Preoperative Decline and Postoperative Recovery of Wearable-Derived Physical Activity Over a Four-Year Perioperative Period in Total Knee and Hip Arthroplasty: Evidence from the All of Us Research Program

This longitudinal study of 238 All of Us participants utilizing four years of Fitbit data reveals that total knee and hip arthroplasty patients experience progressive preoperative activity declines followed by a staged postoperative recovery pattern, with higher immediate preoperative activity levels significantly predicting a greater likelihood of returning to habitual physical activity.

The Gist

Here is an explanation of the study, translated into everyday language with some creative analogies to help visualize the findings.

The Big Picture: Tracking the "Step-Count" Journey

Imagine your daily activity level is like the fuel gauge on a car. For years, this study tracked the fuel gauges of people with severe arthritis (in their knees or hips) who eventually got a new joint (a "transplant" of sorts).

Instead of asking patients, "How do you feel?" (which can be fuzzy and forgetful), the researchers looked at Fitbit data. They watched the "fuel gauge" (daily step count) for two years before the surgery and two years after. This gave them a crystal-clear, objective movie of what happened, rather than just a blurry snapshot.

The Story of the "Decline" (Before Surgery)

Before the surgery, the patients' activity levels were dropping, but the pattern was different for knees versus hips.

• The Knee (TKA) Analogy: Think of a slowly leaking tire. For a long time (about a year), the tire lost air very slowly. But then, about a year before the surgery, the leak got much worse, and the tire went flat very quickly.
• The Hip (THA) Analogy: Think of a tire that holds air well until the very last moment. The hip patients stayed relatively active for a long time, only to experience a sudden, sharp drop in activity in the final few months before surgery.

The Takeaway: By the time they walked into the operating room, both groups were moving much less than they used to, but the "knee" group had been struggling for longer.

The Story of the "Recovery" (After Surgery)

Once the surgery happened, the "fuel gauge" started to rise again. The researchers found that recovery didn't happen in a straight line; it happened in three distinct phases, like climbing a mountain:

1. The Sprint (Weeks 1–6): This was the "rapid improvement" phase. Patients went from barely walking to walking a decent amount very quickly. It was like a rocket launching. Interestingly, the hip patients sprinted faster than the knee patients.
2. The Hike (Weeks 7–20): The rocket fuel ran out, and now they had to hike up the mountain. Progress was still happening, but it was slower and required more effort.
3. The Plateau (Weeks 21–104): They reached the summit. The walking level stabilized. They weren't getting much faster, but they were steady.

The Tricky Part: What Does "Recovered" Mean?

This is the most important lesson from the paper. The researchers asked: "When is a patient truly 'back to normal'?" They tested two different definitions:

1. The "Immediate" Baseline: "Are you back to how you were walking right before the surgery?"
   • Result: Most people hit this mark quickly (about 3 months).
   • The Trap: This is like saying you've recovered because you can run a 5-minute mile again, even though you used to run a 3-minute mile before your injury. You're back to your "sick" level, not your "healthy" level.
2. The "Remote" Baseline: "Are you back to how you were walking two years ago, before the arthritis really started slowing you down?"
   • Result: This took much longer (about 5 months on average) and many people never fully got there.
   • The Reality: This is the true measure of getting your life back.

The Analogy: Imagine you were a marathon runner. Then you broke your leg and could only walk to the mailbox.

• Immediate Recovery: You can walk to the mailbox again. (Great! But you aren't a marathon runner yet.)
• Remote Recovery: You are running marathons again. (This is the real goal, but it's much harder to achieve.)

Who Got Better?

The study found two main things that predicted who would get back to their "marathon runner" self:

1. How much you were moving right before surgery: If you were still moving a decent amount right before the operation, you were more likely to recover fully. If you were already very inactive, it was harder to bounce back.
   • Metaphor: A car with a full tank of gas before a repair is more likely to drive well afterward than a car that was already out of gas.
2. Age: Older patients generally had a harder time reaching that "full recovery" mark.

Why This Matters

For a long time, doctors and patients have been guessing about recovery. They might say, "You'll be back to normal in 3 months." This study says, "Actually, 'normal' depends on which version of normal you mean."

• If you mean "back to how you were just before surgery," you'll be there in 3 months.
• If you mean "back to how you were before the pain started," it might take 5 months, and you might not get all the way there if you were very inactive before the surgery.

The Bottom Line: Wearable devices (like Fitbits) are like having a personal coach who never lies. They show us that recovery is a long, staged journey, not a quick fix. It also tells us that keeping active before surgery is crucial—it builds a "functional reserve" that helps you bounce back faster and stronger afterward.

Technical Summary

Here is a detailed technical summary of the paper "Preoperative Decline and Postoperative Recovery of Wearable-Derived Physical Activity Over a Four-Year Perioperative Period in Total Knee and Hip Arthroplasty: Evidence from the All of Us Research Program."

1. Problem Statement

Total Knee Arthroplasty (TKA) and Total Hip Arthroplasty (THA) are standard treatments for end-stage osteoarthritis, yet a significant proportion of patients experience persistent pain, limited functional improvement, or dissatisfaction post-surgery. Current limitations in assessing recovery include:

• Reliance on Subjective Measures: Traditional assessments depend on Patient-Reported Outcome Measures (PROMs) and intermittent clinical visits, which are prone to recall bias and fail to capture dynamic, real-world activity trajectories.
• Short-Term/Immediate Baselines: Existing wearable studies often cover only short preoperative windows (e.g., <1 month) and define recovery relative to immediate preoperative activity. This baseline may reflect disease-related functional decline rather than the patient's true habitual capacity.
• Lack of Long-Term Characterization: Most studies focus on early recovery (weeks to 6 months), lacking data on long-term stabilization (up to 2 years) and distinct preoperative decline patterns.
• Procedure-Specific Nuances: There is insufficient evidence comparing the distinct biomechanical and recovery trajectories of TKA versus THA over extended periods.

2. Methodology

Data Source & Population:

• Dataset: The study utilized the All of Us Research Program (AoURP) controlled tier dataset (v8), linking Electronic Health Records (EHR) with continuous Fitbit-derived step count data.
• Cohort: 238 participants (147 TKA, 91 THA) with linked EHR and Fitbit data.
• Time Window: A four-year perioperative window (2 years pre-surgery to 2 years post-surgery).

Data Processing:

• Temporal Alignment: Surgery date defined as Day 0. Data aggregated into non-overlapping 7-day intervals (weekly averages).
• Quality Control: Weeks retained only if they contained ≥3 observed days.
• Inclusion Criteria: Participants required valid preoperative data (weeks -104 to -1) and postoperative data (weeks 1 to 104).

Statistical Modeling:

• Trajectory Modeling: Piecewise linear mixed-effects models with participant-specific random intercepts were used to model step count changes.
  • Preoperative: One knot (change point) identified via grid search (minimizing AIC) to detect shifts in decline rates.
  • Postoperative: Two knots identified to define three distinct recovery phases.
  • Covariates: Models adjusted for age, sex, and BMI.
• Recovery Definition: Recovery was defined as the first instance where weekly-averaged step counts met/exceeded a baseline threshold for two consecutive weeks.
  • Remote Baseline: Average steps from weeks -104 to -55 (habitual activity).
  • Immediate Baseline: Average steps from weeks -4 to -1 (pre-surgery decline).
• Time-to-Recovery Analysis:
  • Kaplan–Meier curves estimated median time to recovery.
  • Cox Proportional Hazards (CoxPH) models identified predictors (sociodemographics, BMI, surgery type, baseline activity) associated with recovery speed.

3. Key Contributions

1. Longitudinal Granularity: Provided the first continuous, 4-year characterization of physical activity trajectories surrounding arthroplasty, moving beyond discrete time-point snapshots.
2. Dual-Baseline Framework: Introduced a comparative analysis between "remote" (habitual) and "immediate" (disease-affected) baselines, demonstrating that recovery definitions significantly alter clinical interpretation.
3. Procedure-Specific Trajectories: Quantified distinct preoperative decline patterns and postoperative recovery slopes for TKA vs. THA, highlighting biomechanical differences.
4. Predictive Modeling: Identified that higher immediate preoperative activity predicts better recovery to habitual levels, emphasizing the importance of functional reserve and surgical timing.

4. Key Results

A. Preoperative Decline Patterns:

• Both groups showed progressive decline, but with different temporal profiles:
  • TKA: Exhibited a prolonged, sustained decline starting 55 weeks before surgery (accelerated phase: -28.8 steps/week).
  • THA: Remained relatively stable until 16 weeks pre-op, followed by a sharp, rapid decline in the final months (accelerated phase: -58.5 steps/week).
• By week -1, step counts dropped to ~6,000 (TKA) and ~6,400 (THA) from ~7,500 at week -104.

B. Postoperative Recovery Trajectories:
Recovery followed a consistent three-phase pattern for both procedures:

1. Rapid Recovery (Weeks 1–6): Steep increase in steps. THA showed a steeper slope (+970.3 steps/week) compared to TKA (+668.5 steps/week).
2. Decelerating Gains (Weeks 7–19/20): Slower rate of improvement (~90–110 steps/week).
3. Stabilization (Weeks 20/21–104): Plateau with minimal change.

• Outcome: By week 104, TKA stabilized at ~6,629 steps/week and THA at ~7,014 steps/week.

C. Time-to-Recovery & Predictors:

• Baseline Definition Impact:
  • Recovery to Immediate Baseline: Median time = 13 weeks (THA faster than TKA).
  • Recovery to Remote (Habitual) Baseline: Median time = 22 weeks (No significant difference between TKA/THA).
  • Insight: Relying on immediate baselines overestimates recovery speed by ~9 weeks.
• Predictors (Remote Baseline Recovery):
  • Higher Immediate Pre-op Activity: Strongly associated with faster recovery (HR 1.49 per 1,000 steps).
  • Older Age: Associated with slower recovery (HR 0.96 per year).
  • Higher Remote Baseline: Associated with lower probability of recovery (likely due to a higher threshold to reach).
• Predictors (Immediate Baseline Recovery):
  • Surgery type was the only significant predictor; TKA had a lower likelihood of recovery compared to THA (HR 0.55).

5. Significance and Implications

• Clinical Management: The study validates the use of wearable data to set realistic recovery expectations. The "staged" recovery model (rapid early gain followed by a long plateau) helps clinicians and patients distinguish between normal recovery and stagnation.
• Surgical Timing: The finding that higher immediate preoperative activity predicts better long-term recovery supports the concept of "prehabilitation" and suggests that delaying surgery until functional decline becomes severe may negatively impact outcomes.
• Methodological Shift: The paper argues that defining recovery against a remote baseline is clinically superior to an immediate baseline, as the latter reflects disease severity rather than true functional capacity.
• Future Directions: Highlights the potential of large-scale, passive monitoring (mHealth) to replace subjective PROMs for long-term outcome assessment and the need for larger, diverse cohorts to validate these trajectory models.

Limitations Noted: Potential selection bias (health-conscious users with wearables), lack of concurrent pain scores/PROMs, and the retrospective nature of the data.