Routine bouts of stair climbing can benefit cardio-respiratory fitness and fat mass.1 Health agencies worldwide endorse this freely accessible activity and offer financial support to promote it.2 Frontline health promoters need robust evidence about the efficacy of interventions to plan effective large-scale campaigns. More than 40 studies have tested the effects of introducing message prompts at the ‘point-of-choice’ between stairs and escalators/elevators, reporting absolute increases in stair climbing of 0.3–10.6%.3 Individual intervention characteristics (e.g. content) explain some variation in effects. Nonetheless, by referring to studies in the latest review (n=42), we explain how inconsistent practice among investigators and reviewers exacerbates uncertainty in this field.3
Prudently, the review exclusively incorporates studies which report rates of stair ‘climbing’.3 At least seven other studies published from 2001 to 2010 report stair ‘use’, a combination of ascending/descending movements (references on request). Interventions produce distinct effects on ascent/descent, which are not discernible from stair use. For example, one study observed a marked increase in stair descent (+12%) but no significant change in stair climbing.4 Combining upward/downward data from this study would show a positive but uninterpretable intervention effect. The heightened energy cost of ascent versus decent (9.6 vs. 4.9 METs) makes it the target for promotion. Specific changes in stair climbing should, therefore, be reported in every study. Furthermore, reviews should not discuss changes in stair climbing and stair use interchangeably.5
Frontline practitioners need to know where interventions are likely to succeed. Problematically, some reviews overlook a critical contextual difference between studies.5 Ten studies took place in venues where elevators provided the alternative to stairs (i.e. workplaces), whereas 32 were in public-access settings (e.g. malls), with escalators.3‘Stratifying’ studies, according to context, unearths a critical finding; ‘stair vs escalator’ interventions (28/32) are more frequently successful than ‘stairs vs elevator’ interventions (3/10) [see6 for explanation].3 Context must clearly be considered when reviewing intervention effects.
Several variables moderate stair choice. ‘Pedestrian traffic volume’ describes the number of people using the stairs and elevator/escalator over a given period.6,7 Studies in public-access settings consistently show that as traffic increases, so does stair choice; probably because escalators become saturated, compelling individuals onto stairs.6 Effects are considerable. In malls, for instance, an additional 16 pedestrians/min increases the likelihood of stair choice by >55%.6 Failure to control for traffic could, therefore, confound findings. In one study, traffic was ∼16% higher during the baseline than the intervention.8 Consequently, the full effect of the intervention may have been obscured. Traffic can fluctuate according to seasonality/vacations.6 Nonetheless, only 15/42 studies recorded this variable.3
Stair choice is also moderated by pedestrians’ demographics.4,6–8 Efforts to record variables such as ethnicity are patchy (12/42 studies), with inconsistent coding criteria (e.g. White/Non-White [n=7]; African American/Caucasian/other [n=2]; Black/White [n=2]; Asian/Non-Asian [n=1]).3 Consequently, it is difficult to meaningfully compare and/or meta-analyse studies. A synthesis of mall-based interventions combined only 6/12 relevant studies owing to coding inconsistency.6
Nine of 42 studies assessed behavioural changes between baseline/intervention using univariate analyses, which cannot incorporate additional moderators.3 Conversely, 33/42 studies used multivariate analyses (i.e. logistic regression).3 Under this approach, moderators including demographics, traffic and condition (baseline/intervention) can be simultaneously entered as predictors of behaviour.4,6,7 Crucially, intervention effects are corrected for the influence of other moderators, providing a more accurate impression of efficacy.
A recent synthesis (n=127,221) discusses the minimum sample required to show the effects of key moderators, with adequate power (β=0.80).6 To detect the average intervention effect, 2,420 stair/escalator choices are required with larger samples needed for unevenly distributed variables (e.g. ethnicity; 10,610). Nine of 42 studies feature <10,610 observations.3 Without performing sample size calculations a priori, studies may lack sufficient power to detect important effects on behaviour.
All previous studies used pre-post designs, where behaviour at one site was observed during baseline and subsequent intervention phases.3 Controlling for known moderators, investigators observed dramatic effects which, discounting interventions, are difficult to explain. Some reviews contend that without control/comparison groups, intervention effects remain unproven.9 Instead, they propose quasi-randomised designs, featuring matched control/intervention sites. Whilst desirable, this approach appears problematic.
To exclude competing sources of causality, sites would need to be matched for all moderators of behaviour, except the intervention. Moderators include demographics and traffic (which vary across sites); several spatial factors (e.g. stair width);10 humidity;7 and ‘behavioural context’ (i.e. work/commuter/leisure).5 Behavioural context is highly site-specific and, therefore, difficult to quantify. For instance, malls often serve shoppers and provide short-cuts for commuters. Without matching sites perfectly, benefits of quasi-randomised designs are undermined. If identical sites exist, they are certainly rare and should, perhaps, be preserved for landmark studies.
To summarise, adopting consistent practice when conducting and reviewing stair climbing research could provide a more cohesive and insightful evidence-base for health promoters/policy-makers.