This study was conducted in two phases (a visual overview of these can be found in Appendix). In the first phase, back and forth translation and input for an expert panel (face validity) were used to create a Dutch language version. Subsequently, discriminant validity and inter-rater reliability of the Optimality Index-Netherlands (OI-NL) were assessed.
The study was approved by the ethics committee, Zuyderland Zuyd (16-N-69).
Development of the OI-NL
Following the steps for cross-cultural validation [29], the OI-US was translated from English to Dutch by two of the authors (ST and MN) both of whom have midwifery expertise in the Dutch setting and are fluent in both languages. Some small discrepancies in translation were resolved by discussion before synthesis into one document. The document was then translated back into English by two linguistic experts and this translation was reviewed for faithfulness to the OI-US version by two of the authors (LKL and RdV), both of whom are familiar with the content and context of the OI-US. Additional file 1 contains an overview of the items contained in both the OI-US and the OI-NL.
Face validation
A Dutch language translation was then presented to a panel of 10 experts in Dutch maternity care, consisting of community midwives (n = 2), hospital midwives (n = 2), obstetricians (n = 2), midwifery lecturers (n = 2) and midwife researchers (n = 2). They were requested to review all the items within the OI-NL to determine the relevance and accuracy of the items in relationship to care within the Netherlands. They were also invited to identify items that were not part of the OI-US but were relevant to the Dutch context. The feedback on the PBI (this comprises the first 15 items of the OI-NL) focused on the item pertaining to ethnicity, this was adapted to reflect ethnicity in the Netherlands and simplified for use. Both midwives and obstetricians pointed out the possibility of Body Mass Index (BMI) values being incorrectly scored and suggested collating values for height and weight and calculating the BMI once data was collected in order to reduce errors. The further feedback concerned a number of items which we adapted, it was decided to group previous birth by caesarean section into an item specifying ‘intrapartum complications’, adding ‘instrumental birth’ to the item ‘intrapartum complications’ as they considered these relevant to the obstetric history of multiparous women. They also suggested separating previous fetal death after 16 weeks gestation from previous antenatal complications, as detailed in the OI-US. We also opted to include the item ‘domestic/intimate partner violence’ as a perinatal background factor, adding detail to clarify that this item pertains to both social history and current pregnancy.
For the OI items, the experts suggested the removal of the item ‘non-stress test in pregnancy’ as it is not part of midwife-led antenatal care in the Netherlands. Clarification was requested on the item ‘liquor’ as, without colour classification (clear/meconium stained) it was considered ambiguous. The item ‘delayed cord clamping’ was added, as this is recommended practice in physiological births [30]. We adapted the quantity of blood loss (item postpartum haemorrhage) from 500 to 1000 ml, reflecting the Dutch definition of PPH [31]. A similar adaptation was made for the item on placenta retention, again bringing the duration of time for defining retention in line with Dutch definitions [31]. In addition to gestation at birth, our experts suggested that the index should reflect the certainty of gestation, as first trimester (dating) ultrasonography is a routine part of care in the Netherlands [32]. Following World Health Organization’s (WHO) recommendations for early infant feeding [33], we added early breastfeeding as an item and used the WHO guidelines to define the measure. Further to these suggestions, some minor linguistic feedback was given in order to ensure clarity of wording in the Dutch translation of the OI.
Pilot test
The OI-NL was then pilot tested in a primary midwifery practice in order to determine its feasibility as a data collection instrument in this environment. Three primary care midwives in one midwifery practice used the OI-NL on 15 occasions. They reported both clarity of items and ease of use.
An English or Dutch language version of the OI-NL is available from the authors.
Discriminant validity
Settings and participants
Between September 2016 and January 2017, 161 Dutch midwifery practices linked to the department of midwifery education at Zuyd University of Applied Sciences were sent written information about this study. They were invited to participate in this study and collect data on the women that started their birth under their care.
Data collection
In the midwifery practices that agreed to participate, women were approached for consent either during pregnancy or shortly after birth. Once consent was given, the midwife completed the OI-NL data form in a de-identified manner using information about the pregnancy, labour and birth and early postpartum period recorded in the case notes of each individual client, shortly after each birth. Primiparous and multiparous women who were in midwifery care at the start of labour were included.
Power calculation and data analysis
For assessing discriminant validity of the OI-NL, we conducted our data analysis using completed OI-NL forms.
A sample size calculation indicated that 28 completed OI-NLs per defined group (n = 56 in total) would be adequate to demonstrate an effect size of 0.5 (α = < 0.05, 80% power) using an independent t-test to compare mean OI scores between two independent groups: primiparous and multiparous women.
For each participant, percentages scores were calculated for the total OI-NL and the PBI and OI part of the OI-NL. These were used for analysis. A priori, we established that forms with more than 10% missing would be excluded. When less than 10% of the items were missing the dominator would be reduced, in accordance with the instructions provided by the OI-US [22].
We assessed discriminant validity by testing two hypotheses. We hypothesized that primiparous women in our data set will demonstrate a lower OI score than multiparous women. This assumption is based on national data, which show that primiparous women have more interventions during birth than multiparous women [17]. This finding will demonstrate that the OI is sensitive to these intervention differences between primiparous and multiparous women. Furthermore, we examined the association between birth-setting and the Optimality Index Score, hypothesizing that women cared for in either a home or out-patient (polyclinic) midwife-led setting will demonstrate significantly higher OI scores than women requiring intrapartum transfer to obstetric-led care [34].
Using SPSS version 24, we conducted multiple linear regressions analysis, using the OI percentage score as dependent variable and parity, PBI score and midwife-led births as independent variables. Parity was a dichotomous variable, coded as 0 for primiparous women and 1 for multiparous women, generating results that indicate the effect of parity, corrected for perinatal background factors. Midwife-led care was a dichotomous variable, coded as 0 for obstetric-led care and 1 for midwife-led care. Significance was set at 5% (two-tailed test).
Inter-rater reliability
Percentage agreement was measured in order to determine inter-rater reliability between two raters. The first author of this study (ST) and a midwife working with the OI-NL scored a data set from 25 clients in one midwifery practice to examine reliability of the OI-NL as a whole and at item level. Agreement of more than 80% was considered evidence of reliability [35].