AI-Powered Remote Monitoring for Lower Extremity Wound Management: A Randomized Controlled Trial Protocol

Abstract

Background:

Lower extremity wounds associated with diabetes are a serious global health issue, with diabetic foot ulcers affecting 12% to 25% of adults with diabetes and accounting for 80-90% of all lower extremity amputations in the United States. Comprehensive in-person care for lower extremity wounds is important but can be burdensome for patients and costly for healthcare systems. A cost-effective telehealth model using a smartphone-integrated digital application that remotely analyzes wound status with machine-learning algorithms in real-time could make lower extremity wounds care more accessible to patients. This trial aims to determine if an artificial intelligence (AI)-powered digital remote monitoring is a feasible, patient-centered solution for remote wound monitoring and management compared to standard in-person visits.

Methods:

We will conduct a non-blinded randomized control trial of 120 patients with active lower extremity wounds treated in the Johns Hopkins Hospital Multidisciplinary Diabetic Foot and Wound Clinic in Baltimore, Maryland (ClinicalTrials.gov: NCT05579743). Participants will be randomly assigned 1:1 to receive wound care monitoring using AI-powered remote wound monitoring technology (Healthy.io Ltd.) or standard in-person monitoring for 12 weeks. The primary aim is to establish the feasibility of a novel remote patient-centered monitoring program for the surveillance and monitoring of lower extremity wounds. Secondary aims include evaluating patient and provider satisfaction with remote wound monitoring technology compared to standard in-person monitoring; and generating pilot data on wound healing time and major amputation rates in patients who are monitored remotely compared to patients treated with standard of care.

Conclusion:

This trial will determine whether AI-powered remote digital monitoring is feasible and acceptable as an alternative to standard in-person monitoring for the monitoring and management of patients with active lower extremity wounds.

INTRODUCTION

Lower extremity wounds associated with peripheral arterial disease and diabetes are a global epidemic, with diabetic foot ulcers affecting more than 4 million adults in the US¹. More than 80% of all lower extremity amputations are the result of lower extremity wounds, but the risk of amputation varies according to race and socioeconomic status. Frequent evaluation via regular short-interval clinic visits by a multidisciplinary team has been shown to improve lower extremity wound healing outcomes. However, low socioeconomic status and vulnerable populations tend to struggle with frequent in-person healthcare visits due to a combination of financial and nonfinancial barriers to care, including lack of transportation. As a result, there are substantial socioeconomic and racial disparities in the treatment and outcomes of lower extremity wounds.

We have established a multidisciplinary Diabetic Foot and Wound Clinic that provides cost-free, regular comprehensive foot care for adult lower extremity wounds. One of our goals is to mitigate disparities in major amputation. Our multidisciplinary team treats approximately 200 patients per year, of which 60% are Black adults and 40% are classified as socioeconomically disadvantaged. We have previously shown excellent wound healing outcomes using this model, including reduced wound healing times and lower major amputation rates compared to standard national averages ^2-5. However, the Diabetic Foot and Wound Clinic requires bi-weekly clinic visits to facilitate wound assessment and care plans, which is challenging for many patients due to structural barriers including difficulty with transportation, time off work, and caregiver availability.

We plan to leverage our multidisciplinary Diabetic Foot and Wound Clinic to assess the feasibility of using a novel artificial intelligence (AI)-powered remote patient-centered wound monitoring program in high-risk adults presenting with an active lower extremity wound. In a pilot single-arm feasibility trial, we previously demonstrated that this approach was a feasible means of remote wound monitoring for use by patients and/or their caregivers ⁶. The proposed randomized controlled trial (RCT) is designed to evaluate the feasibility, risks, benefits, and participant satisfaction associated with AI-powered remote monitoring through smartphone digital applications compared to standard in-person monitoring for active lower extremity wound.

METHODS

Study Design:

This will be a pilot non-blinded RCT feasibility trial. We will enroll 120 patients with an active lower extremity wound who present to our multidisciplinary clinic. Patients will be computer randomized 1:1 to receive wound care monitoring using remote monitoring technology or standard in-person monitoring for 12 weeks (Figure 1).

Figure 1:

Study flow of the clinical trial for patients randomized to the remote monitoring program (top row) or standard in-person monitoring program (bottom row).

Study Population:

All the patients referred to the Johns Hopkins Hospital Multidisciplinary Diabetic Foot and Wound Clinic will be screened for inclusion. Patients will be eligible for enrollment if they are >18 years of age and present with an active lower extremity wound requiring a dressing change at least once per week. Patients who do not read or speak English, patients without an open wound, and those not willing to learn and use the remote monitoring app will be excluded.

Informed Consent:

Patients will be screened for study inclusion in the Johns Hopkins Hospital inpatient and outpatient settings based on a recent diagnosis of an active lower extremity wound. Patients will subsequently be enrolled during a regular outpatient clinic visit. Patients will give a written consent to participate in the study and may opt out of participation at any point. If a patient wants to enroll in the study but has concerns about their ability to capture wound images due to wound location, they may participate if they have a regular caregiver who is willing to help them with image capture. Patients of all sexes and races will be approached for enrollment. The mean age of patients treated in the Johns Hopkins Diabetic Foot and Wound Clinic is 64.6 years, and 58.8% are Black race (self-reported) ⁴.

Ethical Considerations:

This study has been reviewed and approved by the Johns Hopkins School of Medicine Institutional Review Board and will be conducted in full conformance with principles of the Declaration of Helsinki. It has been registered at ClinicalTrials.gov (NCT05579743). Patients randomized to the standard in-person monitoring arm will be treated according to the standard of care as previously described for the Johns Hopkins Diabetic Foot and Wound Clinic ². Patients randomized to the remoted wound monitoring program will have planned in-person visits at half the frequency as usually required. The reduction in in-person visits will be offset by the ability to review the wound virtually via the digital monitoring application. However, if there are issues with obtaining images, or provider concern that a wound cannot be properly assessed via the app, the patient will be asked to come in for an unscheduled in-person visit to ensure appropriate evaluation and treatment.

Patient characteristics:

Patient wound size, depth, area, and Wound, Ischemia, Foot Infection (WIfI) classification⁷ and point-of-care hemoglobin A1c will be measured at baseline and the 12-week follow-up visit. Patient age, sex, race/ethnicity, education level, and comorbidities will be collected based on self-report and patient medical records. Area deprivation index, a measure of socioeconomic disadvantage, will be calculated for each patient based on their 9-digit zip code using established methods ⁸.

Intervention:

Patients randomized to the intervention arm will be trained on using the AI-powered remote patient-centered lower extremity wound monitoring program via a smartphone application (Minuteful for Wound, Healthy.io Ltd.). This an FDA-registered technology that uses a smartphone camera to create a visual record of wounds over time. Patients are provided with proprietary calibration stickers to normalize the lighting and color array, which are placed approximately 2 cm from the wound on healthy skin to enable the technology to automatically measure and analyze wounds (Figure 2). All images and data are automatically uploaded and securely stored in a dedicated portal where the care team can monitor the wound progression over time. In addition to calculating wound dimensions and surface area, the technology automatically analyzes and calculates its interior makeup using advanced machine learning algorithms that have been developed based on real wound images evaluated by wound care experts. The result is a reliable, accurate tool that creates a standardized digital record of wound progression over time (Figure 3).

Figure 2:

The Minuteful for Wound (Healthy.io Ltd) AI-powered remote monitoring wound application allows patients to easily capture wound images. Patients are provided with proprietary calibration stickers that allow the technology to automatically measure and document wounds (left panel). The images are automatically analyzed using a machine-learning approach via proprietary software (right panel) and uploaded to a secure cloud-based portal for physician review.

Figure 3:

The remote wound monitoring program automatically analyzes the characteristics of the wound using advanced machine learning algorithms to create a standardized digital record of wound progression over time.

Patients randomized to receive remote wound care monitoring will be given in-person training on how to use the smartphone app and instructed to capture wound images at least once per week during the time of a regular dressing change. Patients will download the app on their smartphone and will be given a box of calibration stickers. Wound assessments will be electronically transmitted to a dedicated portal for physician review, which will occur on a weekly basis. Changes to wound management will be made by a provider based on the wound images. Patients enrolled in the remote wound monitoring technology arm will have scheduled in-person follow-up at enrollment, week 4, week 8, and week 12.

Standard of care:

Patients randomized to receive standard of care will be provided with a wound care plan at the time of enrollment, and then follow-up in clinic on a biweekly basis (week 2, 4, 6, 8, 10, 12) for a wound check and care plan update as needed.

Objectives:

The primary objective of this clinical trial is to determine the feasibility of participants performing remote wound scan at least once per week. The trial is designed with the following hypothesis:

H₀: Patients with a lower extremity wound will successfully capture and submit wound images using the novel technology at least once per week. Providers will access those images to monitor wound progression and adjust treatment recommendations remotely.

The secondary objectives of this clinical trial are as follows:

• To assess patient and provider satisfaction with remote wound monitoring technology compared to standard in-person monitoring

• To generate data on wound healing time and major amputation risk in patients with lower extremity wounds who are monitored remotely compared to patients treated with standard of care

• To assess app useability, including violation of boundary conditions and image scan failures

Sample size:

The mean time to complete wound healing in our multidisciplinary diabetic foot clinic 136.3 ± 7.9 days. At 12 weeks following initial evaluation, 61.6% ± 5.5% lower extremity wounds achieve complete healing⁹. Assuming alpha = 0.05 and ß = 0.80, we would need to enroll 100 patients (50 per group) to detect a 5% difference in wound size reduction at 12 weeks with the remote monitoring vs. standard of care surveillance program. Accounting for an estimated 20% dropout rate ¹⁰, we plan to enroll 120 patients in the study (60 per group). We will not have adequate power to detect differences in major amputation, but we will collect these data.

Data management:

All patient and wound images and data will be uploaded and securely stored in a dedicated portal. The portal infrastructure is hosted and operated over secured platforms while implementing restricted access as part of a defense-in-depth approach, and complies with FDA 21 CFR Part 820, EN ISO 13486:2016, and the EU Medical Device Directive 93/42/EEC. All data is encrypted at-rest and in-transit using strong encryption protocols (at rest, AES256; in-transit, HTTPS TLS1.2). The portal is HIPAA compliant, ISO 27001 certified by an external party, and GDPR compliant. Patient data will be collected in secure, cloud-based REDCap software. The data will be stripped of all personal identifiers and all investigators using these data will sign data use agreements that specifically outline the parameters for data use and the protection of the confidentiality and privacy of participants.

Adverse Events:

Throughout the study, the providers and study team principal investigator will review all Common Terminology Criteria for Adverse Events grade 3 or higher adverse events and report them to the Institutional Review Board. Once reported, an adverse event will be followed until it resolves with or without sequelae, or until the end of study participation, or until 30 days after the participant prematurely withdraws (without withdrawing consent) or is withdrawn from the study, whichever occurs first. A Data and Safety Monitoring Board will convene to review adverse events and study data every 6 months according to a pre-established charter. The study principal investigator, the sponsor, the Johns Hopkins Institutional Review Board, and the Data and Safety Monitoring Board will have the authority to stop or suspend the study and require modifications at any time.

Statistical Methods:

Main analyses will be conducted according to intention to treat principles. Patient engagement and app usability will be reported using descriptive statistics for patients in the remote monitoring arm. Patient engagement >80% will be considered a success. Patient and provider satisfaction will be reported and compared for the remote monitoring vs. in-person wound monitoring groups. Wound size reduction, wound healing time, and major amputation outcomes will be compared between groups using standard survival analysis methods including Kaplan-Meier curves, logistic regression, linear regression, and Cox proportional hazards models. Exploratory subgroup analyses by age, sex, area deprivation index, and race/ethnicity will be conducted. We will also report descriptive statistics for app performance, including the frequency of violations of boundary conditions and image upload failures, for the remote wound monitoring group.

DISCUSSION

Lower extremity wounds associated with peripheral arterial disease and diabetes are a highly prevalent and morbid pathology. Adults with a lower extremity wounds associated with diabetes have 56-times greater risk of hospitalization and a 155-greater risk of amputation compared to adults with diabetes but no diabetic foot wound ¹¹. Similarly, adults with lower extremity wounds related to peripheral artery disease have 1-year amputation and mortality rates as high as 30% and 25–35% mortality, respectively.^12-15. Wound healing is delayed for larger ulcer diameters and ulcers that have an associated infection¹⁶. Frequent evaluation via regular short-interval clinic visits by a multidisciplinary team has been shown to improve lower extremity wound healing outcomes ^{2, 5, 9}.

The Johns Hopkins Hospital Multidisciplinary Diabetic Foot and Wound Clinic, where the trial will be performed, offers an integrated service with a multidisciplinary team of experts⁵. Patients with active lower extremity wounds are typically evaluated in clinic by a multidisciplinary team on a biweekly basis to assess wound healing progress, refine wound care algorithms, and identify and treat infection preemptively. However, this model is expensive and requires a substantial time commitment from both the patients and providers⁹. The need for frequent clinic visits requires repeated days off employment, transportation assistance, and costs associated with travel time, distance, and parking. Nonfinancial barriers to care, including lack of transportation, are common reasons for unmet need or delayed care among US adults and frequently coincide with affordability barriers ¹⁷. Studies have shown that lack of transportation are associated with delaying or skipping medication, rescheduling or missing healthcare appointments, and postponing care altogether ^{18, 19}.

Remote delivery of healthcare has the potential to reduce disparities and burdens in the treatment of patients with active lower extremity wounds, and preliminary patient perceptions of telehealth for the management of diabetes complications are positive^{6, 20}. A recent systematic review reported that remotely delivered healthcare for the management of diabetes-related foot disease is well received by patients and clinicians ²¹. However, the majority of studies evaluating the utility of telehealth programs for lower extremity wounds are single-arm, observational studies that have focused on ulcer prevention rather than ulcer treatment. “Smart” foot insoles that monitor for uneven plantar pressure distribution have been shown to lower the incidence of diabetic foot wounds by 71% ²². Similarly, home foot temperature monitoring has been shown to be predictive of diabetic foot ulcer formation, leading to preemptive interventions such as early podiatric referral and offloading that may reduce ulceration risk ²³. One recent pilot study enrolled 31 patients in a program where wound photos were taken and emailed to a generic email address for review by 2 podiatrists ²⁴. That intervention was met with high patient satisfaction, although images were not of adequate quality to inform real-time treatment decisions. The use of a remote monitoring imaging technology that can provide high-quality images and standardize active lower extremity wound characteristics in a quantitative manner has not been previously described.

The Minuteful for Wound (Healthy.io Ltd) wound monitoring program is an FDA-registered technology that uses a smartphone camera to create a visual record of wounds over time ⁶. This technology has been commercially adopted by several integrated health care systems in the US, UK, and by Israel’s largest Health Maintenance Organization, Maccabi. Healthy.io has participated in several preliminary studies to validate the accuracy of the Healthy.io mobile app semi-automatic area measurements, but its use by patients for regular lower extremity wound surveillance has not been previously explored. The proposed trial leverages patient participation in a high-volume multidisciplinary Diabetic Foot and Wound Clinic to assess the feasibility of using an AI-powered novel remote patient-centered wound monitoring program in high-risk adults presenting with an active lower extremity wound.

The purpose of this study is to replace the need to perform frequent in-person foot examinations to monitor wound healing progress with a standardized at-home approach that will not compromise wound healing outcomes. The increasing utilization of wound photography as a means of tracking healing progress can be difficult to perform in a standardized manner. Clinical photography is highly variable in quality and can lead to inefficiencies due to prolonged upload time to the electronic medical record ²⁵. In addition, the wound information that is typically assessed by a physician at an in-person visit, including a combination of objective data (size, depth) and qualitative data (healing potential, tissue quality), is not available through standard photography. The proposed remote wound-healing program leverages AI technology together with a patient-friendly smartphone application and secure cloud-based data storage in an effort to overcome these barriers.

There are some limitations to the proposed study. It is designed as a single-center RCT, which may restrict the scalability of the remote monitoring technology and limit its accessibility to a broader population beyond the community served by the clinic. In addition, remote monitoring may pose potential risks, including delays in recognizing clinical deterioration for patients who are not receiving regular in-person assessment. To the best of our knowledge, there are currently no established guidelines specifically addressing the use of remote monitoring for lower extremity wounds. Maintaining active communication with participants to ensure they are submitting images for review and ensuring timely review of submitted images by the clinical team will be essential to help mitigate these risks. Enrollment will require access to a smartphone. Patients who are willing to participate but do not have access to a smartphone will be provided with a smartphone with the wound monitoring app pre-installed for the duration of the study. Major amputation rates are low in our clinic, occurring in approximately 10% of patients by one year ⁵. The study is powered to focus on wound size reduction during the study period. The amputation data will be considered exploratory only as we do not anticipate having sufficient power to detect differences in amputation between groups. We will also not be powered to make formal comparisons about digital application performance based on wound etiology, location, or skin tone.

CONCLUSIONS

The planned RCT aims to test the feasibility risks, benefits, and participant satisfaction associated with AI-powered remote monitoring through smartphone digital applications compared to standard in-person monitoring for active lower extremity wounds. The study results will be used to generate preliminary data to support a future larger grant submission that will expand the scope of this work, enabling a greater number of patients to be enrolled and followed for a longer period of time for a broader array of clinical outcomes including amputation. If the remote monitoring program is feasible and positively perceived by patients and physicians, this would justify a larger initiative with extended follow-up to more thoroughly evaluate the wound healing and major amputation outcomes. If the remote monitoring program is deemed not feasible or is negatively perceived by patients or physicians, this would provide us with an opportunity to explore the barriers to adoption and potentially redesign the technology or its utilization toward better ease of use.