Sections of this chapter have been reproduced from Wickramasekera et al.109 This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY 4.0) licence, which permits others to distribute, remix, adapt and build upon this work, for commercial use, provided the original work is properly cited. See: https://creativecommons.org/licenses/by/4.0/. The text below includes minor additions and formatting changes to the original text.
Methods
Design and theoretical/conceptual framework
Discrete choice experiments are an attribute-based measure of benefit; they assume that healthcare interventions, services or polices can be described by their attributes. In a DCE, respondents are required to make trade-off decisions regarding the quantity or quality of a good or service. The resulting choices are analysed to estimate the overall utility (value) and willingness to trade between services. The use of DCEs to identify patient preferences in health and healthcare has increased. Where no clear treatment decision exists, accurate quantification of patient preferences for risks and benefit is crucial.
Development of survey
The survey contained four separate sections. These were:
- patient characteristics and disease history
- a treatment ranking exercise
- DCE-specific tasks
- survey feedback questions.
The ranking exercise and DCE tasks were developed by conducting qualitative interviews with 20 patients. The patients were asked to identify the key attributes and levels that they considered when choosing a treatment (see Chapter 4 for further information). A thematic analysis of the interview data was conducted and identified a list of factors that patients considered important when making treatment decisions. PPI members and clinicians reviewed an initial list of themes and selected attributes considered the most important for inclusion in the DCE and ranking task.
Nine attributes were included. These were:
- type of excision and closure
- type of anaesthetic
- length of hospital stay
- wound care
- pain medication requirement
- infection risk
- healing time
- risk of recurrence
- scarring.
Currently, there are 18 theoretically possible surgical options for PSD.14,110 Clinicians developed a treatment classification of five treatment groups. This was considered important due to the potential cognitive burden of ranking 18 treatment categories. The treatment categories, related descriptions and attributes were informed by the literature, clinical input and PPI piloting (see Appendix 3, Box 1).
Two attributes were included in the DCE: risk of infection/persistence and recovery time. This selection was based on two reasons. Firstly, these two attributes were assessed as most important by patients and clinicians. Secondly, DCE attributes must be independent to avoid presenting implausible combinations of attribute and level profiles. The levels for the DCE attributes were selected based on plausible values published in the literature and additional input from clinicians.
The choice tasks were constructed based on an orthogonal design using a design catalogue.111 The survey contained 16 hypothetical DCE tasks. Participants were asked to choose between two combinations of outcomes with varying levels (see Appendix 3, Box 1). Forced unlabelled choices were presented – ‘treatment A’ or ‘treatment B’. An ‘opt-out’ alternative was not provided for the purposes of realism. A dominant task was included – where one treatment option is logistically superior – to test participant understanding of the task (see Appendix 3, Box 2).
Patient and public involvement
Patient and public involvement representatives were heavily involved in the design and implementation of the DCE. They assisted with the following tasks:
- Reviewed the initial list of themes to finalise the attributes.
- Reviewed the acceptability of the treatment category descriptions.
- Prior to the dissemination of the survey, three PPI representatives reviewed the survey to assess comprehensibility, interpretation and complexity of tasks.
Sampling
Discrete choice experiment sample sizes can vary from 100 to 1000 plus.112 An Orme113 rule of thumb formula was adopted to estimate the minimum sample size. Using this formula – {N = 500 × [4 (maximum number of levels)]/[2 (number of alternatives) × 16 (number of tasks)]} – the estimated sample size was 63.
Recruitment
All participants aged 16 years and above with symptomatic PSD, referred for elective surgery and participating in the PITSTOP cohort study, were invited to take part in the survey. Interested participants were e-mailed a Qualtrics® (Qualtrics, Provo, UT, USA) link which included a participant information sheet, consent form and the questionnaire. Participants with symptomatic PSD and not participating in the PITSTOP cohort could also take part in the DCE by accessing a digital QR code advertised on a study leaflet. The study leaflet was displayed in NHS Foundation Trust colorectal outpatient clinics and disseminated via the PITSTOP Twitter website.
Data analysis
Descriptive statistics were calculated for patient characteristics, disease history and survey feedback variables. DCE responses were modelled using conditional logistic regression where the dependent variable was the preferred treatment choice, and the independent variables were risk of infection/persistence and recovery time. Linearity of the attributes was assessed before deciding to treat risk as a linear variable (Figure 6). Regression coefficients were used to estimate the relative importance of attributes. Maximum acceptable risk (MAR) is the rate at which patients are willing to sacrifice a benefit of one attribute in exchange for an improvement in another. MARs were calculated by dividing the ratio of recovery time coefficients by the infection/persistence coefficient and 95% CIs calculated using the delta method.114 Latent class models were used to analyse individual heterogeneity and to identify subsets of patients with varying preferences. The optimal number of classes was selected using the Bayesian information criterion (BIC) and consistent Akaike information criterion (CAIC) and model parsimony. Data were analysed using Stata® 17 (StataCorp LP, College Station, TX, USA).
FIGURE 6 Regression model 1 results reproduced in a diagram. (Figure not reproduced here; see the original article.)
Results
Participants
One hundred and eleven participants were included in the DCE survey. The completion rate was 74% (423 unique visitors entered the survey, 150 participants consented to take part and 3 participants declined). Table 16 reports the characteristics of the 111 included participants. Of these, 75 (68%) respondents were male and 73 (66%) were between the ages of 17 and 29 years; 89 respondents (80%) were employed and 97 (87%) were white. Except for six respondents, the rest of the sample reported having had at least one surgery for PSD. The respondents had various types of surgeries, including excision of the skin and closure of the wound with stitches (26%); excision of the sinuses only and leave the wound open to heal (23%); excision of skin and leave the wound open (23%); excision of the skin and closure of the wound with a skin flap and stitches (9%); and excision of the sinuses and closure of the wound with glue (19%).
| N = 111 | No. | % |
|---|---|---|
| Sex | ||
| Male | 75 | 68 |
| Female | 36 | 32 |
| Age (years) | ||
| 17–29 | 73 | 66 |
| 30–39 | 24 | 22 |
| 40–49 | 8 | 7 |
| 50–59 | 4 | 4 |
| 60–69 | 2 | 2 |
| Median age, years (range) | 28 (17–65) | |
| Which of the following best describes your main activity? | ||
| Employed | 89 | 80 |
| Retired | 1 | 1 |
| Homemaker | 3 | 3 |
| Carer | 1 | 1 |
| Student | 17 | 15 |
| Ethnicity | ||
| White | 97 | 87 |
| Black | 2 | 2 |
| Asian | 2 | 2 |
| Mixed | 10 | 9 |
| Education | ||
| Primary | 4 | 4 |
| GCSE | 16 | 14 |
| A-level | 48 | 43 |
| Degree | 42 | 38 |
| Prefer not to say | 1 | 1 |
| Previous pilonidal sinus surgeries (including both emergency drainage and previous ‘definitive’ elective repair) | ||
| 0 | 6 | 6 |
| 1 | 67 | 64 |
| 2 | 17 | 16 |
| 3 | 6 | 6 |
| 4 | 5 | 5 |
| 5 | 2 | 2 |
| 6 | 1 | 1 |
| 10 | 1 | 1 |
| Type of previous pilonidal sinus surgeries | ||
| Excision of skin and leave the wound open (e.g. leave open/marsupialisation) | 29 | 23 |
| Excision of the skin and closure of the wound with stitches (e.g. midline closure, Bascom’s cleft closure, Karydakis) | 33 | 26 |
| Excision of the skin and closure of the wound with a skin flap and stitches (e.g. rhomboid, Limberg, Dufourmental) | 11 | 9 |
| Excision of the sinuses and closure of the wound with glue | 24 | 19 |
| Excision of the sinuses only and leave the wound open to heal (e.g. pit picking, EPSiT, laser) | 29 | 23 |
Patient preferences
Appendix 3, Table 34 presents the regression modelling results of the DCE. Patients preferred treatments with lower risk of infection/persistence and this attribute was modelled linearly (see model 2 in Appendix 3, Table 34). Risk of infection/persistence was the most important attribute when patients were choosing a treatment, with an attribute importance score of 70%. Patients also preferred treatments with shorter recovery time; for example, compared to a treatment that takes 12 weeks to recover, a treatment with a 1-, 2- or 6-week recovery period was preferred. However, their preferences were not linear, so this attribute was modelled as a categorical variable. Treatments with shorter recovery time had an attribute importance score of 30%.
Maximum acceptable risk
When choosing a treatment, patients were willing to make trade-offs between risk of infection/persistence and recovery time (Table 17). These trade-offs were measured using MAR, which is the maximum risk of infection/persistence participants are willing to accept to have a treatment with faster recovery times. The highest-risk patients were willing to accept was a 17.08 risk of infection/persistence in return for a treatment with 2-week recovery period compared to a treatment with 12 weeks recovery period. Patients were willing to accept a 10.49 increase in risk of infection/persistence to have a treatment with 6-week recovery period compared to 12-week recovery period. Patients were willing to accept a 6.59 increase in risk of infection/persistence to have a treatment with a faster recovery period (2 weeks compared to 6 weeks).
| Treatment benefit (for selected level changes) | MAR of infection/persistence | 95% CI calculated using the delta method |
|---|---|---|
| Recovery time reduction from 12 weeks to 2 weeks | 17.08a | 14.83 to 19.33 |
| Recovery time reduction from 12 weeks to 6 weeks | 10.49 | 8.76 to 12.22 |
| Recovery time reduction from 6 weeks to 2 weeks | 6.59 | 4.88 to 8.30 |
Preference heterogeneity
Differences in preferences between patients were explored using the latent class modelling approach. This identified two groups of respondents with different preferences (Table 18). The first subgroup of respondents (class 1) were risk-averse and so they were only willing to accept a small risk (1.51–2.15) in exchange for a treatment with faster recovery time (Table 19). The second subgroup of respondents (class 2) showed stronger preferences for treatments with shorter recovery time: they were willing to accept higher risks of infection/persistence (22.35–34.67) to receive treatments with quicker recovery time. Of all the demographic variables used to predict whether a respondent belonged to class 1 or 2, only age of the respondents was statistically significant (see Table 18). The results show that respondents in the 17–29 age group were more likely to belong to class 1 and respondents above the age of 30 were more likely to belong to class 2.
| Class1 | Class2 | |
|---|---|---|
| Week = 12 (reference level) | ||
| Week = 1 | 0.159(0.405) | 2.192***(0.209) |
| Week = 2 | 0.929(0.488) | 2.861***(0.214) |
| Week = 6 | 0.652(0.363) | 1.844***(0.171) |
| Risk (%) | −0.432***(0.049) | −0.083***(0.008) |
| Class membership | ||
| Sex = female | −0.280(0.469) | Reference |
| Age = 17–29 years | 1.365**(0.486) | |
| Activity = employed | 0.375(0.564) | |
| Number of non-emergency surgeries patients had for pilonidal sinus | −0.008(0.163) | |
| Education = degree | 0.450(0.456) | |
| Constant | −1.485*(0.692) | |
| Observations | 3328 | |
| Log-likelihood | −569.55 | |
| BIC | 1206.51 | |
| CAIC | 1220.51 | |
| Treatment benefit (for selected level changes) | Subgroup 1 (class 1) MAR of infection/persistence (95% CI) | Subgroup 2 (class 2) MAR of infection/persistence |
|---|---|---|
| Recovery time reduction from 12 weeks to 2 weeks | 2.15 (−0.03 to 4.34) | 34.67 (28.24 to 41.10) |
| Recovery time reduction from 12 weeks to 6 weeks | 1.51 (−0.11 to 3.13) | 22.35 (17.35 to 27.35) |
Ranking of treatments
Patients ranked the treatments they preferred in order of importance, and Figure 7 shows the results of this ranking task. The best preferred treatment was complex flap (e.g. Limberg, Dufourmental) procedures (27%), followed by excision of the sinuses only (22%), glue (19%), excision of the skin and closure of the wound with stitches (18%) and lastly leave open (14%). The least preferred treatment was leave open (35%), followed by glue (23%), complex flap procedures (18%), excision of the sinuses only (17%) and excision of the skin and closure of the wound with stitches (7%).
FIGURE 7 Ranking of treatments. (Figure not reproduced here; see the original article.)
Participants’ understanding of and engagement with the survey
Table 20 reports the results of the questions used to test the internal validity of the DCE.
| Internal validity (dominance questions) passed or failed? | N | % |
|---|---|---|
| Failed | 18 | 16 |
| Passed | 93 | 84 |
| Always choose the same side (e.g. left profile) in all the DCE questions? | ||
| Yes | 0 | 0 |
| No | 111 | 100 |
| I found the ranking question made sense – please tell us how strongly you agree | ||
| Strongly disagree | 3 | 3 |
| Disagree | 2 | 2 |
| Uncertain | 10 | 9 |
| Agree | 60 | 54 |
| Strongly agree | 36 | 32 |
| I understood the questions about making choices between different treatment options | ||
| Strongly disagree | 1 | 1 |
| Disagree | 1 | 1 |
| Uncertain | 8 | 7 |
| Agree | 59 | 53 |
| Strongly agree | 42 | 38 |
| When choosing options, I needed more information than was provided | ||
| Strongly disagree | 16 | 14 |
| Disagree | 43 | 39 |
| Uncertain | 24 | 22 |
| Agree | 21 | 19 |
| Strongly agree | 7 | 6 |
| I found making a choice between different treatments confusing | ||
| Strongly disagree | 21 | 19 |
| Disagree | 56 | 50 |
| Uncertain | 15 | 14 |
| Agree | 15 | 14 |
| Strongly agree | 4 | 4 |
| Median time to complete survey, minutes (range) | 12 (4–5388) | |
Most of the respondents said that they understood the DCE tasks (91%) and ranking task (86%). Ninety-three (84%) of the respondents correctly answered the DCE question with a logically correct answer (dominance test). None of the respondents were always choosing the same side (left or right) profiles of the DCE tasks. Fewer than 25% said that the DCE task was confusing and that they needed more information.
Discussion
This substudy assessed patient treatment preferences for PSD. When choosing a surgical treatment, patients prioritised risk of infection/persistence relative to recovery time. However, patients were willing to compromise and accept treatments associated with varying degrees of greater risk of infection/persistence in favour of treatments that were associated with quicker recovery. The results provide insight into the type of treatments patients accept and value. In the overall group, patients were willing to accept up to a 17-percentage-point increase in risk of infection/persistence for treatments with a shorter recovery period. This suggests that some patients are willing to accept less invasive treatments with shorter recovery periods and greater risk of infection/persistence (e.g. glue and/or pit picking) over more invasive treatments with longer recovery periods but reduced risk of infection/persistence (e.g. an excise-and-leave-open procedure). In the ranking task, similar results were found: open surgery was ranked as the least favoured treatment option. This is understandable given the impact of prolonged wound care management on psychosocial well-being.115
The results demonstrated preference heterogeneity, which indicates the importance of providing treatments tailored to subgroups of patients with distinct preferences. Patients aged 30 years and over were willing to accept up to a 35-percentage-point increase in risk of infection/persistence for treatments with a shorter recovery period. This suggests that patients within this age bracket would be likely to reject an excise-and-leave-open procedure in favour of a treatment associated with a faster recovery period. In our sample, this age demographic reported that they were either homemakers, retired or had caring responsibilities. Therefore, it is plausible that their personal circumstances may have influenced their preference for a treatment associated with a shorter recovery time. In contrast, patients aged 17–29 years were more risk averse and were only willing to accept a two-percentage-point increase in risk of infection/persistence for treatments with a shorter recovery period. The differences in preference heterogeneity further support the tenet that patients should be involved in making decisions about their surgical care to avoid treatment DR.30 Literature exploring SDM in PSD is growing.30,74,116 Such studies have highlighted the importance of providing patients with sufficient information for each available surgical procedure (including wound care management) to aid treatment decision-making.30,74,116
This substudy is the first to conduct a DCE to assess PSD patient treatment preferences. A robust methodology was employed; qualitative interviews were conducted to inform the development of the survey; experienced clinicians and PPI representatives contributed throughout the design and implementation phases. Relevant interval validity checks were made and identified task comprehension, supporting confidence in the results.
However, this substudy has limitations. The sample size was sufficient to estimate overall modelled preferences, but a larger sample size would have allowed greater confidence in the analyses classifying members to different latent classes. During recruitment, several methods were employed to increase response rate. As PSD typically affects a young, working-age population, engagement barriers can be incurred.117 Currently, there is no consensus on how PSD should be classified.110 The clinicians developed five treatment categories for the ranking task based on their own clinical experience and the literature. A different team of clinicians may have made different classification decisions. In addition, the DCE included two key attributes to avoid presenting implausible combinations of attribute and level profiles. However, in a real-life context, patients may consider other factors not included in the DCE to support treatment decision-making. Finally, 16% of participants failed the internal validity test, demonstrating that the risk information presented was not well understood. In future, including a numeracy test or presenting risk information using pictorial icon ranges may support internal validity.