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

How to cite

Sussman, Kay-leigh; Costa-Font, Joan; Snoyman, Dean; Venter, Stefan; Lagarde, Mylene. 2025. "Personalised Step Thresholds and Physical Activity Incentives: A Randomised Controlled Trial." AEA RCT Registry. AEARCTR-0015589. https://doi.org/10.1257/rct.15589-2.1

Download: BibTeX · RIS — for reference managers without a browser translator. Zotero and most others pick up this page's citation metadata directly.

Kay-leigh Sussman* , Joan Costa-Font , Dean Snoyman , Stefan Venter , Mylene Lagarde

* Primary investigator

Registry ID
AEARCTR-0015589
DOI
https://doi.org/10.1257/rct.15589-2.1
First registered
Last updated
Status
Completed
Trial dates
2023-11-20 – 2024-03-03
Countries
United Kingdom of Great Britain and Northern Ireland (Nationwide, United Kingdom (health insurer members))
JEL codes
I13, I14, I18
Registry record
http://www.socialscienceregistry.org/trials/15589

Abstract

This study evaluates whether personalised step-to-points thresholds improve engagement in an insurer-linked physical activity incentive programme in the United Kingdom. The trial was embedded in a digital app platform operating alongside the insurer’s standard rewards programme.

Eligible adult members with low prior recorded physical activity were invited to join the study app. Participants entered a six-week baseline period followed by a six-week intervention period, with rolling enrolment between November 2023 and February 2024. After post-baseline eligibility confirmation, 714 participants were randomised within insurer-defined engagement strata (Unengaged vs Low engaged).

Randomisation was stratified by engagement group and implemented automatically within the study app. In the Low engaged stratum, participants were allocated to control (CL) or personalised thresholds with bonus (TL). In the Unengaged stratum, participants were allocated to control (CU), personalised thresholds with streak bonus (TU), or personalised thresholds without streak bonus (TU no bonus). Control groups remained on the insurer’s standard step-to-points schedule. Personalised groups started on more attainable step thresholds and could progress toward less generous schedules according to pre-specified platform rules. Participants received a £5 weekly Amazon voucher for meeting the weekly goal (>=12 points), and bonus-eligible arms also received a £15 streak bonus for meeting the goal in all six intervention weeks.

The primary outcome was participant-level mean weekly goal achievement during intervention weeks 0 to 5 (observed weeks only). Secondary outcomes included participant-level mean weekly steps and an indicator for achieving the weekly goal at least once. The primary analysis pooled bonus-treatment arms (TU and TL) versus pooled controls (CU and CL), with a secondary comparison within the Unengaged stratum.

In the main pooled comparison, personalised thresholds with a bonus improved weekly goal attainment but did not produce a statistically detectable increase in mean weekly steps over the six-week intervention period. Follow-up observation in platform data was incomplete and differed by arm, so missing-data sensitivity analyses were conducted and reported.

Intervention

This study evaluated personalised step-to-points thresholds in a UK health insurer-linked trial app operating alongside the insurer’s standard rewards programme. Participants were adults with very low or low prior recorded physical activity (insurer-defined “Unengaged” and “Low engaged” groups) who downloaded the study app and consented to data use. After a six-week baseline period in the trial platform (weeks -6 to -1) used for eligibility confirmation, participants were randomised within engagement strata.

The trial platform awarded points for device-recorded activity (steps and workouts), with a weekly goal of at least 12 points. Participants received a £5 Amazon voucher for each intervention week in which they achieved the weekly goal. In bonus-eligible arms, participants also received a £15 Amazon voucher if they achieved the weekly goal in all six intervention weeks (weeks 0 to 5).

The intervention manipulated the step-to-points schedule (levels 0 to 5). Level 5 was the standard schedule used in the insurer’s usual programme. Personalised arms started on a more generous schedule and could move toward less generous schedules over time according to pre-specified platform rules. Progression required (i) achieving the weekly goal in the preceding week and (ii) meeting a rolling six-week average points threshold for the next level. Participants who failed to achieve the weekly goal for two consecutive weeks moved back one level (to a more generous schedule).

Trial arms (stratified by engagement group) were:

Low engaged stratum:
- CL (control, low engaged): standard schedule (level 5 throughout)
- TL (treatment, low engaged): personalised thresholds with streak bonus; started on level 3 and could transition across levels under the platform rule; eligible for the £15 streak bonus

Unengaged stratum:
- CU (control, unengaged): standard schedule (level 5 throughout)
- TU (treatment, unengaged): personalised thresholds with streak bonus; started on level 1 and could transition across levels under the platform rule; eligible for the £15 streak bonus
- TU no bonus (treatment, unengaged): same personalised threshold progression as TU, but not eligible for the £15 streak bonus

No group started on level 0. The intervention period lasted six weeks (weeks 0 to 5), following the six-week baseline period.

Experimental design

This study is a randomised controlled trial embedded in a UK health insurer-linked digital rewards environment. Participants were members with low prior recorded physical activity, identified using insurer engagement classifications (“Unengaged” and “Low engaged”), invited by email, and enrolled in a study app. Participants were enrolled on a rolling basis between November 2023 and February 2024. Study time was indexed to each participant’s intervention start date. The design included a six-week baseline period (weeks -6 to -1) and a six-week intervention period (weeks 0 to 5).

After consent and app download, six weeks of trial-platform baseline data were used to confirm eligibility and exclude participants whose observed baseline activity was inconsistent with the intended low-activity sample. Randomisation was then implemented automatically within the study app after baseline eligibility confirmation. Randomisation was stratified by insurer engagement stratum (Unengaged vs Low engaged). Intended allocation was 1:1 in the Low engaged stratum (CL vs TL) and 1:1:1 in the Unengaged stratum (CU vs TU vs TU no bonus). Allocation was computer-generated and concealed prior to assignment. Outcomes were recorded automatically by the trial platform. Participants could view their own targets and rewards in the app but were not informed of arm labels or other participants’ schedules.

The intervention varied the step-to-points schedule used to earn weekly points. Control groups (CL and CU) remained on the standard schedule (level 5). Personalised groups began on more generous schedules and could progress toward less generous schedules under pre-specified level transition rules. All groups were eligible for a £5 Amazon voucher for achieving the weekly goal (>=12 points) in a given week. TU and TL were additionally eligible for a £15 bonus for achieving the weekly goal in all six intervention weeks; TU no bonus was not bonus-eligible.

The primary endpoint is the participant-level mean weekly goal-achievement indicator over intervention weeks 0 to 5 (observed weeks only). Secondary endpoints are participant-level mean weekly steps over intervention weeks 0 to 5 (observed weeks only) and an indicator for achieving the weekly goal at least once during weeks 0 to 5. The main ITT comparison pools bonus-treatment arms (TU and TL) versus pooled controls (CU and CL). A secondary comparison is conducted within the Unengaged stratum (TU vs TU no bonus vs CU). Analyses are estimated at the participant level using assignment-based (ITT) models; missing platform observation is characterised separately and addressed in sensitivity analyses.

Design details

Randomization method
Randomisation was implemented automatically within the study app using a computer-generated allocation procedure after six-week baseline eligibility confirmation. Randomisation was stratified by insurer engagement stratum (Unengaged vs Low engaged), with intended allocation 1:1 in the Low engaged stratum and 1:1:1 in the Unengaged stratum. Allocation was concealed prior to assignment. Outcomes were recorded automatically by the trial platform.
Randomization unit
Individual participant (randomisation occurred at the participant level within engagement strata).
Clusters
Not applicable (individual randomisation; no clusters)
Observations
Individual-level randomised sample: 714 participants (final realised sample). The study design includes up to 12 participant-weeks per participant (6 baseline + 6 intervention), implying up to 8,568 participant-week windows before outcome-specific non-observation. Analyses are primarily conducted at the participant level using participant-level summary outcomes over intervention weeks 0 to 5
Arms
Final randomised sample (individuals): 714 participants. Randomised arm sizes (realised): CL (control, Low engaged): 205 TL (personalised targets + streak bonus, Low engaged): 189 CU (control, Unengaged): 117 TU (personalised targets + streak bonus, Unengaged): 118 TU no bonus (personalised targets, no streak bonus, Unengaged): 85 Participants were randomised within engagement strata. Intended allocation was 1:1 in Low engaged (CL vs TL) and 1:1:1 in Unengaged (CU vs TU vs TU no bonus). Realised allocation differed slightly from intended ratios due to observed enrolment and eligibility within strata. App entrants before post-baseline eligibility confirmation: 997; excluded after baseline eligibility confirmation: 283.
Primary outcomes
Mean weekly goal achievement during intervention weeks 0 to 5 (primary endpoint). Participant-level mean of a weekly indicator equal to 1 if the participant accumulated at least 12 trial-platform points in that study week, and 0 otherwise, averaged over observed intervention weeks only (weeks 0 to 5). Weeks with missing weekly goal achievement are treated as unobserved in the main specification. This is the primary endpoint for the main pooled ITT comparison (TU + TL vs CU + CL).
Secondary outcomes
Mean weekly steps during intervention weeks 0 to 5 (observed weeks only). Participant-level mean weekly steps over intervention weeks 0 to 5, calculated using observed weekly steps only. This outcome is defined for participants with at least one observed weekly steps value during weeks 0 to 5. Any weekly goal achieved at least once during intervention weeks 0 to 5. Indicator equal to 1 if the participant achieved the weekly goal (>=12 points) in at least one intervention week (weeks 0 to 5), and 0 otherwise, among participants with at least one observed intervention week for goal achievement. Platform observation / non-observation patterns (descriptive and sensitivity-related). Week-by-week platform observation rates for goal achievement and steps, endline observation (week 5), and mean number of observed intervention weeks, by arm. These are used to characterise missingness/non-observation in trial-platform outcomes.

Cited by

No citation data available for this registration yet.

Keywords