YOOMI: Effect of AI-Guided Gamified Physical Therapy Exercise Software on Inpatient Mobility.

Demonstration of YOOMI Software in Use. Thirty percent of hospitalized adults over age 70 frequently develop new impairments in Activities of Daily Living (ADLs) during their hospitalization. This often leads to long-term dependence and the need for post-acute care. Bed rest interferes with the function of major body organs, with adults losing more than 1% of lean muscle mass per day of hospitalization and up to 12% per week. While hospital mobility protocols help reduce atrophy and improve mobilization, such as initiatives to get patients to sit in a chair or ambulate patients with assistance, staffing limitations, time constraints, and safety concerns often hinder compliance. Patient rooms can be equipped with TV monitors featuring exercise programs, but the exercises cannot be modified to match the patient's individual ability in difficulty, lack engagement via gamification, and lack monitoring. We conducted, to our knowledge, the first exploratory pilot trial to evaluate the feasibility, patient satisfaction, and impact on outcomes of YOOMI, a bedside, staff-independent, AI-driven, interactive exercise software device. YOOMI is a compact, wheeled system with a large screen and patient-facing camera that delivers real-time feedback during gamified, in-bed physical therapy sessions. Patients perform guided, full-body exercises designed around everyday activities and recovery milestones. Configurable by staff with a single click, YOOMI provides an autonomous, engaging, and accessible approach to promoting mobility and rehabilitation during hospitalization (see Figure 1). We conducted a prospective, randomized-controlled pilot trial on patients admitted to the medicine hospitalist services at RWJUH-NB and enrolled in the Hospitalized Elder Life Program (HELP) program. Participants were recruited and randomized into two groups: routine care (control) or routine care plus YOOMI. Inclusion criteria: age 65 or older, English fluency, and capacity to consent. Exclusion criteria: unstable psychiatric illness, critical illness requiring ICU-level care, postsurgical devices/drains, or surgery during the hospitalization. Outcomes included: Activity Measure for Post-Acute Care (AMPAC) scores (performed by nurses) which indicate physical activity level during hospitalization, length of stay, and discharge disposition. AMPAC scores are measured on a scale from 6 to 24, with scores ≥ 17 reflecting more functional independence, supporting discharge home. Statistical analysis was conducted using SAS version 9.4 (SAS Institute Inc., Cary, NC). Descriptive statistics were obtained through stratified univariate analyses. For continuous variables, including change in AMPAC score and length of stay, generalized linear models (GLM) were applied to evaluate group differences and statistical significance. For the binary outcome of discharge disposition, differences between groups were assessed using logistic regression. All tests were two-sided, and a significance level of α = 0.05 was used to determine statistical significance. The study was approved by Rutgers' IRB. Thirty patients were enrolled with randomization stratified by gender. Five intervention patients and one control patient were discharged before any participation or data collection and were excluded, resulting in 10 intervention patients and 14 controls in the analysis. In the final cohort (n = 24), mean age (age: 77.1 ± 6.6 intervention vs. 79.8 years ±8.6 control), gender (male: 50.0% intervention vs. 57.1% control), and racial composition (White: 80.0% intervention vs. 85.7% control), Charlson Comorbidity Index scores (5.5 ± 1.3 intervention vs. 5.07 ± 1.9 control), and baseline AMPAC scores (18.3 ± 5.3 intervention vs. 17.5 ± 4.6 control) were similar between groups. Patients completing the YOOMI activity participated for an average of 33.5 ± 11.0 min, completing 112 ± 79.1 exercise repetitions over 1–2 sessions. Mean change in AMPAC scores were +0.6 ± 1.3 in the intervention group and −0.07 ± 1.4 in the control group (GLM regression, p value 0.25). Discharge home occurred in 90% of intervention patients versus 57% of controls (RR = 4.3, 95% CI = 0.6–30.3, p = 0.09) with non-home discharges to subacute rehabilitation or nursing facilities. Average length of hospital stay was 6.3 ± 3.1 days in the control group, compared to 4.1 ± 2.02 days for intervention (GLM regression, p = 0.07), see Table 1. Nine of 10 patients in the intervention arm completed satisfaction surveys (90% response rate). Eighty-nine percent of intervention patients reported enjoying their workout and believed they physically benefited from the activity. One-hundred percent of survey respondents said they would do the activity again while inpatient, and 50% said they would do the activity at home [1-7]. This exploratory pilot trial showed that patients using the YOOMI device had trends of increased AMPAC scores, lower length of stay, and a higher rate of discharge to home compared to the control group, though these differences were not statistically significant. Patient satisfaction was very high, with most participants enjoying the activity and perceiving improved mobility, showing that the geriatric population will engage with software solutions. Limitations include generalizability due to a small sample size and single-center design. This study supports the feasibility and potential value of AI-guided gamified therapy in mitigating functional decline. Larger trials are warranted. Jay Naik and Ibiyonu Lawrence conceived the study. Jay Naik, Ibiyonu Lawrence, Dhyana Veliez, and Karthik Kota helped co-develop and customize the software with the YOOMI team. Daniel Schaer performed statistical analysis. Michael B. Steinberg helped design the study. All authors helped interpret results, revised the manuscript for important intellectual content, approved the final version, and agreed to be accountable for the work. Special thanks to Ben Catania and Joseph Whelan from YOOMI device and software development, and Dr. Jeffrey Carson MD for his guidance and support. The authors have nothing to report. There was no sponsor or funding provided for the study. YOOMI team provided the YOOMI equipment and software but had no role in the design, methods, subject recruitment, data collection, analysis, or preparation of the manuscript. The authors declare no conflicts of interest. Video S1: YOOMI Software to improve inpatient mobility demo video. Everyone seen in the video is an employee, faculty member, or volunteer at the medical school. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.