Time points

Participants were given six months in which to set up the BeST group sessions. Therefore, clinical competency was assessed within six months of completing the training. All other

outcome measures were assessed at baseline and/or immediately following the training (Table 14).

Table 14. Outcome measures collected with corresponding time points

Domain in Kirkpatrick’s model Measures Assessment points Baseline Immediately after training Within 6 months of training Level three Implementation of BeST  CTS-R-Pain  Level two Knowledge test 

Attitudes and beliefs  

Self-efficacy  Level one Satisfaction  Learner analytics   Baseline demographics 

5.8 Other training

There was a possibility that participants could have received additional training of rel evance to the assessment of competency in delivering the BeST intervention. Whilst this information was not formally captured outside of those being interviewed, participants were informally asked about any additional training they might have received.

5.9 Randomisation

Participants were randomised on an individual basis using a computer generated random number sequence to determine the allocation of the interventions. An independent statistician generated the random number sequence in Microsoft Excel (Dipesh Mistry).

Randomisation type

The randomisation sequence was stratified by centre to ensure that therapists from the same centre were allocated equally to the intervention and control arms.

Allocation

An external researcher (Catherine Lawrence) held the allocation sequence, which was concealed with sequential opaque envelopes, and was responsible for allocating

participants to the intervention and control groups. Following allocation, Helen Richmond (HR) then informed the participant of their allocation and sent out the relevant information according to that allocation.

Blinding of intervention allocation

Whilst the allocation sequence was concealed from the study researcher (Helen

Richmond), they were responsible for the management of the trial and all data collection and therefore, it was not possible to blind the study researcher to the allocated

intervention following randomisation. However, several of the outcome measures were self-reported and were therefore completed by participants independently of the study researcher.

5.10 Sample Size

This was an exploratory study to explore the efficacy of i-BeST. No formal sample size calculation was performed, however, based on a published literature, a total sample of 30 participants was considered to be sufficient to assess variability between participants and provide an estimate of possible effect (151, 152). This was a small study, so the size of the intervention effect was unlikely to be estimated with great certainty. Therefore, this exploratory study aimed to recruit a 15 participants to each group. To allow for a 15-20% drop out rate, a total of 35 physiotherapists were recruited.

5.11 Statistical Analysis

Statistical reporting followed the latest CONSORT guidelines for reporting of parallel group randomised trials (116). All statistical analyses were conducted using SPSS (IBM; version 21).

Descriptive analyses

Descriptive analyses reported the number of participants approached, the number of participants meeting the eligibility criteria, numbers agreeing to randomisation, the number completing training, and the length of time taken to complete the online training. Descriptive statistics were also reported from the learner analytics (chapter 6a).

Statistical analyses

This was a small exploratory trial and therefore statistical analyses were of an exploratory nature. The following methods were used to explore between group differences:

• For continuous outcome measures, data were plotted using histograms to assess proximity to the normal distribution (76) and equality of variance was measured using Levene’s test (114). When these assumptions were met, mean differences between groups were analysed using the Students t-test due to the small sample of participants (76). Although the sample size was small, where data met the

assumption of being normally distributed, parametric methods were used for these analyses. This approach is advocated by Bland and Altman (153), who warn that non-parametric methods do not have sufficient power to detect potential

significant differences in small samples. Therefore, they advocate using parametric methods, suggesting that the most likely effects of using these methods in small samples are a loss of power to detect any significant differences and the

production of wide confidence intervals, as opposed to spurious significant findings (148).

Data with a skewed distribution were transformed to establish a closer resemblance to the normal distribution (76, 114). Where normal or transformed data did not meet the

assumptions of the t-test, one of two strategies were employed:

 Equivalent non-parametric methods were used, in this case, the Man Whitney-U test (76).

Consistent with the CONSORT guidelines, all estimates of effect were provided with 95% confidence intervals (154). An alpha of 0.05 was used as guide for interpreting the statistical significance of effect estimates (76). Selecting a smaller alpha level would have further reduced the risk of incorrectly rejecting the null hypothesis and committing a type 1 error; however, this would also have increased the risk of falsely accepting the null

hypothesis (type 2 error). Bland (76) advises that an alpha of 0.05 is the conventional compromise for managing the risk of committing either error, and was therefore used in this thesis.

Standardised mean differences (SMDs) were reported with 95% CIs to aid comparison of the groups and give an indication of effect size (60, 112). These were both calculated manually using the formula proposed by Hedges and Olkin (155) in Yang and Dalton (156) (appendix 17). Where outcome measures were collected at baseline and again at follow-up, the mean change in scores were adjusted to account for participants baseline score. This was achieved by using this baseline score as a covariate in an analysis of covariance (ANCOVA) (114).

Due to the small sample size and low expected cell count (cells with an expected count of less than five), categorical outcomes were analysed using Fishers exact test for association (76, 114). Where a statistically significant association was found, each category was

Planned sub-group analyses

Participants prior training preference could have impacted upon their engagement with the training programme (137, 139). Therefore, the results from outcome measures were stratified according to those allocated to their preference, those not receiving their

preference and those with no preference. For greater clarity, these categories were further grouped into: those receiving their preference or with no preference, and those not receiving their preference. In both cases the sub groups were summarised with descriptive statistics (mean and standard deviation).

5.12 Ethical Considerations

Since this was a small exploratory study involving physiotherapists training, there was no Trial Steering Committee or Data Monitoring and Ethics Committee. The study had a trial management group, consisting of the researcher’s supervisory team (Professor Sallie Lamb and Dr David Davies).

Ethical arrangements

Ethical and governance approval was granted from the University of Warwick’s Biomedical and Scientific Research Ethics Committee (BSREC; reference number 244-10-2012) and from all hospital research, development and innovation (RD and I) departments (8 NHS Trusts) via the Integrated Research Application System prior to study commencement. To ensure confidentially, all study documentation identified participants by a unique study number, and were otherwise anonymised. Recordings, transcripts and other data capture forms were kept in a locked filing cabinet in a locked office in compliance with the Data Protection Act (1998) and the Standard Operating Procedures of the Warwick Clinical Trials Unit.

Good Clinical Practice

The trial was conducted in full conformance of the principles of the “Declaration of

Helsinki” (1964) (as amended in Tokyo, Venice, Hong Kong, South Africa and Scotland), the Medical Research Council (MRC) Good Clinical Practice Guidelines, and applicable UK legislation.

5.13 Sponsor

The study was sponsored by the University of Warwick.

5.14 Trial administration

The trial was coordinated from WCTU. The researcher (Helen Richmond) was responsible for all trial management duties and for the day to day running of the trial.

5.15 Trial Registration

This study was registered with the Controlled Clinical Trials database (number: ISRCTN82203145).