Introduction

Whilst many women report the transition to motherhood positively, for some, it can present psychological, social, physical, and emotional challenges. The transition to motherhood has been suggested as a potential period of psychological vulnerability [1], with anxiety and depression often highlighted as examples of poor mental health in the postnatal period.

Preterm birth represents a significant public health issue. The World Health Organization [WHO] estimate that over 13 million infants were born too early in 2020, with preterm birth now the leading cause of child mortality worldwide [2]. Adverse consequences include respiratory diseases [3], and increased likelihood of neuro-developmental delay comparative to term infants [4].

Research consistently shows mothers who give birth prematurely are more likely to develop depression [5], post-traumatic stress disorder [6], and anxiety [7], when compared to mothers of term infants. Postpartum anxiety in particular, may be exacerbated in mothers of premature infants compared to mothers of term infants, perhaps due to the sudden and unexpected nature of the birth [8], and uncertainty surrounding infant health [9].

Postpartum anxiety can be characterised by excessive, irrational worries occurring during the postpartum. From an evolutionary perspective, some anxieties after childbirth are adaptive, so there is a need to evaluate the presence of anxieties occurring after birth that may be considered maladaptive [10], through effective screening and identification. Despite high prevalence rates (up to 40%) [11], postpartum anxiety remains under-recognised and under-diagnosed [12]. This may perhaps be, in part, due to the lack of appropriate measurement tools for postpartum women. Most studies measuring postpartum anxiety use tools validated in general adult populations, which have been extrapolated for use in mothers as there has historically been no suitable alternative [13]. This can lead to spurious findings which may be unreliable [14].

Recently, the Postpartum Specific Anxiety Scale [PSAS] has been developed for research purposes and validated as a 51-item long-form [15], and as 16-item (PSAS-RSF) [16], and 12-item (PSAS-RSF-C) [17] short-forms. The PSAS has not been widely used in mothers of premature infants; to date only one study has investigated preterm birth and anxiety using the measure [7]. It is hypothesised however, some items contained within the PSAS may not be interpreted in the same manner in mothers of premature infants as with mothers of term infants. Items particularly pertinent to mothers of preterm infants may produce inflated findings given the increased likelihood of health concerns in this population [18], irrespective of anxiety. Similarly, items assessing anxieties surrounding infant development and maternal self-efficacy may also be elevated, given these are common anxieties for mothers of premature infants [19, 20].

Given the vast social, economic, and psychological consequences of preterm birth, it is essential to investigate if using a postpartum-specific tool performs in a similar manner in mothers of premature infants as it does in mothers of term infants to ensure accurate measurement. This study therefore aims to undertake a measurement invariance analysis [21], to assess if anxiety has the same meaning across groups, using the PSAS-RSF in an English-speaking population of mothers of preterm and term infants in the first postpartum year.

Materials and methods

The current study forms part of a wider survey investigating postpartum anxiety and gestational age. The present analysis comprises datasets from two studies, which received full ethical approval from the University of Liverpool Institute of Population Health Research Ethics Committee (Study 1 ref:- 10606; Study 2 ref:- 3616). Written informed consent was gained via an electronically signed consent form, and participants received a full debrief upon completion.

Participants were recruited on-line, via social media and by word-of-mouth. Participants who were under 18 years of age, whose infant was born < 24 week’s gestation, or who were currently in the Neonatal Intensive Care Unit [NICU], were not eligible to participate. Participants were eligible to take part in the study if they were over the age of 18, had an infant born > 24 week’s gestation. If participants had an infant in the NICU, they had to have been discharged and well for at least three months to take part. Study 1 data were collected at two separate time points: February to August 2022 and November 2022 to March 2023. Study 1 data were combined with Study 2 data, which had been collected between January and February 2023, rendering 320 participants total. This allowed for the creation of two groups of equal number (premature [n = 160]; and term [n = 160]). There is no specific power calculation for measurement invariance, but the sample size was determined following the rule of 10 cases per item [22], rendering 160 participants for each group as there are 16 items on the PSAS-RSF. Participants completed demographic questions and psychometric scales including the PSAS-RSF [16].

Measures and analysis plan

Demographics

Several demographic questions, including age, ethnicity, and educational attainment were asked. Infant demographics, including gestational age (week of birth), and birth order were also gathered.

Postpartum Specific Anxiety Scale - Research Short Form (PSAS-RSF)

The PSAS-RSF [16] is a 16-item scale developed in 2021 measuring the frequency of maternal-and-infant anxieties occurring during the previous seven days, scored on a Likert scale from 1 (not at all) to 4 (almost always). The scale can be used in mothers of infants from 0 to 12 months of age. It features four subscales, consisting of four items each, scored separately: Maternal Competence and Attachment Anxieties, Practical Infant Care Anxieties, Infant Safety and Welfare Anxieties, and Psychosocial Adjustment to Motherhood. The scale has excellent reliability in the current study (McDonald’s ωh = 0.75).

Method of analysis

All data analyses were conducted in R version 4.2.1, using lavaan [23], dplyr [24], corrplot [25], semTools [26], semTables [27], and psych [28] packages. Measurement invariance of the scale was tested across the two groups (mothers of term and mothers of premature infants). Diagonally weighted least squares estimation was used as individual responses on the PSAS-RSF are ordinal, scored on a scale of one-to-four [29].

Configural invariance was tested to investigate if the factor structure held across the two groups, with a random effect of group in the model. Model fit was assessed using several indices. The Tucker-Lewis Index [TLI] and Comparative Fit Index [CFI] were computed, with values of ≥ 0.95 considered excellent [30]. A Root Mean Square Error of Approximation [RMSEA] was also computed, with values of < 0.05 considered good, and values of 0.08 considered fair [30, 31]. Finally, the Standardised Root Mean Square Residual [SRMR] was also computed, with values of < 0.08 considered good [30].

The configural model was then compared to a metric invariance model. The model structure remained the same, but factor loadings across groups are fixed, which allows interpretation as to whether each individual item on the scale performs in the same way across the two groups. The validity of this model was tested using CFI differences (ΔCFI < 0.01), RMSEA differences (ΔRMSEA < 0.015), and SRMR differences (ΔSRMR < 0.03) as cut-offs [32]. If problematic, that is: values are above these figures, loadings across items were allowed to vary, to attempt to reach metric invariance. Problematic items were identified by assessing differences in standardised regression coefficients between the two groups.

The metric model is then compared to the scalar invariance model whereby factor loadings and intercepts are assumed to be equal across the two groups, so means of factor loadings can be compared. The cut-offs remain the same as the metric invariance model, but the SRMR cut-off is stricter (ΔSRMR < 0.015).

Results

Participants

Mothers in both the premature (n = 160) and term (n = 160) groups were predominantly from the United Kingdom of Great Britain and Northern Ireland and the Republic of Ireland (52% vs. 59%), married (63% vs. 81%), primiparous (66% vs. 76%); with both groups demonstrating similar levels of education to either undergraduate or equivalent degree level (34% vs. 39%). See Tables 1 and 2 for full demographic characteristics.

Table 1 Maternal demographic characteristics, split by the premature and term groups
Full size table
Table 2 Infant demographic characteristics, split by the premature and term groups
Full size table

Configural invariance

The model showed a good fit for the data (CFI = 0.99, TLI = 0.98, RMSEA = 0.03 [90%CIs 0.000 to 0.048], SRMR = 0.07), showing the factor structure was consistent across groups.

Metric invariance

The PSAS-RSF did not reach cut-offs for metric invariance (ΔRMSEA = 0.01, ΔCFI = 0.02, ΔSRMR = 0.01). Problematic items were identified by assessing differences in standardised regression coefficients between groups in the configural model (see Table 3).

Table 3 Standardised regression coefficients for the premature and term group
Full size table

Metric invariance was reached after the slope for one item was allowed to vary (I have felt frightened when my baby is not with me; ΔRMSEA = 0.01, ΔCFI=-0.01, ΔSRMR = 0.06). This model was also tested for scalar invariance, with loadings and intercepts for the problematic item being free to vary, and this was achieved (ΔRMSEA = 0.009, ΔCFI=-0.015, ΔSRMR = 0.006).

However, it is also notable that a comparison of the regression slopes between groups for three items with high differences in standardised regression coefficients were significant: I have worried I will not know what to do when my baby cries (p <.001); I have worried about my baby being accidentally harmed by someone or something else (p <.001); I have worried more about my finances than before my baby was born (p =.023), indicating these items may also be problematic.

Discussion and conclusions

The aim of this study was to conduct a measurement invariance analysis to assess the performance of the PSAS-RSF in mothers of premature and term infants. Results demonstrate the PSAS-RSF achieves configural invariance, and so retains the original four-factor structure. However, it is clear there are several problematic items in the current version of the PSAS-RSF, which are being answered differently in mothers of premature infants.

The item “I have felt frightened when my baby is not with me” is being interpreted differently in mothers of premature infants on the PSAS-RSF. Extended periods of maternal-infant separation are relatively commonplace with premature infants, particularly at or shortly after the point of birth [33], and may contribute to feelings of anxiousness when separated. Admission to the neonatal intensive care unit [NICU] can contribute feelings of physical separation from the infant [34]. It may also be the environment of the NICU itself leads to feelings of separation, with parents often reporting the physical space around the infant is dominated by medical equipment, meaning they find it difficult to remain close to the infant [35]. Furthermore, the constraints of hospital policies mean mothers can feel further apart from their new baby [36]. This is important as prolonged separation, particularly considering the NICU, can lead to impairments in attachment and bonding [37, 38], which can increase anxiety [39].

Parental concerns for preterm infants also extend beyond their baby’s physical health. Financial concerns following the birth of a preterm baby are inversely associated with gestational age [40], and financial burdens of caring for a preterm baby are further exacerbated if the infant is in hospital for a prolonged period after birth, and can lead to a reliance on support networks [41]. Notwithstanding the financial burdens associated with hospital care immediately after birth, significant costs can persist right across the first year postpartum. For example, infants with severe health complications as a result of their birth may require heightened medical care once at home, including specialised equipment required to further support development, and follow-up appointments, which may lead to time off work extending beyond parental leave [42]. Furthermore, the NICU has been associated with significant out-of-pocket expenses, including childcare for other siblings, and direct medical costs which cause significant concern even in mothers who have paid parental leave [43]. It is also important to consider, as this study was international, the cost of healthcare may also be a contributing factor. For example, estimates suggest privately insured families in the United States may still incur out-of-pocket costs relating to NICU care in excess of $10,000 [44]; and in the UK, costs can reach approximately £75,000 for the earliest preterm infants, with all of this cost absorbed to the National Health Service [NHS] [45].

Anxieties surrounding infant health and accidental harm, as well as concerns surrounding infant care may also be increased in women with babies in NICU. They have already been found to be more prevalent for mothers with the earliest preterm infants [7]. The environment and constraints of the NICU, including handing over routine care of the infant to medical professionals can lead mothers to feel powerless and disconnected from their infant [19], which can further contribute to anxiety, leading mothers to feel uncertain about their parental role. This can lead to feelings of over-protection towards the infant [45], including fears regarding accidental harm and health. Mothers often feel uncertain about their parental role [46], and so may feel unsure and alone when the environment of the NICU is removed [47].

Limitations

This study is the first to conduct a measurement invariance analysis of the PSAS-RSF across mothers of term and premature infants. However, the sample were predominantly White, married women from the United Kingdom of Great Britain and Northern Ireland and the Republic of Ireland. Future research should make a concerted effort to recruit women from diverse backgrounds, particularly as anxiety levels amongst mothers of premature infants can differ by race [48], and women from minority ethnic groups are less likely to engage with perinatal mental health services [49]. The results of the current study can inform modification of the current version of the PSAS-RSF for use in mothers of premature infants. Adaptation of the scale can be further informed by stakeholders and mothers of premature infants, before ensuring validity and reliability of any new measure.

Conclusion

We conducted a measurement invariance analysis of the PSAS-RSF across mothers of term and premature infants. Results support the retention of the four-factor structure found in other variations of the PSAS as the measure achieved configural invariance, however, potentially problematic items include those around infant care, safety, and psychosocial adjustment to motherhood. Future research should now modify the PSAS-RSF for use in this specific population of mothers of premature infants.