ICONS: Identifying Continence OptioNs after Stroke: An evidence synthesis, case study and exploratory cluster randomised controlled trial of the introduction of a systematic voiding programme for patients with urinary incontinence after stroke in secondary care

Narrative review search additions

The search undertaken for the full review was also used to identify studies for the narrative review. The identification of studies relating to predictors was not found to be reliable during filtering because it did not include studies testing predictors of adherence for single behavioural interventions, or trials including subgroup analysis. Therefore, all trials included in existing systematic reviews of single behavioural interventions identified by the original search were checked. The review of predictor variables by Goode61 was also used to trace further studies, with forward and lateral citation searching for all included studies.

Review of effectiveness

Review of acceptability and feasibility

Review of predictors of treatment adherence or outcome

Studies of predictors of adherence or treatment outcome of CBIs, or studies of predictors of adherence to single behavioural interventions were included. Predictors of adherence for single interventions were included because they were thought to be generalisable to behavioural adherence to combined interventions. Predictors of treatment outcome of single interventions were not included because they were not judged to be reliably predictive of treatment outcome for combined interventions, due to the potential for differences in physiological mechanisms of action.

Study designs included were:

• prospective longitudinal cohort studies or clinical trials

• RCTs that include subgroup analysis of factor(s) influencing adherence/outcome

• retrospective cohort or cross-sectional studies.

To be included, studies had to include a description of the method of data analysis, and provide data on the relationship between the predictor and outcome based on individual study participants (other than the baseline value of that variable). For full data extraction of results for predictor variables, studies had to identify independent predictors using multivariate analysis. Studies using univariate analysis were only partially data extracted, for the identification and listing of potential predictor variables.

The dependent variables included were any of the following:

• intention to adhere/short- or long-term adherence

• treatment failure/non-response

• cure

• improvement

• psychological status/QoL.

Any time points for outcome measurement were considered.

Review of effectiveness

Data relevant to the pre-stated outcome measures, characteristics of the study, interventions and participants were extracted. The elements of the voiding intervention were categorised based on a previous meta-study. 53,54 The categorisation of the client behaviour change intervention was based on a taxonomy of behaviour change techniques. 62

Assessment of methodological quality was undertaken using the Cochrane Collaboration Risk of Bias Tables to include assessment of adequate sequence generation; allocation concealment; blinding of outcome assessors; incomplete data addressed; freedom from selective reporting; and freedom from other bias (see Appendix 4).

Where appropriate, data were quantitatively combined using meta-analysis to determine the typical effect of the intervention. Intention-to-treat analysis was used, where participants are analysed in the group to which they are randomised. Trial data were processed as described in the Cochrane Collaboration Handbook63 using the Cochrane Collaboration statistical package RevMan 4.2.8 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark).

For individual clinical indicators, a fixed-effects model was used to calculate pooled estimates of treatment effects with 95% confidence intervals (CIs) using:

• relative risk (RR) for binary data

• weighted mean difference (WMD) for continuous data using similar measurement

• standardised mean difference (SMD) for continuous data from different measurement sources

• standardised effect (SE) when generic inverse variance (GIV) was used to pool binary and continuous outcomes.

For trials with missing data, primary analysis was based on observed data, without imputation. Assessment of heterogeneity of intervention effects was made using the I² statistic. If substantial heterogeneity of treatment effects was evident (I² ≥ 50%), a random-effects model was used.

A priori subgroup analyses were planned as follows.

Client group factors:

• type of incontinence – SUI only, UUI only, mixed urinary incontinence (MUI)

• sex – female only, male only, mixed

• age – mean age < 65 years, ≥ 65 years

• cognitive status – people with cognitive incapacity excluded/not excluded.

Intervention factors:

• intervention content – BT primary, PFMT primary, PV primary

• intervention level – basic (i.e. the delivery of multiple strategies aimed at increasing the effectiveness of urinary function activities); enhanced (i.e. the addition of strategies aimed at tailoring an intervention to the specific needs of the individual or enhancing adherence or commitment to practice/activities, e.g. goal-setting, reminder systems, coaching)

• intervention duration – i.e. length of time in weeks in contact with intervention delivery (≥ 8 weeks, > 8 weeks)

• intervention intensity – i.e. number of contacts with the person providing the intervention for content delivery or monitoring/feedback (at least weekly, less than weekly).

A priori sensitivity analyses were planned for type of comparison group (no treatment vs. another treatment), study quality to include allocation concealment (adequate, unclear/not adequate) and loss to follow-up (≤ 20%, > 20%).

Subgroup and sensitivity analyses were performed using a chi-squared test for heterogeneity (via the decomposition of the Q-statistic).

Review of acceptability and feasibility

Review of predictors of treatment adherence or outcome

Data were extracted for (independent) predictor variables relating to characteristics of the client group as follows:

• sociodemographic variables, i.e. age, ethnicity, sex, education/income

• physiological variables, i.e. gynaecologic/obstetric status and history, weight/body mass index (BMI), urodynamic variables, prior treatment, type, duration of UI, severity of UI/symptoms

• health/functional variables, i.e. general health status/comorbidities, self-care ability, functional ability, cognitive status, mental health

• psychological variables, i.e. health/treatment perceptions; perceived QoL, self-efficacy/esteem, attributions of control, prior adherence, knowledge/skill, motivation/attitude, goal orientation

• social variables, i.e. social influences and demands.

A coding frame for the definition and classification of predictor variables was used (see Appendix 4). Using a standardised protocol, data extraction for studies using multivariate analysis included:

• research design classification (prospective cohort/clinical trial, RCT, retrospective cohort/cross-sectional study)

• client group classification (age range, sex, UI type, cognitive status)

• intervention classification (combined, PFMT, BT, PV)

• selection and measurement of independent variables (hypothesis/model for selection, definition, who/how measured, timing, validity and reliability of measurement)

• measurement of outcome variables (who/how measured, timing, validity and reliability of measurement, definition of outcome categorisation)

• statistical analysis method

• variables entered into univariate analysis

• variables entered into multivariate analysis

• statistically significant results from univariate analysis not subsequently confirmed as an independent predictor

• statistically significant results for independent predictor variables for:

  • intention to adhere/adherence behaviour

  • treatment failure

  • cure

  • improvement

  • psychological status

  • QoL.

Descriptive data extraction was undertaken by one reviewer and checked by a second reviewer. Data extraction of predictor and outcome variables and quality appraisal was undertaken by two reviewers independently. Quality assessment was based on quality criteria for observational studies67 and for regression studies,68 and included criteria related to participant selection and representativeness, predictor and outcome variable selection, definition and measurement, adequacy of sample size, follow-up and analysis (see Appendix 4).

Review Management Group

The Review Management Group was composed of the named authors on the review. They met quarterly during the review process and their input included:

• discussion of studies referred by reviewers where inclusion was unclear, with subsequent refinement of the criteria for inclusion and exclusion

• feasibility testing of the classification structures for the review and review of the data extraction proforma

• checking back to the original study data from the results to comment on robustness of interpretation

• reading and commenting on all review outputs.

Service User Group

The PPC involvement group were involved at three stages for consultation on the review:

• to advise on the parameters and scope of the review, and the included interventions, comparisons and outcomes

• to consider the draft results of the review and comment on their perceptions and priorities for the components of the intervention and mediating factors

• once the review was completed, to assist in the translation of the findings into practical products for implementation.

Trial Management and Steering Groups

The review findings were presented to the Identifying Continence OptioNs after Stroke (ICONS) Trial Management and Steering Groups, who then made suggestions for:

• the content of the behavioural intervention for UI to be used with people after stroke

• optimal conditions of implementation on which to base the tailoring of the intervention to client groups and settings

• hypotheses about potential mediators and moderators for consideration in the design of the pilot trial.

Their suggestions were then used to adapt the intervention and data collection protocols for use in the case study.

Included studies

Of the 21 studies identified for the effectiveness review, eight were excluded (reasons are given in the table of excluded studies in Appendix 7). Three studies are ongoing. 82,103,104 There were no unpublished studies. The remaining 10 studies are detailed in Table 3.

Description of urinary incontinence interventions

Table 4 summarises details of the interventions used in the 10 trials, including the components of the intervention, method of delivery, and the duration and intensity of contact with professionals. Some of these details were provided by contact with study authors.

All of the trials included PFMT, albeit to various degrees. All of the trials except one33 included BT, with one trial73 including BT or PV, depending on the cognitive status of the individual. Six trials included the teaching of either urge strategies (e.g. distraction) or stress strategies (e.g. muscle clamping), with three trials teaching both31,33,89 and three trials teaching one or the other. 71,81,90 However, description and labelling of the techniques used tended to be inconsistent. Three trials included other strategies, such as advice about alterations to diet and/or fluid intake. 73,76,90

Interventions in two trials were delivered to groups. 90,100 The delivery format was unclear in two trials81,87 and the remainder were delivered to individuals. Eight out of 10 interventions were delivered by nurses, with another intervention predominantly delivered by nurses but including other professions. 90 One intervention was delivered by physical therapists. 81

Most of the interventions ran over 6–12 weeks, with interventions in two related trials running over a minimum of 6 weeks and a maximum of 24 weeks. 73,76

Three trials had weekly contacts with a health-care professional for the duration of the intervention,71,89,100 with four trials having at least bi-weekly contact. 31,33,87,90 The number of contacts was stated as being in the range of 2 to 40 contacts in an intervention lasting a minimum of 6 weeks and a maximum of 24 weeks in one pilot trial. 73 This is unstated, but likely to be similar in the related trial. 76 Most of the trials stated a requirement for practice of the techniques between contacts.

Intensity of intervention was defined as the ratio of the number of contacts with a person delivering the intervention to the length of the intervention period. An intensity of 1 is weekly contact. Intensity could not be derived for the two trials that did not specify the exact number of contacts. 73,76 Three trials had at least weekly contact71,89,100 and four trials had at least bi-weekly contact. 31,33,87,90 Only one trial had less than bi-weekly contact. 81

Features of urinary incontinence intervention components

Table 5 details the features of the BT and PFMT interventions in the included trials. A dash in the table means that the feature was not stated in the paper. The level of description of the interventions was variable and lack of description cannot be interpreted as absence of the feature in practice.

Features of behavioural intervention components

Table 7 details and categorises the behavioural component of the interventions, as per the taxonomy of behavioural interventions described by Abrahams and Michie. 62 The categorisation was based only on the published accounts given of the interventions. The level of description of intervention components was variable. Given the restrictions on length of publication, absence of description of a feature may not constitute its absence in practice.

Allocation of interventions

Not all participants received the same interventions in all trials. In two trials73,76 participants received intervention components dependent on need. Participants were allocated to a self-monitoring phase if they had problematic fluid or caffeine intake, excessive daytime void intervals, nocturia or constipation. Participants were then allocated to BT, or to PV if functionally or mentally dependent on a caregiver. Finally, participants progressed to PFMT if insufficient progress had been made in earlier stages. How many participants progressed through each phase of the intervention is not reported for Bear et al. ,73 but Dougherty et al. 76 report that out of 94 people in the intervention group, the number progressing through each phase was as follows: self-monitoring (n = 41), BT (n = 89), PFMT (n = 45).

In the trial by McDowell et al. ,89 urge strategies were taught to participants who reported involuntary urine loss following a strong urge to void (85/105), stress strategies were taught to those who reported leaking urine with sudden increases in abdominal pressure (44/105), and only participants who reported frequent voiding got BT. However, in a related paper Engberg et al. 107 reported that many participants had high urinary frequency.

Outcomes

The 10 trials used a range of outcome measures, measurement statistics and time intervals for follow-up. Outcomes measured are detailed in Table 8 and summarised below.

Outcome measurement timing

Post-treatment measurement timing was variable (Table 9), with post-treatment measurement at 6 weeks in one trial,100 8–10 weeks in four trials,33,71,89,90 3 months in three trials31,81,87 and 6 months in two trials. 73,76

Follow-up timing also varied, with the most common being 6 months, which was included in seven trials. 31,71,81,87,89,90,100 Long-term follow-up of ≥ 12 months was included in four trials. 76,81,89,90

Quality of included effectiveness studies

The quality of the included studies was assessed against the Cochrane criteria of adequate sequence generation and allocation concealment; completeness of data reporting and blinding for each main class of outcome measure; selective outcome reporting; and any other sources of bias.

Of the 10 studies, three trials had adequate description of the random sequence generation procedure;33,89,100 five trials stated that sequence allocation was random but did not describe the procedure;31,73,76,87,90 and two trials used non-random sequence generation processes, i.e. alternate allocation. 71,81

Allocation was judged to be adequately concealed in one trial,100 unclear in seven trials31,33,73,76,87,89,90 and not adequately concealed in two trials. 71,81

In the three trials that used objective measures of urine loss, blinding of analysts to the results was judged to be unclear in two trials71,73 and adequate in one trial. 76 Blinding of analysts to the results of the bladder diary was judged adequate in two trials,33,100 unclear in five trials71,73,76,87,89 and not adequate in three trials. 31,81,90 In the seven trials that used subjective outcome measurements, blinding of outcome assessors to the results was deemed unclear in two trials76,87 and not adequate in five trials. 31,33,81,90,100

Objective outcome data were judged to be complete in two trials71,73 and not complete in one trial. 76 Data from bladder diaries were judged to be complete in four trials,33,73,89,90 unclear in two trials31,71 and not complete in three trials. 76,81,100 Data from subjective measurements were judged to be unclear in three trials31,33,87 and not complete in four trials. 76,81,90,100

Outcome reporting was judged to be free of the suggestion of selective outcome reporting in five trials,31,33,76,81,87 unclear in one trial89 and selective in four trials. 71,73,90,100

Two trials had over 20% loss to follow-up. 73,90

Generalisability of the included effectiveness studies

All except one of the studies were limited to females, with the remaining study being 10% male. 89 The majority of studies were also limited to older women.

All types of incontinence were included, but there are more data relating to participants with UUI (46%) than with MUI (40%) or SUI (14%). Most of the studies were undertaken with participants who had established and moderate to severe incontinence. Only one study related to people with mild or very mild symptoms. 90 This was the only study to recruit mainly from community populations rather than from clinical settings, although a few other trials did include community advertising as an additional route of recruitment.

In general, people with cognitive impairments were excluded, although two trials did not exclude people with cognitive impairment if a carer was available and willing to be involved. 73,76 One specifically targeted older people who were homebound89 and one was undertaken in a nursing home,71 suggesting that behavioural UI interventions could at least be feasible with frailer client groups.

Most of the studies have been undertaken in a North American or Middle Eastern setting, with only one early study in a European setting. However, given that these are mostly clinic or home delivered interventions, there is no reason to believe that they are not transferable.

Description of studies of client experience

Table 10 lists the studies identified by the search. Eight studies were identified: two studies were unpublished and therefore no data extraction was undertaken. 108,109

Of the remaining six studies, three were completed in the USA,80,83,94 one in Canada,93 one in New Zealand79 and one in the UK. 92

Description of studies of staff experience

Six studies elicited the opinions of staff about aspects of delivering behavioural interventions for UI (Table 11).

One of the studies80 is also included in the client experience section. Five studies were completed in long-term care (LTC) facilities in the USA77,80,86,97,98 and one in acute care in the UK. 75 Two of the studies were within the last 3 years,75,98 four studies were conducted between 12 and 16 years ago. 77,80,86,97

Two studies used questionnaires to collect data;86,97 four studies used group interviews or focus groups, with two of these studies also using a small number of individual interviews. 75,80 All of the studies collected data from nurses, four studies collected data from mixed grades of nursing staff, with two studies specific to nursing assistants (NAs). 77,86 Two studies86,97 were undertaken to elicit the views of staff about a PV intervention that they had participated in; the remaining four studies were about views on aspects of general continence care that could include behavioural intervention. 75,77,80,98

Predictors of treatment adherence

Two linked studies undertaken in the Netherlands measured predictors of different aspects of adherence in the same sample (n = 129) at different time points. 69,70

Predictors of treatment outcome

Six studies were included: one from more than 20 years ago,95 four over 10 years old,31,74,89,100 and one from the past 10 years. 102 Five studies were from the USA31,74,89,100,102 and one from Australia. 95 There were 814 participants in the six studies for whom data were available. Two studies89,100 do not report separate results according to type of incontinence (n = 257); the remaining studies reported results separately for participants with SUI (n = 205) and urinary urgency/UUI or detrusor instability (n = 352).

Number of people remaining incontinent (post treatment)

Three trials31,33,90 gave results for the proportion of participants remaining incontinent at post treatment, by self-report in 1- or 2-week bladder diaries (i.e. not achieving 100% reduction in incontinent episodes). The pooled results are presented in Figure 1.

FIGURE 1.

Number of people remaining incontinent: post treatment. ^aBurgio 199833 = another treatment comparison (drug); ^bBurgio 199833 = attention control comparison; ^aWyman 199831 = combined intervention vs. BT. df, degrees of freedom.

For two trials33,90 with no treatment comparisons (n = 275, data available for 85%), the pooled effect was statistically significant, favouring the CBI (RR 0.81, 95% CI 0.70 to 0.94).

For two trials31,33 with alternative treatment comparisons (n = 267, data available for 96%), the results were marginally statistically non-significant (RR 0.87, 95% CI 0.75 to 1.01; p = 0.06).

Number of people remaining incontinent (follow-up to 12 months)

Wyman et al. 31 (n = 204, data available for 91%) reported the proportion of participants not achieving continence at 3 months post treatment (6 months post baseline). Results were not statistically significant (RR 0.87, 95% CI 0.72 to 1.05).

Number of incontinent episodes per week (post treatment)

Eight trials31,33,73,76,82,89,90,100 including nine intervention comparison pairs reported the number of episodes of incontinence per week at post treatment (between 6 and 12 weeks), measured by self-report in 1- or 2-week bladder diaries (except two trials73,76 where 3-day diaries were used). Bear et al. 73 did not provide data suitable for pooling. The results of the remaining eight studies are summarised in Figure 2.

FIGURE 2.

Number of incontinent episodes: post treatment. ^aBurgio 199833 = another treatment comparison (drug); ^bBurgio 199833 = attention control comparison; ^aWyman 199831 = combined intervention vs. BT. df, degrees of freedom.

Five trials33,76,89,90,100 included a no treatment comparison (n = 750, data available for 79%). Pooled results show a statistically significant mean reduction in episodes of incontinence per week in CBI trials (WMD –3.57, 95% CI –5.52 to –1.62).

Three trials (n = 309, data available for 95%) included comparison against another intervention. 31,33,81 Pooled results were marginally statistically non-significant (WMD –2.18, 95% CI –4.53 to 0.17; p = 0.07).

Number of incontinent episodes per week (follow-up to 12 months)

Three trials31,76,81 reported follow up data (Figure 3). One trial with a no treatment comparison76 reported statistically significant results favouring the CBI at 6 months post treatment (12 months post baseline) (WMD –5.60, 95% CI –9.92 to –1.28).

FIGURE 3.

Number of incontinent episodes: follow-up to 12 months. ^aWyman 199831 = combined intervention vs. BT. df, degrees of freedom.

Two trials31,81 with comparisons against other treatments (n = 179, data available for 88%) report results for 3 months post treatment (6 months post baseline). The pooled effect was not statistically significant (WMD –1.40, 95% CI –4.59 to 1.79).

Proportion of people achieving 75% or more reduction in incontinent episodes (post treatment)

Three trials31,33,90 with four relevant intervention comparison pairs gave results for the proportion of participants achieving 75% or more reduction in incontinent episodes at post treatment, by self-report in 1- or 2-week bladder diaries (see Figure 4).

FIGURE 4.

Seventy-five per cent or more reduction in incontinent episodes: post treatment. ^aBurgio 199833 = another treatment comparison (drug); ^bBurgio 199833 = attention control comparison; ^aWyman 199831 = combined intervention vs. BT. df, degrees of freedom.

For two trials33,90 including no treatment comparisons (n = 275, data available for 85%), the pooled effect was statistically significant favouring the CBI (RR 2.16, 95% CI 1.58 to 2.95).

Two trials31,33 included another treatment comparison. The pooled effect (n = 265, data available for 97%) was statistically significant favouring the CBI for the comparison including the Wyman et al. 31 BT comparison group (RR 1.40, 95% CI 1.12 to 1.75; as illustrated in Figure 4); but statistically non-significant with the inclusion of the Wyman et al. 31 PFMT comparison group (RR 1.60, 95% CI 0.94 to 2.73; p = 0.08).

Proportion of people achieving 75% or more reduction in incontinent episodes (follow-up to 12 months)

One of the trials31 also reported results for 75% or more reduction in incontinent episodes at 6 months post baseline (3 months post treatment). The effect size (n = 204, data available for 92%) was statistically significant favouring the CBI for the comparison with BT (RR 1.79, 95% CI 1.16 to 2.78); but not statistically significant for the comparison with PFMT (RR 1.32, 95% CI 0.92 to 1.91).

Post treatment

Two trials31,33 with three relevant comparison groups gave results for the proportion of participants who classified their incontinence as ‘much better’ at post treatment, by self-report on four- or five-point scales. Results are summarised in Figure 5.

FIGURE 5.

Subject perceptions of improvement: post treatment. ^aBurgio 199833 = another treatment comparison (drug); ^bBurgio 199833 = attention control comparison; ^aWyman 199831 = combined intervention vs. BT. df, degrees of freedom.

One trial33 reported comparison against a placebo control group (n = 130, data available for 85%), with a statistically significant effect size favouring the CBI (RR 2.75, 95% CI 1.72 to 4.42).

Two trials31,33 included comparisons against another treatment (n = 265, data available for 91%). The pooled effect was also statistically significant favouring the CBI (RR 1.42, 95% CI 1.12 to 1.81).

Follow-up to 12 months

One trial31 comparing a CBI with either BT or PFMT (n = 204, results available for 90%) gave results for patient perception of UI as ‘much better’ at 6 months post baseline (3 months post treatment). The effect size was just statistically significant for the BT comparison group favouring the CBI (RR 1.53, 95% CI 1.00 to 2.32; p = 0.05); but not statistically significant for the PFMT comparison group (RR 1.43, 95% CI 0.96 to 2.12; p = 0.08).

Grams of urine lost in 24 hours (post treatment)

Four studies used a pad test to evaluate severity of urine loss31,71,73,76,109 (Wyman et al. 31 reported in Elser et al. 109). However, Elser et al. 109 did not report data separately for each treatment group, and Bear et al. 73 reported mean grams of urine loss per day, but no SDs. Results for the remaining trials using no treatment comparison groups71,76 are illustrated in Figure 6.

FIGURE 6.

Severity of incontinence (grams of urine loss in 24 hours): post treatment. df, degrees of freedom.

Results were pooled using GIV and a random-effects model given the substantial heterogeneity of treatment effect (I² = 77.9%) (see Figure 6). The result was not statistically significant (SE –0.70, 95% CI –2.41 to 1.01).

Dougherty et al. 76 also provided a subjective measure of severity of urine loss, rated 1–7, with 7 defined as ‘the best bladder control you can imagine’ and 1 defined as ‘the worst bladder control you can imagine’. Treatment effects were significantly different between experimental and control conditions favouring the CBI (SMD –1.21, 95% CI –0.86 to –1.56).

Grams of urine lost in 24 hours (follow-up to 12 months)

The same two trials included follow-up data: Dougherty et al. 76 at 12 months post baseline (6 months post treatment) and Aslan et al. 71 at 6 months post baseline (4 months post treatment).

The pooled effect (Figure 7) using GIV and a fixed-effects model was statistically significant favouring the CBI (SE –0.43, 95% CI –0.80 to –0.06). However, by using a random-effects model to facilitate comparison at post treatment and follow-up, the pooled effect was not statistically significant (SE –0.60, 95% CI –1.47 to 0.26).

FIGURE 7.

Severity of incontinence (grams of urine loss in 24 hours): follow-up to 12 months. df, degrees of freedom.

Urinary frequency (post treatment)

Five studies provided data on the number of voids during the day (Figure 8). Four studies (n = 579, data available for 75%) compared a CBI with no treatment. 71,76,90,100 GIV was used to combine dichotomous outcomes from Aslan et al. 71 with the continuous outcome data from the other three trials. Using a random-effects model given the substantial heterogeneity of treatment effects (I² = 74.9%), the pooled result was statistically significant favouring the CBI (SE –0.55, 95% CI –0.97 to –0.13).

FIGURE 8.

Frequency (number of voids during the day): post treatment. df, degrees of freedom.

The result for one quasi-randomised trial81 comparing a CBI against GIV was used to combine dichotomous outcomes from Aslan et al. 71 with the continuous outcome another treatment (n = 44, data available for 82%) was not statistically significant (SE –0.04, 95% CI –0.70 to 0.62; WMD –0.10, 95% CI –1.83 to 1.63).

Urinary frequency (follow-up to 12 months)

Three studies reported follow-up results for frequency of micturition during the day. Two studies (n = 282, data available for 57%) compared combined behavioural training against a no-treatment control (Figure 9).

FIGURE 9.

Frequency (number of voids during the day): at follow-up to 12 months. df, degrees of freedom.

Aslan et al. 71 reported data for 6 months post baseline (4 months post treatment). Dougherty et al. 76 reported data for 12 months post-baseline (6 months post treatment). Using a random-effects model given the substantial heterogeneity of treatment effects (I² = 74.3%), the pooled effect was not statistically significant (SE –0.63, 95% CI –1.48 to 0.22).

In a quasi-experimental study comparing a CBI against medication, Kafri et al. 81 reported a statistically significant treatment effect favouring the CBI for 6 months post-baseline (3 months post treatment) (SE –0.71, 95% CI –1.39 to –0.03; WMD –1.70, 95% CI –3.26 to –0.14).

Nocturia (post treatment)

Six trials reported results for the number of voids during the night. The results for five trials33,71,76,90,100 using a no treatment comparison are illustrated in Figure 10.

FIGURE 10.

Nocturia (number of voids during the night): post treatment results for no treatment comparisons. ^aBurgio 199833 = another treatment comparison (drug); ^bBurgio 199833 = attention control comparison. df, degrees of freedom.

Generic inverse variance was used to combine dichotomous and continuous outcomes, and substantial heterogeneity of treatment effect (I² = 89.5%) necessitated the use of a random-effects model. Pooled results (n = 709, data available for 72%) were not statistically significant (SE –0.33, 95% CI –0.95 to 0.29).

Two trials33,81 compared a CBI against another treatment (drug therapy). Results (n = 176, data available for 73%) were statistically significant favouring the CBI (SE –0.46, 95% CI –0.81 to –0.11; WMD –0.36, 95% CI –0.67 to –0.04).

Nocturia (follow-up to 12 months)

Three trials reported data for nocturia at follow-up. 71,76,81

Two trials reported no treatment comparisons: Aslan et al. 71 reported results for 6 months post baseline (4 months post treatment); Dougherty et al. 76 reported results for 12 months post baseline (6 months post treatment). Results are illustrated in Figure 11.

FIGURE 11.

Nocturia (number of voids during the night): follow-up to 12 months for no treatment comparisons. df, degrees of freedom.

Pooled results (n = 282, data available for 59%) were not statistically significant (SE –0.97, 95% CI –3.30 to 1.37).

One quasi-randomised trial81 compared combined behavioural training with another treatment. Results (n = 44, data available for 82%) at 6 months post baseline (4 months post treatment) were statistically significant favouring the CBI (SE –0.89, 95% CI –1.59 to –0.19; WMD –1.00, 95% CI –1.75 to –0.25).

Urgency (post treatment)

Only one quasi-randomised trial71 reported results for urinary urgency (n = 64, data available for 78%). The number of people reporting that symptoms of urgency were unchanged or worse was statistically significant post-treatment favouring the CBI, compared with a no treatment control group (RR 0.57, 95% CI 0.37 to 0.89).

Urgency (follow-up to 12 months)

Aslan et al. 71 also reported results for urinary urgency at 6 months post baseline (4 months post treatment). The number of people (n = 64, data available for 78%) reporting that symptoms of urgency were unchanged or worse was statistically non-significant (RR 0.67, 95% CI 0.41 to 1.07; p = 0.09).

Post treatment

Two different aspects of QoL were measured in the trials: scales measuring the impact of incontinence; and scales measuring symptom distress. Owing to the difference in the underlying concepts, these measures were not pooled and results are presented separately.

Impact of incontinence (post treatment)

Five trials included a measure of disease-specific QoL but two of these trials did not report data in a form suitable for pooling. Aslan et al. 71 used the King’s Health Questionnaire, but post-treatment data were not reported. McFall et al. 90 used the Short Form questionnaire-36 items (SF-36), but do not report data separately for treatment and control group, other than mean scores on one subscale.

Three trials reported data suitable for pooling. Dougherty et al. 76 and Wyman et al. 31 used the IIQ (lower score = improvement). Kafri et al. 81 used the I-QOL (higher score = improvement). To harmonise the direction of scores, results for Kafri et al. 81 were entered as negative. Substantial between trial heterogeneity (I² = 75.8%) necessitated the use of a random-effects model for pooling. Results are presented in Figure 12.

FIGURE 12.

Quality of life (impact of incontinence): post treatment. ^aWyman 199831 = combined intervention vs. BT. df, degrees of freedom.

Impact of incontinence (follow-up to 12 months)

The same three trials measured impact of incontinence at follow-up: Dougherty et al. 76 at 12 months post baseline (6 months post treatment); and the other two trials31,81 at 6 months post baseline (3 months post treatment) (Figure 13).

FIGURE 13.

Quality of life: impact of incontinence: follow up to 12 months. ^aWyman 199831 = combined intervention vs. BT. df, degrees of freedom.

Dougherty et al. 76 compared a CBI with no treatment control. The effect (n = 218, data available for 51%) was marginally statistically non-significant (SMD –0.36, 95% CI –0.74 to 0.01; p = 0.06).

A random-effects model was used due to the heterogeneity of treatment effect (I² = 85.7%) to pool the two trials using comparison against another treatment. 31,81 The pooled effect size (n = 179, data available for 86%) was not statistically significant (SMD –0.57, 95% CI –1.62 to 0.49).

Symptom distress (post treatment)

Two studies used measures of symptom distress or impact. Burgio et al. 33 used the Symptom-Checklist-90-Revised; Wyman et al. 31 used the UDI. Results are summarised in Figure 14.

FIGURE 14.

Quality of life (symptom distress): post treatment. ^aBurgio 199833 = another treatment comparison (drug); ^bBurgio 199833 = attention control comparison; ^aWyman 199831 = combined intervention vs. BT. df, degrees of freedom.

One trial33 compared combined behavioural training with a placebo control group (n = 130, data available for 61%). Results showed no statistically significant difference (SMD –0.05, 95% CI –0.44 to 0.34).

A random-effects model was used to pool the results of the two trials using comparisons against another treatment because of substantial heterogeneity in treatment effects (I² = 92.5%). Results (n = 265, data available for 89%) showed no statistically significant between-groups difference in treatment effects (SMD –0.46, 95% CI –1.41 to 0.50).

Symptom distress (follow-up to 12 months)

One trial31 assessed symptom distress at 6 months post baseline (3 months post treatment). The treatment effect (n = 135, data available for 89%) was marginally statistically non-significant for the BT comparison group (SMD –0.36, 95% CI –0.72 to 0.01; p = 0.06); but not statistically significant for the comparison with PFMT (SMD –0.24, 95% CI –0.59 to 0.12).

Post treatment

Table 15 shows a summary of pooled effect sizes per outcome.

Follow-up to 12 months

Table 16 shows a summary of pooled effect sizes per outcome.

Quality of results

The results presented above need to be considered in the light of the quality of evidence to support them. Tables 15 and 16 illustrate that not many trials contributed data to each outcome. Trials were also judged to be of good, moderate, or poor quality as follows:

• good quality (++): studies where the results are unlikely to be affected by any weaknesses in study design or conduct

• moderate quality (+): studies where weakness in study design or conduct has the potential to impact on the validity or reliability of the results

• poor quality (–): studies were the results are likely to be affected by weaknesses of study design or conduct.

For each outcome, results that are supported by trials of moderate or good quality will be summarised, together with any other quality issues (number of respondents, per cent of respondents data are available for, heterogeneity of pooled treatment effect) that could influence interpretation of the quality of the evidence. For each outcome, results for comparison with another treatment will be presented first, followed by results for no treatment comparisons.

Generalisability of results

Type of incontinence

Two trials included only people with UUI as the predominant pattern. 33,81 Four trials presented results for combinations of UI type (i.e. MUI, SUI or UUI) without subgroup data. 76,89,90,100 One trial presented subgroup data for people with SUI or MUI. 31 Results are presented in Figure 16.

FIGURE 16.

Subgroup analysis: type of incontinence. ^aBurgio 199833 = another treatment comparison (drug); ^bWyman 199831 = combined intervention vs. PFMT. df, degrees of freedom.

Age

Two trials had younger samples (i.e. with a mean age of less than 65 years)31,81 and five trials had relatively older samples (i.e. aged ≥ 65 years). 33,76,89,90,100 Results are presented in Figure 17.

FIGURE 17.

Subgroup analysis: age. ^aBurgio 199833 = another treatment comparison (drug); ^bWyman 199831 = combined intervention vs. PFMT. df, degrees of freedom.

No statistically significant difference in treatment effects was found between the three types of UI (p = 0.34).

No statistically significant difference in treatment effects was found for younger versus older age groups (p = 0.50).

Type of intervention

Four intervention–comparison pairs31,76,90,100 were judged to have an initial or primary emphasis on BT and four intervention–comparison pairs were judged to have an initial or primary emphasis on PFMT. 31,33,81,89

Results are presented in Figure 18. No statistically significant difference in treatment effects was found for type of intervention, but there was a trend towards BT (p = 0.08).

FIGURE 18.

Subgroup analysis: type of intervention. ^aBurgio 199833 = another treatment comparison (drug); ^aWyman 199831 = combined intervention vs. BT; ^bWyman 199831 = combined intervention vs. PFMT. df, degrees of freedom.

Level of intervention

Trials were classified dependent on whether they were judged to be basic (i.e. delivery of behavioural strategies aimed at increasing the effectiveness of urinary function activities) or enhanced (i.e. additional behavioural strategies aimed at tailoring an intervention to the specific needs of the individual or enhancing adherence or commitment to practice/activities, e.g. goal-setting, reminder systems, coaching).

Two trials were judged to be focused on basic delivery of strategies aimed at voiding function. 81,100 The remaining five trials were judged to have at least some enhancement to basic delivery of a voiding function intervention,31,33,76,89,90 although the relative emphasis on additional behavioural strategies varied. Results are illustrated in Figure 19.

FIGURE 19.

Subgroup analysis: level of intervention. ^aBurgio 199833 = another treatment comparison (drug); ^bWyman 199831 = combined intervention vs. PFMT. df, degrees of freedom.

No statistically significant difference in treatment effects was found for level of intervention (p = 0.92).

Duration of intervention

Trials were categorised into those with 8 weeks or less intervention delivery, or more than 8 weeks. Three trials had an intervention delivery period of 8 weeks or less. 33,89,100 Four trials had intervention delivery periods of more than 8 weeks. 31,76,81,90 Results are presented in Figure 20.

FIGURE 20.

Subgroup analysis: duration of intervention. ^aBurgio 199833 = another treatment comparison (drug); ^bWyman 199831 = combined intervention vs. PFMT. df, degrees of freedom.

No statistically significant difference in treatment effects was found for duration of intervention (p = 0.71).

Intensity of intervention

Intensity of intervention was defined as the ratio of the number of contacts with a person delivering the intervention to the length of the intervention period, with subgroups defined as contact at least weekly or less than weekly. Five trials had less than weekly contact,31,33,76,81,90 and two had at least weekly contact. 89,100 Results are illustrated in Figure 21.

FIGURE 21.

Subgroup analysis: intensity of intervention. ^aBurgio 199833 = another treatment comparison (drug); ^bWyman 199831 = combined intervention vs. PFMT. df, degrees of freedom.

No statistically significant difference in treatment effects was found for duration of intervention (p = 0.48).

Type of comparison group

Five trials had no treatment intervention–comparison pairs. 33,76,89,90,100 Three trials had intervention versus another treatment comparison groups. 31,33,81 Results are illustrated in Figure 22.

FIGURE 22.

Sensitivity analysis: type of comparison group (no treatment, another treatment). ^aBurgio 199833 = another treatment comparison (drug); ^bBurgio 199833 = attention control comparison; ^bWyman 199831 = combined intervention vs. PFMT. df, degrees of freedom.

No statistically significant difference in treatment effects was found for type of comparison group (p = 0.48).

Research aims

Table 17 details the stated aims of the six included studies: two studies mainly referred to factors influencing choice/uptake of UI treatments in older people in residential care;80,94 two studies provided information about factors influencing participation and adherence to behavioural therapies;79,93 and two studies focused on reasons for dropout from a UI treatment programme. 83,84,92

Choice/uptake

Table 18 details the results of the two studies considering factors impacting on choice or uptake of behavioural treatments for UI. 80,94 Both of these studies considered the treatment preferences of older adults in LTC facilities in the USA. Results suggest that clients may have a higher tolerance for symptoms and a lower tolerance for disturbance, with a preference for interventions promoting independence and comfort, and resistance to any invasive intervention. Behavioural interventions such as PV can be viewed as embarrassing and resulting in dependence on others, with residents in care facilities disliking the subsequent reliance on nursing staff.

Both of these studies did not limit data collection to respondents with UI. In one study, other problems with elimination were not differentiated94 and in the other study, respondents without UI were used as proxies. 80

Participation/adherence

Table 19 details the results of two studies considering factors impacting on participation in and adherence to behavioural interventions. Hay-Smith et al. 79 was specific to PFMT, whereas Milne and Moore93 referred to client factors impacting on adherence to both BT and PFMT.

For BT, a barrier to adherence was increased fear of accidents, whereas for both BT and PFMT, respondents identified difficulty with developing a routine and fitting the intervention into daily life, but a feeling of mastery and control if successful. Enablers included realistic goals and adaptation of daily routines.

There were negative perceptions of PFMT, including the difficulty of learning the exercises and knowing whether or not they were done correctly. Respondents valued feedback and follow-up. Contextual features that impacted on adherence included the requirement for privacy. Both of the studies were conducted with women, with one study specific to women with SUI. 79

Withdrawal/dropout

Two studies considered women’s reasons for withdrawal from behavioural UI programmes. PFMT was a major component of both programmes. The findings are detailed in Table 20.

Women cited other health problems, competing pressures, the inconvenience of attending clinics and negative perceptions of PFMT as barriers. Due to the difficulty of knowing whether or not practice was successful, feedback was viewed as helpful by some respondents in both studies. Both of these studies were completed on non-attenders of established continence clinics, so the results may not be generalisable beyond these specific examples.

Quality of findings

As well as the number of studies supporting a particular finding, the quality of the original study needs to be considered. Table 22 takes the main category of findings from Table 21 and attaches levels of evidence. The number of studies are identified, with a quality classification as follows:

• studies where the credibility of the findings are unlikely to be affected by any weaknesses in study design or conduct (++)

• studies where weaknesses in study design or conduct have the potential to impact on the credibility of the findings (+)

• studies were the credibility of the findings is likely to be affected by weaknesses of study design or conduct (–).

Researcher conclusions and implications for practice

As well as extracting the original data, we extracted any suggestions by the researchers on their conclusions about suggested ways of improving practice. Although not primary data originating from clients, these conclusions are useful to researchers who are designing future interventions. The data was sourced from the conclusions and implications for practice sections of research reports. The suggestions were simply classified as either relating to the structure of health care (e.g. resources, staff training), or the process of health care (i.e. what should be done). Suggestions were identified as relevant to the stage of informed choice and assessment of suitability for an intervention, or encouraging adherence and preventing dropout.

Research aims

Table 25 details the stated aims of the six included studies. One study mainly referred to factors influencing choice/uptake of UI treatments for older people in residential care. 80 Three studies detailed the factors influencing the provision of continence care: one in acute care settings75 and two in LTC settings. 77,98 Two studies (from the last decade) focused on staff perceptions of delivering a PV intervention in LTC settings. 86,97

Although the studies had slightly different aims, they all included factors that potentially impacted on the methods selected and used for the promotion of continence by staff, including factors relating to clients, interventions and context. Factors are presented as relating to generic continence promotion, except for those factors specific to an intervention, which are labelled as such in the tables.

Client factors

The included studies identified characteristics of the client that would affect the continence promotion strategies that staff used, or their chances of success, as detailed in Table 26.

From a staff perspective, the success of continence promotion strategies was affected by clients’ views on UI and their past experiences; their functional, cognitive and communication abilities; their motivation; and whether or not their continence improved. Ensuring functional ability to participate and the appropriate assessment of clients’ suitability for participation were seen as enablers to appropriate continence promotion.

Intervention-related factors

Table 27 identifies the factors that staff identified as potentially influencing the use of specific interventions.

Contextual factors

Contextual factors were most frequently identified by staff as impacting on their ability to promote continence, as detailed in Table 28.

In summary, factors which could act as a barrier to continence promotion by staff included:

• views on UI in older people

• different views on aims of UI therapy than clients or family

• referral and admission routes

• nursing assessment procedures

• staff motivation and education

• lack of staff and conflicting work priorities

• the requirements of manual handling

• perceptions of treatment effectiveness

• scheduling conflicts.

Factors that could act as enablers to continence promotion included:

• education

• teamwork

• adequate staffing

• methods of work allocation

• sufficient and appropriate equipment and supplies

• experience of success.

The two studies specific to the NA role in the promotion of continence or the implementation of PV also identified methods to improve management of the NA contribution to care, including:

• regular assignments77

• inclusion in the plan of care, and in reports on mobility and functional status77

• increased accountability for adhering to a toileting plan98

• more autonomy and freedom to prioritise, work as a team77

• recognition and reward of contribution to continence promotion and management77,98

• identification of role models. 98

Quality of findings

The quality of the studies related to staff views was poor overall, with only one study of good quality (1 ++)98 and two studies of moderate quality (1 +). 75,77 However, all of the findings were supported by at least one study of moderate quality, except those that are given in italics relating to the differences between staff, clients and family preferences, and some specific enablers for PV implementation such as assessment, allocation and staff monitoring.

Generalisability of findings

The generalisability of most of the findings related to barriers is confirmed in both acute and LTC settings, in the USA and the UK. However, the generalisability of findings related to enablers is mostly confined to NAs working in LTC settings in the USA.

Researcher conclusions and implications for practice

Researcher suggestions to improve the general delivery of continence interventions extracted from the discussion or conclusions included:

• changing the philosophy from one of accepting incontinence, and the use of self-efficacy-based motivational interventions to help staff and residents believe continence can be improved98

• identifying residents most likely to benefit from routine toileting77,97

• staff education97 including ways in which behavioural interventions may help improve urinary control98

• staff skills in promoting mobility and continence, the experience of caregiving and strategies for mutual support86

• presenting realistic expectations of outcome to staff

• supervising/monitoring nursing staff performance86,97 and providing appropriate incentives to ensure adherence to behavioural-based continence care programmes by staff, patient and family98

• sufficient staff77,86 and a team approach to continence care77,98

• appropriate infrastructure/organisational environment86,98

• new technologies to facilitate documenting continence care and new technologies such as bladder scanners98

• examination of routines that promote or hinder productivity. 86

The two studies specific to the NA role also suggested:

• consistency of NA assignments to allow the development of relationships77

• a substantial role for NAs in developing continence care plans for residents for whom they are responsible77,97

• including NAs in the daily nursing report77

• nursing recognition and commendation of the contributions of NAs to successful continence care. 77

Predictors of adherence

Two papers relating to one study reported multivariate analyses of predictors of adherence. Alewijnse et al. 69 tested predictors of intention to adhere to a behavioural programme, prior to the intervention starting. At 1-year post treatment, Alewijnse et al. 70 measured predictors of long-term adherence (i.e. up to 1 year post treatment) to a behavioural exercise regime in the same cohort of women, in a RCT.

Three factors were found to be independent predictors of intention to adhere: severity of UI (more urine loss per wet episode); self-efficacy difficulties (perceived ability to perform exercises as required); and self-efficacy abilities (perceived ability to perform required exercises in various situations).

At 1 year post treatment, severity of UI was again a predictor of long-term adherence behaviour. Two other variables were independent predictors of long-term adherence: not having sex education at school; and adherence behaviour during treatment.

Predictors of improvement in urinary incontinence

Four studies tested predictors of improvement in UI. 74,89,95,100 Improvement was defined and measured in various ways and at different time points. Studies also presented results for different subgroups, and one study included two regression models. 95

Predictors of urinary incontinence cure

Two studies measured cure (defined as 100% reduction in UI episodes). Wyman et al. 31 found no significant association between rates of cure and severity and type of UI in their RCT of CBIs, at post treatment, or at 3 months post treatment. In women with UUI, Burgio et al. 74 found a positive association between rates of cure and previous surgery; severity of UI as measured by baseline diary (but not as measured by self-report); use of garment protection; and a lower number of years of education.

Other outcomes

Wyman et al. 31 found a positive association between type of UI and QoL measures. At post treatment, women with SUI reported less life impact (IIQ) and women with urge incontinence reported less symptom distress (UDI). No significant associations were found at 3 months post treatment.

Tadic et al. 102 identified history of depression to be a predictor of QoL (as measured by the Urge Impact Scale) in older women with UUI.

Univariate analyses

Thirteen studies report a univariate analysis. 69,70,72,74,78,84,88,89,95,99–102 The number of times variables have been included in a univariate analysis is identified in Table 31. Variables included in the three studies measuring predictors of adherence are presented separately to the ten studies measuring predictors of treatment outcome.

Multivariate analyses

Independent predictors of intention to adhere and treatment adherence were measured in one study. 69,70 Severity of UI and self-efficacy were found to be predictors of intention to adhere, but only severity of UI was also a predictor of adherence at 1 year post treatment, together with lack of sex education at school and treatment adherence behaviour.

Table 32 summarises the number of times predictors have been tested against treatment outcome in a multivariate analysis. Results are summarised for each category of predictor variable in terms of how many studies have included the variable in multivariate analysis, and the results across studies. Associations between variables are described as positive or negative.

Quality of evidence

The previous section identified that many of the variables have only been included in one or two studies, with the most tested variable (severity of UI) included in three studies. Evidence for each predictor variable is therefore relatively weak, but the quality of the study also has to be taken into account in interpreting the strength of evidence. The description of the quality of included studies identified two studies as of reasonably good quality (++),31,100 three studies of moderate quality (+),74,89,102 and one study of poor quality (–). 95 The final table (Table 33) summarises the strength of evidence for each predictor of adherence or treatment outcome of behavioural interventions for UI.

Generalisability of evidence

There was variation in the client groups included in the studies, so variables may only be confirmed predictors in specific client groups. Results for predictors of adherence are generalisable to women with self-reported UI. 69,70 Severity of UI was a predictor of worse outcome in women with SUI or UUI. 74,95 Previous treatment was a predictor of worse outcome in women with UUI. 74,95 Sex, functional and social status variables were predictors of outcome in older housebound people. 89 Fewer psychological problems were also predictors of improvements in UI in older housebound people and improvements in perceived QoL in women. 102 Type of UI was also found to be related to QoL in women. 31

Modelling predictor variable relationships

Although individual studies can provide information about individual predictor variables, none of the studies tested predictive models, so there is little information available about how predictor variables might interact.

Figure 23 summarises the independent predictor variable relationships from at least one study of moderate quality (dotted line). Black lines indicate negative impact, i.e. worse outcomes, while blue lines indicated positive impact. Dark green dotted lines indicate where evidence for the direction of correlation is mixed.

FIGURE 23.

Draft model of predictor variable relationships.

The only correlation to be confirmed in more than one study (illustrated by a solid line) is the link between adherence and improved outcome.

Primary outcome

Results for comparisons with another treatment in the number of people remaining incontinent at post treatment were marginally not statistically significant (RR 0.87, 95% CI 0.75 to 1.01) in two trials of moderate quality. 31,33 Follow-up results had a similar estimate of effect size but were not statistically significant for comparison with another treatment (RR 0.87, 95% CI 0.72 to 1.05). Results for no treatment comparisons were statistically significant favouring the CBI (RR 0.81, 95% CI 0.70 to 0.94) from two trials33,90 (one of moderate quality33), with no measurements available at follow-up.

Secondary outcomes

Results for number of incontinent episodes at post treatment were marginally statistically non-significant (WMD –2.18, 95% CI –4.53 to 0.17) in comparisons with another treatment in three trials,31,33,81 two of which were of moderate quality. 31,33 Follow-up results were also not statistically significant in two trials,31,81 one of these was of moderate quality31 (WMD –1.40, 95% CI –4.59 to 1.79). Results were statistically significant favouring the CBI (WMD –3.57, 95% CI –5.52 to –1.62) in five trials with no treatment comparisons, four of which were of moderate quality. Follow-up results were also statistically significant favouring the CBI (WMD –5.60, 95% CI –9.92 to –1.28) in one trial of moderate quality.

Results for comparison with another treatment were statistically significant favouring the CBI for improvement in UI at post treatment in terms of subject perceptions of improvement (RR 1.42, 95% CI 1.12 to 1.81) in two trials of moderate quality. 31,33 At follow-up to 12 months the effect was of similar magnitude but was not statistically significant (RR 1.43, 95% CI 0.96 to 2.12). In comparison with another treatment, the effect size for 75% or more reduction in incontinent episodes was not statistically significant at post treatment (RR 1.60, 95% CI 0.94 to 2.73) or follow-up (RR 1.32, 95% CI 0.92 to 1.91). The no treatment comparison was statistically significant post treatment favouring the CBI (RR 2.16, 95% CI 1.58 to 2.95), but was not measured at follow-up.

Results for severity of UI for comparison with another treatment were not measured at post treatment or follow-up. No treatment comparison in two trials71,76 (one of moderate quality76) was not statistically significant at post treatment (SE –0.70, 95% CI –2.41 to 1.01) or follow-up (SE –0.60, 95% CI –1.47 to 0.26).

From a single trial,81 the effect for symptoms in terms of urinary frequency was not statistically significantly different from that of another treatment in one trial (SE –0.04, 95% CI –0.70 to 0.62), but was statistically significantly different from the no treatment comparison favouring the CBI (SE –0.55, 95% CI –0.97 to –0.13) in four trials,71,76,91,100 two of which were of moderate quality. 76,100 Results for nocturia were statistically significant favouring the CBI (SE –0.46, 95% CI –0.81 to –0.11) in comparison against another treatment in two trials,33,81 one of which was of moderate quality. 33 Results for nocturia were also statistically significant at follow-up favouring the CBI (SE –0.71, 95% CI –1.39 to –0.03) in comparison with another treatment, in one study of poor quality. 81 Urinary urgency was not measured at post treatment in any trial using comparison against another treatment, but results were statistically significant for a no treatment comparison favouring the CBI (RR 0.57, 95% CI 0.37 to 0.89) in one trial of poor quality. Effect differences for urgency at follow-up were not statistically significant (RR 0.67, 95% CI 0.41 to 1.07).

Results for QoL were not statistically significant for either impact of incontinence or symptom distress in comparison against another treatment at post treatment or at follow-up. For a no treatment comparison, reduction in impact of incontinence on QoL was statistically significant at post treatment favouring the CBI (SMD –0.47, 95% CI –0.80 to –0.14) in one trial of moderate quality,76 and was marginally statistically non-significant at follow-up (SMD –0.36, 95% CI –0.74 to 0.01) in the same trial.

The chance of satisfaction with treatment was statistically significantly different favouring the CBI (RR 1.41, 95% CI 1.18 to 1.68) when compared against other treatments in two trials of moderate quality. 31,33

Results for number of adverse events were also just statistically significant favouring the CBI (WMD –1.20, 95% CI –2.40 to 0.00), with more adverse events in the drug comparison group in one trial of poor quality. 81

In summary, there is evidence that, in comparison against no treatment, CBIs show gains in the number of people cured, objective and subjective measures of the degree of improvement in UI, reduction in some symptoms and impact of incontinence on QoL, but not severity of incontinence or symptom distress.

In comparison with other treatments, CBIs seem to be more advantageous for subjective perceptions of improvement and satisfaction with treatment. CBIs are possibly more advantageous than other treatments in terms of reducing the number of incontinent episodes and nocturia, but there is insufficient evidence for gains in rates of cure, frequency of micturition, or improvements in QoL.

There is evidence that in comparison with usual care, treatment effects on the number of incontinent episodes, the severity of incontinence and the impact of incontinence on QoL, can persist into the longer term. In comparison with other treatments, the magnitude of effect at follow-up is similar to the post-treatment effect, but only achieves statistical significance favouring the CBI for urinary symptoms of nocturia and urgency.

Review of effectiveness: quality of the evidence

In general, results were supported by at least one trial of moderate quality. 73 However, out of the 10 trials, one did not provide data suitable for pooling and two quasi-experimental trials reported some large treatment effects that were not consistent with other trial results. 71,81 Although eight31,33,73,76,87,89,90,100 out of 10 trials reported random allocation, only three provided an adequate description of the procedure;33,89,100 and allocation concealment was judged adequate in only one trial. 100 However, judgements about the quality of trials are based on the trial reports and not on contact with trials authors, so may refer more to the quality of trial reporting. This effectiveness review is the basis for a Cochrane review, and trial authors will be contacted as part of that process.

Although most of the results are supported by at least one trial of moderate quality, there are some inconsistencies in the results, which need exploring. These are in part due to the small number of trials contributing to each outcome, meaning that the direction and magnitude of effect does not necessarily show consistency across no treatment and another treatment comparisons, or across post treatment and follow-up results for the same outcomes. This may be in part due to the fact that different trials contribute to the comparisons, with variation in trial quality, client groups and timing of outcome measurement. The statistical significance of an effect is often variable across different outcome measurement time points, whereas the effect size is relatively similar.

Rates of cure, number of people achieving 75% of more reduction in incontinent episodes and number of incontinent episodes were all sourced from the same data in bladder diaries. In comparing CBIs against another treatment, the difference in the number of incontinent episodes and the number of people cured were marginally statistically non-significant, whereas the number of people gaining 75% or more reduction was not statistically significant. However, the general pattern is for statistically significant difference favouring the CBI in the number of incontinent episodes when compared against no treatment, but marginal or statistically non-significant differences of smaller size when compared against other treatments.

Subject perceptions appear to be more positive than relatively more objective data such as bladder diaries. One explanation might suggest increased contact with a health professional affecting subjective perceptions. However, this explanation would not be consistent with the evidence. The review of PFMT identifies average time of intervention as 8–12 weeks,29 whereas the average time of CBI delivery is also 8–12 weeks. 79 There could be increased intensity of contact within the programme duration, but the PFMT review identified weekly clinic contacts as the average, whereas CBI interventions had slightly more than weekly contact, on average. Alternatively, the differences seen could be an artefact of comparing severely right-skewed distributions for which the SDs are very large relative with the means (effectively making the mean for number of incontinent episodes insensitive to change).

Results for urinary symptoms are at times inconsistent and also subject to considerable heterogeneity, which is in some part attributable to the results from two quasi-experimental trials. 71,81 For urinary frequency, effects are heterogenous for no treatment comparisons at post treatment, mainly due to the large treatment effects from one quasi-experimental trial. 71 The inconsistency in statistically significant results favouring the CBI for nocturia in comparisons with another treatment, but not in no treatment comparisons, could be because the significant difference is sourced only from a small quasi-experimental trial. 81 However, there is also a high degree of heterogeneity in the treatment effects of the no treatment trials, attributable to large treatment effects from Aslan et al. 71 as before, but also results favouring the control group from two trials. 90,100 In both trials, this was attributable to a rise in nocturnal micturitions in the treatment group. Neither author offers a possible explanation. Last, only one quasi-experimental trial contributes to the statistically significant treatment effects favouring the CBI seen for urinary urgency. 71 Owing to the inconsistency and heterogeneity of treatment effects without plausible clinical explanation, results for urinary symptoms therefore appear unreliable.

Results for QoL are more consistent, in that there is no statistically significant difference seen in either impact of incontinence or symptom distress on QoL for two trials using comparisons against another treatment,31,81 but results for impact of incontinence are statistically significant favouring the CBI in one trial using comparison with no treatment,76 and marginally statistically non-significant with smaller effect size at follow-up for the same trial. However, there was significant heterogeneity of treatment effect observed for both impact of incontinence and symptom distress, in some part attributable to large treatment effects observed in one quasi-experimental trial. 81 Owing to the small number of trials contributing data and the heterogeneity of treatment effects, results for QoL also appear unreliable.

Results for satisfaction with treatment are based on two trials of moderate quality,31,33 but results for adverse effects are based on only one quasi-experimental trial,81 so should be interpreted with caution.

Review of effectiveness: overall completeness and applicability of evidence

The results would suggest that, in terms of effectiveness immediately after treatment, the choice between CBIs and other treatments (mainly drugs or single interventions) is based on greater subject perceptions of improvement, greater satisfaction with treatment, and the potential for more people to achieve greater levels of improvement (75% or more) in UI. Results for symptom reduction or improvements in QoL are inconsistent. We found no trials that included a comparative assessment of the cost of delivering CBIs compared with other interventions, but the contact time and patterns do not appear significantly different to PFMT trials.

The results for effectiveness over the longer term are less clear, with no clear benefit over other treatments, other than in symptom reduction, where the evidence is potentially unreliable. The effect size estimates at follow-up are generally consistent with post-treatment effect sizes, with only a small estimated reduction over time. However, the small number of trials and the smaller sample size at follow-up makes interpretation very difficult.

Few trials used objective measures of UI, such as a pad test for grams of urine lost per day. Most trials included the number of incontinent episodes, so could report degree of improvement and number of people cured, but only three trials reported on the primary outcome of cure. Requests for further outcome data were made to some authors, with no success.

Overall, the results are more relevant to people with UUI or MUI, as only 14% of participants were classified as having genuine SUI only. In the main, results are applicable to people with established, moderate to severe UI, with only one trial including people with mild UI. 90 The results are also mainly applicable to women, with only a few men included in one trial. 89 Participants had a mean age > 55 years in 9 out of 10 trials, > 60 years in eight trials and > 70 years in three trials. Most participants were cognitively able.

The review documents the degree of diversity in the content, main focus, duration and intensity, enhancement, and allocation of interventions, and in what they are compared against. However, there is sufficiently similar content in terms of main components and methods of delivery that CBIs can be easily differentiated from single interventions and their results feasibly combined. However, there are not sufficient trials using the same outcome measure in any one comparison of main features to clarify what might be the dominant mechanisms of effect. None of the subgroups of intervention characteristics showed any significant difference and, in truth, it was difficult to allocate interventions to such broad brush categories as basic or enhanced delivery. The benefit of such an in-depth description of intervention content may lie more in delineating potential mechanisms of action to be tested in future trials.

In terms of acceptability and safety of the intervention, there was no evidence of adverse effect, although few trials explicitly monitored them. 33,81 Although trials have mostly been conducted in US settings,31,33,73,76,89,90,100 there is no reason to think that they might not be acceptable to either staff or clients in the UK, or feasible in a UK health-care context. However, only one trial used a truly community derived sample,76 and most participants were self-selecting after being in contact with health-care services. Three trials included people who were older and home-bound or in nursing homes,71,73,89 so interventions seem feasible with this client group, although loss to follow-up was higher, and there were aspects of the interventions that were not fully accepted, such as the internal examination for PFMT.

Review of effectiveness: potential limitations of the review process

For the purpose of gaining an overall picture, the review structure has leaned towards combining the results of trials, but the degree of heterogeneity in treatment effects for some outcomes suggests that the pooled results may be unreliable. Subgroup analyses were only possible on one outcome, and were in the main not statistically significant, so did not provide much information about potential sources for the observed heterogeneity. However, the explanation is likely to be in the combination of differences in client group, intervention type, comparison group, setting and study quality in the included trials. The detailed description of the included interventions may help to detect potential mechanisms of action for exploration in future trials.

The alternative to pooling is to consider each trial in isolation, and to undertake narrative review for each outcome. In effect, the review has also done this, in terms of a detailed description and comparison of the interventions, and narrative consideration of likely sources of heterogeneity for each outcome. Another alternative would be to structure the review comparisons differently; in particular, to use types of incontinence as subgroups rather than type of comparison group. However, even though the sensitivity analysis for type of comparison group was not statistically significant and there is therefore no evidence for not pooling results, the review group prefers to maintain the distinction. Only three trials offer the potential to analyse results by type of incontinence,31,33,81 and the effect of this has been checked in a subgroup analysis.

The choice of primary outcome measure was a matter for considerable discussion, as the review group felt that cure was not necessarily the most appropriate outcome to be expected of a behavioural intervention, although it might be the most desirable. Behavioural interventions were thought more likely to incur improvement in continence and urinary symptoms. There was also discussion about whether or not subject perceptions of improvement should be preferenced over relatively less subjective measures such as bladder diaries. If subject perceptions had been chosen as the primary outcome, CBI would be viewed as successful, whereas less benefit is observed in more objective measures such as number of incontinent episodes, or number of people regaining continence. The measure of cure was eventually chosen as the primary outcome because it is the end target of treatment for urinary continence, and as such is a gold standard against which all treatments can be compared. There remains the potentially valid criticism that we may have chosen the wrong outcome measure as primary, and that an alternative comparison structure should have been used.

The choice of post-treatment timing and up to 12 months follow-up are not perhaps the most desirable choice, as long-term continence would be the most valuable end point. However, few trials measure longer-term outcomes, and those that do measure at variable time points. Post-treatment scores were thought likely to be the most comparable. In this review, a number of trials measured medium-term (i.e. up to 12 months) outcomes,76,89,90 but only three trials provided outcome data past 12 months follow-up. 76,81,90

Client views

Six studies of the experiences of clients were found: five qualitative studies (n = 105)79,83,92–94 and one postal survey (n = 79). 80

Factors identified by at least one study of moderate quality that could act as a barrier to continence promotion for clients included:

• perceptions of UI as a problem and level of tolerance for symptoms

• preference for interventions that promote independence and avoiding reliance on staff in residential care

• increased fear of being wet with BT, negative attitudes to and experiences of PFMT and difficulty of doing exercises properly

• competing interests/demands, and difficulty developing routines and fitting them into daily life

• convenience and cost of treatment

• delivery mode (e.g. group vs. individual).

Factors identified by at least one study of moderate quality that could act as an enabler to continence promotion for clients included:

• having realistic goals and expectations, and gaining a sense of mastery and control if successful

• adapting daily routines to include PFMT with feedback on correct performance of exercises and professional follow-up

• availability of an accessible and clean bathroom in residential care.

Additional suggestions by researchers to facilitate treatment choice and adherence included:

• considering the timing, siting and labelling of interventions, and the training of staff

• eliciting and honouring client preferences and their goals and expectations for treatment

• including an assessment of self-efficacy and barriers to uptake of treatment in the initial stages

• the provision of adequate written information

• consideration of the use of strategies such as reminders or motivational intervention, with early follow-up and alternative options for people who fail to attend

• consideration of group teaching as a strategy.

Staff views

Six studies of the views of nursing staff were also included about continence care including behavioural strategies in general (n = 273);75,77,80,98 or about involvement in PV interventions (n = 154). 86,97

Factors identified by at least one study of moderate quality that could act as a barrier to continence promotion for clients included:

• views of staff about the aims of treatment

• likelihood of continence assessment affected by routes of admission or referral

• lack of suitable assessments and involvement of the multidisciplinary team (MDT)

• staff motivation and education

• conflicting work priorities, lack of staff, requirements of manual handling, scheduling conflicts

• perceptions of treatment effectiveness.

Factors identified by at least one study of moderate quality that could act as enabler to continence promotion for clients included:

• education

• teamwork, adequate staffing, methods of work allocation

• sufficient and appropriate equipment and supplies

• experience of success.

Specific suggestions relating to enablers for the involvement of NAs in the promotion of continence included:

• regular allocation of clients

• inclusion in planning and reporting of care

• increased accountability for adherence, more autonomy and freedom to prioritise

• recognition and reward for contribution

• identification of role models.

Additional suggestions by researchers to facilitate continence promotion by staff included:

• changing philosophy of care away from accepting continence

• the use of self-efficacy-based motivational interventions for both staff and clients

• targeting clients most likely to benefit

• providing realistic expectations of outcome to staff

• supervision and monitoring or staff performance

• incentives to adherence to behavioural programmes

• technology to support continence care (e.g. bladder scanners).

Narrative review of acceptability and feasibility: quality of the evidence

The quality of studies was mixed, with the potential for bias originating in the main from poor description of methods of data analysis, or methods of testing the robustness of the findings. However, most of the qualitative studies were descriptive rather than in-depth studies, so the identification and listing of potential barriers and facilitators to the delivery or uptake of behavioural interventions is unlikely to be problematic, in terms of interpretation or synthesis. Only a small number of data was extracted from the descriptive component of the two surveys,80,97 such as the frequency of agreement with barriers to the use of behavioural methods for continence promotion, or additional barriers identified in free-text responses. However, most of the findings were supported by at least one study of moderate quality.

Narrative review of acceptability and feasibility: overall completeness and applicability of the evidence

Although there were 12 studies included in the narrative analysis focusing on broadly the same topic, there was relatively little overlap between their specific focus, with three studies of treatment choice;80,93,94 three of treatment adherence;79,93,94 two of treatment withdrawal,83,92 two of PV implementation;86,97 two of continence promotion in LTC;80,94 and one in acute care. 75 Only one study79 collected detailed information about clients’ experiences of a particular behavioural therapy (i.e. PFMT), so there is in fact very little in-depth exploration of client responses to behavioural therapies.

The available evidence for clients relates to cognitively able women in the main, but apart from that similarity, the samples differed. Two of the studies were restricted to women with SUI,79,92 and the samples in these two studies were younger. Two of the studies related to older women in residential care,80 and one study was a community sample. 93 The mix of samples and the different focus of the studies means that the barriers and enablers identified should be viewed as context specific, rather than generalisable to any client group or setting.

There is no information relating to client experiences of behavioural therapies other than PFMT, and none of the studies were longitudinal, so not much is known about how views might change and develop over time. There is very little information about particular subgroups, for example those with severe or mild incontinence, or about men’s experience of behavioural therapies.

The available evidence for staff all relates to nursing staff, predominantly those working in long-term residential care, although one study was undertaken in acute care. 75 The samples of four of the studies included a mix of grades from NA through to charge nurse and director of nursing. The identified barriers and enablers related in the main to the direct delivery of care and although two studies did take a rather wider view that included some system features, there is not a strong whole systems approach overall. The focus on behavioural therapies in the general studies of continence care was rather small, with the impression being that nurses did not know much about them. There were no studies in a community setting, and none relevant to the delivery of a specific behavioural therapy other than PV.

Narrative review of acceptability and feasibility: limitations of the review process

The synthesis of qualitative data is not a straightforward mechanical process, and some issues were encountered that need to be taken into consideration when reading the synthesis.

Review of predictors of adherence or outcome: quality of the evidence

Only two studies were of good quality,31,100 and these two studies only measured the two major variables of type and severity of UI, and did not confirm either variable as an independent predictor of improvement, although type of UI was correlated to QoL in one study. 31 The remaining studies were of moderate quality overall, although all had some weakness in the definition or measurement of predictor or outcome variables, and none had adequate blinding of assessors.

As most of the predictors have only been confirmed in one study, the results need to be treated with caution. In particular, results for prior treatment or severity of UI are mixed, and are probably related to the type of UI. The only result that can be viewed with any level of confidence is the relationship between adherence and treatment outcome.

Review of predictors of adherence or outcome: completeness and applicability of the evidence

Although many variables have been included in univariate analyses, the majority of variables tested in multivariate analysis have been physiological, with few studies including social or psychological variables. Only three studies have included mental or cognitive status,89,95,102 and only one study has considered functional status. 89

The studies also cover different populations, with one study specific to older housebound people, and the remainder specific to either older versus younger women, or women with different types of UI. Women made up the majority of the study samples, with only a few males included in one study. 89

Setting

An 18-bedded acute stroke unit in a large trust serving a population of 370,000 and with teaching and foundation trust status.

Subjects and sampling

Patient inclusion criteria:

1. aged ≥ 18 years with a diagnosis of stroke based on the World Health Organization (WHO) criteria138 (no upper age limit)

2. UI as defined by the ICS139 as ‘involuntary loss of urine’

3. conscious (defined as a Glasgow Coma Score140 of ≥ 12)

4. medically stable as judged by the clinical team

5. incontinence classified as SUI, UUI, MUI or ‘functional’ UI. 141

Health professional and clinical leader inclusion criteria:

1. health professionals either delivering the intervention or linking with the intervention in any capacity.

Participant consent process

All patients admitted to the unit during the intervention period were screened for eligibility; patients were recruited as early as possible following admission. Informed consent to collect baseline and outcome data was sought from all patients meeting the inclusion criteria. Nursing staff asked these patients whether or not their name could be given to the research team. If the patient agreed, a member of the research team visited the patient to explain the project, answer any questions the patient (and their families/carers) had and provided a participant information sheet (see Appendix 16). Patients were given at least 24 hours to consider participation and were visited by a member of the research team after this period; patients choosing to participate signed the consent form at this stage.

For patients unable to consent for themselves, a person able to advise on the presumed wishes of the patient was approached to act in the role of consultee. This is in line with the recommendations of the Mental Capacity Act 2005 (www.legislation.gov.uk/ukpga/2005/9/contents). 142

Health professional and clinical leader consent process

Health professionals and clinical leaders were identified by the research team in collaboration with senior nurses, medical staff and other health professionals working in the stroke service. Health professionals and clinical leaders were approached by a member of the research team. An initial appointment was made with each potential participant, where the study was explained and a participant information sheet provided (see Appendix 17). Health professionals and clinical leaders indicated they may like to take part by returning a card to the research team. A further appointment was made with those indicating they might like to take part, where signed consent was obtained.

Identifying organisational context for embedding the systematic voiding programme

A soft systems approach143 was used to identify systems affecting the likelihood of the SVP becoming embedded in mainstream stroke care. The stroke pathway can be conceptualised in terms of the patient’s journey through constituent services which vary in purpose, structure, location and workforce. The patient’s experience of UI is similarly longitudinal, and has the capacity to span the entire pathway. This poses a challenge to achieving continuity in all aspects of incontinence care and associated patient management, which is similarly multifaceted drawing on different interventions delivered by different professional groups in different contexts. These contexts include, among others, different settings (acute stroke unit, rehabilitation unit, community, care home) and different practice paradigms (emergency/acute, rehabilitation, continuing, palliative).

The ICONS study clinical intervention can be conceptualised as one component of a whole incontinence care system and our aim was to ensure that the intervention could be embedded within this. The first step was to understand the incontinence care system. This information was used to inform subsequent stages of the research which focus on implementation of the new intervention, and the degree to which it is embedded in systems and clinical practice. To ensure this phase of the research programme was manageable within planned resources, the analysis focused in detail on the stroke unit and primary discharge destinations, and considered other components of the pathway in general.

In systems theory, services are considered as complex human activity systems143 which can be understood in terms of the relationships between their structure, process and outcomes. The aim of systems analysis was to describe relationships and use them to generate a definition of how the service or, as in this context, groups of services worked. This is known as the root definition of the service and describes the relationship between six factors:

• customers (patients and their family carers)

• actors (providers of UI care)

• transformations (key changes occurring as a result of the service)

• world view (the value system and policy context in which the service operates or the justification for the service)

• owners or stakeholders of the system (agencies involved in commissioning the service)

• environment (local conditions that enable or limit UI care).

Although not a hypothesis to be tested in this research programme, understanding the whole system within which the intervention was delivered may have enhanced the degree to which implementation was successful. In addition, when led by theory, this ‘diagnostic analysis’ information may help in the design of transferable implementation strategies by highlighting barriers and facilitators to implementation, workforce education and training issues, and process factors such as requirements for clinical documentation.

A purposive sample of staff engaged in managing and delivering the incontinence systems were approached for inclusion in this study component. Subjects were selected to ensure breadth of coverage in terms of the range of health professionals who provide UI interventions in both primary and secondary care.

The aim of the systems analysis was to evaluate CATWOE (customers, actors, transformations, worldview, ownership and environment) elements and use them to generate a definition of how continence provision worked as a clinical system, highlighting the barriers and facilitators that were anticipated to impede or enhance implementation of the SVP. Qualitative data relating to the evaluation of organisational context were analysed using a directed content analytic approach. 144 This involved drawing deductively on categories derived from pre-existing theory to guide the analysis of interview data within the soft systems framework. 145 The purpose of analysis was to develop new and more specific insights into implementation context by extending the theoretical position within the particular clinical focus of the case study.

Practically, interviews were coded using the major concepts of the framework. Each data set was analysed by two people independently. At regular stages of the analysis, agreement was explored, disagreements discussed and coding frame guidance revised accordingly as more nuanced understandings of the concepts in the study context emerged. During the final stages of analysis, a site summary was created by merging findings across interviews, linking related text chunks together for key concepts and condensing to remove overlap.

Stroke unit staff views of embedding the systematic voiding programme

Six audio-taped focus or group interviews were conducted with health professionals involved in delivering the programme at 1 to 2 monthly intervals throughout the case study. Interviews explored views of the behavioural approaches and their acceptability and feasibility, and the perceived impact of different components of care (e.g. focused education, individualised voiding regimens).

Normalisation process theory55–58 provided the theoretical framework for implementing and evaluating the SVP. The model is designed to facilitate understanding of the practical issues involved in embedding complex interventions into routine practice (e.g. ease of use and integration), and has been used in a range of settings. 55,146 It provides a theoretical framework for both implementation and its evaluation and our intervention provided a good fit with May et al. ’s definition of complex interventions as comprising ‘multiple behavioural, technological, and organisational components’. 57 In addition, the model’s view of change as resulting from collective, rather than individual, action57 is in line with our aim of bringing about change through group activity. The NPT also has a similar emphasis on organisational context. It comprises 16 dimensions in four categories, illustrated in simplified form147 in Table 36.

Initially, focus or group interviews were coded using the NPT framework by two people independently; at the completion of coding each interview, agreement was explored, disagreements discussed and coding frame guidance revised accordingly (see Appendix 18). Multiple comments relating to the same point were condensed for each NPT dimension to remove overlap and redundancy. Comments were labelled by respondent to facilitate identifying any variation in findings by staff grade (i.e. HCA for health-care assistant, or Q for qualified staff).

Second, a site summary was created by merging the findings across the six interviews, linking related text chunks together for each NPT dimension and condensing down to remove overlap and redundancy, taking care to avoid loss of meaning or viewpoint. Findings were categorised and reported as either barriers or facilitators to implementation of the SVP. Implications for the trial phase were built from these by the research team and discussed at Steering Group meetings. These were used to refine the SVP intervention, facilitation manual and implementation plan.

Patient outcome and factors affecting discharge continence status