Personalised Step Thresholds and Physical Activity Incentives: A Randomised Controlled Trial

Intervention Description
This study examines whether personalised step goal progression can enhance engagement in an incentive-based physical activity (PA) programme run by a private health insurer. The intervention targets adults with low or inconsistent PA engagement, defined as individuals who have linked an activity tracker but have not consistently met weekly PA goals in the preceding three months.

Participants were randomly assigned to one of five groups:

Control Group (Standard Goals)

Participants received the standard fixed step goals and incentive structure provided by the insurer’s wellness programme.
Treatment 1-1: Variable Step Threshold – Fast Progression

Participants started with a low daily step goal (2,000 steps) that increased every two weeks:
Weeks 1-2: 2,000 steps per day (3 points)
Weeks 3-4: 3,000 steps per day (3 points)
Weeks 5-6: 5,000 steps per day (3 points)
Weeks 7-8: 7,000 steps per day (3 points)
Treatment 1-2: Variable Step Threshold – Slow Progression

Similar to Treatment 1-1, but step targets increased at a slower rate:
Weeks 1-3: 3,000 steps per day (3 points)
Weeks 4-6: 5,000 steps per day (3 points)
Weeks 7-8: 7,000 steps per day (3 points)
Treatment 2-1: Variable Step & Point Threshold – Fast Progression

Participants followed the same fast step progression as Treatment 1-1, but earned higher rewards at early stages:
Weeks 1-2: 3,000 steps per day (5 points)
Weeks 3-4: 5,000 steps per day (5 points)
Weeks 5-6: 5,000 steps per day (3 points)
Weeks 7-8: 7,000 steps per day (3 points)
Treatment 2-2: Variable Step & Point Threshold – Slow Progression

Participants followed the same slow step progression as Treatment 1-2, but with an adjusted points structure:
Weeks 1-3: 5,000 steps per day (5 points)
Weeks 4-6: 5,000 steps per day (3 points)
Weeks 7-8: 7,000 steps per day (3 points)
Incentive Structure:

Participants in all groups were eligible for a weekly £5 Amazon voucher upon achieving their PA goals.
If participants failed to meet their assigned step goal, they remained at the same threshold or moved backward in step targets.
This intervention seeks to evaluate whether adaptive step thresholds—with varying progression speeds and incentive structures—improve PA engagement, adherence, and habit formation compared to a fixed goal approach.

2023-04-01

2023-06-01

Daily Step Count (Continuous Variable)
Total number of steps recorded per participant per day.
Measured using activity tracking devices throughout the 8-week intervention period.

Weekly Physical Activity Goal Achievement (Binary Variable)
Whether a participant meets their assigned step goal for at least 4 out of 7 days in a given week.
Assessed weekly over the intervention period.

Step Goal Adherence Over Time (Continuous & Binary Variable)
Continuous: Number of days per week participants meet their assigned step goal.
Binary: Whether step goals are met for at least 4 out of 7 days in the final two weeks of the study.

Programme Dropout Rate (Binary Variable)
Whether a participant stops engaging before the end of the 8-week period.
Defined as no recorded steps for ≥7 consecutive days.

1. Post-Intervention Physical Activity Levels (Continuous & Binary Variables)
Continuous: Average daily step count during the four weeks following the intervention.
Binary: Whether a participant maintains ≥5,000 steps per day for at least 4 days per week post-intervention.
2. Step Goal Progression Patterns (Continuous Variable)
Number of times a participant progresses, stagnates, or regresses in step goal levels throughout the intervention.
3. Response to Step Goal Increases (Continuous Variable)
Change in daily step counts following an increase in assigned step thresholds.
4. Engagement with Incentive Structure (Continuous & Binary Variables)
Continuous: Number of weeks participants earn the £5 voucher for goal achievement.
Binary: Whether a participant claims the incentive at least four times during the intervention.
5. Habit Formation Metrics (Binary & Continuous Variables)
Automaticity of Physical Activity Behavior
Proportion of participants who continue meeting step goals without financial incentives (measured in post-intervention weeks).
PA Consistency
Reduction in daily step count variability (i.e., fewer extreme fluctuations in step levels).
Persistence of Routine
Whether participants continue walking at similar times and intensities post-intervention.
6. Heterogeneous Treatment Effects (Interaction Variables)
Differential treatment effects by:
Baseline PA levels (low, medium, high).
Demographics (age, gender).
Device type (smartphone vs. wearable).
Dropout risk (propensity score-matched subgroups).

This study is a randomised controlled trial (RCT) evaluating the impact of progressively increasing step thresholds on physical activity (PA) engagement. The intervention is delivered through a private health insurer’s wellness platform and targets adults with low or inconsistent PA engagement, defined as individuals who have linked an activity tracker but have not consistently met weekly PA goals in the preceding three months.

Study Design & Methodology
This is a prospective, parallel-arm RCT in which participants are randomly assigned to one of five experimental arms. The trial assesses whether gradually increasing step goals, with and without adjusted point thresholds, improve engagement, adherence, and habit formation compared to a fixed-goal approach.

1. Randomisation & Stratification
Participants were randomly assigned using block randomisation stratified by:
Baseline PA level (low, medium, high)
Demographics (age, gender)
This ensures balanced treatment allocation across key participant characteristics.
2. Intervention Arms
Participants are assigned to one of five groups:

Control Group (Standard Goals)

Receives fixed daily step goals as per standard programme conditions.
Treatment 1-1: Variable Step Threshold – Fast Progression

Step goal increases every two weeks: 2,000 → 3,000 → 5,000 → 7,000 steps/day.
Treatment 1-2: Variable Step Threshold – Slow Progression

Step goal increases at a slower rate: 3,000 → 5,000 → 7,000 steps/day over 8 weeks.
Treatment 2-1: Variable Step & Point Threshold – Fast Progression

Same fast step progression, but higher points in early stages:
Weeks 1-2: 3,000 steps/day, 5 points
Weeks 3-4: 5,000 steps/day, 5 points
Weeks 5-6: 5,000 steps/day, 3 points
Weeks 7-8: 7,000 steps/day, 3 points
Treatment 2-2: Variable Step & Point Threshold – Slow Progression

Same slow step progression, but with an adjusted points structure:
Weeks 1-3: 5,000 steps/day, 5 points
Weeks 4-6: 5,000 steps/day, 3 points
Weeks 7-8: 7,000 steps/day, 3 points
3. Incentive Structure
Participants in all groups are eligible for a weekly £5 Amazon voucher upon meeting step goals.
If participants fail to meet their assigned goal, their step target remains the same or moves backward.

Causal Identification Strategy
To ensure robust causal inference, the study employs multiple estimation techniques:

Intention-to-Treat (ITT) Analysis

Estimates the impact of assignment to treatment groups, regardless of adherence.
Treatment-on-the-Treated (TOT) Analysis

Adjusts for non-compliance and dropout using inverse probability weighting (IPW).
Difference-in-Differences (DiD)

Compares pre- and post-intervention step counts between treatment and control groups.
Heterogeneous Treatment Effects

Explores differences by baseline PA level, demographics, and dropout risk.

Data Collection & Measurement
Data Source: Automatically collected via wearable fitness devices linked to the wellness platform.
Primary Outcomes:
Daily step count (measured continuously).
Weekly goal achievement (binary outcome).
Step goal adherence over time (measured in early vs. late intervention periods).
Programme dropout rate (binary outcome).
Secondary Outcomes:
Post-intervention PA levels (assessing sustained behaviour change).
Engagement with incentives (tracking participation in rewards scheme).
Habit formation indicators (measuring consistency and persistence of PA behaviour).

Long-Term Analysis & Post-Intervention Effects
To evaluate the sustained impact of progressive step thresholds, the study includes a long-term follow-up phase analyzing post-intervention behavior.

1. Post-Incentive Physical Activity (Long-Term Outcomes)
Continuous: Average daily step count in the four weeks following the intervention.
Binary: Whether a participant maintains ≥5,000 steps per day for at least 4 days per week post-intervention.
2. Habit Formation & Retention Analysis
Persistence of Routine: Evaluates whether participants maintain similar PA levels without financial incentives.
PA Consistency: Assesses reduction in daily step count variability (i.e., fewer extreme fluctuations).
Automaticity of PA Behavior: Measures whether step targets are met without ongoing intervention prompts.
3. Causal Identification for Long-Term Effects
Event Study Approach:

Estimates dynamic treatment effects at multiple time points post-intervention.
Compares step counts before, during, and after the experiment to determine whether effects persist or fade.
Tests for placebo effects using pre-treatment activity data.
Panel Regression with Individual Fixed Effects:

Controls for time-invariant individual characteristics (e.g., baseline fitness levels).
Models post-intervention changes in PA relative to pre-treatment trends.

Randomization was conducted using a computer-generated process.
Participants were randomly assigned to treatment arms at the time of enrollment.
The original study plan proposed block randomization stratified by baseline step counts and demographic characteristics, but final implementation details should be verified

individual

No

20,200

Planned number of observations: Up to 1,131,200 individual-day observations (20,200 participants tracked daily for 8 weeks).

Total sample size: 20,200 participants.
Participants were randomly assigned to one of five treatment arms. Exact allocation per arm will be updated once internal records are verified.
Based on the study design, we anticipate approximately equal allocation (~4,040 per arm), but final numbers may vary due to data constraints.