The Childbirth Fear Questionnaire: A New Measure of Fear of Childbirth

Abstract


Page 3/21
In our opinion, one or two item measures are insu cient to produce a stable estimate of childbirth fear, or encompass the possible range of concerns experienced by people who are pregnant or may become pregnant (e.g., fear of harm, medical interventions, or pain).Further, evidence suggests that fear of childbirth is multidimensional (4,6,30) .Among the longer scales developed (4,18,21) each either fails to assess key domains of childbirth fears (e.g., pain, harm to self or infant), includes non-fear relevant items, or under-samples content domains (i.e., have only one or two items for a particular content domain, whereas a minimum of three is needed to produce a stable measure).
By far the most commonly used measure of fear of childbirth is the Wijma Delivery Expectancy/Experience Questionnaire Version A (W-DEQ-A) (18).The W-DEQ-A has been used in several countries (12,(31)(32)(33)(34)(35)(36)(37)(38)(39).The W-DEQ-A has been found to possess good psychometric properties (12,18,40).Although the psychometric properties of the W-DEQ-A have been well established (12,18), it is not limited to an assessment of fear, but rather assesses a wide range of perceptions of labour and delivery (e.g., During labour and delivery, do you think you will feel: lonely; strong; con dent; afraid; deserted; weak; safe; independent; desolate; tense; happy, etc.).In factor analytic studies of the W-DEQ-A, fear has been found to emerge as one of four or one of six factors, strongly suggesting that the W-DEQ-A is not only a measure of fear (12,18).Further, many aspects of fear of childbirth are not addressed in this measure (e.g., pain; perceptions of social embarrassment; pressure to receive/avoid pain medication; mother's safety; changes to the body and sexual function; fear of medical interventions).The W-DEQ-A may limit researchers' ability to assess whether speci c aspects of fear of childbirth are predicted by different life experiences and/or result in differing outcomes.At the same time, the broad nature of W-DEQ-A items may still capture participants' experience of fear of childbirth albeit in a general way (i.e., still report fear, even if the exact content of their fear of childbirth would not be known).
Accurate measurement of fear of childbirth is important in correctly identifying those experiencing high levels of fear of childbirth, as well as identifying targets for treatment.At present, currently available measures of fear of childbirth do not fully meet this standard.

The Present Studies
We sought to develop a self-report measure of fear of childbirth that recognizes the complexity of childbirth fear, and assesses fear of birth regardless of the planned or preferred mode of delivery.The purpose of this paper is to describe the psychometric development of the Childbirth Fear Questionnaire (CFQ).In study one, we conducted an exploratory factor analysis of the proposed items, and identi ed the items and subscales for Study 2. We also assessed the reliability, and convergent and discriminant validity of the measure.In Study 2, we evaluated the replicability and generalizability of the factor structure of the CFQ and conducted further reliability, and convergent and discriminant validity evaluations.

Participants
We recruited English speaking, pregnant people who were over the age of 18 and who were residing in an English language country to participate, via online forums frequented by pregnant people (e.g., pregnancy-related web sites and blogs).We planned for a sample size of approximately 500 individuals, following the recommendations of MacCallum et al. (41) for the sample size needs of an exploratory factor analysis (EFA).Our nal sample consisted of 643 pregnant people.

Procedures
Prospective participants were directed to the online survey via the study advertisement.In order to proceed and complete the survey, participants were required to acknowledge that they had read the study cover sheet/consent form and agreed to participate.For each survey completed, $0.50 was donated to the Children's Health Foundation of Vancouver Island, British Columbia.The study was approved by the Behavioural Research Ethics Board of the University of British Columbia.

Measures
Background Questions.Participants completed a set of demographic questions (e.g., age, marital status, education, income, country of residence, and race and ethnicity), questions about the current pregnancy (e.g., method of conception, and number of fetuses), and previous pregnancies (if applicable) (e.g., the number of prior pregnancies, births, miscarriages, and vaginal and cesarean deliveries).
Birth Preferences.Using a 7-point Likert-type scale (ranging from a very strong preference for a vaginal birth to a very strong preference for a cesarean birth), participants were asked about mode of delivery preference (i.e., vaginal versus cesarean).
Childbirth Fear Questionnaire (CFQ) -Initial Item Pool.The initial pool for the CFQ comprised 49 items.We initially developed 49 items covering the following domains of potential childbirth-related fear: social embarrassment (e.g., fear of losing control), pain (i.e., fear of pain), pain medication (e.g., fear of not receiving the pain medication one is hoping for), mode of delivery (e.g., fear of a cesarean delivery), baby's and mother's physical safety (e.g., fear that one's infant may be harmed or die during labour/delivery), changes to one's body (e.g., scarring), sexual functioning (e.g., enjoying sexual activity less following delivery), and medical interventions (e.g., fear of having an episiotomy).CFQ items are scored on a 0 (not at all) to 4 (extremely) point, Likert-type scale.
CFQ items and item domains were developed by a team of perinatal researchers from the elds of psychology, midwifery and nursing, and based on earlier work in this area, including our own (42) The CFQ includes dimensions of fear that are commonly reported by pregnant people, such as fear of pain and fear that harm might come to the baby (30,43).In earlier work (42), maternal complications, feelings of embarrassment, fear of medical interventions/surgery, scarring, sexual functioning, and body damage, were also identi ed as areas of childbirth related fear and concern, and were considered when developing the items for the CFQ.
Our group of investigators collaborated in generating items for inclusion in the CFQ, reviewing item wording, and ensuring that all domains of fear deemed relevant to childbirth had been included in the initial pool of items.In recognition of the likelihood that the CFQ would include both a total scale score and subscale scores, we developed multiple items for each fear domain (a minimum of three items per subscale are needed to ensure a reasonable degree of internal consistency reliability).In order to be able to reduce the overall number of items, and develop subscales with high internal consistency reliability, each content domain initially included ve or more items.
Wijma Delivery Expectancy Questionnaire (W-DEQ-A).The W-DEQ-A is a 33-item questionnaire, with items scored on a 0-5 Likert type scale, and a range of possible scores from 0 to 165.The psychometric properties of this assessment tool have been well established (12,18).In the current sample, the internal consistency reliability for the W-DEQ-A was 0.92.In addition to the W-DEQ-A total score, there is also data to support the use of a 6-item fear scale (44).
In this study, in error, we administered the W-DEQ-A using a 0-4 Likert type scale (rather than the usual 0-5 scale).We then prorated the W-DEQ-A scores to the more standard 0-5 point scale as follows: original W-DEQ-A score divided by four, multiplied by 5. Our rescaled mean W-DEQ-A (M = 55.9)score is consistent with those found in the literature which range from 52.9 to 68.3 (9).Our mean scores for both nulliparous (60.7) and multiparous people (50.2) were consistent with those reported in the literature (i.e., 54.1 to 68.51 and 50.3 to 60.7, respectively), as was the percentage of participants scoring above 85 (i.e., 7.5% to 15.6% in the literature, and 9.8% in the current study) (9).We are con dent that our prorated W-DEQ-A scores are a valid estimate of correctly scaled W-DEQ-A items, and are valid to use as the main measure of convergent validity.
Edinburgh Postnatal Depression Scale (EPDS).The EPDS is a 10-item self-report screening tool for pre and postnatal depression.The sensitivity and speci city of the EPDS are in acceptable ranges (65% -100%, and 49% -100%, respectively) (45).The EPDS is the most widely used screening tool for perinatal depression (46).It was included in this study as a measure of discriminant validity for the CFQ; childbirth fear should be no more than moderately correlated with depression.In the current sample, the internal consistency reliability of the EPDS was 0.88.Mutilation Questionnaire (MQ).The MQ is a 30-item measure of blood and injury fears.Internal consistency for the MQ ranges from 0.75 to 0.86 (47).In the current sample, the internal consistency reliability was 0.87.High MQ scores are associated with fainting at the sight of blood and injury (47,48).The MQ was included as a second measure of discriminant validity for the CFQ; blood-injury fears should be no more than moderately correlated with fear of childbirth.
Exploratory Factor Analysis.We followed the recommendations of Sakaluk and Short (54) and others (e.g., (55)) for conducting exploratory factor analysis on all CFQ items.Speci cally, all solutions extracted common factors via maximum likelihood estimation, in order to facilitate the calculation of model t indices.We determined the number of factors to retain through a combination of criteria, including: (1) parallel analysis (56); (2) the minimum average partial (MAP) criterion (57); (3) interpretations of absolute and relative indexes of model t (the root mean square error of approximation [RMSEA], and Tucker-Lewis Index [TLI], respectively) (58); (4) interpretations of the Bayesian Information Criterion (BIC); (5) nested model comparisons using likelihoodratio tests between competing models; and (6) factor solution interpretability.All solutions were rotated to achieve simple structure and estimate factor correlations, using the oblique Oblimin method.
We considered items that loaded onto a factor at ≥ .35 to be substantive indicators of the underlying latent construct.Items that did not load onto any factor beyond this threshold were determined to be poor indicators, and would be removed from the nal version of the CFQ.
Reliability and Validity Analyses.The remaining analyses involved descriptive data (means, standard deviations, and percentages), Cronbach alpha reliability coe cients, correlations, and independent-samples t-tests.Differences between correlations were tested using a test of the difference between two dependent correlations with one variable in common (52).

Demographics
Participant demographic and reproductive information is presented in Table 2.Note that complications in the current pregnancy were reported by participants (22.9%), and ranged broadly in severity (e.g., early spotting, anemia, pre-eclampsia).In this sample, 21.3% of participants scored a 12 or greater (common cut-score for depression) on the EPDS.
Exploratory Factor Analysis of the CFQ Results of our parallel analysis (see Figure 1) suggested a maximum of 13 factors ought to be retained, whereas the MAP test suggested 9 factors was su cient.We then proceeded to evaluate indexes of model t, information criteria, and nested model comparisons for 1-13 factor solutions (see Table 3).
Each additionally extracted factor signi cantly improved the t of our model.Adequate model t based on the RMSEA was achieved from a 6-factor model onward, whereas adequate model t based on the TLI was achieved near the 9-to 10-factor models.The BIC indicated that our models became unnecessarily complex after the 11-factor solution, and our 13-factor solution failed to converge.We therefore examined the pattern matrixes of loadings for the 9-, 10-, and 11-factor solutions.
The 9-factor solution was supported by the MAP test and had acceptable t according to the RMSEA, and near-acceptable t according to the TLI 1 ; it was also the most conceptually interpretable of the three solutions we investigated in detail.The 10-factor solution, though acceptably tting according to both the RMSEA and TLI, yielded a tenth factor that was not conceptually coherent.The 11-factor solution, nally, was also acceptably tting, but the extracted eleventh factor had no substantially loading items.We therefore selected the 9-factor solution as the best tting and conceptually interpretable model of the CFQ items.
Factor loadings for the nal nine factors are presented in Table 4, and represent: (1) Fear of loss of sexual pleasure / attractiveness (SEX), (2) Fear of pain from a vaginal birth (PAIN), (3) Fear of medical interventions (INT), (4) Fear of embarrassment (SHY), (5) Fear of harm to baby (HARM), (6) Fear of cesarean birth (CS), (7) Fear of mom or baby dying (DEATH), (8) Fear of insu cient pain medication (MEDS), and (9) Fear of body damage from a vaginal birth (DAMAGE).The nine CFQ factors were generally weakly-to-moderately correlated (see Table 4), with most correlations below r = .60.The exceptions were the Fear of loss of sexual pleasure / attractiveness and Fear of body damage from a vaginal birth (r = .62),Fear of harm to baby and Fear of mom or baby dying (r = .78),and the Fear of cesarean birth and Fear of medical interventions (r = .60)factors, which were strongly associated.
Descriptive, Reliability, and Validity Analyses Descriptive Analyses.Means and standard deviations for the 9 subscales, and the CFQ Total scale scores are presented in Table 6.
Reliability Analyses.The Cronbach alpha for the overall 40-item scale was 0.94.Cronbach alphas for the individual subscales ranged from .76 to .94.Speci cally, Cronbach's alphas were .93 for Fear of loss of sexual pleasure/attractiveness, .94for Fear or pain from a vaginal birth, .82for Fear of medical interventions, .84 for Fear of embarrassment, .93 for Fear of harm to baby, .85 for CS, .86 for Fear of mom of baby dying, .76 for Fear of insu cient pain medication, and .85 for Fear of body damage from a vaginal birth.
Convergent/Discriminant Validity.The convergent/discriminant validity of the CFQ was assessed by comparing the relationship between the CFQ and the W-DEQ-A, and the relationships between the CFQ and measures of depressed mood (the EPDS) and blood and injury fears (the MQ).We predicted that the CFQ would correlate more strongly with both the W-DEQ-A (another measure of fear of childbirth) full scale and the W-DEQ-A fear scale than with either the MQ or EPDS (44).The correlations between the CFQ and the W-DEQ-A (full and fear scales) were 0.41 (p < 0.001) and 0.57 (p < 0.001) respectively.The correlation between the CFQ and the EPDS was 0.35 (p < .001),and the correlation between the CFQ and the MQ was 0.28 (p < 0.001).The CFQ-W-DEQ-A (full scale) correlation was signi cantly greater than the CFQ-MQ correlation, z = 2.73, p = 0.006, but not the CFQ-EPDS correlation, z = 1.60, p = 0.109.The CFQ-W-DEQ-A (fear scale) correlation was signi cantly greater than both the CFQ-MQ correlation, z = 7.17, p < 0.001, and the CFQ-EPDS correlation, z = 6.61, p < 0.001.Birth Preferences.A second approach to assessing the validity of the CFQ was to compare participants who reported a preference for a vaginal delivery to those who reported a preference for a cesarean delivery.We predicted that participants who reported a preference for a cesarean delivery would also report higher levels of fear of pain from a vaginal birth, fear of harm to baby, fear of mom or baby dying, fear of insu cient pain medication, and fear of damage to one's body from a vaginal birth, but lower levels of fear of cesarean delivery and fear of medical interventions, compared with those who reported a preference for a vaginal birth.
The majority of the people in our sample indicated a strong or a very strong preference for a vaginal childbirth (83.8%, n = 539), whereas only a small proportion indicated a strong or a very strong preference for a cesarean delivery (5.1%, n = 33).Consistent with the above hypotheses, compared with those who strongly preferred a vaginal birth, people who strongly preferred a cesarean delivery reported higher scores on the Fear of pain from a vaginal birth , t (34.08) = -2.83,p = .008,d s = 0.68, Fear of harm to baby , t (570) = -2.84,p = .005,d s = 0.51, Fear of mom or baby dying , t (570) = -2.81,p = .005,d s = 0 .50,and Fear of insu cient pain medication , t (33.50) = -5.54,p < .001,d s = 1.53, subscales of the CFQ, but lower scores on the Fear of cesarean birth, t (37.22) = 6.64, p < .001,d s = -1.07,and the Fear of medical interventions, t (570) = 2.15, p = .032,d s = -0.39,subscales of the CFQ.However, mean subscale scores did not differ signi cantly between participants who indicated a preference for a vaginal birth compared with those who indicated a preference for a cesarean birth for the Fear of loss of sexual pleasure/attractiveness, d s = 0.23, Fear of embarrassment, d s = 0.05, or the Fear of body damage from a vaginal birth, d s = 0.04, subscales of the CFQ.The means and standard deviations for these subscale differences are presented in Table 7.
Parity.Nulliparous and multiparous participants differed signi cantly on six of the nine CFQ subscales, and the CFQ Total scales.In each case, nulliparous participants scored higher than multiparous participants.Speci cally, nulliparous participants scored higher than multiparous participants on the CFQ factors: Fear of loss of sexual pleasure/attractiveness, t (639.7. Country.Canadian and American participants differed on only two of nine CFQ subscales, Fear of medical interventions, t (605) = -2.40,p = .017,d s = 0.21, and Fear of cesarean birth, t (605) = -3.00,p = .003,d s = 0.26.In both instances, American participants reported higher levels of fear, though the magnitude of these nationality differences were generally smaller than those between birth preference and parity groups.Means and standard deviations by country for Canada and the US are presented in Table 7.

Summary
In our development/validation study of the initial 49 CFQ items, involving 643 pregnant people, exploratory factor analysis resulted in a 9-factor scale, supported by MAP test with acceptable t based on RMSEA.The resulting 9 factors represent: (1) Fear of loss of sexual pleasure / attractiveness (SEX), (2) Fear of pain from a vaginal birth (PAIN), (3) Fear of medical interventions (INT), (4) Fear of embarrassment (SHY), (5) Fear of harm to baby (HARM), (6) Fear of cesarean birth (CS), (7) Fear of mom or baby dying (DEATH), (8) Fear of insu cient pain medication (MEDS), and (9) Fear of body damage from a vaginal birth (DAMAGE).Subscales were weakly to moderately correlated with a small number of strong correlations (Fear of loss of sexual pleasure/attractiveness with Fear of body damage from a vaginal birth, Fear of harm to baby with Fear of mom or baby dying, and Fear of cesarean birth with Fear of medical interventions).Cronbach alpha coe cients for the total scale and individual subscales were all above .76.Strong evidence of convergent/discriminant validity was found when comparing the 9-factor CFQ with another measure of fear of childbirth and measures of blood, injury injection fears and depressed mood.The CFQ subscale means were also compared across subgroups (e.g., preferred mode of delivery) with hypothesized differences supported by the data.

Participants
Participants were also English-speaking, pregnant people living in Canada.Eight hundred and eighty-one (N = 881) participants completed an online survey between 11-and 46-weeks' gestation with an average of 35-weeks.

Procedures
We recruited English speaking, pregnant people who were over the age of 18 and who were residing in Canada, via Facebook and online forums frequented by pregnant people (e.g., pregnancy-related web sites and blogs).Consenting participants completed an online survey.Participants were eligible to win one of seven $150 prizes.The research was approved by the Behavioural Research Ethics Board of the University of British Columbia.

Measures
Participants completed an online survey.Similarly to Study 1, the online survey included the same background and demographics questions, the 40 CFQ items retained from Study 1, the W-DEQ-A (without the scoring error described in Study 1), and the EPDS.We also administered a measure of PTSD (see below) (59).The MQ was not administered to this sample.
Posttraumatic Diagnostic Scale for DSM-5 (PDS-5).The PDS-5 is a self-report tool used to assess post-traumatic stress disorder (PTSD) based on the DSM-5 diagnostic criteria.The PDS-5, one of the most used self-report measures of PTSD, has been found to show good sensitivity and speci city, internal consistency and test-retest reliability, and convergent and discriminant validity (59,60).A signi cantly elevated PDS-5 score (i.e.≥28) yields a sensitivity of 79% and speci city of 78%, allowing for probable prediction of a PTSD diagnosis (59).

Data Analysis Strategy
Con rmatory Factor Analysis and Invariance Testing.We used con rmatory factor analysis (CFA) to test the replicability of our exploratory measurement model from Study 1; we also speci ed two additional methods factors that we anticipated shared variance on account of the repeated use of the terms "vaginal" (items 8, 19, 20, 31, 35, and 37) and "caesarian" (items 9, 21, and 34).Given the limited number of response options for the CFQ and that our indicators failed to meet assumptions of multivariate normality (Multivariate Skewness p < .001,Multivariate Kurtosis p < .001,univariate nonnormality for all indicators p < .001)for the default maximum-likelihood estimator, we opted instead to use a robust unweighted least squares estimator (ULSM).The model was identi ed and the scale of latent variables was set using a xed-factor method, whereby latent variances were xed to a value of 1 and all loadings were freely estimated.We evaluated models using conventional recommended cut offs for absolute and relative indexes of model t (58,61), including the RMSEA and standardized root mean square residual (SRMR; both recommended to be < .08),and the TLI and Comparative Fit Index (CFI; both recommended to be > .90),being mindful of how model reliability can impact the appropriateness of these cut offs (see (62).We conducted our CFA using the lavaan() package (v.0.5-23.1097)(63) for R (64).
Multi-Group Measurement Invariance Testing.We then tested the generalizability of our CFA model by examining measurement invariance across participants based on their experience of pregnancy as primiparous (n = 208) or multiparous (n = 683) mothers.Establishing measurement invariance is a necessary precursor to group comparisons of factor correlations or means, in order to rule out the possibility that differences from such comparisons simply re ect divergences in the way groups think about the constructs under consideration (61,65).Speci cally, groups must demonstrate the same number of factors and general pattern of loadings (i.e., con gural invariance) and factor loadings of comparable magnitude (i.e., weak invariance) in order for group comparisons of correlations involving the factors to be valid.Moreover, groups must demonstrate con gural and weak invariance, alongside intercepts of comparable magnitude (i.e., strong invariance) in order for group comparisons of factor means to be valid.
We began the process of testing measurement invariance by tting and evaluating a con gural invariance model.We then used a combination of nested model comparisons and examining the change in model t indexes in order to determine whether the constraints imposed by the subsequent levels of invariance (i.e., weak and strong) were supported (e.g., (66); (67).We t and evaluated our invariance models using the semTools() package (v.0.4-14) (68) for R (64), using the same scale-setting, and identi cation selections from our CFA analysis.
However, because of convergence issues with the USLM estimator for our invariance testing, we reverted to using a robust maximum-likelihood estimator for specifying invariance models.As a consequence, our invariance models appear worse tting than they would have been under the more appropriate USLM estimator (e.g., there was nearly a .10CFI difference between base models depending on estimator selection).We think this compromise is acceptable, given that with these invariance tests we were primarily concerned with relative changes in model t as we imposed more stringent invariance constraints.
Taxometric Analyses.Our nal analysis regarding the measurement structure of the CFQ involved examining whether-as our factor analysis models presumed-the CFQ was best understood as re ecting some continuous dimension(s) or rather, some number of discrete categories, using taxometric analyses (69) (for reviews see ( 70) and ( 71)).In essence, taxometric analyses function by calculating indexes that ostensibly evidence continuity vs. categorical-ness for a set of observed indicators (e.g., Mean Above Minus Below A Cut, MAMBAC), and then comparing the values of those indexes against those of the same indexes when coming from simulated populations in which a dimensional or categorical structure is speci ed.Speci cally, a comparison curve t index (CCFI) is computed as the ratio of the degree of mis t for the observed data to a dimensional population compared to a categorical population, with CCFI values less than .45evidencing support for a dimensional model, values greater than .55evidencing support for a categorical model, and values in between indicating an ambiguous outcome.Further, multiple taxometric indexes can be used to compute CCFIs in this fashion; in fact, it is recommended to do so as a form of consistency testing, in order to ensure interpretations are robust to the idiosyncrasies of each index (70) We therefore evaluated CCFIs from three standardly reported taxometric indexes: MAMBAC, MAXEIG (maximum eigenvalue), and L-MODE (latent mode).
In order to conduct a taxometric analysis, we had to determine two additional analytic features: the indicators that we would include in the analysis, and the plausible size of a taxon (i.e., the rst extracted category) underlying the CFQ, were a categorical solution to be supported (70).Unlike other forms of latent variable modeling, taxometrics works best when using an e cient (i.e., limited, non-exhaustive) non-redundant (i.e., spanning the conceptual breadth of the construct) set of indicators from a larger measure; in particular Ruscio et al. (72) recommends somewhere between 3-5 indicators (as cited in ( 71)).As the CFQ contains many more items, we therefore conducted our taxometric analyses three times, using a different sampling of items across the subscales of the CFQ for each instance (Analysis 1: items 2, 8, 34, 39; Analysis 2: items 13, 21, 24, 26, and 40; and Analysis 3: items 3, 7, 9, and 37), which provided us another opportunity to evaluate the consistency of our analyses across different analytic speci cations.
Next, taxometric model tting requires the speci cation of a plausible taxon base rate (see (70)) in order to compute the desired CCFI values.As a recently developed alternative to subjectively determining this base rate (e.g., by consulting previous literature, guestimating, etc.,) Ruscio et al. (72) developed a method of creating CCFI pro les, in which taxometric analyses are performed iteratively across a range of speci ed base-rates.The CCFI pro le method, though computationally more intensive, is advantageous in that it provides a reliable means of determining whether the underlying measurement model is dimensional or categorical, and, when in fact categorical, CCFI pro les provide the most accurate estimate of the true underlying base rate.We therefore used the CCFI pro le method, evaluating of CCFI values (and their average) across the broadest range of taxon base rates (2.5% to 97.5%).
Descriptive, Reliability, and Validity Analyses.The remaining analyses involved descriptive data (means, standard deviations, and percentages), Cronbach alpha reliability coe cients, and correlations.Differences between correlations were tested using a test of the difference between two dependent correlations with one variable in common (52).

Demographics
Participant demographic and reproductive information is presented in Table 2.
Parity.Using structural equation modeling and comparing against the intercept-invariance model of the CFQ, we again found differences in CFQ scores based on parity, c 2 (9) = 60.73,p < .001.As in Study 1, nulliparous participants scored higher than multiparous participants on seven of the nine CFQ factors, including the Fear of loss of sexual pleasure / attractiveness, z = 4.78, p < .001,d s = 0.52, Fear of pain from a vaginal birth, z = 2.47, p = .01,d s = 0.19, Fear of medical intervention, z = 2.68, p = .007,d s = 0.26, Fear of embarrassment, z = 3.77, p < .001,d s = 0.41, Fear of harm to baby z = 3.67, d s = 0.32, Fear of mom or baby dying, z = 2.27, p = .02,d s = 0.20, and Fear of body damage, z = 4.50, p < .011,d s = 0.37.Nulliparous and multiparous participants did not differ signi cantly on the Fear of caesarian birth, d s = 0.09, Fear of insu cient pain medication, d = 0.05.
In sum, our CFA-related analyses suggest that the measurement model of the CFQ is both replicable, and generalizable across parity groups, indicating that it is appropriate for use within (and comparisons between) samples of participants who are expecting with different levels of pregnancy experience.
Taxometric structure of the CFQ Our three selected samplings of CFQ items generally exhibited excellent properties for candidate indicators in taxometrics analyses, in terms of distributional characteristics, validity coe cients, and within-taxon and within-compliment correlations (see (70,71) All three CCFI pro les (see Figure 2) strongly supported a dimensional structure for the CFQ and its subscales, as all individual CCFIs (with the exception of one CCFI from in one CCFI pro le, MAXEIG in CCFI Pro le 2) and their averages, across all three analyses, unambiguously supported dimensional structure (CCFIs < .45).We take the consistency of these effects as compelling evidence that the fear of childbirth is best understood as interrelated factors on which individuals differ in degree, not kind (69).
Descriptive, Reliability, and Validity Analyses Descriptive Analyses.Means and standard deviations for each of the 9 subscales, and the CFQ Total scale scores are presented in Table 6.
Reliability Analyses.The Cronbach alpha for the overall 40-item scale was 0.94, and the Cronbach alphas for the individual subscales ranged from 0.71 to 0.94 (i.e., Fear of loss of sexual pleasure/attractiveness = .90;Fear of pain from a vaginal birth= .94;Fear of medical interventions = .78;Fear of embarrassment = .79;Fear of harm to baby = .84;Fear of cesarean birth = .87;Fear of mom or baby dying = .88;Fear or insu cient pain medication = .71;Fear of body damage from a vaginal birth= .85).
Convergent/Discriminant Validity.The convergent/discriminant validity of the CFQ in Study 2 was assessed by comparing the relationship between the CFQ and the W-DEQ-A with the relationships between the CFQ and measures of depressed mood (the EPDS) and symptoms of posttraumatic stress disorder (the PDS-5).We predicted that the CFQ would correlate more strongly with both the W-DEQ-A (another measure of fear of childbirth) full scale and the W-DEQ-A fear scale (44) than with either the EPDS or the PDS-5.The correlations between the CFQ and the W-DEQ-A (full and fear scales) were 0.58 (p < 0.001) and 0.62 (p < 0.001) respectively.The correlation between the CFQ and the EPDS was 0.34 (p < .001),and the correlation between the CFQ and the PDS-5 was 0.24 (p = 0.001).The CFQ-W-DEQ-A (full scale) correlation was signi cantly greater than the CFQ-EPDS correlation, z = 7.59, p < 0.001, and the CFQ-PDS-5 correlation, z = 6.97, p < 0.001.The CFQ-W-DEQ-A (fear scale) correlation was signi cantly greater than both the CFQ-EPDS (z = 9.00, p < 0.001), and the CFQ-PDS-5 (z = 7.97, p < 0.001) correlations.See Table 5 for a full list of correlations.

Summary
Study two supported the 9-factor structure of the CFQ, and provided evidence of measurement invariance across parity groups.Speci cally, those who had previously given birth understood and responded to CFQ items in the same way as participants who had not previously given birth.Additionally, further tests of the CFQ's latent structure strongly supported a dimensional structure.Thus, fear of childbirth is a construct on which individuals differ in degree rather than in kind (i.e., higher fear of childbirth is not qualitatively different from a lower fear of childbirth).As in study one, the CFQ demonstrated good reliability, and convergent and discriminant validity.As was true in study one also, overall, nulliparous participants scored higher on CFQ subscales compared to multiparous participants.

Discussion
The purpose of this research was to develop a new measure of fear of childbirth that would encompass the breadth of pregnant people's childbirth fears and overcome some of the limitations of commonly used measures of fear of childbirth.Exploratory factor analysis of the CFQ resulted in a 40-item, nine-f actor questionnaire.Our nine-factor model was supported by a MAP test, exhibited reasonable model t and good simple structure, and our factors were readily conceptually interpretable.The 9-factor structure of the CFQ was further supported in Study 2, in a larger sample of pregnant participants.The CFQ total scale and the nine subscales demonstrated good internal consistency, and convergent and discriminant validity across both studies.
The taxometric analyses strongly supported a dimensional structure.Thus, fear of childbirth is a construct on which individuals differ in degree rather than in kind (i.e., higher fear of childbirth is not qualitatively different from a lower fear of childbirth).It also suggests that multiple causal in uences with small additive effects may best explain more intense fear of childbirth (i.e., rather than a larger single causal factor).While diagnostic categories are frequently used in psychology and might be helpful for clinicians and health authorities to prioritize individuals' access to treatment, diagnostic categories hypothesize a categorical latent structure.Given the dimensional latent structure of fear of childbirth, it will be important to bear in mind that any cut-off score will be arbitrary and result in a loss of information.It is thus better for future studies to keep the full continuum of scores and respect the dimensional latent structure of the data (74).
Second, our nine-factor model showed initial evidence of measurement invariance between parity groups.Thus, those who had previously given birth understood and responded to CFQ items in the same way as participants who had not previously given birth.Any variations in responses between those two groups will be due to real world differences, rather than to a misspeci cation of the measurement model for one or the other group.This is especially important, provided our nding that nulliparous participants scored higher than multiparous participants on several of the nine CFQ factors.
Data from both studies provide excellent support for the convergent and discriminant validity of the CFQ.As predicted, the CFQ correlated most strongly with another measure of fear of childbirth (W-DEQ-A), and less so with measures of depressed mood (EPDS), trauma symptoms (PDS-5) and blood, injury, injection fears (MQ).In addition, correlations with the CFQ were stronger for the W-DEQ-A (fear subscale) compared with the W-DEQ-A (full scale).This was expected and supports our contention that the W-DEQ-A is not strictly a measure of fear (i.e., includes multiple items more relevant to feelings of depressed mood and other positive and negative emotions).Although the W-DEQ-A fear subscale contains 6 items, only three truly re ect fear (i.e., afraid, tense and panic).The other three (hopelessness, pain and lose control of myself) are not speci cally fear items.The weaker correlation between the CFQ and the W-DEQ-A (full scale) provides support for the CFQ as a novel measure of fear of childbirth, with an emphasis on fear, and distinct from the W-DEQ-A.
Furthermore, the nine factors of the CFQ have the potential to signi cantly add to our knowledge about fear of childbirth.For example, the nine subscales of the CFQ, identi ed through factor analysis, make it evident that pregnant people's concerns about childbirth encompass a broad range of potential fears, and that pregnant people who prefer a cesarean birth have different concerns than those who prefer a vaginal birth.Findings regarding the association of CFQ domains and mode of delivery preferences are consistent with our prediction that those who strongly prefer a cesarean birth are especially fearful of (a) the pain from a vaginal birth, and the possibility that they may not receive su cient pain medication during labour/delivery, and (b) the possibility that something may go terribly wrong during labour/birth, and they or their infant may be harmed or die.Conversely, pregnant people who strongly prefer a vaginal birth are, as expected, more fearful of cesarean delivery and labour/birth related medical interventions in general.The same pattern of results can also be observed in the intercorrelations among the CFQ subscales.
In both studies, the lowest CFQ-W-DEQ-A subscale correlation was for the Fear of cesarean birth subscale.The Fear of medical interventions subscale also correlated weakly with the W-DEQ-A in both studies, although less so in study two (i.e., r = .33and .38 in study two, and r = .10and .03 in study one).This weak relationship between the W-DEQ-A and these two CFQ subscales is likely a function of the fact that the CFQ is unique among measures of fear of childbirth in its assessment of fears related to operational (i.e., cesarean) delivery.
These nding highlight an interesting phenomenon, not easily assessed by previously available measures of fear of childbirth: that some pregnant people strongly prefer a cesarean birth and are predominantly fearful of the perceived pain and/or danger associated with vaginal delivery, whereas others strongly prefer a vaginal birth and are predominantly fearful of medical interventions in general, and cesarean birth in particular.The link between fear of labour pain and a preference for cesarean delivery is well-documented (42,75).Very little is known about fears speci c to those who prefer vaginal birth.In this regard, the CFQ lls an important gap in our knowledge of the childbirth fears most relevant to pregnant people who prefer a vaginal birth.
The only between country (Canada and the United States) differences were for the Fear of medical interventions and the Fear of cesarean birth subscales.This is an expected nding as childbirth is a more medicalized experience in the United States in comparison with Canada (Shaw et al., 2016; Stoll et al., 2020).Previous research has shown that those who experience pregnancy in a more medicalized birth culture report heightened fear of interventions and other fears that are speci c to hospital settings (76).Consequently, one would expect pregnant people's fears of medical experiences in childbirth to be heightened.
Another example, and a novel and important aspect of the CFQ, is the inclusion of subscales measuring (a) pregnant people's fears about negative, childbirth-related changes to their appearance and sexual functioning, including Fear of loss of sexual pleasure/attractiveness, and (b) Fear of embarrassment as a result of events occurring during labour/delivery (e.g., fear of urinating in front of others).It is well known that becoming a parent has a signi cant, and oftentimes negative, impact on one's romantic relationship, including one's sexual relationship (77,78).That this is a concern for pregnant people appears well captured by the CFQ.That fears about a loss of sexual pleasure and attractiveness are associated with a fear of embarrassment during labour/delivery is not surprising in that both involve potential negative judgments by others, and potentially being seen in ways that are perceived as unattractive by typical standards.Our ndings demonstrate that fears regarding embarrassment and sexual functioning/appearance are closely related to fears about childbirth pain and bodily damage in the context of a vaginal delivery, as well as harm or death to mom and baby during childbirth.It appears that pregnant people associate pain from a vaginal birth with vaginal damage, and correspondingly with negative changes to their sexual functioning and appearance, and embarrassing aspects of labour/delivery.

Clinical Implications
Current measures of fear of childbirth fail to assess the full spectrum of perinatal people's childbirth related fears.Given that fear of childbirth has been associated with a number of negative medical and social outcomes, an accurate assessment of these fears is important and has implications for pregnant individual's reproductive and mental health.The development of an effective self-report measure of fear of childbirth will facilitate: (a) the provision of appropriate treatment for those with these fears, (b) assessment of speci c aspects of perinatal people's childbirth related fears, and (c) identi cation of fear of childbirth as a potential psychosocial indication for a cesarean delivery.The new CFQ will help to identify pregnant people's speci c childbirth concerns, which may be amenable to education or a psychosocial intervention if more extreme.

Limitations
This study is limited by the fact that we collected data from two convenience samples of pregnant people.We did not collect prospective data, nor did we collect data from reproductive-aged people who are not pregnant.A further limitation is the fact that our sample was English-speaking only and predominantly Caucasian.Generalizability of the measure is, until further validation in other cultural contexts, limited to Caucasian pregnant people living in predominantly English-speaking countries.Finally, online survey administration prevents the calculation of response rates.

Future Directions
Future research would bene t from an evaluation of the CFQ among reproductive aged people who are not pregnant but may one day become pregnant or give birth, those who are gender diverse, as well as reproductive-aged people who are biologically male.The attitudes of biologically male people towards birth and fears concerning childbirth have been shown to in uence decision-making around mode of delivery (79).Further, the validity of the measure should be assessed in other cultural contexts beyond predominantly Caucasian, English-speaking countries.
In our opinion, the most important next steps in the development of the CFQ are to: (a) evaluate the test-retest reliability and sensitivity to change of the CFQ, and (b) assess the CFQ as a screening tool for speci c phobia of fear of childbirth (speci c phobia is the diagnostic category which has been put forth as the most appropriate classi cation of fear of childbirth).

Conclusion
The Childbirth Fear Questionnaire (CFQ) is a promising new instrument for the multi-factorial assessment of fear of childbirth.Evidence of its reliability and validity has been presented.We hope this new measure proves useful to identify pregnant people with elevated fear of childbirth, and for future research into the fear of childbirth.

Figures
Figures

Table 7 CFQ
Subscale Means and Standard Deviations: By Delivery Preference and Country.
Note.All scores are mean item scores with a possible range of 0 (not at all) to 4 (extremely).*p < 0.05; ** p < 0.01; *** p < 0.001, based on t-tests for independent samples comparing women who: (a) prefer a vaginal birth to those who prefer a Cesarean birth, (b) are nulliparous to women who are multiparous, and (c) are resident of Canada to those who are resident of the United States of America.Note.χ 2 is Yuan-Bentler corrected version, based on robust MLR estimation; Δ χ 2 is therefore computed using scaled Satorra-Bentler (2001) method