Effect of two behavioural ‘nudging’ interventions on management decisions for low back pain: a randomised vignette-based study in general practitioners

Abstract

Objective‘Nudges’ are subtle cognitive cues thought to influence behaviour. We investigated whether embedding nudges in a general practitioner (GP) clinical decision support display can reduce low-value management decisions .MethodsAustralian GPs completed four clinical vignettes of patients with low back pain. Participants chose from three guideline-concordant and three guideline-discordant (low-value) management options for each vignette, on a computer screen. A 2×2 factorial design randomised participants to two possible nudge interventions: ‘partition display’ nudge (low-value options presented horizontally, high-value options listed vertically) or ‘default option’ nudge (high-value options presented as the default, low-value options presented only after clicking for more). The primary outcome was the proportion of scenarios where practitioners chose at least one of the low-value care options.Results120 GPs (72% male, 28% female) completed the trial (n=480 vignettes). Participants using a conventional menu display without nudges chose at least one low-value care option in 42% of scenarios. Participants exposed to the default option nudge were 44% less likely to choose at least one low-value care option (OR 0.56, 95%CI 0.37 to 0.85; p=0.006) compared with those not exposed. The partition display nudge had no effect on choice of low-value care (OR 1.08, 95%CI 0.72 to 1.64; p=0.7). There was no interaction between the nudges (OR 0.94, 95% CI 0.41 to 2.15; p=0.89).InterpretationA default option nudge reduced the odds of choosing low-value options for low back pain in clinical vignettes. Embedding high value options as defaults in clinical decision support tools could improve quality of care. More research is needed into how nudges impact clinical decision-making in different contexts.

Nudging clinicians: A systematic scoping review of the literature

“…Most of the studies across the two healthcare settings used a pre‐post design. Of the 11 community‐based studies, five (45%) were randomized controlled trials, 38,45‐48 one was cluster‐randomized, 43 one was quasi‐randomized (alternate weeks), 39 and the remaining four (36%) were pre‐post studies40‐42,44: two single‐site and two multi‐site with controls (one with randomized allocation to intervention, the other unclear). Of the 11 hospital‐based studies, seven were single site pre‐post comparisons, 49‐52,54,56,59 one was a two site pre‐post comparison without a control, 53 one was a multi‐site pre‐post with one site non‐randomly allocated to intervention, 57 one was a multi‐site with multiple interventions rolled out at different times (non‐uniformly) across sites, 58 and one was a qualitative study 55 .…”

“…Seven of the salience studies used defaults and default alerts with reported absolute changes ranging from 4% to 56% (all statistically significant); the largest change was from Malhotra et al 40 who explored the shift in default to prescription of generic versions of medicines. The remaining two studies were simulation studies using partition menus; results for Soon et al 47 were found to be not statistically significant, while Tannenbaum et al 48 showed a reduction in low value antibiotic prescribing by 12%.…”

“…The studies in our review commonly melded interventions types in order to exploit multiple cognitive biases or staged their study through phases of intervention add‐ons to measure changes in outcomes. Soon et al refer to this as a “double nudge.” 47 Default alert to increase compliance with best practice ventilation for patients in ICU increased in effectiveness, for example, when paired with audit and feedback implementation. We note, however, that education modules introduced prior to implementing a nudge intervention were not always accounted for in the assessment of outcomes or defined as a control 54,59 …”

A randomised on-line survey exploring how health condition labels affect behavioural intentions

“…Using participants from survey panels may introduce selection bias as participants are reimbursed in some way for their time. However, reimbursing participants is common and many studies have relied on recruitment agencies [e.g., 26 , 28 ]. Recruiting participants in this manner affords a broad geographical reach increasing the study’s generalisability and tests the “labelling effect” with a large cross-cultural sample.…”

Nudging patients with chronic kidney disease at screening to visit physicians: A protocol of a pragmatic randomized controlled trial

“…Additionally, these studies have been conducted on some limited issues, including smoking and obesity. In behavioral economics, an increasing number of studies has investigated the effect of the nudge approach, especially in the United States and Europe [16,[21], [22], [23]]. However, many of these studies have also focused on smoking and obesity, or when studying the other clinical issues, the study subjects have usually been limited to patients in a certain hospital.…”