Computer Vision App Improves Hospital Cleaning Consistency Across Multiple Campuses

A prospective interventional study conducted across a four-campus tertiary hospital system in Southwest China has demonstrated that a computer vision-assisted smartphone application can significantly improve the thoroughness and consistency of environmental surface cleaning. Operating within a technology-empowered Plan-Do-Check-Act quality framework, the digital monitoring approach replaced conventional subjective inspections with automated image evaluation, elevating overall cleaning performance and reducing variation between separate hospital facilities over a 17-week intervention period.

Maintaining uniform environmental cleaning standards is widely recognized as essential for preventing healthcare-associated infections, yet institutional consistency often proves challenging to sustain across geographically dispersed medical campuses. Traditional surveillance approaches, including routine manual visual inspections and conventional fluorescent gel marking, rely heavily on individual human observation.

To establish an objective monitoring benchmark, researchers implemented a mobile application that automates the evaluation of fluorescent markers placed on high-touch environmental surfaces. The digital platform analyzed ultraviolet fluorescence images captured immediately before and after routine cleaning cycles, applying an automated algorithm to calculate the fluorescent marker removal rate. This automated image analysis generated standardized quantitative performance scores and provided real-time operational feedback directly to staff, eliminating reliance on human visual estimates.

The technical deployment was embedded directly into a technology-empowered Plan-Do-Check-Act model, known as T-PDCA, designed to drive iterative operational improvement across hospital sites. The protocol incorporated standardized training for environmental services teams, blinded placement of fluorescent markers, algorithm-driven performance grading against predefined removal thresholds, and continuous data-guided feedback. The 17-week trial took place from June through October 2025 across all four campuses, capturing data through routine shifts.

Over the course of the intervention, investigators collected and evaluated 2,457 valid fluorescent marker samples across the participating medical centers. In the first week of implementation, the overall weekly mean fluorescent marker removal rate stood at 83.91 percent. Following the introduction of algorithm-derived feedback and structured reviews, overall removal efficiency rose steadily, crossing and remaining above the 90 percent mark from approximately the fourth week onward. Nonparametric statistical analysis confirmed a significant upward trajectory throughout the evaluation period, recording a Spearman rank correlation coefficient of 0.615 with a significance value of P = 0.01.

In addition to raising overall cleaning scores, the digital intervention substantially mitigated performance discrepancies between the four hospital campuses. The weekly standard deviation of marker removal rates decreased from an initial 29.72 percent down to 9.59 percent by the end of the trial, reflecting a pronounced increase in cleaning homogeneity. Campuses with lower baseline performance exhibited substantial and rapid gains, whereas campuses that entered the study with high baseline cleaning adherence maintained consistent, stable performance levels over time.

Despite broad system-wide improvements, the researchers identified significant disparities in cleaning performance when stratifying data by clinical unit type and physical surface characteristics. Specifically, intensive care units and outpatient departments demonstrated lower overall marker removal rates and higher outcome variability compared to other clinical spaces. Material composition also influenced results, with surfaces manufactured from matte plastic or natural wood exhibiting lower cleaning efficiency and greater performance fluctuations than other environmental substrates.

The study authors concluded that embedding a computer vision-assisted mobile monitoring tool within a structured T-PDCA quality improvement model provides an effective, objective mechanism to enhance environmental surface cleaning and promote operational consistency across multi-campus healthcare systems.

Subject of Research: Computer vision mobile applications for hospital environmental cleaning monitoring and quality improvement

Article Title: Improving environmental cleaning quality and consistency in a multi-campus hospital using a computer vision-assisted mobile application: a prospective interventional study

Article References: Wang, P., Luo, J., Chen, L., Xiang, Z., Wang, Y., He, Y., Sun, Q., Chen, J., & Xiang, Q. (2026). Improving environmental cleaning quality and consistency in a multi-campus hospital using a computer vision-assisted mobile application: a prospective interventional study. Antimicrobial Resistance & Infection Control, 15(1), Article 139. https://doi.org/10.1186/s13756-026-01803-x

Image Credits: AI Generated

DOI: 10.1186/s13756-026-01803-x

Keywords: hospital environmental cleaning, computer vision, mobile health applications, infection control, healthcare-associated infections, fluorescent marker removal rate, T-PDCA framework, quality improvement, multi-campus hospitals, ultraviolet fluorescence imaging, real-time feedback, surface disinfection, intensive care units, outpatient departments, surface material variations, prospective interventional study