Analytical validation, construct validity and longitudinal interpretation of digital mobility outcomes.
Evaluate each publication against the proposed context of use: cohort spectrum, reference quality, acquisition setting, endpoint definition and inferential target. The summaries separate measurement evidence from clinical association and intervention effects, with design features and limitations retained for appraisal.
Design and findings. Analytical comparison in 20 healthy adults across four walking conditions, with a separate feasibility cohort of 12 patients following hip or knee arthroplasty. Reliability varied by comparison and parameter; agreement analysis identified relative underestimation of spatiotemporal outputs against motion capture. Postoperative assessments at 2 and 10 weeks detected change.
Appraisal. Transportability is constrained by cohort size, task spectrum and metric-specific performance. The postoperative component establishes feasibility and observed responsiveness, not complication detection or treatment effectiveness. Celloscope partly funded the study and company-affiliated authors are disclosed.
Design and findings. A home-context study of 25 participants with Parkinson’s disease and 25 controls used wearable sensors and video during unscripted activities. Before/after-medication observations examined whether gait-signal characteristics captured motor-state differences under less constrained conditions.
Appraisal. Hand position affected wrist and pocket signals, making context part of the measurement model rather than incidental noise. The study informs ecological validity and feasibility; it does not establish a transferable dosing rule or universal performance across sensor configurations.
Design and findings. Pooled observational analysis across nine older-adult cohorts. Baseline gait speed carried prognostic information, with a reported hazard ratio of 0.88 per 0.1 m/s increment (95% CI 0.87–0.90).
Appraisal. The coefficient characterises a population association conditioned on the analysis model. It cannot be interpreted as the mortality benefit of increasing an individual’s speed, nor transferred directly to a free-living sensor endpoint without demonstrating comparability.
Design and findings. The V3 framework separates sensor verification, analytical validation and clinical validation. It structures the evidence chain linking acquisition performance, derived measurements and a specified clinical concept.
Appraisal. For appraisal, identify which layer a publication addresses and which remain unresolved. Technical accuracy cannot substitute for a clinically interpretable endpoint, and clinical association cannot repair an analytically unreliable measure.
Rozanski, Delgado & Putrino · Front Rehabil Sci · 2023
Design and findings. Cross-sectional analysis of 132 rehabilitation participants, relating smartphone-derived spatiotemporal parameters to Lower Extremity Functional Scale strata. Group differences in pace and temporal organisation provide evidence of association with perceived lower-limb function.
Appraisal. Known-group separation does not establish individual-level interchangeability with LEFS, longitudinal responsiveness or a clinically important change threshold. Assess cohort selection, within-group dispersion and potential spectrum effects before using a parameter for classification.
Verghese et al. · J Neurol Neurosurg Psychiatry · 2007
Design and findings. A prospective community cohort examined whether quantitative gait domains were associated with subsequent cognitive decline and dementia. The study supports gait phenotyping as a contributor to longitudinal neurological risk research.
Appraisal. Prediction in an ageing cohort is distinct from concurrent diagnosis. Consider outcome ascertainment, comorbidity, domain-specific effects and external validation before translating an association into individual risk classification. The findings do not validate mood inference from gait.
Design and findings. V3+ adds usability validation to the original framework, recognising intended-user operation as a condition of fit-for-purpose measurement. This is particularly relevant where adherence, accessibility and home deployment determine data completeness.
Appraisal. Evaluate usability in the target population and setting rather than inferring it from interface simplicity. Burden and failed use can change the sampled population and observation process, with consequences for both analytical and clinical interpretation.
Match the reported estimand to the intended decision. Prioritise error distributions, agreement across the relevant measurement range, repeated-measure performance and external replication. Examine whether the validation cohort includes the slow gait, walking aids, transitions and comorbidities expected in deployment. V3 evaluation framework.
How to use the ratings and topic labels
Stars indicate editorial relevance to clinicians and researchers using this library: 5 = foundational or broadly applicable; 4 = directly useful for a defined question; 3 = specialist or emerging application; 2 = indirect relevance; 1 = background context. They are not journal impact factors, evidence-certainty grades, product comparisons or estimates of clinical benefit. Lists place higher-relevance entries first; chronological steps and test protocols retain their required sequence. Topic labels use the same colors throughout the library, with text identifying every subject.
