  • High antibiotic consumption does not automatically translate into higher resistance.

  • Predictive AMR models may overestimate the burden in certain contexts.

  • Real-time laboratory surveillance (RWD) provides a more accurate and actionable picture.

  • Routine-care AMR data can guide empirical treatment and stewardship programs.

Ethical Considerations

To ensure strict compliance:

  • No original data from the published study are reused.

  • Only synthetic, simulated, or aggregated datasets are used.

  • No identifiable patient information is present.

  • The project is solely for educational and portfolio purposes.

  • The analytic framework is recreated, not the dataset.

This aligns with GDPR, HIPAA, and good RWE practice.

Limitations

  • Simulated data cannot perfectly replicate the distribution of real isolates.

  • No patient-level linkage (mortality, comorbidities).

  • Results cannot be used for clinical decision-making.

  • Geographic specificity limits generalizability.

Conclusion

This RWE reproduction successfully reconstructs the methodological framework and analytical insights from the 2020 Scientific Reports article using modern RWE standards and simulated datasets.

It demonstrates:

  • Expertise in large-scale AMR surveillance

  • Mastery of RWE methodologies

  • Reproducible analytic workflows

  • Ability to translate published research into data-driven insights

Skills Demonstrated

  • Real-World Evidence (RWE) methodology

  • Surveillance epidemiology

  • Handling large microbiology datasets

  • Statistical analysis of resistance trends

  • Implementation of DTR definitions

  • Data visualization for AMR

  • Scientific communication

Publication (Referenced)

Diallo OO, Baron SA, Dubourg G, et al. Major discrepancy between factual antibiotic resistance and consumption in South of France. Scientific Reports 2020;10:18262.

Ousmane Diallo, MPH-PhD – Biostatistician, Data Scientist & Epidemiologist based in Chicago, Illinois, USA. Specializing in SAS programming, CDISC standards, and real-world evidence for clinical research.

Back to top