Decisions were defined as ‘‘a verbalised choice or deferment of
choice that could alter the patient’s current or planned manage-
ment.’’ This was based on Braddock’s original definition, but adapted
to include deferment as a valid outcome [2,7,8]. Care was taken to
distinguish between decision-making and information giving.
Information giving occurred when women asked a question and
the doctor responded with advice, but there was no commitment to
a particular course of action and hence a decision was not made. In
the absence of any known system for classifying obstetric decisions
by topic, the authors developed a categorisation system de novo.
While listening to the consultations, decisions were identified and
sorted into sub-categories, until no new decisions were found. Each
time a decision was identified, it was either matched with an
existing decision (and coded similarly) or it was added as a new
decision under a sub-category. For the purposes of analysis, the
subcategories were grouped into five decision categories (DCs):
Delivery; Care Structure (appointments, referrals and admissions);
Investigations; Medication and other therapy; and Other. Each
consultation was reported as a binary variable (category discussed
or not discussed) for each of these five categories.
The number of different options that were offered and available
to be chosen were counted per decision. Where the option of doing
nothing was offered, this was counted. However, options that were
excluded by the healthcare professionals as impractical or unsafe
did not count (see Fig. 1). Each decision was converted into a binary
variable indicating that a choice between multiple options was or
was not offered. The number of decisions with choice was divided
by the total number of decisions in the consultation and reported
as a percentage. This percentage was called the decisions with
choice (DWC) variable and formed the study’s primary outcome.
The secondary outcome measures included the number of
decisions per consultation, who initiated the decision discussion
and the range of decisions made. They also included the proportion
of decisions where the option to do nothing or defer the decision
was offered, and the proportion of decisions that were deferred.
Basic descriptive statistics (percentages, means and standard
deviation) were used to analyse and present the decisions made.
Agreement between the two raters was assessed using a Bland–
Altman plot. This scatter plot compares the pair mean with the pair
difference for each observation [15]. In doing so, it uses the pair
mean as an estimate for the unknown true value and the mean
difference as an estimate of bias [16,17]. The Bland-Altman
compares points in relation to a horizontal line at a mean
difference of 0, which indicates perfect agreement. In normally