<- Back to digest

JOA - 2026-10-01 - Journal Article

Using Early Gait Data From a Smart-Enabled Total Knee Arthroplasty to Identify Patient Function and Activity at 90 Days Postoperative.

Schwab JM, Raynor ME

retrospective cohortLOE IIIn = 4,281 (training n=2,809; validation n=1,472)90 days

Topics

arthroplasty
PMID: 41611104DOI: 10.1016/j.arth.2026.01.063View on PubMed ->

Key Takeaway

A smart TKA implant using gait data from postoperative days 8–21 predicted 90-day recovery class with 70.4% accuracy and 78.0% sensitivity in a validation cohort of 1,472 patients.

Summary Depth

Choose how much analysis to show on this article page.

Summary

This study asked whether passive kinematic data from a smart TKA implant during the first three postoperative weeks could predict 90-day functional recovery class (green/yellow/red). Seven daily gait parameters were reduced via PCA to composite function and activity scores; multivariate logistic regression was trained on 2,809 patients and validated on 1,472. The validation model achieved sensitivity 0.780, specificity 0.612, PPV 0.712, NPV 0.692, and overall accuracy 0.704, with highest sensitivity in women under 65 (0.801).

Key Limitation

The outcome classification system (green/yellow/red ellipses) is proprietary and device-specific, with no correlation reported to validated patient-reported outcome measures, making it impossible to determine whether predicted 'red' patients have clinically meaningful functional deficits.

Original Abstract

BACKGROUND

A subset of total knee arthroplasty (TKA) patients experiences suboptimal recovery manifested by a decline in function, activity, or both. Smart implantable devices capable of continuous, adherence-independent kinematic monitoring offer an opportunity for early risk identification. This study aimed to determine whether early gait data from a smart implant could predict 90-day recovery outcomes reported as a three-class measure (green, yellow, and red).

METHODS

Kinematic data were obtained from 4,281 devices implanted in TKA patients. Data from 2,809 smart implantable devices formed the training dataset, and 1,472 formed the validation dataset. There were seven daily gait parameters from postoperative days eight to 90 that were reduced to two composite scores, function (tibial range of motion, knee range of motion, and stride length) and activity (step count, walk speed, distance, and cadence), based on principal components analyses. "Starting" (days, 8 to 21) and "outcome" (days, 77 to 90) ellipses were calculated for each patient, and a three-class outcome measure was assigned based on the proportion of the 95% confidence ellipse area within predefined function/activity quadrants. Multivariate logistic regression models using starting ellipse parameters were tested to predict the three-class outcomes at 90 days.

RESULTS

The distribution of outcomes was consistent across datasets (training: 42 green, 42 yellow, and 15% red; validation: 45 green, 40 yellow, and 16% red; P = 0.2). Using early gait data, the model achieved a sensitivity of 0.780, specificity of 0.612, positive predictive value of 0.712, negative predictive value of 0.692, and overall accuracy of 0.704 in the validation cohort. Sensitivity was highest in women under the age of 65 years (0.801).

CONCLUSIONS

Passively collected implant-based gait data within the first three postoperative weeks can reasonably predict 90-day recovery class after TKA. Early kinematic modeling may enable timely, targeted interventions to improve functional recovery and satisfaction.