Leveraging Artificial Intelligence in Infection Prevention and Control

Article appeared in Healthcare Hygiene Magazine May/June 2026 issue.

By:  Hillary Hei, MPH LSSGB CIC FAPIC, Senior Infection Preventionist, ECRI

Artificial intelligence (AI) has become a major disruptor across global industries, and the field of infection prevention and control (IPC) is no exception. The creation and implementation of AI technologies represent a major alternative to traditional reactive workflows. Traditionally, IPC has relied on manual, retrospective reviews—a process where infection preventionists (IPs) spend hours scouring charts to identify infections and detect healthcare outbreaks. However, AI changes this process by acting as a proactive partner that can monitor data in real-time, predict risks before they occur, and automate the most labor-intensive parts of surveillance. This shift is essential given that healthcare-acquired infections (HAIs) remain a significant threat to patient safety, contributing to increased morbidity, mortality, and billions of dollars in excess healthcare expenditures annually.¹

Understanding the AI Toolbox

AI is a large umbrella term that encompasses a diverse array of computational approaches. The most common form seen in published applications in IPC is Machine Learning (ML),^2,3 which uses statistical models to find patterns in large datasets, such as identifying which patients are most likely to develop a certain infection. A more advanced branch, Deep Learning, mimics the human brain’s neural networks to process complex information like medical images or high-resolution signals.

Another critical tool is Natural Language Processing (NLP). This technology allows computers to “read” and understand human language, including the unstructured clinical notes and imaging reports that often contain the most important clues about a patient’s condition. Finally, Large Language Models (LLMs) are generative AI models that provide humanlike responses to users’ prompts, such as chatbots. These are being tested for their ability to summarize clinical cases and help decide if a patient meets the official definitions for an HAI.

The New Standard for Surveillance

One of the most immediate impacts of AI is in automating the surveillance of infections like central line-associated bloodstream infections (CLABSI) and catheter-associated urinary tract infections (CAUTI). Traditional surveillance is labor-intensive, time-consuming, and requires extensive training to ensure reliable objectivity.

Recent studies show that AI-assisted reviews are consistently more efficient than manual review alone for correctly identifying CLABSI⁴ and CAUTI.⁵ LLMs can correctly identify CLABSI and CAUTI when provided with complete clinical information, assisting with surveillance and education of newer IPs.^6,7 While AI tools can correctly identify infections when given clear clinical descriptions, they still face challenges with missing data. This highlights a recurring theme: AI is best used as a supportive tool that complements, rather than replaces, human clinical judgment.^3,8,9

Surgical site infections (SSI) are also a major area of focus, as identification is complex and labor-intensive. The integration of ML and NLP allows for the automated harvesting from both structured and unstructured electronic health record (EHR) sources. While structured data like laboratory results (e.g., positive wound cultures) are easily captured, the clinical context required to confirm an SSI is often buried in unstructured free-text clinician notes or surgical summaries. NLP models can be trained to recognize specific clinical criteria, such as purulent draining and erythema. Researchers have successfully used ML and NLP to assist with identifying SSIs with high accuracy.^10-13 In an innovative twist, some models are now even assessing potential SSIs using photos submitted by patients through electronic portals, providing a way to monitor wounds after the patient has left the hospital.¹⁴ Advanced time-series models are even beginning to predict the onset of an SSI an average of one day before a clinical diagnosis is made.¹⁵

Predicting Risks and Preventing Harm

Beyond just finding infections, AI helps predict who is at risk before the infection occurs. This predictive analytics allows for risk-stratification, where resources and more vigilant care can be directed toward the most at-risk patients. Prospective risk scores have been explored using AI tools for predicting sepsis, SSI, hospital-onset bacteremia (HOB), Clostridioides difficile infections (CDI), urinary tract infections, ventilator-associated pneumonia, CLABSI, and multi-drug resistant organisms.^3,8,9 The number of published studies has increased exponentially over the years. Newer models are even looking beyond the individual patient. One group demonstrated improved risk predictions for HOB by including "non-patient" features such as the history of the previous occupant of a patient’s hospital bed and the number of healthcare workers per day.¹⁶

Most studies include performance metrics such as sensitivity, specificity, and accuracy to demonstrate predictive power. However, to build clinician trust, some researchers are integrating explainable AI (XAI) methods, such as SHAP values, that pull back the curtain on "black box" models to show exactly why a patient is flagged, such as an extended duration of an indwelling catheter or a high number of unique caregivers.^10,17,18 These XAI systems enable prompt notification to bedside teams to immediately revisit and reinforce prevention bundles during the patient's most vulnerable windows of care. This evolution moves AI from a retrospective reporting tool into a prospectively actionable partner, allowing for evidence-based interventions that can stop an infection before it starts.

Real-Time AI-Based Alerting and Monitoring Systems

AI is increasingly utilized to provide real-time feedback to IPC teams and healthcare workers regarding compliance with essential IPC protocols. These systems use a combination of computer vision, wearable sensors, and Internet of Things (IoT) devices to monitor behavior in the clinical environment, which has historically been difficult to track accurately through human observation. AI applications for tracking hand hygiene include video analytics, badge-based sensors, and alcohol handrub dispenser sensors.² Similarly, AI-powered computer vision can detect the appropriate use of personal protective equipment (PPE), ensuring that staff are adequately protected when entering rooms of patients on transmission-based precautions.¹⁹ Lastly, AI can assist IPs in detecting undiscovered healthcare outbreaks. Sunderman et al.²⁰ pairs AI with whole genome sequencing to identify clusters of similar pathogen strains and potential transmission routes.

AI Outside the IPC Department

The influence of AI extends into other fields adjacent to IPC. In ensuring environmental disinfection, industry leaders in supplemental disinfection technologies are utilizing AI-powered robotic automation to plan their own paths and disinfect surfaces in the healthcare environment. In the sterile processing department, AI-powered cameras are being trained to identify thousands of different surgical instruments, ensuring that sets are complete and reducing human error during prep and pack.²¹ For high-level disinfection, flexible endoscopes are notoriously difficult to clean. AI is now being applied to imaging software that looks into the narrow channels of the scope to identify microscopic damage, residual moisture, or biological residue.²²

The Road to Universal Adoption: Barriers and Recommendations

Leading organizations in the IPC domain now recommend using technologies like AI and NLP to educate, train, and test clinical staff.²³ However, several hurdles remain before these tools become universal.

• Data Integrity and Integration: AI is only as good as the data it receives. Incomplete records, non-standard vocabulary, and delayed data integration are persistent barriers. In order for the AI tools to be effective, they must be integrated into a facility’s electronic health record system and its adjacent databases.
• Generalizability and Bias: Most AI models are trained at a single hospital. A model that works perfectly in a large academic center in the U.S. might not work at all in a small rural clinic or a different country. Furthermore, there is a risk of "dataset bias," where an algorithm might inadvertently learn to provide unequal care based on how the data was originally collected.
• Cost and Complexity: The high setup and maintenance costs of these systems can be a significant deterrent, especially for under-resourced facilities. Additionally, tools may require continued validation and troubleshooting, as advances in modeling may require continued optimization.

AI offers opportunities to strengthen IPC workflows, but applications must be thoughtfully planned prior to fully integrating into practice. The misuse of AI chatbots in healthcare is the top concern in ECRI’s Top 10 Hazards for 2026 due to the potential for users to take outputs as fact. With any AI tool, users must recognize the limitations of all AI-enabled health technologies. For IPC teams ready to implement AI technologies, Gastaldi et al2 created a structured, evidence-based checklist to help with planning, implementation, and risk mitigation.

Conclusion

For stakeholders across the IPC spectrum, the integration of AI is not just a technological upgrade but rather a new way of working. By partially automating routine surveillance and providing advanced warnings of patient risk, AI allows IPs to spend more time on the front lines of patient safety. As the technology continues to mature, the focus must remain on ensuring these tools are transparent, equitable, and seamlessly integrated into the daily workflows of healthcare.

References:

^{About HAIs. Centers for Disease Control and Prevention. Updated June 27, 2025. Accessed March 1, 2026, 2026. https://www.cdc.gov/healthcare-associated-infections/about/index.htmlGastaldi S, Tartari E, Satta G, Allegranzi B. Advancing infection prevention and control through artificial intelligence: a scoping review of applications, barriers, and a decision-support checklist. Antimicrob Steward Healthc Epidemiol. 2025;5(1):e317. doi:10.1017/ash.2025.10191Cozzolino C, Mao S, Bassan F, et al. Are AI-based surveillance systems for healthcare-associated infections ready for clinical practice? A systematic review and meta-analysis. Artif Intell Med. Jul 2025;165:103137. doi:10.1016/j.artmed.2025.103137Morgan DJ, AlShanqeeti S, Coffey KC, et al. Using generative artificial intelligence to identify central line-associated bloodstream infections. Clin Infect Dis. Nov 19 2025;doi:10.1093/cid/ciaf636Alshanqeeti S, Coffey K, McDermott K, et al. Comparing Generative Artificial Intelligence vs Experts for Detection of Catheter-Associated Urinary Tract Infection. Clin Infect Dis. Feb 9 2026;82(1):148-150. doi:10.1093/cid/ciaf486Wiemken TL, Carrico RM. Assisting the infection preventionist: Use of artificial intelligence for health care-associated infection surveillance. Am J Infect Control. Jun 2024;52(6):625-629. doi:10.1016/j.ajic.2024.02.007Rodriguez-Nava G, Egoryan G, Goodman KE, Morgan DJ, Salinas JL. Performance of a large language model for identifying central line-associated bloodstream infections (CLABSI) using real clinical notes. Infect Control Hosp Epidemiol. Oct 30 2024;46(3):1-4. doi:10.1017/ice.2024.164Barbati C, Viviani L, Vecchio R, et al. Artificial intelligence use and performance in detecting and predicting healthcare-associated infections: A systematic review. Artif Intell Med. Feb 2026;172:103321. doi:10.1016/j.artmed.2025.103321van der Werff SD, van Rooden SM, Henriksson A, et al. The future of healthcare-associated infection surveillance: Automated surveillance and using the potential of artificial intelligence. J Intern Med. Aug 2025;298(2):54-77. doi:10.1111/joim.20100Celik U, Liu F, Kobayashi K, et al. Machine Learning-Enhanced Surveillance for Surgical Site Infections in Patients Undergoing Colon Surgery: Model Development and Evaluation Study. JMIR Form Res. Oct 1 2025;9:e75121. doi:10.2196/75121Pierce L, Pak C, Stanley K, et al. Retrospective validation study of a large language model approach to screening for intraabdominal surgical site infections for quality and safety reporting. BMJ Qual Saf. Dec 21 2025;doi:10.1136/bmjqs-2025-019493Chakraborty A, Lybarger K, Estebane JAO, et al. Development and Evaluation of Machine Learning Models for the Identification of Surgical Site Infection in Electronic Health Records. Surg Infect (Larchmt). Sep 2025;26(7):474-481. doi:10.1089/sur.2024.266Perkins L, O'Keefe T, Ardill W, Potenza B. Modernizing Surgical Quality: A Novel Approach to Improving Detection of Surgical Site Infections in the Veteran Population. Surg Infect (Larchmt). Sep 2024;25(7):499-504. doi:10.1089/sur.2024.013Muaddi H, Choudhary A, Lee F, et al. Imaging-based Surgical Site Infection Detection Using Artificial Intelligence. Ann Surg. Sep 1 2025;282(3):419-428. doi:10.1097/SLA.0000000000006826Kiser AC, Shi J, Bucher BT. An explainable long short-term memory network for surgical site infection identification. Surgery. Jul 2024;176(1):24-31. doi:10.1016/j.surg.2024.03.006Vazquez Guillamet MC, Zhang J, Bewley A, et al. Integrating Nonindividual Patient Features in Machine Learning Models of Hospital-Onset Bacteremia. JAMA Netw Open. Jul 1 2025;8(7):e2518815. doi:10.1001/jamanetworkopen.2025.18815El Arab RA, Almoosa Z, Alkhunaizi M, Abuadas FH, Somerville J. Artificial intelligence in hospital infection prevention: an integrative review. Front Public Health. 2025;13:1547450. doi:10.3389/fpubh.2025.1547450Ashrafi N, Abdollahi A, Alaei K, Pishgar M. Enhanced prediction of ventilator-associated pneumonia in patients with traumatic brain injury using advanced machine learning techniques. Sci Rep. Apr 2 2025;15(1):11363. doi:10.1038/s41598-025-95779-0McLuckey RJ, Chowdhury Z, Goodman KE, et al. Exploring AI-assisted cameras to assess use of contact precautions. Infect Control Hosp Epidemiol. Feb 23 2026:1-3. doi:10.1017/ice.2026.10408Sundermann A, Chen J, Saul M, et al. Artificial intelligence enhances genomic surveillance in healthcare outbreak investigations. Infect Control Hosp Epidemiol. Nov 14 2025:1-5. doi:10.1017/ice.2025.10355Haider SA, Ho OA, Borna S, et al. Use of Multimodal Artificial Intelligence in Surgical Instrument Recognition. Bioengineering (Basel). Jan 15 2025;12(1)doi:10.3390/bioengineering12010072Barakat MT, Girotra M, Banerjee S. Initial application of deep learning to borescope detection of endoscope working channel damage and residue. Endosc Int Open. Jan 2022;10(1):E112-E118. doi:10.1055/a-1591-0258Talbot TR, Baliga C, Crapanzano-Sigafoos R, et al. SHEA/APIC/IDSA/PIDS multisociety position paper: Raising the bar: necessary resources and structure for effective healthcare facility infection prevention and control programs. Infect Control Hosp Epidemiol. Apr 28 2025;46(7):1-19. doi:10.1017/ice.2025.73}