Shop-to-Stop Hypertension Statistical Analysis Plan
Discuss this preprint
Start a discussion What are Sciety discussions?Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background
High blood pressure (BP) is the leading global risk factor for death. In Australia, BP control has stagnated, with approximately 50% awareness among those with high BP. Shop-to-Stop Hypertension evaluates whether retail-based screening plus digital nudges can improve detection and follow-up. This document extends the already-published study protocol by pre-specifying the planned analyses.
Design and Setting
Multi-centre, parallel-group, cluster-randomised clinical trial across 30 Bunnings stores in New South Wales, Australia. SiSU Health Stations will be installed in the participating Bunnings stores to screen the public, with stores randomised to whether participants at that store receive text message nudge to retest or not.
Outcomes
The primary outcome is proportion of patients having a repeat BP check at a SiSU Health Station in the text-message based nudge group versus usual care. Secondary outcomes also compare between these two groups and include examining change in BP control (continuous and binary), change in weight, body mass index, and body fat percentage (all continuous). Medication use, health service use, number of BP checks, and awareness of BP risks will all also be examined, as will factors associated with BP control and whether certain groups are more responsive to text message-based nudges.
Planned analyses
All analyses will be intention-to-treat with alpha=0.05 (two-sided), presented with 95% confidence intervals, and conducted using R software version 4.5. Categorical outcomes including primary will be presented as relative risks and analysed using robust Poisson mixed models with random intercepts for store and fixed effects include treatment arm, stratification variables (socioeconomic decile and rurality), and baseline SBP. Continuous outcomes will be analysed similarly using linear mixed models. In the case of convergence issues, cluster-level stratification fixed effects will be omitted with fallback to generalised estimating equation modelling. Multiplicity across six secondary efficacy endpoints will be controlled using Holm-Sidak and potential selection bias will be examined using tipping point analyses. Subgroup and sensitivity analyses include stratifying or controlling for key prognostic variables e.g., age, sex, site rurality, BMI, pregnancy, BP meds at baseline, and self-reported diabetes.
