Research report: Documented patient consent for ambient video monitoring technologies in an academic hospital: a cross-sectional analysis of consent outcomes and demographic differences

Of the patients approached, 184 actively refused or later revoked permission and 213 had consent recorded as missing or unobtainable. The result shows that camera- and sensor-based monitoring can be integrated into ordinary hospital admission workflows with high documented uptake, but it does not show that every agreement was informed, voluntary or durable. It also identifies modest differences associated with race and age that could become more consequential as ambient surveillance expands across healthcare.

The study was conducted at a tertiary academic medical center in the southeastern United States and included every patient admitted during the study period, with no exclusions. Ambient monitoring systems in patient rooms may include cameras and sensors capable of audio and video monitoring for patient safety, detection of environmental risks, movement monitoring and remote clinical support. The consent form stated that participation was optional, declining would not affect care, and permission could be revoked at any time.

It also disclosed that data could be used for clinical care, healthcare operations, education and technology development, including de-identified use by internal or external partners, and that collected data were not incorporated into the legal medical record. Operational integration was central to the findings. Consent was generally presented through an electronic or paper acknowledgment during registration in the emergency department or inpatient admission process, with status stored in the electronic health record.

Integration with the monitoring system automatically enabled devices when consent had been documented and disabled them when consent was refused, incomplete or revoked. Patients could revoke permission through direct communication with staff or a revocation form, which triggered both a record update and immediate deactivation. Staff received training, patients received standardized educational materials, and in-room indicators signaled when monitoring was active.

This configuration reduced reliance on manual checks and was designed to prevent activation without documented permission. Researchers divided consent records into three operational categories: received, refused and missing or unable. “Received” meant that consent had been documented.

“Refused” included an explicit refusal and later revocation, representing active nonacceptance. “Missing/unable” combined absent consent, inability to obtain consent and forms not received, representing passive or incomplete capture. The primary statistical analysis converted the outcome into documented acceptance versus refusal or missing status, then used multivariable logistic regression to account simultaneously for age, gender and race.

A secondary model treated age in four groups—18–44, 45–64, 65–74 and 75 years or older—and sensitivity analyses tested whether alternative treatment of revoked or incomplete records changed the results. Documented acceptance did not differ significantly by gender. It was 95.6% among 4,770 female patients and 96.1% among 4,752 male patients.

After adjustment for age and race, the odds ratio for acceptance among male patients relative to female patients was 1.11, with a 95% confidence interval of 0.91–1.36 and a p-value of 0.30, providing no evidence of an independent gender association. The 18–44 age group had an acceptance rate of exactly 95.0%, while acceptance was 95.8% among those aged 45–64, 96.7% among those aged 65–74 and 95.4% among those aged 75 or older. Continuous age was not associated with acceptance in the primary adjusted model, although the categorical analysis identified some age-related variation that was sensitive to how missing consent was classified.

The age pattern was more about how consent was not obtained than a broad rejection of monitoring. Patients aged 18–44 had the highest active-refusal rate, at 3.2%, whereas those aged 75 or older had the highest missing-or-unobtainable rate, at 3.3%. In the categorical model, patients aged 18–44 and those aged 75 or older each had lower adjusted odds of acceptance than patients aged 65–74, with adjusted odds ratios of 0.70 and corresponding 95% confidence intervals of 0.49–0.95 and 0.53–0.93.

When the analysis was restricted to patients who had made an explicit yes-or-no decision, however, older age was associated with greater odds of acceptance. The investigators could not determine whether missing records among older patients reflected illness, delirium, cognitive impairment, communication barriers, timing or the availability of a surrogate. Age itself, they emphasized, is not equivalent to decisional incapacity.

Race was associated with documented acceptance in the adjusted model. Acceptance was 96.1% among 7,847 White patients, 94.2% among 1,017 Black or African American patients, 98.4% among 245 Asian patients and 93.2% among 413 patients categorized as Unknown, Declined or Other. Compared with White patients, Black or African American patients had lower adjusted odds of acceptance, with an odds ratio of 0.66 and 95% confidence interval of 0.49–0.88, while the Unknown/Declined/Other group had an adjusted odds ratio of 0.56 and 95% confidence interval of 0.37–0.83.

The point estimate for Asian patients indicated higher odds, but its wide confidence interval of 0.90–6.59 included no difference, and the p-value was 0.079. The sensitivity analyses preserved the main findings, including lower adjusted odds among Black or African American patients. These race-related differences were modest in absolute terms and should not be read as evidence that race itself caused reluctance or that any group inherently rejects digital monitoring.

Race was an administrative social category, and the models did not account for socioeconomic conditions, education, health literacy, prior encounters with the health system or the quality of communication during consent. Possible explanations include differences in institutional trust, perceptions of respect, awareness of inequities in digital health, language, the person explaining the technology and whether privacy protections were credible to the patient. Those explanations remain hypotheses rather than demonstrated mechanisms because the study collected no patient interviews or comprehension measures.

Prospective qualitative research would be needed to determine why patients made the choices they did and whether implementation practices contributed to the observed pattern. The high aggregate rate is therefore best understood as an implementation finding, not a verdict on the ethics or effectiveness of ambient monitoring. Documented acceptance may reflect a clear clinical rationale, confidence in human and institutional oversight, trust in the hospital, social desirability, routine-form completion, limited attention during acute illness or concern that refusal could affect care despite written assurances that it would not.

Consent was also authorized for future hospitalizations unless revoked. Although few revocations were recorded, low revocation cannot establish durable endorsement: patients may not know that permission remains active, may have difficulty locating the revocation pathway or may receive little follow-up. The study did not evaluate whether the technology improved outcomes, caused privacy events or produced a better patient experience, nor did it compare the consent process with any alternative form or workflow.

The analysis also has important boundaries. It came from one hospital with a predominantly White patient population, so results may differ across regions, clinical units, hospital types and technologies. Race categories were consolidated, potentially obscuring heterogeneity within smaller groups, and the dataset lacked information on capacity, delirium, illness severity and surrogate involvement.

The retrospective cross-sectional design describes associations at one point in the consent process but cannot establish causality or track changes over time. The findings support the operational feasibility of documenting and enforcing consent, while leaving unresolved whether patients understood the systems, felt free to decline, valued their benefits, experienced their burdens and would maintain the same preference under changed circumstances. Future evaluations should measure comprehension and motivation directly, track revocation and patient experience prospectively, and determine whether consent, trust and the distribution of benefits and burdens remain equitable across populations.

Subject of Research: Patient consent outcomes for ambient video monitoring in an academic hospital

Article Title: Documented patient consent for ambient video monitoring technologies in an academic hospital: a cross-sectional analysis of consent outcomes and demographic differences

Article References: Godsey, E. L., Galloway, T. M., Crews, J. S., & Helgeson, S. A. (2026). Documented patient consent for ambient video monitoring technologies in an academic hospital: a cross-sectional analysis of consent outcomes and demographic differences. Frontiers in Digital Health, 8, Article 1969218. https://doi.org/10.3389/fdgth.2026.1969218

Image Credits: AI Generated

DOI: 10.3389/fdgth.2026.1969218

Keywords: ambient video monitoring, patient consent, hospital technology, digital health, health equity, ambient intelligence, privacy, artificial intelligence, cross-sectional study, demographic differences, Documented, patient