AI-Enabled Indoor Air Quality Monitoring for a U.S. Children's Hospital

Introduction

Paediatric healthcare settings combine two conditions that make indoor air quality unusually consequential: occupants with developing respiratory systems and compromised immunity, and a building operating continuously with high turnover of people, cleaning chemistry and clinical consumables.

This trial evaluated whether continuous AI-interpreted monitoring could hold both threads at once, infection-relevant signals and chemical exposure, inside a working children's hospital.

Methodology

The study applied the same four-pronged approach used across the trial programme: data collection via sensor deployment, pattern recognition using the AI engine, near real-time reporting through dashboards, and risk assessment with mitigation recommendations.

Measured parameters were temperature, humidity, particulate matter (PM1, PM2.5, PM4, PM10 in µg/m³), carbon monoxide, CO₂ in ppm, VOCs, and viral pathogen detection through spectral signatures.

• Phase 1, telemetry collection establishing baseline metrics
• Phase 2, risk assessment analysing Viral Risk Index, Pathogen Risk Index and Vapour Pressure Deficit
• Phase 3, forensic analysis identifying contamination sources

Results: carbon dioxide

CO₂ ranged from 600 ppm to 1,600 ppm, with peaks concentrated between 1,200 and 1,400 ppm. The upper end of that range indicates periods where ventilation was not keeping pace with occupancy, relevant both to comfort and to the dilution of airborne infectious particles.

Results: particulate matter

As in the other facilities studied, coarse particle control performed well while the finest fraction, the fraction that reaches deepest into a child's lower airway, was the one showing excursions.

• PM1, high peaks detected
• PM2.5, elevated
• PM4 and PM10, within normal ranges

Results: volatile organic compounds

VOC concentrations ranged from 407 to 500 ppb, with peaks indicating potential off-gassing. In a clinical environment the likely contributors are disinfectant chemistry, consumables and furnishings, all of which are substitutable once identified.

Results: viral pathogen detection

RSV, COVID-19 and influenza were successfully identified through spectral analysis. Across the monitoring period the Viral Spread Indicator remained stable and the Pathogen Risk Indicator normal, meaning viral load stayed within prescribed thresholds rather than that pathogens were absent.

Environmental KPIs

• Viral Spread Indicator, stable
• Pathogen Risk Indicator, normal
• IAQ Index, variable, with concerning fine particles and elevated VOCs

Limitations

• Single-facility study
• Specific VOC compounds not identified individually
• No evaluation of mitigation impact within the study window
• Seasonal and occupancy pattern analysis not undertaken

Conclusions and recommendations

The study establishes that the technology monitors paediatric healthcare environments effectively, distinguishing infection-relevant signals from chemical exposure in the same data stream.

• Expand sensor deployment across the estate
• Implement HEPA filtration
• Move to low-VOC disinfectants
• Enhance ventilation in zones showing CO₂ excursions
• Integrate real-time dashboards
• Run follow-up studies evaluating mitigation effectiveness