Infectious Diseases and Antimicrobial Stewardship

Infectious diseases combine clinical complexity with rich data: microbiology cultures, genomic sequencing, antimicrobial susceptibility testing, and population-level epidemiology. AI enhances antimicrobial stewardship through real-time bug-drug mismatch detection, though alert fatigue remains a challenge. Genomic surveillance platforms like Nextstrain automate phylogenetic analysis, tracking pathogen evolution in real time. But outbreak prediction remains limited to short-term nowcasting. Long-range epidemic forecasting consistently fails.

Learning Objectives

After reading this chapter, you will be able to:

  • Evaluate AI systems for antimicrobial stewardship and antibiotic selection
  • Understand outbreak prediction and genomic surveillance AI applications
  • Assess rapid pathogen identification and resistance prediction tools
  • Navigate AI applications in infection control and hospital epidemiology
  • Recognize the role of AI in pandemic preparedness and response
  • Apply evidence-based frameworks for infectious diseases AI adoption

The Clinical Context: Infectious diseases combine clinical complexity (host factors, pathogen factors, antimicrobial pharmacology) with rich data sources (microbiology, genomics, epidemiology). AI applications span individual patient care to population-level outbreak response.

What Works Well:

Application Evidence Level Key Benefit
Antimicrobial stewardship alerts Moderate Reduces inappropriate antibiotic use
Genomic surveillance platforms Strong Real-time tracking of pathogen evolution
MALDI-TOF identification Strong Rapid species-level identification

What’s Emerging:

Application Status Notes
Genotype-to-phenotype resistance prediction Research Promising but inconsistent clinical validation
Outbreak prediction Variable Nowcasting works; long-range prediction does not
HAI risk prediction Moderate Implementation studies ongoing

The Bottom Line: AI enhances antimicrobial stewardship through decision support and genomic surveillance through automated phylogenetics. Outbreak prediction remains challenging except for short-term nowcasting. Resistance prediction from sequencing is promising but requires prospective validation. Alert fatigue affects stewardship tools as it does all clinical decision support.

Introduction

Antimicrobial resistance kills an estimated 1.27 million people globally each year, yet stewardship programs still struggle with inappropriate prescribing and alert fatigue. Genomic surveillance platforms like Nextstrain have transformed real-time pathogen tracking, but outbreak prediction beyond short-term nowcasting consistently fails despite substantial investment. The gap between what AI can do at scale (flag bug-drug mismatches, automate phylogenetics) and what it cannot (forecast epidemic trajectories months ahead) defines the current landscape.

Part 1: Antimicrobial Stewardship AI

The Stewardship Imperative

Antimicrobial resistance kills an estimated 1.27 million people globally each year (Murray et al., 2022). Stewardship programs reduce inappropriate antibiotic use by 10-30% and decrease C. difficile infections (CDC Core Elements, 2019). AI augments stewardship by processing microbiology data at scale.

Decision Support Applications

Current AI stewardship tools:

  1. Bug-drug mismatch detection
    • Real-time alerts when prescribed antibiotic lacks coverage for cultured organism
    • Integration with lab information systems
    • Performance: High sensitivity but alert fatigue with low positive predictive value
  2. Duration of therapy monitoring
    • Flags antibiotic courses exceeding guideline recommendations
    • Supports de-escalation and IV-to-oral conversion
  3. Empiric therapy recommendations
    • Antibiotic selection based on local antibiogram, patient factors, indication
    • Updates dynamically as resistance patterns shift

IDSA/SHEA Stewardship Guidelines

IDSA/SHEA Guidelines on Antimicrobial Stewardship

The joint IDSA/SHEA guidelines on antimicrobial stewardship programs (Barlam et al., 2016) provide the framework for AI implementation:

Core recommendations:

  • Preauthorization and prospective audit with feedback remain gold standards
  • Clinical decision support should be integrated into EHR workflows
  • Local antibiograms should drive empiric therapy recommendations
  • Stewardship interventions require ID or pharmacy oversight

Implications for AI:

  • AI tools should support, not replace, prospective audit processes
  • Automated recommendations require human stewardship review
  • Local validation essential: national models may not reflect institutional resistance patterns
  • Alert frequency must balance sensitivity with actionability

2024 Update: SHEA released position statements on pandemic preparedness addressing data infrastructure and surveillance systems (SHEA, 2024).

Implementation Evidence

Study Intervention Outcome
Timbrook et al., 2017 Rapid diagnostic + stewardship CDS Reduced mortality with mRDT + ASP
Corbin et al., 2022 ML personalized antibiograms Improved empiric coverage prediction

Alert Fatigue in Stewardship

The fundamental challenge: stewardship alerts compete with dozens of other clinical alerts. Studies show clinicians override 49-96% of antibiotic alerts (Slight et al., 2013).

Mitigation strategies:

  • Tiered alerting (hard stops for critical mismatches, soft alerts for duration)
  • Pharmacist review before alert delivery
  • Bundled recommendations rather than individual alerts

Part 2: Genomic Surveillance and Phylogenetic Analysis

The Genomic Surveillance Revolution

COVID-19 transformed pathogen genomics from research tool to clinical necessity. The infrastructure built for SARS-CoV-2 surveillance now applies to other pathogens.

Key Platforms

Nextstrain:

The Nextstrain platform (Hadfield et al., 2018) provides real-time genomic epidemiology through:

  • Automated phylogenetic tree construction
  • Geographic and temporal visualization
  • Clade and variant assignment
  • Open-source analysis pipelines

Nextstrain processes thousands of sequences daily for pathogens including SARS-CoV-2, influenza, Ebola, and monkeypox.

GISAID:

The Global Initiative on Sharing All Influenza Data (GISAID) hosts the largest repository of SARS-CoV-2 sequences (over 16 million sequences). Machine learning classification systems automatically assign:

  • Pango lineages
  • WHO variant designations
  • Clade membership

Africa CDC Pathogen Genomics Initiative:

Regional surveillance demonstrates AI-augmented genomics at scale. The Nextstrain Africa CDC builds track pathogen evolution across the continent, with automated quality control and lineage assignment (Wilkinson et al., 2021).

Automated Phylogenetic Analysis

AI accelerates phylogenetic workflows:

  1. Sequence quality control
    • Automated detection of frameshifts, premature stops, contamination
    • Flagging for manual review
  2. Lineage assignment
    • Pangolin (SARS-CoV-2), Nextclade, custom classifiers
    • Milliseconds per sequence vs. hours for manual curation
  3. Transmission cluster detection
    • Genetic distance thresholds for outbreak definition
    • Integration with epidemiologic data
  4. Recombination detection
    • Identification of mosaic genomes
    • Critical for SARS-CoV-2 and influenza

Resistance Gene Detection

Whole-genome sequencing enables genotypic resistance prediction:

Current capabilities:

Pathogen Resistance Prediction Clinical Use
M. tuberculosis INH, RIF, FQ resistance Clinical (Cepheid, Hain)
S. aureus mecA for MRSA Clinical
Enterobacteriaceae Carbapenemase genes Clinical for CPE
N. gonorrhoeae Research stage Not yet clinical

Limitations:

  • Novel resistance mechanisms not in databases
  • Phenotype-genotype discordance (expression levels, epistasis)
  • Turnaround time still exceeds culture-based AST for most pathogens

Machine Learning for Resistance Prediction: Validation Challenges

A 2025 Lancet Infectious Diseases Series assessed machine learning approaches to antimicrobial resistance prediction across pathogens (Miglietta et al., 2025). The findings reveal substantial gaps between research promise and clinical readiness.

Current validation status by pathogen:

Pathogen ML Approach Validation Status Clinical Use
M. tuberculosis WGS-based DST Most validated; WHO-endorsed catalogues Clinical in reference labs
E. coli (ESBL) Genotypic prediction Moderate validation; geographic variation Research/emerging clinical
K. pneumoniae Carbapenemase prediction Variable; resistance mechanism diversity Limited clinical
S. aureus MRSA prediction Strong for mecA; limited for other resistance Clinical (mecA only)
P. aeruginosa Multi-drug resistance Poor; complex resistance mechanisms Research only

Validation remains inconsistent across settings and populations. Key barriers include:

  • Geographic diversity: Models trained on high-income country data perform poorly in regions with different resistance epidemiology
  • Breakpoint variability: CLSI, EUCAST, and regional standards define resistance differently
  • Expression-level effects: Gene presence does not guarantee phenotypic resistance
  • Novel mechanisms: Databases lag behind emerging resistance determinants

The Series emphasizes that LMIC settings face the greatest diagnostic burden yet have the least validation data. Resistance prediction tools developed for European and North American populations require prospective validation in the settings where diagnostic capacity is most constrained.

Structured Data vs. Clinical Notes for AMR Prediction

Recent evidence challenges assumptions that clinical narratives improve AMR prediction. In a study of nearly 30,000 sepsis episodes across 10 BJC Healthcare hospitals, deep learning models using structured EHR data (demographics, vitals, labs, prior infection history) achieved AUROC 0.85 for predicting ceftriaxone-resistant Gram-negative infections in community-onset sepsis, while LLMs analyzing history and physical notes achieved only AUROC 0.74 (Hixon et al., 2025). Combining the approaches provided no improvement over structured data alone.

For AMR prediction, the signal lies in structured clinical and microbiologic history rather than narrative documentation. This finding aligns with broader evidence that gradient boosting and deep learning on tabular EHR data often outperform language models for clinical prediction tasks (see AI Fundamentals).

Part 3: Healthcare-Associated Infection Surveillance

AI in Infection Prevention

Infection preventionists face increasing surveillance requirements with limited staffing. AI promises to automate case finding and risk stratification.

SHEA/APIC Position on Digital Surveillance

SHEA on Digital Infection Prevention

SHEA and related organizations have addressed digital surveillance and infection prevention through various guidance documents:

Endorsed applications:

  • Automated HAI surveillance from EHR data
  • Real-time syndromic surveillance dashboards
  • Hand hygiene monitoring systems (with privacy considerations)
  • Antimicrobial use tracking and benchmarking

Implementation requirements:

  • Validation against manual chart review (sensitivity, specificity, PPV)
  • Transparent algorithms with audit capability
  • Integration with existing infection control workflows
  • Privacy protections for patient and staff data

Cautions:

  • AI predictions should not replace clinical judgment for isolation decisions
  • False positives consume investigation resources
  • Equity review required (do algorithms perform equally across patient populations?)

HAI Risk Prediction Evidence

Infection Type Model Performance Clinical Impact
Surgical site infection AUC 0.72-0.85 Targeted surveillance studies ongoing
CAUTI AUC 0.68-0.78 Mixed results on prevention
CLABSI AUC 0.70-0.82 Bundle compliance improvement when coupled with alerts
C. difficile AUC 0.65-0.75 Enhanced isolation protocols studied

Performance gap in resource-limited settings: The 2025 Lancet Infectious Diseases Series found that HAI prediction models developed in high-income settings show 10-20% lower sensitivity when deployed in LMICs (Miglietta et al., 2025). Contributing factors include differences in baseline HAI epidemiology, documentation practices, and the clinical workflows that generate training data. Models require local validation before deployment, particularly in settings with distinct pathogen distributions and infection control resources.

Key challenge: Predicting HAI is inherently difficult because it requires distinguishing colonization from infection and accounting for interventions that occur after prediction.

Automated Surveillance Systems

Commercial systems (Premier, Theradoc, Vigilanz) incorporate:

  • NHSN criteria-based case finding
  • Antibiotic exposure tracking
  • Line and device day counting
  • Automated reporting generation

Evidence: Automated systems achieve 85-95% sensitivity compared to manual review but with lower specificity (Woeltje et al., 2014).

Part 4: Rapid Diagnostics and Resistance Prediction

AI-Enhanced Microbiology

MALDI-TOF Mass Spectrometry:

Matrix-assisted laser desorption ionization-time of flight (MALDI-TOF) MS revolutionized clinical microbiology. AI pattern recognition provides species identification in minutes rather than days (Seng et al., 2009).

Current performance: - 95-99% accuracy for common pathogens - Lower accuracy for unusual species, mixed cultures - FDA-cleared (Vitek MS, MALDI Biotyper)

Gram Stain AI:

Computer vision analysis of Gram stain images:

  • Research accuracy: 80-93% for morphology and Gram reaction
  • Challenges: slide quality variation, mixed flora, unusual organisms
  • Limited clinical deployment; most systems in validation phase

Resistance Prediction from Spectral Data:

Emerging research uses MALDI-TOF spectra to predict: - MRSA vs. MSSA - Carbapenemase production - Extended-spectrum beta-lactamase (ESBL)

Performance varies widely (AUC 0.65-0.95) and remains mostly research-stage.

Molecular Diagnostics AI

Syndromic panels (BioFire, Verigene) use algorithmic interpretation:

  • Multiplexed pathogen detection
  • Automated resistance gene reporting
  • Integration with stewardship pathways

Clinical impact: Syndromic panels reduce time to appropriate therapy by 24-48 hours (Patel et al., 2018) but may increase broad-spectrum antibiotic use if not coupled with stewardship.

Part 5: Outbreak Prediction and Response

What Works: Nowcasting

Nowcasting estimates current disease burden from incomplete, lagged data. AI approaches:

  • CDC FluSight: Ensemble models forecast influenza 1-4 weeks ahead
  • COVID-19 Forecast Hub: Aggregated models for hospitalizations and deaths
  • Performance: Short-term forecasts reasonably accurate; accuracy degrades beyond 2-4 weeks

Key insight: Nowcasting outperforms traditional surveillance by 1-2 weeks, valuable for resource planning.

What Doesn’t Work: Long-Range Prediction

Despite substantial investment, AI cannot reliably predict:

  • When and where novel outbreaks will emerge
  • Pandemic timing or severity
  • Long-range seasonal disease patterns

Why prediction fails:

  • Rare events (low base rate problem)
  • Complex, stochastic determinants (human behavior, weather, pathogen evolution)
  • Data quality and timeliness limitations
  • Non-stationary dynamics (patterns change over time)

The Conceptual Framework for Prediction Failure

Beyond technical limitations, a 2025 Lancet Infectious Diseases framework explains why outbreak prediction faces fundamental, not merely computational, barriers (Odone et al., 2025).

Four fundamental barriers to prediction:

  1. Stochastic emergence: Spillover events are governed by rare, random interactions between pathogens, hosts, and environments that defy forecasting
  2. Behavioral feedback loops: Human responses to predictions change the outcomes being predicted (reflexivity)
  3. Pathogen evolution: Mutation and recombination introduce unpredictable changes mid-outbreak
  4. Data latency: By the time surveillance detects a signal, the intervention window has often closed

This framework reframes expectations: AI excels at nowcasting current burden from incomplete data, not predicting future emergence. Nowcasting provides 1-2 weeks of lead time for resource planning. Long-range prediction remains beyond current capabilities regardless of model sophistication.

The practical implication for ID specialists: invest in surveillance infrastructure that enables rapid response rather than prediction systems that promise early warning.

Syndromic Surveillance

AI-powered syndromic surveillance:

  • Emergency department chief complaint monitoring
  • OTC medication sales tracking
  • Social media and search trend analysis

Performance: Anomaly detection works for large outbreaks. Sensitivity for small, localized events remains poor.

Part 6: Professional Society Positions

IDSA Position on Digital Health

IDSA on Digital Health and Telemedicine

The Infectious Diseases Society of America has addressed digital tools through position statements on telemedicine (Young et al., 2019):

Telemedicine and digital consultations (IDSA Telehealth Resources):

  • Remote ID consultation supports stewardship in facilities without on-site ID
  • Telehealth platforms can provide educational opportunities, share tools, and allow antimicrobial use review with feedback
  • Documentation and liability considerations apply to technology-assisted consultations

Antimicrobial stewardship implications:

  • AI decision support should align with IDSA/SHEA stewardship guidelines
  • Local validation required before deployment
  • Human oversight of AI recommendations is essential

Genomic diagnostics:

  • WGS-based diagnostics require interpretive expertise
  • Automated interpretation tools should not replace ID physician review
  • Turnaround time for genomic results must meet clinical needs

Note: IDSA has not released a formal position statement specifically on AI. The principles above reflect general guidance applicable to clinical decision support.

CDC and WHO Positions

CDC Core Elements of Hospital Antibiotic Stewardship Programs (2019) recommend decision support systems but do not specifically address AI. Updated guidance anticipated.

WHO Global Action Plan on Antimicrobial Resistance includes surveillance and diagnostics strengthening but limited AI-specific guidance. The WHO AWaRe classification (Access, Watch, Reserve antibiotics) informs AI stewardship tools.

Clinical Scenarios

Case: A 68-year-old man with bacteremia is receiving piperacillin-tazobactam. Blood cultures grow E. coli susceptible to ceftriaxone, ampicillin-sulbactam, and TMP-SMX. The stewardship AI generates a “de-escalation recommended” alert for the attending physician.

The alert states: “Current therapy: piperacillin-tazobactam. Organism susceptible to narrower-spectrum agents. Consider de-escalation to ceftriaxone.”

Question: How should the ID consultant approach this alert?

Discussion

Key considerations:

  1. Source control: Has the source of bacteremia been identified and addressed?
  2. Clinical trajectory: Is the patient improving on current therapy?
  3. Polymicrobial risk: Are cultures from other sites pending?
  4. Patient factors: Allergies, prior resistant organisms, immunocompromise?

The AI’s limitation: The alert is based solely on the susceptibility report. It lacks context about clinical status, imaging findings, or pending workup.

Appropriate response:

  • Review the chart before acting on the alert
  • Document rationale if continuing broad-spectrum therapy
  • De-escalate when clinically appropriate, not simply because of an alert
Teaching point: Stewardship AI identifies opportunities for de-escalation. The decision to de-escalate requires clinical judgment that integrates information the AI does not have.

Case: A 45-year-old woman with recurrent UTIs has a positive urine culture. The lab performs rapid molecular testing that detects blaCTX-M (ESBL gene). Conventional susceptibilities are pending.

The EHR displays: “ESBL gene detected. Consider carbapenem therapy pending phenotypic susceptibilities.”

Question: How should this information guide empiric therapy?

Discussion

Genotype-phenotype considerations:

  1. Gene detection confirms ESBL production likely but expression levels vary
  2. Carbapenems are reliable for ESBL-producers but represent overtreatment if alternatives work
  3. Cephalosporins may fail but MIC testing will clarify
  4. Nitrofurantoin and fosfomycin may have activity regardless of ESBL status

The nuanced approach:

  • For severe infection: carbapenem empirically while awaiting phenotypic susceptibilities
  • For uncomplicated cystitis: can await full susceptibilities given low severity
  • Review prior cultures and susceptibility patterns
Teaching point: Genotypic resistance detection is faster than phenotypic testing but requires interpretation. The presence of a resistance gene informs but does not dictate therapy.

Case: The infection prevention team identifies three patients with carbapenem-resistant K. pneumoniae (CR-KP) on the same unit over two weeks. The hospital has WGS capability. Genomic analysis shows the isolates are within 5 SNPs of each other.

Question: What does the genomic data contribute to the investigation?

Discussion

Interpretation of genetic distance:

  • <10 SNPs for K. pneumoniae strongly suggests recent transmission
  • The three cases are likely part of a transmission cluster
  • Transmission may have occurred on the unit or in a shared exposure setting

What genomics does not tell you:

  • Who was the index case
  • The specific transmission route (hands, environment, devices)
  • Whether there are undetected carriers

Investigation steps:

  1. Enhanced contact precautions for known cases
  2. Review common exposures (procedures, staff, equipment)
  3. Consider point prevalence surveillance for additional carriers
  4. Environmental cultures if transmission route unclear
Teaching point: Genomic epidemiology confirms or refutes transmission clusters with high confidence. It accelerates outbreak response but does not replace shoe-leather epidemiology.

Case: An ID consultant receives a page because a “sepsis AI alert” fired on a 72-year-old patient with pneumonia on the medicine floor. The patient’s vital signs are stable, and lactate is normal. The AI alert cites elevated white count and recent antibiotic initiation as risk factors.

Question: How should the ID consultant interpret and respond to this alert?

Discussion

Context on sepsis prediction AI:

Sepsis prediction models have significant limitations (see Critical Care chapter): - High false positive rates (>80% in some implementations) - Low PPV: many alerts do not represent true sepsis - The Epic Sepsis Model, widely deployed, showed 33% sensitivity and 83% specificity in external validation (Wong et al., 2021)

This specific case:

  • Stable vital signs and normal lactate are reassuring
  • Antibiotic initiation and elevated WBC are expected for pneumonia treatment
  • The alert adds little to clinical assessment

Appropriate response:

  1. Brief chart review to confirm clinical stability
  2. No action required based solely on the AI alert
  3. Continue planned pneumonia management
  4. Document assessment if institution requires alert acknowledgment
Teaching point: Sepsis AI alerts have high false positive rates. Clinical judgment supersedes AI predictions. The goal is not to respond to every alert but to identify the subset representing true clinical deterioration.

Key Takeaways for ID Specialists

Practical Guidance

Antimicrobial stewardship AI:

  • AI tools identify de-escalation opportunities; clinical judgment determines appropriateness
  • Local validation essential: national resistance data may not reflect your institution
  • Alert fatigue is real: advocate for tiered, high-value alerts

Genomic surveillance:

  • Phylogenetic analysis confirms or refutes transmission clusters
  • Automated lineage assignment enables rapid variant tracking
  • Genotypic resistance prediction is fast but requires phenotypic confirmation

Outbreak prediction:

  • Nowcasting (1-4 week forecasts) provides actionable lead time
  • Long-range prediction remains unreliable
  • AI augments surveillance; it does not predict novel emergence

HAI prevention:

  • Automated surveillance improves case finding sensitivity
  • Risk prediction models require prospective validation
  • Alert fatigue affects infection prevention as it does other clinical domains

Professional standards:

  • IDSA/SHEA guidelines apply to AI-assisted stewardship
  • Human oversight of AI recommendations is required
  • Local validation and equity review are essential before deployment