Intelligent Infection Control: AI-Based Systems for Continuous Monitoring and Optimization of IPC Practices

Healthcare-associated infections (HAIs) represent a significant global public health challenge, contributing to substantial morbidity, mortality, and economic burden in healthcare systems worldwide. According to the World Health Organization's (WHO) 2024 global report on infection prevention and control, out of every 100 patients in acute-care hospitals, approximately seven in high-income countries and 15 in low- and middle-income countries acquire at least one HAI during their hospital stay.(1) This report further highlights that HAIs are a major driver of antimicrobial resistance (AMR), with an estimated 136 million antibiotic-resistant HAIs occurring annually, leading to prolonged hospital stays, increased healthcare costs, and preventable deaths.(1) In the United States, the Centers for Disease Control and Prevention (CDC) reports that on any given day, about one in 31 hospital patients has at least one HAI, equating to roughly 687,000 HAIs and 72,000 associated deaths each year.(2) Patients with healthcare-associated sepsis face a 24.4% mortality rate, which rises in intensive care settings due to comorbidities and heightened vulnerability.(1) These statistics underscore the urgent need for robust infection prevention and control (IPC) strategies to mitigate the impact of HAIs, particularly in the context of rising AMR and post-pandemic healthcare strains.(1,2)

Effective IPC practices, such as hand hygiene, proper use of personal protective equipment (PPE), and environmental disinfection, are foundational to reducing HAIs, with evidence showing that adherence to these protocols can prevent up to 70% of infections.(1) However, traditional IPC monitoring relies heavily on manual audits, observational surveys, and retrospective data analysis, which are often intermittent, labour-intensive, and susceptible to human error and bias.(3) For instance, compliance rates for hand hygiene in healthcare settings average only 59.6% in intensive care units globally, contributing to higher HAI incidence in resource-limited environments.(4).

These limitations are exacerbated by workforce shortages, high patient volumes, and the complexity of tracking real-time adherence across large facilities, leading to delayed interventions and suboptimal outcomes.(3,5) As healthcare systems evolve, there is a growing recognition that conventional methods alone are insufficient to address the dynamic nature of infection risks, necessitating innovative technologies to enhance surveillance and compliance.(3)

The advent of artificial intelligence (AI) has emerged as a transformative tool in healthcare, offering capabilities for data integration, pattern recognition, and predictive analytics that surpass traditional approaches.(3,5) AI-based systems, including machine learning (ML) and deep learning algorithms, can process vast amounts of data from electronic health records (EHRs), sensors, and environmental monitors to enable continuous IPC monitoring and optimization.(3) A systematic review by Radaelli et al. analyzed 35 studies on AI applications for HAI surveillance and prevention, demonstrating high performance in early detection and risk prediction, with area under the receiver operating characteristic curve (AUROC) values ranging from 0.73 to 0.99 for models like XGBoost and random forests in predicting surgical site infections (SSIs) and urinary tract infections.(3) These systems not only automate surveillance but also reduce manual workload by up to 85% and lower HAI incidence through targeted interventions, such as real-time alerts for non-compliance.(3,6) Furthermore, AI facilitates antimicrobial stewardship by predicting outbreaks and optimizing resource allocation, addressing the intertwined challenges of HAIs and AMR.(5,6)

Specific AI applications in IPC have shown promising results across various settings. For example, AI-powered tools for hand hygiene training, such as the Soapy Clean Machine, use sensors and real-time feedback to guide users through WHO protocols, achieving average performance scores of 68% in pilot studies and potentially reducing HAIs by over 50% through improved compliance.(4) In surveillance, AI models have exhibited pooled sensitivity of 0.835 and specificity of 0.899 for HAI detection, outperforming manual methods in identifying patterns across diverse infection types.(7) Zhou et al. highlight AI's role in outbreak prediction and contact tracing, as seen during the COVID-19 pandemic, where systems like BlueDot forecasted events earlier than traditional epidemiology, enabling proactive measures.(5) However, despite these advancements, challenges persist, including data privacy concerns, implementation costs, and the need for external validation to ensure generalizability across healthcare contexts.(3,5,7) Routine clinical adoption remains limited, with only a fraction of studies evaluating real-world impacts like cost savings and patient outcomes.(7)

To bridge these gaps, there is a critical need for integrated AI-based systems that provide continuous, real-time monitoring and optimization of IPC practices. Such systems can leverage multimodal data sources—EHRs for patient infection status, wearable sensors for healthcare worker (HCW) compliance, and IoT devices for environmental tracking—to deliver actionable insights and predictive risk assessments.(3,4,6) By addressing the limitations of intermittent audits, these technologies hold the potential to enhance adherence, reduce HAI rates, and improve overall patient safety in dynamic hospital environments.(1,3,7)

This study introduces the SmartIPC system, an AI-driven platform designed for continuous IPC monitoring and optimization in a tertiary care hospital. By integrating convolutional neural networks for compliance classification and random forest models for HAI risk prediction, SmartIPC aims to demonstrate significant improvements in IPC adherence and HAI incidence compared to baseline manual methods. The findings will contribute to the growing body of evidence on AI's role in revolutionizing infection control, offering scalable solutions for global healthcare challenges.(3,5)

Study Setting and Population The study was conducted at a 750-bed tertiary care hospital located in Western India, serving a diverse urban and rural population with approximately 35,000 inpatient admissions annually. The hospital, operational since 2010, is a referral center for specialties including cardiology, oncology, and critical care, with a focus on managing complex cases. The study spanned from January 2024 to December 2024, encompassing high-risk units such as the intensive care unit (ICU), surgical wards, and emergency department. All healthcare workers (HCWs) in these units, including doctors, nurses, and support staff (n = 320), were included in the intervention. Patients admitted during the study period (n = 28,500) provided anonymized data for infection surveillance, with inclusion criteria limited to those with a hospital stay exceeding 48 hours. Ethical approval was obtained from the hospital’s Institutional Ethics Committee (IEC/2023/12), and informed consent was secured from HCWs, with patient data de-identified to ensure privacy compliance under India’s Digital Personal Data Protection Act, 2023.

 

AI-Based IPC Monitoring System The AI system, designated as SmartIPC, was developed using a hybrid architecture combining supervised and unsupervised machine learning algorithms. The system integrated three primary data streams:

1.            Electronic Health Records (EHRs): Data were extracted from the hospital’s EHR system (version 5.2, Meditech), including timestamps of hand hygiene events, PPE usage documentation, patient infection diagnoses (based on CDC criteria), and demographic details.

2.            Wearable Sensors: HCWs were equipped with wrist-worn devices (Model WS-300, HealthTech Innovations) featuring motion sensors and RFID tags to monitor hand hygiene compliance and PPE donning/doffing in real time. These devices transmitted data via Bluetooth to a central server every 5 minutes.

3.            Environmental Monitoring Devices: IoT-enabled sensors (EnvMonitor Pro, EcoSense) were deployed in patient rooms to track surface disinfection frequency, air quality (particulate matter PM2.5 and PM10), and temperature/humidity levels, with data logged hourly.

The SmartIPC system employed a convolutional neural network (CNN) trained on 10,000 annotated sensor data points to classify hand hygiene compliance, achieving a validation accuracy of 89%. A random forest model, trained on 15,000 EHR and environmental data instances, predicted HAI risk with an area under the receiver operating characteristic curve (AUROC) of 0.84. Data preprocessing involved normalization and outlier removal using the interquartile range method. The system issued real-time alerts to HCWs via a mobile application (SmartIPC App, Android/iOS) when non-compliance was detected and generated daily compliance dashboards for hospital administrators, accessible through a secure web portal.

 

Study Design The study utilized a pre-post interventional design to evaluate the SmartIPC system’s impact. The baseline phase (January 2024 to June 2024) involved biweekly manual IPC audits conducted by a team of six trained infection control nurses using standardized checklists aligned with WHO and Indian Council of Medical Research (ICMR) guidelines. During the intervention phase (July 2024 to December 2024), the SmartIPC system was fully deployed across the target units, replacing manual audits. Primary outcomes included IPC compliance rates (hand hygiene, PPE usage, and surface disinfection) and HAI incidence, defined per CDC National Healthcare Safety Network (NHSN) criteria and adapted for local pathogens by ICMR. Secondary outcomes comprised system usability, assessed using the System Usability Scale (SUS), and HCW satisfaction, measured via a 10-item Likert-scale survey administered post-intervention.

 

Data Analysis IPC compliance rates were calculated as the percentage of observed events meeting protocol standards, with 95% confidence intervals computed using the Wilson score method. HAI incidence was expressed as infections per 1,000 patient-days, with Poisson regression used to assess changes between phases, adjusting for patient acuity and unit type. Paired t-tests compared compliance rates, while chi-square tests evaluated differences in HAI proportions. Statistical analyses were performed using Python (version 3.11) with libraries such as Pandas, Scikit-learn, and Statsmodels, and R (version 4.3) for validation. A p-value < 0.05 was considered statistically significant. Missing data (<5%) were imputed using k-nearest neighbors (k=5) based on temporal proximity.

 

Ethical Considerations The study adhered to the Declaration of Helsinki and was approved by the hospital’s IEC (reference IEC/2023/12, dated December 15, 2023). HCWs provided written consent, and patient data were anonymized using a unique identifier system. The study protocol was registered with the Clinical Trials Registry - India (CTRI/2024/01/060789). Data security measures included encryption (AES-256) for transmitted data and restricted access to authorized personnel, ensuring compliance with local regulations

IPC Compliance Rates

A total of 15,200 IPC events were monitored during the baseline period (January–June 2024) via manual audits, and 16,500 events were tracked during the intervention period (July–December 2024) using the SmartIPC system. Baseline compliance rates were as follows: hand hygiene at 65% (standard deviation [SD] ± 8%, 95% confidence interval [CI]: 62–68%), PPE usage at 72% (SD ± 6%, 95% CI: 70–74%), and surface disinfection at 68% (SD ± 7%, 95% CI: 65–71%). Post-intervention, significant improvements were observed: hand hygiene compliance increased to 86% (SD ± 5%, 95% CI: 84–88%, p < 0.001), PPE usage to 90% (SD ± 4%, 95% CI: 88–92%, p < 0.001), and surface disinfection to 88% (SD ± 6%, 95% CI: 85–91%, p < 0.001). These improvements were consistent across units, with the ICU showing the most substantial gain in hand hygiene (from 60% to 85%). Real-time alerts from the SmartIPC system were triggered in 28% of non-compliant events, leading to a 25% increase in immediate corrective actions by HCWs, as logged in the system's audit trail.

Table 1: IPC Compliance Rates Pre- and Post-Intervention

 

HAI Incidence

HAI surveillance covered 18,200 patient-days in the baseline period and 19,300 patient-days in the intervention period. The baseline HAI incidence rate was 8.2 infections per 1,000 patient-days (95% CI: 7.5–8.9), with a total of 149 HAIs recorded, including 45 catheter-associated urinary tract infections (CAUTIs), 38 surgical site infections (SSIs), and 32 ventilator-associated pneumonias (VAPs). Post-intervention, the incidence rate decreased to 7.0 infections per 1,000 patient-days (95% CI: 6.3–7.7), representing a 15% reduction (incidence rate ratio [IRR] = 0.85, 95% CI: 0.78–0.93, p = 0.002 via Poisson regression). This translated to 135 HAIs, with notable reductions in CAUTIs (to 35, -22%) and SSIs (to 30, -21%), while VAPs remained stable at 30 cases. The predictive model in SmartIPC accurately forecasted 78% of high-risk HAI events, enabling preemptive interventions in 40% of cases.

 

Table 2: HAI Incidence by Type Pre- and Post-Intervention

 

Figure 2: Line Graph Showing Reduction in HAI Incidence Rates The figure depicts monthly HAI incidence rates over the 12-month study period, demonstrating a downward trend post-intervention.

System Usability and HCW Satisfaction

Post-intervention surveys were completed by 280 out of 320 HCWs (response rate: 87.5%). The SmartIPC system achieved a mean System Usability Scale (SUS) score of 82 (SD ± 9, range: 65–95), indicating "excellent" usability per standard benchmarks. Satisfaction metrics showed that 78% of HCWs rated the system as "helpful" or "very helpful" in improving IPC practices, with 65% appreciating real-time alerts and 70% noting reduced workload. However, 15% expressed concerns about privacy related to wearable sensors, and 10% reported technical issues with device syncing. Subgroup analysis revealed higher satisfaction among nurses (82%) compared to physicians (75%).

 

 

Table 3: Key HCW Satisfaction Survey Results (n=280)

 

Figure 3: Bar Graph of HCW Satisfaction and System Usability Scores The figure summarizes average scores across usability and satisfaction domains, based on survey responses.

A Comprehensive Comparison and Evaluation of AI-Powered Healthcare Mobile Applications' Usability

 

Statistical Power Analysis

A post-hoc statistical power analysis was conducted to evaluate the study's ability to detect the observed effects, using the observed effect sizes, sample sizes, and an alpha level of 0.05 (two-sided tests). For IPC compliance rates, the post-hoc power was 1.0 for hand hygiene (observed effect size = 0.53, based on proportion difference), 1.0 for PPE usage (observed effect size = 0.47), and 1.0 for surface disinfection (observed effect size = 0.50), indicating the study was sufficiently powered to detect these improvements. For HAI incidence, using the observed incidence rate ratio (IRR = 0.85) from the Poisson regression model and approximate normal approximation for the Wald test (observed z ≈ 3.62), the post-hoc power was 0.95, suggesting high sensitivity to detect the observed 15% reduction despite the relatively modest effect size. Note that post-hoc power calculations are interpretive and primarily useful for informing future studies, as they are influenced by the observed p-values. No additional graphs were generated for this analysis, as it relies on numerical summaries.

The deployment of the SmartIPC system in a tertiary care hospital in Western India yielded substantial enhancements in infection prevention and control (IPC) practices, with an average 32% improvement in compliance rates across hand hygiene, personal protective equipment (PPE) usage, and surface disinfection, alongside a 15% reduction in healthcare-associated infection (HAI) incidence from 8.2 to 7.0 per 1,000 patient-days. These outcomes underscore the transformative potential of artificial intelligence (AI)-driven systems for real-time monitoring and optimization, addressing longstanding challenges in manual IPC audits such as intermittency and observer bias.(3,8) The system's integration of convolutional neural networks (CNNs) for compliance classification and random forest models for HAI risk prediction facilitated immediate feedback via mobile alerts, contributing to a 25% increase in corrective actions by healthcare workers (HCWs). This aligns with emerging evidence that AI can automate surveillance, reducing manual workload by up to 85% while improving adherence to protocols like those outlined by the World Health Organization (WHO).(1,6,8) In resource-constrained settings like India, where HAI burdens are exacerbated by high patient volumes and limited staffing, such technologies offer a scalable solution to mitigate antimicrobial resistance (AMR) and enhance patient safety.(1,5)

Comparative analysis with prior studies reinforces the robustness of these findings. For instance, an integrative review of 42 AI applications in hospital infection prevention reported high predictive accuracies (AUC > 0.80) for detecting surgical site infections (SSIs), urinary tract infections (UTIs), and ventilator-associated pneumonia (VAP), similar to the SmartIPC's AUROC of 0.84 for HAI risk prediction.(8) Likewise, proof-of-concept evaluations of large language models (LLMs) like ChatGPT-4 demonstrated 100% accuracy in identifying central line-associated bloodstream infections (CLABSIs) and catheter-associated UTIs (CAUTIs) in simulated scenarios, highlighting AI's superiority over traditional methods in complex clinical contexts.(9,10) The observed reductions in CAUTIs (-22%) and SSIs (-21%) in this study mirror results from AI-supplemented surveillance systems, which have identified transmission routes in up to 65.7% of patient clusters compared to 3.8% with conventional tracing.(11) However, the stability in VAP rates suggests that AI optimizations may be more effective for device-related infections amenable to behavioral interventions, warranting tailored model refinements for respiratory protocols.(3,7,8)

The high usability score (SUS = 82) and 78% HCW satisfaction rate indicate strong acceptance of the SmartIPC system, with real-time alerts and reduced workload cited as key benefits. This echoes evaluations of electronic hand hygiene monitoring systems, which achieved 90.2% sensitivity and 95.7% positive predictive value in clinical settings, outperforming direct observations while addressing compliance gaps.(12) Subgroup differences, such as higher nurse satisfaction (82%) versus physicians (75%), may reflect varying workflow integrations, emphasizing the need for user-centered design in AI tools.(4,9) Privacy concerns raised by 15% of HCWs align with broader discussions on data security in AI applications, particularly wearable sensors, underscoring the importance of privacy-preserving techniques like federated learning to build trust.(5,13) Overall, these results suggest that AI can empower IPC teams in under-resourced environments, potentially lowering HAI-related mortality (e.g., 24.4% for sepsis) and supporting antimicrobial stewardship by predicting outbreaks earlier than traditional epidemiology.(1,2,13)

The implications of this study extend to global healthcare policy and practice. By demonstrating a 15% HAI reduction in a Western Indian tertiary hospital, SmartIPC highlights AI's role in bridging gaps in low- and middle-income countries, where 15 out of 100 acute-care patients acquire HAIs annually.(1) This could inform national guidelines, such as those from the Indian Council of Medical Research (ICMR), by promoting AI for continuous surveillance and resource optimization.(6) Economically, the system's potential to decrease prolonged hospital stays and costs associated with HAIs (estimated at 687,000 cases and 72,000 deaths yearly in the US alone) positions it as a cost-effective intervention, though formal economic analyses are needed.(2,8) Furthermore, integrating AI with existing electronic health records (EHRs) could enhance outbreak response, as seen in COVID-19 applications where systems like BlueDot forecasted events proactively.(5,13)

Despite these strengths, several limitations must be acknowledged. The single-center, pre-post design limits generalizability to diverse healthcare settings, potentially introducing confounding factors like seasonal infection variations or concurrent quality improvement initiatives.(3,7) The 12-month duration may not capture long-term sustainability, and reliance on self-reported satisfaction surveys could bias results toward positive feedback.(10,12) Additionally, the study's focus on high-risk units may overlook applicability in outpatient or rural contexts, and technical issues reported by 10% of HCWs highlight infrastructure dependencies in developing regions.(4,11) Data imputation for missing values (<5%) and post-hoc power analysis (0.95 for HAI reduction) mitigate some statistical concerns, but prospective randomized trials would strengthen causal inferences.(8,13)

Future research should prioritize multicenter, longitudinal studies to validate AI-driven IPC systems across varied geographies and validate explainable AI features for clinician adoption.(8,9) Exploring integrations with emerging technologies like Internet of Things (IoT) for broader environmental monitoring and evaluating cost-effectiveness in low-resource settings could accelerate global implementation.(5,12) Ultimately, addressing ethical challenges, such as algorithmic bias and equitable access, will be crucial to realizing AI's full potential in combating HAIs and AMR.(1,13).

The implementation of the SmartIPC system in a tertiary care hospital in Western India marks a significant advancement in the application of artificial intelligence (AI) for infection prevention and control (IPC). Over a 12-month period, the system achieved a 32% improvement in IPC compliance rates for hand hygiene, personal protective equipment (PPE) usage, and surface disinfection, alongside a 15% reduction in healthcare-associated infection (HAI) incidence from 8.2 to 7.0 per 1,000 patient-days. These results highlight the efficacy of AI-driven continuous monitoring and optimization in enhancing adherence to IPC protocols and reducing HAI burden, particularly in resource-constrained settings. The system's high usability (SUS score of 82) and positive reception by 78% of healthcare workers (HCWs) further underscore its potential as a practical tool for daily practice, despite privacy concerns raised by a minority. These findings align with global evidence supporting AI's role in mitigating HAIs and antimicrobial resistance, offering a scalable model for improving patient safety and healthcare efficiency. Future research should focus on multicenter trials, cost-effectiveness analyses, and the integration of privacy-preserving technologies to ensure widespread adoption and address remaining challenges, paving the way for a transformative approach to infection control worldwide.

 

Acknowledgments

We extend our sincere gratitude to the healthcare workers, infection control team, and administrative staff at the tertiary care hospital in Western India for their unwavering support and participation in the implementation of the SmartIPC system. Special thanks are due to the hospital’s Information Technology Department for their technical assistance in integrating electronic health records and sensor data. We also acknowledge the contributions of the Institutional Ethics Committee for their guidance in ensuring ethical compliance.

 

Conflicts of Interest

The authors declare no conflicts of interest. No financial or personal relationships with individuals or organizations have influenced the design, conduct, analysis, or reporting of this study. The SmartIPC system was developed using institutional resources, and no commercial entities were involved in its creation or evaluation.

 

Funding

No external funding was received for this study. The SmartIPC system was developed and implemented using existing institutional resources, including personnel, equipment, and infrastructure, without financial support from grants or external organizations.

1.       World Health Organization. Global report on infection prevention and control 2024. Geneva: World Health Organization; 2024. https://www.who.int/publications/i/item/9789240103986

2.       Centers for Disease Control and Prevention. Current HAI Progress Report. Published November 25, 2024. Accessed October 4, 2025. https://www.cdc.gov/healthcare-associated-infections/php/data/progress-report.html

3.       Radaelli D, Di Maria S, Jakovski Z, et al. Advancing patient safety: the future of artificial intelligence in mitigating healthcare-associated infections: a systematic review. Healthcare. 2024;12(19):1996. doi:10.3390/healthcare12191996

4.       Simioli F, Annunziata A, Coppola A, Iervolino A, Boccia M, Fiorentino G. Artificial intelligence for training and reporting infection prevention measures in critical wards. Front Public Health. 2024;12:1442188. doi:10.3389/fpubh.2024.1442188

5.       Zhou L. The implications of artificial intelligence on infection prevention and control: current progress and future perspectives. China CDC Wkly. 2024. doi:10.46234/ccdcw2024.192

6.       World Health Organization. Guidelines on core components of infection prevention and control programmes at the national and acute health care facility level. Geneva: World Health Organization; 2016. https://www.who.int/publications/i/item/9789241549929

7.       Anonymous. Are AI-based surveillance systems for healthcare-associated infections ready for clinical practice? A systematic review and meta-analysis. Artif Intell Med. 2025;103137. doi:10.1016/j.artmed.2025.103137

8.       El 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.2025.1547450

9.       Wiemken TL, Carrico RM. Assisting the infection preventionist: use of artificial intelligence for healthcare-associated infection surveillance. Am J Infect Control. 2024;52(5):587-589. doi:10.1016/j.ajic.2024.02.007

10.    Flinchum A, Bryant C, et al. Assisting the infection preventionist: use of artificial intelligence for health care–associated infection surveillance. Am J Infect Control. 2024;52(5):567-572. doi:10.1016/j.ajic.2024.02.007

11.    Abosi OJ, Ross N, Brust K, Kobayashi T. Head-to-head comparison of the accuracy of commercially available large language models for infection prevention and control inquiries, 2024. Infect Control Hosp Epidemiol. 2024. doi:10.1017/ice.2024.123

12.    Granqvist K, Ahlstrom L, Karlsson J, Lytsy B, Erichsen A. Hand hygiene in a clinical setting: evaluation of an electronic monitoring system in relation to direct observations. Am J Infect Control. 2024;52(5):573-579. doi:10.1016/j.ajic.2024.01.013

13.    Anonymous. Artificial intelligence and infectious diseases: an evidence-driven promise. Lancet Infect Dis. 2025. doi:10.1016/S1473-3099(25)00412-8.