Women’s preferences for inpatient and outpatient priming for labour induction: a discrete choice experiment

The OPRA trial ([23]; http://​www.​anzctr.​org.​au) was a 3-year, two-centre randomised controlled trial to compare the effectiveness of : (1) inpatient priming for induction of labour, and (2) outpatient priming with maternal/fetal assessment and CTG monitoring, followed by discharge home, then reassessment the following morning. Trial participants included women scheduled for an induction of labour for social reasons or prolonged pregnancy and who had a healthy term, singleton pregnancy with no medical conditions that would necessitate continuous maternal or fetal monitoring after cervical priming. Cervical priming for both groups was performed with prostaglandin E2 vaginal gels. Trial participants recruited in the second half of the trial received the DCE survey as part of their routine 7-week post-natal questionnaire. In addition, we also sought pregnant women volunteers from the two hospital’s antenatal clinics. Previous evidence has suggested that health care experience can influence preferences, increasing the value that people place on the service [24] and this has been particularly demonstrated in intrapartum care [20, 22]. It was therefore important to include a range of experiences with pregnancy and maternity care.

Women’s preferences were assessed using a discrete choice experiment (DCE) and followed good research practice [15, 16, 18]. The DCE approach is based on the idea that goods and services, including health care services can be described in terms of a number of separate attributes or factors and individuals’ value of the service depends on the levels these attributes take. The levels of attributes are varied systematically in a series of questions and respondents answer by choosing the option that they most prefer. The choices selected by the individual result from them weighing up the difference in attribute levels of each choice. From these choices we can estimate how respondents value different attributes across the range of levels presented. Other data collected in the survey, including attitudinal questions and sociodemographic information, can also enter the value functions as explanatory variables.

Based broadly upon the treatment arms in the OPRA trial, alternative priming options were described in terms of six attributes: 1) physical environment while waiting for gels to work, 2) availability of pain relief and sleeping medication 3) the professional who checks while waiting for the gels to work 4) how well the midwife is known to the woman 5) numbers of trips to hospital 6) travel time to nearest hospital offering the service (Table 1). The alternatives are described by this set of attributes, some of which have generic or common levels across outpatient and enhanced inpatient options (travel time and how well the midwife is known) and other which have alternative specific attribute levels (all other attributes) that feature key differences across the alternatives. Specific policy-relevant attributes were developed following qualitative interviews with trial participants who had experienced inpatient or outpatient care [25], the published literature, and an expert panel workshop with obstetricians, midwives and researchers.

Attribute levels of priming care were presented in three options. These were described as: Option A, an outpatient care option, Option B, an enhanced inpatient care option (including aspects such as private bathrooms and familiar clinicians), and Option C, a fixed basic inpatient care option that described the minimum level of standard care currently available. It was thought not to be clinically appropriate to present an opt-out option. To ensure plausibility of the scenarios, some attribute levels were specific for the alternative presented (Table 1) and we also ensured that the combinations of levels were realistic for the alternatives presented.

Prior to the main study, a pilot survey was given to 28 pregnant women awaiting clinic appointments and to ten midwives. The pilot survey included 25 questions, with each question offering three options of care (as above). Women reported no difficulties answering the scenarios; some pre-survey instructions were simplified and additional pre-survey explanation was provided for the main study survey. No changes were made to the DCE attributes and levels.

Utilising the parameter estimates from analysis of the pilot study, a statistically efficient [26] fractional factorial design of 50 different choice combinations, segmented into 2 blocks of 25 choices was created using NGENE (http://​www.​choice-metrics.​com) (design d-error of 0.024). Desirable statistical properties were maintained within each block. Some constraints were applied to the design, for example, non-drug pain relief methods such as warm baths were only available when the inpatient accommodation option included a private bathroom. In the final survey, women were presented with 25 choice sets, each with the three alternative options. An example of a choice set is presented in Table 2.

The final survey included a description and explanation of the attributes, a practice question to familiarise respondents with the format and interpretation of the attributes, followed by the discrete choice questions and sociodemographic characteristics.

Written informed consent was obtained from all participants in the OPRA trial; for volunteers, consent to participate was implied by completion and return of the questionnaires. The study was approved by the Human Research Ethics Committees of Flinders Medical Centre (Ref: 131/08) and the Women’s and Children’s Hospital (Ref: REC2034/2/11) in Adelaide, Australia.

The data set comprised 25 discrete choice questions of three alternatives for each respondent. A multinomial logit (MNL) model was used to analyse preferences [15, 16, 18]. The dependent variable indicates the choice from three alternatives in each scenario; this approach is consistent with other discrete choice analyses in antenatal care [19‐22]. Utility functions were specified as follows:

Models were evaluated for goodness of fit using the likelihood ratio Chi-square statistic for the global test of zero model coefficients, the McFadden’s pseudo R-squared, and Akaike’s information criterion (AIC). The final model was selected on the basis of AIC after testing a number of different model specifications.

Model results are presented as beta parameters, odds ratios and 95% confidence intervals of odds ratios. Trade-offs between attributes were calculated by taking ratios of beta parameters; categorical variables were effects coded. All analyses were conducted using NLOGIT Version 4.0. (Econometric Software, Australia, Castle Hill, NSW, Australia). We also calculated the overall value of alternative priming strategies, using the utility functions and beta parameters from the models, the characteristics of the services as they would be delivered in clinical practice, and representative patient characteristics.

The majority of pregnant women volunteers (approximately 90%, 102/114) approached agreed to complete the questionnaire while waiting for their clinic appointment. The overall response rate for women in the OPRA trial was 50% (260/515). This yielded a total sample of 362 women (9050 choices), giving an overall response rate of 58%. There were no significant differences in the preference structure of trial participants, compared to clinic women, or across women who were 1) randomised to, or 2) received the different trial arms, so all 362 respondents were included in the analysis [16]. Women were similar to the women in the OPRA trial in terms factors such as of treatment they were randomized to, age, university education and distance from hospital, and the demographic profile was similar to all women who were being induced for prolonged pregnancy at both hospitals (available from authors). Demographic characteristics of the respondents and their experiences of pregnancy and maternity care are shown in Table 3.

Preferences were significantly influenced by attributes, and also by some sociodemographic characteristics (see Table 4).

As would be expected, the ability to return to their own home strongly influenced women’s preferences for outpatient priming. Women were significantly (OR 1.8, 95% CI 1.4 to 2.1, p < 0.0001).

more likely to choose the outpatient care over basic inpatient care when they were able to return home (as opposed to being given off-site accommodation close to the hospital), and when the midwives were more familiar to them (compared to just having rostered midwife care). As the number of extra trips to hospital increased, or as the travel time to the closest hospital providing the service increased, women were less likely to choose the outpatient alternative over basic inpatient care.

The larger negative (beta) coefficient for travel time in the outpatient option indicates that each extra minute of travel time had a larger negative influence on women’s preferences for outpatient care compared with basic inpatient care. The type of pain relief and sleeping medication provided, and the availability of a doctor to talk to on the phone did not significantly influence preferences.

Demographic characteristics of the women, and their past care experiences also influenced women’s preferences for outpatient priming Women whose previous model of care was an obstetric clinic with doctors were significantly less likely to prefer outpatient priming over basic inpatient care, as were women who had previously experienced induction. Older women, women with a university level education and women for whom it was their first pregnancy were significantly more likely to choose outpatient priming over basic inpatient care; while women who came from a non-English speaking background were significantly less likely to prefer outpatient priming over basic inpatient care.

The inpatient environment while waiting for priming gels to work influenced women’s preferences. When a private room, with a private bathroom was available, women were significantly (OR 1.6, 95% CI 1.5 to 1.8, p < 0.0001) more likely to choose the enhanced inpatient option over basic inpatient care. Women were also significantly more likely to choose the enhanced inpatient option when the midwives were more familiar to them. When non-drug pain relief methods (hot baths) were not available, the availability of stronger pain relief also became important. As the travel time to the closest hospital providing the service increased, women were less likely to choose the enhanced inpatient option.

Demographic characteristics of the women also influenced women’s preferences for enhanced inpatient priming. Women whose previous model of care was an obstetric clinic with doctors were significantly less likely to prefer enhanced inpatient priming. Older women, and women for whom it was their first pregnancy, were significantly more likely to choose the enhanced inpatient priming option. Women who came from non-English speaking backgrounds were significantly less likely to choose enhanced inpatient priming over basic inpatient care; while education level and previous experience of induction did not significantly influence preferences.

The trade-offs between potential benefits and downsides of the alternative models of care (compared with basic inpatient care) are shown in Table 5. For outpatient services, women were willing to accept an extra 1.42 trips to hospital (2.42 trips total) and a travel time of 30.6 minutes per trip (73.3 minutes total) to be able to return to their own home while waiting for the priming to work. For enhanced inpatient services, women were willing to accept a travel time of 54.7 minutes (for their one trip to hospital) to have a private room with a private bathroom while waiting for the priming to work. Thus, women were more prepared to be inconvenienced in terms of travel time to have their own home environment, rather than a private room with a private bath (73 minutes vs 55 minutes).

Similarly, improved familiarity with the midwife was more highly valued (in terms of extra travel time women were willing to accept) in the enhanced inpatient option.

The overall value based on patient characteristics and beta parameter values (from Table 6) for outpatient priming was 3.63; for enhanced inpatient priming was 3.59, and for basic inpatient care was 2.89; indicating that outpatient primary was slightly more preferred than enhanced inpatient priming for the typical respondent in our survey. Both were preferred over basic care.

The purpose of this study was to determine the preferences of women for inpatient as compared with outpatient priming for induction of labour. Our study found that women prefer outpatient priming and are willing to accept a total travel time of 73 minutes to be able to return to their own home. Other DCE findings have also indicated stronger preference for home-like factors in the delivery of obstetric care [27]. Qualitative studies suggest that women favour outpatient priming [25], that midwives generally have positive views towards outpatient priming [28] and that it is associated with slightly higher levels of satisfaction with care [1, 12, 29].

Our study also provides additional information on the particular aspects of care, and the trade-offs between these aspects that are likely to be acceptable to women; we have also gained an understanding of the patient characteristics that are likely to drive women’s preferences for alternative priming strategies.

The DCE provides information on the likelihood (as an odds ratio) of choosing a particular option, given changes in attribute levels. However, women are prepared to make trade-offs between levels of attributes, and their own personal circumstances will also influence their decision of what they “really want”. The calculation of a value score for each management option uses the DCE results to try and answer this question, by combining information on the value of individual attribute levels as well as patient characteristics into a single value. The higher overall value score for outpatient priming, suggests that it was slightly favoured over enhanced inpatient care for this group of women. However the differences in scores were not large, and will also be dependent on the characteristics of the women themselves.

As might be expected, preferences for outpatient priming increased when women could return to their own home, and decreased with more trips to hospital and longer travel time. Women’s preferences for inpatient priming increased when a private room with a private bath was available, and decreased with longer travel time.

Sociodemographic characteristics of the women also influenced preferences: older women, and those with a university level education preferred outpatient priming to basic inpatient care. However women whose model of care had been an obstetric clinic run by doctors, and women from a non-English speaking background were significantly less likely to choose both outpatient priming and inpatient priming over basic inpatient care. It is not entirely clear why basic inpatient care would be preferred by these women. It does not appear to be driven by an overall preference for inpatient management, because the enhanced inpatient option was also less preferred than basic care. It may represent some form of status quo preference [30], even though the options were unlabelled; either women had experienced something like this type of care before and preferred what they knew, or perceived it to be usual care, and preferred what they believed was the ‘usual way’ of doing things. Women being cared for by doctors may have a stronger belief in a medicalised model of pregnancy care with inpatient care for direct medical supervision, especially those from non-English speaking backgrounds. It is also possible that some women may have been making assumptions about other factors that were not presented in the scenarios, for example some women may have assumed that the basic inpatient care option was free of charge, and the alternative options were associated with higher costs, even though costs were not included as an attribute. Although our pilot testing indicated that attributes and scenarios were understandable, women who came from a non-English speaking background may have interpreted some of the attributes in unexpected ways. Additional work is needed to better understand these issues, and to ensure that the information provided to women is adequate for them to make informed choices about alternative priming services.

As we did not, contrary to expectations, find differences between women’s preferences as respondents who volunteered compared to those who participated in the trial, it meant the data could be pooled and focus placed on other sources of preference variation. However, it is worth reflecting on possible reasons for why no difference was found. One explanation is that given the nature of the intervention, cervical priming was not required in approximately half the women in each trial arm who gave birth spontaneously. Thus even though women were enrolled in the trial, they did not actually experience induction of labour, and thus the scenarios were presented in the DCE were really hypothetical scenarios rather than experienced.

This is the first DCE to examine women’s preferences for cervical priming strategy for labour induction. We used a robust method to generate a statistically efficient DCE design and conducted the survey with a large number of participants. We obtained a good mix of respondents with different experiences of labour, induction or other priming strategies by recruiting from trial participants and obstetric clinics at the two hospitals. Although trial participants may not have been representative of women who might use induction services (as indicated by a 50% response rate) we were reassured by the 90% response rate from our clinic volunteers, who did not significantly differ from our trial women concluding that any issues around response rates in trial participants may not be as important in this particular study. It is also less likely that these results are directly applicable to rural and remote patients, where multiple trips to and from hospital for outpatient priming may not be feasible or practical.