Use of evidence based leaflets to promote informed choice in maternity care: randomised controlled trial in everyday practice

Participants

We identified two samples of women. The first sample was all women who reached 28 weeks' gestation during a six week period and were receiving antenatal care in any setting (antenatal sample). Women were identified through hospital computer systems and the records of midwives and clerks in hospital and community antenatal clinics. The second sample was all women who delivered during a six week period (postnatal sample). Women were identified through child health computer records and hospital and home delivery registers. Women were excluded if they lived outside the catchment area of the hospital or in areas where antenatal care was provided by midwives from other hospitals, were under 16 years old, or had miscarried or if their baby had died or was seriously ill.

We identified an antenatal sample and a postnatal sample before the introduction of the leaflets and again nine months after they were introduced. We assessed outcomes using a postal questionnaire sent to women in these four different samples (figure). Women in the antenatal samples received the questionnaire at 28 weeks' gestation, and women in the postnatal samples received the questionnaire eight weeks after delivering their babies. Up to two reminders were sent at intervals of three weeks. The second reminder for the women in the antenatal samples was a shorter questionnaire that covered only key questions.

Study design

• Download figure
• Open in new tab
• Download powerpoint

Outcome measures

The primary outcome was the proportion of women who answered “yes” to the question “Have you had enough information and discussion with midwives or doctors to make a choice together about all the things that happened during maternity care?” with the options “yes,” “partly,” “no,” “there was no choice,” and “did not apply.” As informed choice is a difficult concept to measure2 we also asked women about the role they took when choices were made, with six options ranging from “active” to “passive.”5

Secondary outcomes were the “components” and the “consequences” of informed choice. The components measured were women's levels of knowledge of the 10 topics covered by the leaflets; satisfaction with information and with how choices had been made; and views of whether they had had sufficient discussion with health professionals. The consequences measured were the actions taken or services used by women. We also collected data on sociodemographic factors, parity, and women's preferences for involvement in decision making.5

Validation and quality assurance

We developed and piloted some measures specifically for the study on the basis of interviews with women in three maternity units not involved in this trial.6 Other measures were modified from published sources. The questionnaires were piloted by post outside the study area. Ethical approval was obtained for the study in each area.

Sample size and statistical analysis

Assuming an intraclass correlation coefficient of 0.007 and an average cluster size of 200 we calculated we would need a sample size of 925 in the intervention units and 925 in the control units for each of the antenatal and postnatal samples. This would give us an 80% chance of detecting an increase from 50% to 60% in the proportion of women reporting that they made informed choices in the intervention units compared with no change in the control units, at a two tailed significance level of 5%. We paired maternity units to ensure balance and undertook an unmatched analysis.7 We compared sociodemographic variables of respondents in all four antenatal samples and in all four postnatal samples at the individual level using analysis of variance and χ² tests. We compared any changes in the intervention units with changes over the same period in the control units using multilevel modelling. 8 9 Analysis was by intention to treat. We adjusted for women's age, age at leaving full time education, parity, and preference for involvement in decision making in the analyses. Intraclass correlation coefficients were 0.01 and 0.002 for the change in informed choice in antenatal and postnatal samples respectively.

Limitations in design

Possible limitations of this study are response bias, poor definition of “informed choice,” and lack of power. The response rate of 64% may have introduced some bias, with under-representation of non-white women and women with manual occupations in both intervention and control groups. The pilot study of two of the leaflets suggested that women with manual occupations might benefit more from the leaflets,10 suggesting that we may have underestimated the true effect of the leaflets. However, the main study found no relation between social class and effect of the leaflets.6

The question used to measure informed choice may have been insensitive. However, we used two different questions and neither showed change. Additionally, we found little change in the components of informed choice, which was consistent with a genuine lack of change in informed choice. Although we recruited fewer women than planned, the analysis for postnatal women was adequately powered due to a smaller intraclass correlation coefficient than estimated for the sample size calculation. The analysis for antenatal women had a power of about 65%. However, observed changes were small, and, although low power can explain the lack of significance, it cannot explain the size of the observed effect. Overall, it is unlikely that the study failed to detect any important change.

Everyday practice

We carried out this trial in everyday practice. We included thousands of women who might have received the 10 leaflets, but only 70% reported receiving one of them. Studies reported in the systematic review of decision aids were explanatory trials, with the implicit assumption that all patients received the intervention.3 One conclusion might be that the systematic review showed that decision aids can be effective under certain circumstances but that our study showed that they are not necessarily effective in the real world. The pragmatic nature of our design may have affected the outcome, but that outcome represents a true picture of the impact of introducing the leaflets into clinical practice.

Quality of implementation

There is some evidence from the trial that there were difficulties with the implementation of the intervention. The accompanying qualitative study shows that the Informed Choice leaflets were not introduced to women as special and different from other leaflets.4 Additionally, not all women reported that they had been given any of the leaflets during the intervention. This limited implementation of decision aids is by no means unique to either the Informed Choice leaflets or maternity care.11

Generalisability of findings

The results are generalisable to maternity units that use Informed Choice leaflets to promote informed choice but not to health professionals purchasing the leaflets to guide evidence based practice or individual enthusiastic health professionals purchasing the leaflets to help meet the information needs of their clients.