Abstract
The intense emotional and psychological consequences of pregnancy loss have been studied for decades. With the growing body of literature regarding women's experiences of pregnancy loss, there is an increased need for high quality evidence that underpins perinatal grief and bereavement support interventions. In this paper, we describe and analyze the psychometric properties of existing tools (the Perinatal Grief Scale, the Perinatal Bereavement Scale, the Perinatal Grief Intensity Scale) developed specifically to measure grief following perinatal loss. The psychometric properties of these instruments are presented along with an assessment of their trustworthiness. Of the three perinatal grief instruments, the PGS is the most well-established measure of perinatal grief. The PBS, although promising, has not been as extensively tested as the PGS. The PGIS has compared favorably with the PGS and is also a good choice for use in perinatal bereavement studies despite its shorter history of use and less extensive psychometric testing.
Keywords
Introduction
Despite major technological advances in women's health care, approximately 23 million pregnancies end unexpectedly each year worldwide. Over 10% of women experience miscarriage within their lifetime, and there are 44 pregnancy losses occurring every minute (Quenby et al., 2021). Further, the incidence of pregnancy loss has increased during the COVID-19 pandemic (Sacinti et al., 2021). Given the frequency of such losses and awareness of the impact on short- and long-term mental health, perinatal loss research has increased substantially over that past several decades. As a result, researchers have focused on elucidating many of the emotional and social consequences of perinatal loss.
Following unexpected pregnancy loss, women report feelings of grief, guilt, anger, anxiety, depression, shock, emotional pain, ruminative thoughts, and maladaptive cognitions (Camacho-Ávila et al., 2019; Gausia et al., 2011; Gozuyesil et al., 2021; Kokou-Kpolou et al., 2018; Ozgen et al., 2022; Punaglom et al., 2022; Sutan & Miskam, 2012; Testoni et al., 2020). Besides the emotional aspects of loss, consequences of perinatal loss include strained social relationships with spouse and the spousal family, changed view of subsequent pregnancies, social isolation, and social pressure (Furtado-Eraso et al., 2020; Gausia et al., 2011; Roberts et al., 2021). Socio-cultural norms and religious beliefs strongly influence maternal grief and loss rituals (Furtado-Eraso et al., 2020; Markin & Zilcha-Mano, 2018; Punaglom et al., 2022; Roberts et al., 2021; Sutan & Miskam, 2012).
Grief responses are inarguably intertwined with palliative care research and practice; the two are “mutually inclusive” (Moon, 2013). In recent years, US and international palliative care organizations have placed greater emphasis on grief and bereavement support (National Coalition for Hospice and Palliative Care, 2018; World Health Organization, 2020). In fact, the 2022 World Hospice and Palliative Care Day centered on the theme of grief (Hospice Palliative Care Association of South Africa, 2022; Worldwide Hospice & Palliative Care Alliance, 2022). With awareness increasing, the need for high quality evidence is essential to inform palliative care interventions that support grief and bereavement. Importantly, the evidence generated from clinical interventions is only as good as the tools used to assess the multiple dimensions of grief (Coomarasamy et al., 2021).
The purpose of this paper, then, is to describe and analyze the psychometric properties of existing tools developed specifically to measure grief following perinatal loss. Within this paper, the term perinatal loss will be used to refer to all unexpected losses of pregnancy, regardless of gestational age. The reason for using this broad term for both early and late pregnancy losses is to encompass all types of pregnancy loss, and to give a broader scope to the review of the literature. However, in discussing the extant literature, terminology used by original researchers will be retained in an effort to remain true to the work of others, and to highlight the diversity of terminology used to describe pregnancy loss.
Methods
A review of literature was conducted in electronically in 2022 using several databases, including CINHAL, Medline, PubMed, PsychInfo., and Health and Psychosocial Instruments. Search terms were chosen and used in various combinations to reflect grief, perinatal grief, grief and miscarriage, perinatal loss, maternal grief, maternal bereavement, grief instruments, and grief scales. Articles were chosen for their appropriateness to the discussion of instrument development and psychometrics. The reference lists of pertinent articles also were reviewed for instruments or articles that might have been overlooked in electronic searches. However, no additional instruments were found. We identified three commonly used instruments (Hutti et al., 1998; Theut et al., 1989; Toedter et al., 1988) and will present each in chronological order, with a focus on each instrument's development and psychometric properties (see Table 1). Instruments that were modified for the purpose of measuring perinatal grief (Nikcevic et al., 1999), or developed to measure only certain dimensions of perinatal grief (Ritsher & Neugebauer, 2002) were excluded.
A Brief Overview of the Psychometrics of Perinatal Grief Instruments.
Findings
Perinatal Grief Instruments
The Perinatal Grief Scale
The Perinatal Grief Scale (PGS) (Toedter et al., 1988) was developed as part of the Perinatal Loss Project (1984), a longitudinal study of the factors that affect perinatal grief. Because perinatal grief had been shown in previous work to have unique features compared to grief stemming from other types of losses, the researchers forewent use of standard grief scales, such as the well-established Texas Grief Inventory Scale. To develop the PGS, researchers incorporated antecedents and consequences of perinatal grief from the literature, such as previous loss, previous birth, length of gestation, quality of the marital relationship, as well as demographic information. Other important variables such as mental health of the mother, fertility, religiosity, the woman's perception of her own health, and the likelihood of pregnancy after the loss were included.
The researchers established construct validity of the 104-item instrument using the three-step process set forth by Carmines and Zeller (1979): 1.) specifying theoretical relationships among concepts, 2.) examining empirical relationships among concepts, and 3.) clarifying construct validity using empirical evidence. The instrument was then administered along with a marital relationship scale, a researcher-developed religiosity scale, the Symptom Checklist-90 (SCL-90), and two single items: a global indicator of the participant's health and the likelihood of future pregnancy. One hundred thirty-eight women and 56 partners referred through physician's offices completed the instruments during semi-structured interviews at 6–8 weeks, 1 year, and 2-years post-loss. In order to establish baseline information, the researchers used the uncommon retrospective pretest method to randomly select pregnant women participants who were then matched to the post-loss participants according to the trimester of the pregnancy in which the loss occurred. This procedure was done in order to address concerns that a woman's perception of the pregnancy would be altered as a result of the loss.
To support internal consistency of the PGS, the researchers reported that the PGS had an alpha of 0.90; however, 20 items had a corrected item-total correlation of less than 0.20. The item-total correlation, also called the item-scale correlation, is a measure of how well an individual item correlates with the other items on the scale, excluding itself (DeVellis & Thorpe, 2021). Since item-total correlations exclude the perfect correlation between an item and itself, they give a clearer picture of the instrument's internal consistency. Toedter et al. (1988) noted that after the 20 items with low correlations were dropped, the mean corrected item-total correlation was 0.52, and the alpha coefficient increased to 0.97. Such high alpha coefficients may indicate redundancy among items indicating that the large number of items may have artificially inflated the alpha coefficient (Carmines & Zeller, 1979).
To test the theoretical framework, the researchers first performed multiple regression, using a forward-stepwise solution with the beta weight set at p ≤ 0.05. The four variables found to have the greatest impact on perinatal grief were: (1) overall physical health of the mother (partial correlation = – 0.258), (2) gestational age (partial correlation = 0.361), (3) marital relationship (partial correlation = −0.265), and (4) pre-loss mental health symptoms (partial correlation = 0.221). Secondly, the researchers used multiple regression analysis to develop a model of the direction in which each variable affected grief scores. Notably, all of the relationships fit the theoretical predications.
To determine the factor structure, the researchers used varimax rotation and attempted both a 3-factor and 6-factor rotation. The amount of variance explained by the model did not greatly improve with the 6-factor solution, so the 3-factor solution was retained, which also made more sense theoretically. The three factors were named Active Grief, Difficulty Coping, and Despair and within each factor, items with a factor loading of 0.50 or greater were retained. The subscale alphas were: 0.59, 0.55, and 0.51 respectively, indicating that all of the items in the subscales measure the same construct, which helped to further support scale reliability. The instrument showed strong preliminary reliability and validity and was later used in numerous studies of perinatal grief. But, because the original PGS included 84 items and was burdensome to administer and score, the researchers sought to reduce the number of items in the instrument, while maintaining reliability and validity. Participant burden, particularly given the phenomenon under study, was not originally addressed during PGS testing, but was addressed by developing the newer PGS Short-Form (SVPGS).
The Perinatal Grief Scale - Short Form (SVPGS)
Using the data obtained in the original PGS study, Potvin et al. (1989) performed several analyses to determine whether the PGS could be condensed without sacrificing psychometric adequacy. To determine which items should be deleted, the researchers analyzed the inter-item correlations, and dropped items that had low correlations with others in the same subscale; the cut-off was not reported. Then, maximum-likelihood factor analysis was used to determine which items did not belong within a single factor. Netemeyer et al. (2003) explained maximum likelihood factor analysis as a form of confirmatory factor analysis that should be done only after exploratory factor analysis. Since the researchers conducted exploratory factor analysis in the original study, they used confirmatory factor analysis to validate the factor structure of the revised scale to reduce the number of items (Netemeyer et al., 2003).
To further evaluate psychometric stability, Potvin et al. (1989) performed five additional analyses to assess: reliability of the subscales and the total instrument using the alpha coefficients, factor structure with the reduced number of items, distribution of subscale scores, consistency of results between the long and short versions, and reliability of the instrument. In the first analysis, the alpha coefficients revealed that the revised 33-item scale closely paralleled the 84-item scale, with overall alphas of 0.95, and subscales alphas of 0.92 (Active), 0.91 (Difficulty Coping), and 0.86 (Despair). During the second analysis, factor structure was explored using the 11 items retained in each subscale. The three-factor solution using the 33-item PGS accounted for 49.8% of the total variance, but since the percent of variance explained by the factors was not reported in the original instrument development study, the researchers were not able to determine if the lower number of scale items were consistent with and comparably represented each of the original PGS factors.
As a means of assessing reliability of the 33-item scale, the researchers examined test-retest reliability. Potvin et al. (1989) reported that because the original study using the 84-items PGS was longitudinal (using the retrospective pre-test), test-retest reliability could be supported by using correlations between scores obtained at 12- and 15-months post-loss for the portion of the sample (112 of 138 women) who were retained. The researchers acknowledged that test-retest is generally a measure of reliability for stable traits, and that grief scores were expected to decrease over time, but cited the correlations of 0.59 to 0.66 (p < .001) as evidence of temporal and factorial stability. However, Deyo et al. (1991) described the ability of an instrument to measure changes in a construct over time as the instrument's responsiveness, noting that responsiveness is measured using effect size or paired t-tests, which could have been calculated using the data obtained in the PGS studies.
Even so, the researchers demonstrated a thorough attempt to support the reduction of items included in the PGS from 84 to 33. Evidence was presented to demonstrate that the 33-item version reflected the dimensions of perinatal grief as adequately as the 84-items version. According to Netemeyer et al. (2003), the next step to support psychometric stability involves evaluating the confirmatory factor analysis, which Toedter et al. (2001) undertook through meta-analysis.
Meta-Analysis of PGS - Short Form (SVPGS)
To begin the meta-analysis of studies that had used the PGS short form, Toedter et al. (2001) excluded all studies that were published using the original sample from the Perinatal Loss Project, which included the introduction of the short-form of the PGS. They found that twenty-two studies, many of which were doctoral dissertations, had been completed in four countries (U.S., Netherlands, Germany, and England), with a total sample of 2,485 participants. In those studies, the PGS had been used to measure grief responses to various types of perinatal losses.
Toedter et al. (2001) compared the results of the 22 studies to gather psychometric data, establish norms for the PGS scores, and compare the results across samples. Based on the results of the meta-analysis, normative data were established using the mean scores and standard error for the data collected in all of the studies. From the analysis, the researchers (Toedter et al., 2001) noted that a total PGS score above 91 would reflect a high degree of grief and that overall PGS scores progressively declined across subscales, indicating that fewer participants experienced despair than active grief.
The conclusions drawn from the meta-analysis have been bolstered with data from numerous studies using the PGS in the past two decades. The PGS has been used in research to measure perinatal grief following pregnancy loss, including grief during a pregnancy subsequent to perinatal loss (Barr & Cacciatore, 2008; Beauquier-Maccotta et al., 2022; Kulathilaka et al., 2016). It has also been used to determine the relationships among perinatal grief and factors such as number of pregnancy losses, gestational age at time of loss, religiosity, ruminative thoughts, posttraumatic symptoms, and posttraumatic growth (Cowchock et al., 2010, 2011; Gozuyesil et al., 2021; Krosch & Shakespeare-Finch, 2017; Ozgen et al., 2022; Purandare et al., 2012; Penelo et al., 2017). In experimental studies, the PGS has been used to assess the effectiveness of clinical interventions (Fernlund et al., 2021; Johnson & Langford, 2015). The PGS has been translated to Spanish, Turkish, Arabic, Chinese, Persian, Italian, and Greek (Al-Maharma et al., 2016; Capitulo et al., 2010; Könes et al., 2017; Lai et al., 2013; Maniatelli et al., 2018; Ravaldi et al., 2020; Siadatnezhad et al., 2018; Yan et al., 2010), has demonstrated good reliability and validity in all translations, and has been widely used internationally.
The widespread use of the PGS and the consistency of the factor structure support the construct validity of the instrument. Convergent evidence from all of the studies that used the PGS showed that perinatal grief declines over a period of two years, regardless of the length of gestation. Also, all studies using the PGS reported that social support and strong marital relationships were factors that consistently decreased PGS scores. Lastly, poor mental health of a mother prior to the pregnancy loss, rumination, and multiple losses were strong predictors of high PGS scores (Gozuyesil et al., 2021; Purandare et al., 2012; Toedter et al., 2001).
The tool has an established cutoff score and can be used to assess the efficacy and effectiveness of formal bereavement support interventions (Johnson & Langford, 2015). The PGS is undoubtedly a strong research measure and can be used in many populations of bereaved parents, including those experiencing pregnancy after pervious perinatal loss.
The Perinatal Bereavement Scale (PBS)
The Perinatal Bereavement Scale (PBS) (Theut et al., 1989) was developed to measure and compare grief responses of parents following miscarriage, stillbirth, or neonatal loss following the birth of a subsequent child to identify unresolved grief. The instrument consisted of 26 Likert-scale items scored on a 4-point scale ranging from A (almost all of the time) to D (almost never). The items are scored from 1 (D) to 4 (A), meaning scores can range from 26 to 104, with higher scores indicating more severe grief responses. The instrument was distributed to 25 pregnant women who had experienced perinatal loss within the previous two years, and their spouses. The participants were asked to complete the instrument during the eighth month of a subsequent pregnancy and again six weeks after birth. A follow-study was conducted at 16 months post-partum and provided further validation of the instrument as well as new insights into perinatal bereavement (Theut et al., 1990).
The alpha coefficients were computed for each group (mothers and fathers) at both times (pre- and post-natal). Pre-natal alphas were 0.88 (mothers), and 0.84 (fathers) and post-natal alphas were 0.91 (mothers) and 0.83 (fathers). Typically, alpha coefficients over 0.70 for the entire instrument indicated that the scale items are all measures of the chosen construct, in this case perinatal grief. The researchers, therefore, reported that the scale items adequately represented the construct.
Other psychometric properties of the instrument were not reported, such as the establishment of content validity, face validity, or instrument reliability. But, the alpha coefficient points toward internal consistency of scale items, which helps to establish a case for instrument reliability. Factor analysis is often used to help support reliability (Carmines & Zeller, 1979), but whether this was attempted was not reported. The instrument was included in the initial report, and could be utilized with other samples to help establish validity, which researchers suggested is an avenue for future inquiry (Theut et al., 1989, 1990)
Only one study was identified in which perinatal grief was measured using the PBS. In a sample of three bereaved mothers, Rosa (1996) determined that grief responses are individualistic and vary over time. Although generalizable conclusions cannot be drawn with a sample of only 3 participants, all of the participants reviewed and endorsed the PBS, lending further evidence of face validity of the tool. However, a weakness of the PBS is the lack of a cutoff score (Rosa, 1996), which limits use of the PBS in clinical decision-making.
The Perinatal Grief Intensity Scale (PGIS)
Realizing that existing perinatal bereavement instruments had been developed using literature that was limited in some aspects and conflicting in others, Hutti (1986) developed a tool to measure the intensity of perinatal grief using participant data (e.g., words describing loss) from qualitative research studies (Hutti, 1986, 1992). In contrast to the PGS, which was intended to measure perinatal grief following several types of losses, the PGIS was specifically designed to measure grief following early miscarriage. However, the tool was later tested in samples that included women who had experienced stillbirth and neonatal losses, demonstrating acceptable reliability and validity in all groups.
Face validity of the 36-item, Likert-type instrument was established by 10 maternity nurses and 10 women who had experienced miscarriage and items were modified based on their feedback. The instrument was then administered to 186 women who had experienced miscarriage in the prior 12–18 months. Factor analysis was performed using pairwise and oblique rotations because the researcher expected correlation between the scales. Of the original 36 items, 22 items were dropped after factor analysis. The remaining 14 items loaded onto three factors, using 0.40 as the cut off value. The eigenvalues for the factors were 4.53 (Reality of the pregnancy and the baby within), 3.11 (Ability to confront others), and 1.46 (Congruence between the actual experience and the standard of the desirable) (Hutti, 1986). The three factors accounted for 65% of the variance in the original development and validation study and 66.94% in subsequent instrument testing (Hutti et al., 2013).
Initial alpha coefficients for the subscales were 0.89 (Reality), 0.84 (Ability to confront), and 0.71 (Congruence), and the overall alpha coefficient for the PGIS was 0.82, which is considered quite good, especially on initial development. Subsequent research with pregnant women who had experienced miscarriage, stillbirth, or neonatal loss further supported internal consistency of the tool with subscale alpha coefficients of 0.82 (Reality), 0.80 (Ability to confront), and 0.80 (Congruence) and an overall Cronbach's alpha for the PGIS of 0.75.
Construct and face validity for the PGIS were initially demonstrated through the original qualitative data; however, this is admittedly a weak method of supporting validity (Trochim, 2001). Later, validity was tested through concurrent use with anxiety (Pregnancy Outcome Scale) and depression (Center for Epidemiologic Studies Depression Scale) measures in a study of 227 women who were experiencing pregnancy after previous perinatal loss. In that study, the PGIS “demonstrated construct validity via significant and appropriate directional relationships with pregnancy-specific anxiety [and] depression symptoms (Hutti et al., 2015, p. 49).
Reliability of the instrument was initially supported through factor analysis, which also supported model congruence. However, the instrument also has demonstrated reliability through its use and excellent performance in several subsequent studies (Hutti et al., 2013, 2015, 2018). When compared to the PGS, which has well established reliability and validity, the PGIS performed comparably and has the advantage of lower response burden (Hutti et al., 2018). In comparing the two scales, one critique of the PGIS was that a cutoff score had not been established. But, a recent study with 103 women who had experienced miscarriage, stillbirth, or neonatal loss indicated that the optimal cutoff score for the PGIS is 3.53, with higher scores indicating more intense grief responses (Hutti et al., 2018). Further, at the cutoff score of ≥ 3.52, the PGIS showed 97.9% sensitivity and 29.6% specificity for predicting depression symptoms 3 to 5 months after loss. At the same cutoff score, the PGIS predicts intense anxiety at 3 to 5 months post loss with 95.2% sensitivity and 56.2% specificity.
Discussion
In this review, the psychometric properties of three commonly-used instruments designed specifically to measure grief following perinatal loss were presented along with an assessment of their trustworthiness. Over time, the development and use of the PGS, SVPGS, PBS, and PGIS has provided further psychometric support for each instrument as well as a deeper understanding of the consequences of perinatal loss. Additionally, the evidence gathered through use of the instruments in perinatal bereavement research has provided empirical evidence regarding the effectiveness of clinical support interventions (Wang et al., 2022).
Based on the overwhelming data presented in this paper, it is clear that the PGS, (including its shortened version, the SV-PGS), has been the most widely used and translated perinatal grief instrument, and has shown stable results in several countries for over four decades. Therefore, of the three perinatal grief instruments, the PGS is the most widely-accepted and well-established measure of perinatal grief. The PBS, although promising, has not been as extensively tested as the PGS. However, the PGIS has compared favorably with the PGS and is also a good choice for use in perinatal bereavement studies despite its shorter history of use and less extensive psychometric testing.
One limitation of all of the instruments presented is the lack of stakeholder input in the initial development and subsequent use. Although the PGIS was developed using qualitative data, it is unclear how or if the participants were involved in helping to shape the tool throughout its development. People with first-hand experience are important stakeholders because they can provide important feedback to researchers regarding the content of survey instruments, and cultural adaption of tools, including linguistic nuances, which is especially critical when researching sensitive topics such as grief (Moore et al., 2022). Other stakeholders such as organizations that serve and/or advocate for those who have experienced perinatal loss as well as clinicians who work with perinatally-bereaved parents can provide nuanced insights regarding support interventions and needed resources. Even so, it can be challenging for researchers to identify and engage stakeholders, and to appraise the value of their contributions to the research project (Forsythe et al., 2016).
Beyond the challenges of obtaining broad stakeholder involvement, there are several general challenges related to measuring grief and bereavement such as: 1.) recall bias, 2.) participant burden, which can result in reduced completion rates and missing responses and 3.) disentangling grief responses from other issues affecting mental health such as depression, anxiety, or posttraumatic responses (Drews et al., 1990; Hvidtjørn et al., 2018; Kersting & Wagner, 2012; Rolstad et al., 2011). Perinatal loss research specifically entails unique challenges since there is the potential that different types of perinatal losses (i.e., early loss, late loss, loss of one twin, loss after infertility) may result in different grief manifestations. Therefore, psychometric testing of perinatal grief instruments should be conducted with populations whose experiences represent all types of losses.
Another useful contribution to the perinatal loss literature would be administer the PGS, PBS, and PGIS instruments (a total of 83 items) concurrently. Such a study would provide convergent and divergent validity testing and ascertain whether the scale items overlap, or if certain items could be combined. The results could support the development of one unified instrument grounded in decades of research and informed by those with direct perinatal loss experiences. Throughout the instrument development and testing process, the involvement of a stakeholder panel would be essential. The result would be a parsimonious, valid, and reliable tool with expanded applications to perinatal loss research.
Footnotes
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
