The impact of choosing words carefully: an online investigation into imaging reporting strategies and best practice care for low back pain

Outcome measures

The primary outcome measure was a composite score derived from three numeric rating scale (NRS) responses (see Supplemental Information 2). The summed outcome score was labelled the Back-Related Perceptions (BRP) score, with higher scores indicating more positive perceptions. We considered that ‘patient’ reassurance (reduced fear and concern) would be reflected by higher BRP scores. Reassurance is variably defined and measured; we considered that a change in reassurance would be more likely to be detected by a composite outcome measure than a single NRS question.

The two secondary outcome measures were the Tampa Scale of Kinesphobia (TSK-11) (Tkachuk & Harris, 2012) and a patient satisfaction rating (see Supplemental Information 3).

Demographics and baseline screening

Self-report data were collected on age, gender, country of residence, language, and education level attained. Participants completed the Back Pain Attitudes Questionnaire (Darlow et al., 2014) (at baseline) and were asked to indicate “yes” or “no” responses to questions relating to previous back pain, recent back pain, current ‘chronic’ back pain and history of back scans.

Sample size calculation

The study was powered to detect a small effect size, (Cohen, 1998) with a power of 80% and an alpha value of 0.05. Assuming a correlation 0.7, the minimum sample size required for each of the three groups was 195 participants.

Statistical analysis plan

To address each of the primary research aims an analysis of covariance (ANCOVA) was conducted to determine if there was a significant effect of group on the primary outcome (BRP scores) after controlling for the effect of baseline BRP scores. Tukey corrected post-hoc pairwise comparisons (with bootstrapped significance and confidence intervals) investigated whether there was a significant difference between the best practice group and either (or both) of the imaging groups, and between the two imaging groups.

To address the secondary research aims (i) and (ii) we performed one-way analysis of variance (ANOVA) to determine the relationship between the group variable and the secondary outcome measures. If a significant effect was identified, Tukey’s post-hoc tests with Bonferroni correction were conducted to identify the between-group differences.

To address secondary research aim (iii) we used ANCOVA to investigate whether there was a significant effect of group after controlling for the effect of baseline BRP scores. The level of significance for this sub-study was set at P ≤ 0.025 to adjust for multiple comparisons. The level of significance for all other studies was set at P ≤ 0.05. Statistical analyses was undertaken using SPSS Statistics software (v22.0.0.0, IBM Corporation, New York).

‘Pre-information’ (+ scan) vs ‘no pre-information’ (+ scan) vs ‘best practice’

ANCOVA was conducted to examine the effect of group on BRP whilst controlling for baseline BRP scores. Levene’s test and normality checks were carried out and assumptions were met. After adjusting for baseline BRP scores there was a significant difference between the three groups in post intervention BRP scores $\left(F\left(2,74\right)=10.4,p<0.001,{\eta }_{\mathrm{p}}^2=.04\right)$. Post-hoc pairwise comparisons revealed significant differences between the ‘no pre-information’ group and the ‘best practice’ group (p = 0.001, d = .33), and between the ‘pre-information’ group and the ‘best practice’ group (p = 0.002, d = .47). There was no significant difference in BRP between the 2 groups who received scans (p = 0.202). For both of the significant comparisons, mean BRP scores (indicating more positive perceptions) were higher for the group who received best practice care.

One-way ANOVA revealed a significant relationship between group and kinesiophobia $\left(F\left(2,553\right)=3.0,p=0.050,{\eta }_{\mathrm{p}}^2=.01\right)$. Post-hoc pairwise comparisons revealed a significant difference between the ‘no pre-information’ group and the ‘best practice’ group (p = 0.044, d = .29). Mean kinesiophobia scores were lower for the group who received best-practice care.

There was a significant relationship between group and satisfaction ratings $\left(F\left(2,553\right)=7.5,p=0.001,{\eta }_{\mathrm{p}}^2=.03\right)$. Post-hoc pairwise comparisons revealed significant differences between the ‘no pre-information’ group and the ‘best practice’ group (p = 0.009, d = .34), and between the ‘pre-information’ group and the ‘best practice’ group (p = 0.001, d = .33). Mean satisfaction scores were higher for those who received best practice care.

Standard reporting vs enhanced reporting

There was a significant effect of group on BRP after controlling for the effect of baseline BRP scores ($F\left(1,275\right)=13.06,p<0.001,{\eta }_{\mathrm{p}}^2=.05$). Mean BRP scores were higher for those who received enhanced reporting.

The individual effects of the two components of enhanced reporting were also examined. When compared with standard reporting, the effect of altering the summary terminology (alone) was significant $\left(F\left(1,221\right)=7.70,p=0.006,{\eta }_{\mathrm{p}}^2=.03\right)$ as was the effect of both altering the summary terminology and including epidemiological information $\left(F\left(1,207\right)=8.40,p=0.004,{\eta }_{\mathrm{p}}^2=.04\right)$. There was no statistically significant difference between the two versions of enhanced reporting practices (p = 0.85).