AVAL: A Validation Pipeline for AVA Mobility
ForesightCares, Inc.Charlotte, NCPublished
13 pages · 12 figures ·4 tables · PDF 2.1 MB
The question · Report 1 of 3 in the AVA Evidence Program
“Is AVA accurate in real community settings?”
Key findings
6.0%
TUG timing error (MAPE)
Mean absolute percentage error of AVA’s Timed Up and Go times relative to the mean of two physical-therapist timings, across 62 assessments.
Source: Abstract; Table 2, p. 8
0.71 s
Mean absolute TUG error
Average size of AVA’s TUG timing error against the PT reference; by comparison, the MR4 overestimated TUG time by 1.59 s on average.
Source: Abstract; Table 2, p. 8
98.5%
Chair-stand counts matched exactly
AVA’s count matched the PT count for 64 of 65 participants, and all 65 were within one repetition.
Source: Section 7.2, p. 10
88.1%
TUG mobility-category agreement
Share of 59 participants AVA placed in the same low, fair, or high category as the PTs, versus 71.2% for the MR4 on the same participants.
Source: Abstract; Table 3, p. 8
Abstract
Mobility tests such as the 30-second chair stand and the Timed Up and Go (TUG) are widely used to identify older adults at elevated risk of falling. Both tests are simple, but they depend on a trained observer to count repetitions or run a stopwatch, which limits how often and how widely they can be given. AVA Mobility is a vision-based assessment system that automates these tests using a single iPad camera. It performs pose estimation and movement-event detection entirely on the device, with no wearable sensors, external hardware, or cloud processing. AVA performs well in controlled testing, but its accuracy in the community settings where it is meant to be used had not been formally measured.
This report introduces AVAL, a validation pipeline that compares AVA’s results against measurements made by physical therapists (PTs) and against a commercial wearable motion-capture system (Noraxon MR4). Data were collected from 66 older adults across four community sites between June and October 2025. AVAL links each participant’s video to their reference measurements, runs AVA on the video, reconstructs the test outcome, and computes error metrics. TUG timing is evaluated with MAPE-based accuracy, mean absolute error, root mean squared error, and 95th-percentile error. Chair-stand counts are evaluated by the share of predictions that match the clinical count exactly or within one repetition. We also compare the low, fair, and high mobility categories each system assigns using CDC STEADI thresholds.
On the TUG, AVA’s timings had a mean absolute percentage error (MAPE) of 6.0% relative to PT timing, with a mean absolute error of 0.71 s; the MR4 overestimated TUG time by 1.59 s on average. On the chair stand, AVA matched the clinical count exactly for all but one participant, and every prediction was within one repetition. AVA also agreed more closely than the MR4 with PT-assigned mobility categories on both tests (TUG: 88.1% vs. 71.2%; chair stand: 100% vs. 87.5%, on the same participants). These results indicate that AVA can deliver accurate, scalable, and privacy-preserving mobility assessment in real-world community settings.
Keywords: fall risk · mobility assessment · pose estimation · Timed Up and Go · 30-second chair stand · STEADI · validation
Why it matters
The Timed Up and Go and the 30-second chair stand are short, widely used tests for identifying older adults at elevated risk of falling. Both normally need a trained observer to count repetitions or run a stopwatch. This report tested whether AVA Mobility, which runs these tests with a single iPad camera, matches physical therapists’ measurements in everyday community spaces rather than a lab. Data came from 66 older adults at four community sites, and the same tests were also recorded with a commercial wearable motion-capture system (the Noraxon MR4) for comparison.
AVA’s Timed Up and Go times differed from the therapists’ times by 0.71 seconds on average, with errors centered close to zero, and the pattern was similar at all four sites. On the chair stand, AVA’s count matched the therapists’ count for 64 of 65 participants, and the one remaining count was off by a single repetition. When results were sorted into low, fair, and high mobility categories, AVA agreed with the therapists for 88.1% of Timed Up and Go results and 100% of chair-stand results, compared with 71.2% and 87.5% for the motion-capture system, which tended to report longer times.
Clinicians usually act on the mobility category rather than the raw time or count, so category agreement is the result most relevant to screening decisions. The motion-capture system needs sixteen body-worn sensors, calibration, and a trained operator, roughly 30 to 40 minutes of setup per participant; AVA ran the same assessments on one iPad in under five minutes, with processing done on the device. The study measured agreement with therapists, not fall outcomes, and its sample was moderate in size and drawn from one region.
Explore the results
Interactive views of numbers published in the report. Hover or focus a mark for details.
Agreement with physical-therapist mobility categories
Percentage of participants AVA and the MR4 placed in the same low, fair, or high mobility category as the physical therapists, on the same participants. Chair-stand categories follow CDC STEADI normative values; TUG categories (≤11 s high, ≥15 s low) were defined for this study.
- AVA
- Noraxon MR4
View data tableHide data table
| Assessment | AVA | Noraxon MR4 | Notes |
|---|---|---|---|
| Timed Up and Go | 88.1% | 71.2% | n = 59; AVA 52/59, MR4 42/59; Cohen’s κ 0.79 vs. 0.55 |
| 30-second chair stand | 100% | 87.5% | n = 56; AVA 56/56, MR4 49/56; Cohen’s κ 1.00 vs. 0.78 |
Selected figures
From the report. Select a figure to enlarge it, or jump to its page in the full report.
Full report
The PDF is the version of record. Search, follow internal links, or jump to any page.
Limitations
- Very dim or very bright lighting, people or objects passing between the camera and the participant, incorrect iPad positioning, and clothing that blends into the background can affect measurement quality.
- Some recordings were excluded because of calibration failures or missing data in the reference measurements.
- The two PT observers’ start and stop cues and their blinding to AVA and MR4 outputs were not documented, and AVA and the PTs applied different rules to the final chair-stand repetition.
- The sample of 66 participants is moderate in size, weighted toward women (48 of 66), and drawn from four sites in one region.
- The 4-stage balance test was recorded but not analyzed.
Ethics, funding & disclosures
The study was approved by the Advarra Institutional Review Board (protocol Pro00080873) and conducted in accordance with the Declaration of Helsinki. All participants gave informed consent before taking part, including consent to video recording and to wearing the MR4 sensors.
This report was prepared by ForesightCares, Inc., which develops AVA.
Responsible use
AVA supports mobility awareness and screening. It does not diagnose conditions, replace clinical evaluation, or serve as a standalone clinical decision tool. Results describe the study population and settings reported here.
Cite this report
ForesightCares, Inc. (2026). AVAL: A Validation Pipeline for AVA Mobility (Technical Report No. FC-TR-2026-01). https://insights.foresightcares.com/papers/aval-v1/
@techreport{foresightcares2026aval,
title = {AVAL: A Validation Pipeline for AVA Mobility},
author = {{ForesightCares, Inc.}},
institution = {ForesightCares, Inc.},
address = {Charlotte, NC},
type = {Technical Report},
number = {FC-TR-2026-01},
year = {2026},
month = oct,
url = {https://insights.foresightcares.com/papers/aval-v1/},
keywords = {fall risk, mobility assessment, pose estimation, Timed Up and Go, 30-second chair stand, STEADI, validation}
}
TY - RPRT AU - ForesightCares, Inc. TI - AVAL: A Validation Pipeline for AVA Mobility PY - 2026 DA - 2026/10/16 PB - ForesightCares, Inc. CY - Charlotte, NC M1 - FC-TR-2026-01 M3 - Technical Report UR - https://insights.foresightcares.com/papers/aval-v1/ L1 - https://insights.foresightcares.com/papers/aval-v1/aval-v1.pdf AB - Mobility tests such as the 30-second chair stand and the Timed Up and Go (TUG) are widely used to identify older adults at elevated risk of falling. Both tests are simple, but they depend on a trained observer to count repetitions or run a stopwatch, which limits how often and how widely they can be given. AVA Mobility is a vision-based assessment system that automates these tests using a single iPad camera. It performs pose estimation and movement-event detection entirely on the device, with no wearable sensors, external hardware, or cloud processing. AVA performs well in controlled testing, but its accuracy in the community settings where it is meant to be used had not been formally measured. This report introduces AVAL, a validation pipeline that compares AVA’s results against measurements made by physical therapists (PTs) and against a commercial wearable motion-capture system (Noraxon MR4). Data were collected from 66 older adults across four community sites between June and October 2025. AVAL links each participant’s video to their reference measurements, runs AVA on the video, reconstructs the test outcome, and computes error metrics. TUG timing is evaluated with MAPE-based accuracy, mean absolute error, root mean squared error, and 95th-percentile error. Chair-stand counts are evaluated by the share of predictions that match the clinical count exactly or within one repetition. We also compare the low, fair, and high mobility categories each system assigns using CDC STEADI thresholds. On the TUG, AVA’s timings had a mean absolute percentage error (MAPE) of 6.0% relative to PT timing, with a mean absolute error of 0.71 s; the MR4 overestimated TUG time by 1.59 s on average. On the chair stand, AVA matched the clinical count exactly for all but one participant, and every prediction was within one repetition. AVA also agreed more closely than the MR4 with PT-assigned mobility categories on both tests (TUG: 88.1% vs. 71.2%; chair stand: 100% vs. 87.5%, on the same participants). These results indicate that AVA can deliver accurate, scalable, and privacy-preserving mobility assessment in real-world community settings. KW - fall risk KW - mobility assessment KW - pose estimation KW - Timed Up and Go KW - 30-second chair stand KW - STEADI KW - validation ER -
See AVA in your setting
Talk with our team about bringing camera-based mobility assessment to your clinic, community, or health plan, or get notified when we publish the next report.