Psychometric assessment of the US person-centered prenatal and maternity care scales in a low-income predominantly Latinx population in California | BMC Women’s Health

In this paper we describe the psychometric properties of the PCMC-US and PCPC-US scales in a low-income predominantly Latinx-identifying population in California. The findings provide support for the construct, criterion, and know-group validity, as well as the reliability, of the scales in this population. The average inter-item correlations were between 0.20 and 0.40, for both scales, which falls within the ideal range for items in a scale, suggesting that while the items are reasonably homogenous, they contain sufficiently unique variance to not be isomorphic with each other.

Although a few items have inter-item correlations and factor loadings below recommended cut-offs, exclusion on these items do not substantially change the Cronbach alpha values which are all well above the recommended levels of 0.7. Further, the 20 item-version of the PCPC-US scale and 30-item version of the PCMC-US scale, which include only items with the best fit, are strongly correlated with the full 26- and 35-item versions used in this analysis, suggesting these sub-set of items are adequate to capture the overall levels of PCPC-US and PCMC-US if a smaller sub-set of items is desired. Moderate correlations between the PSQ and MORI suggest that while they measure related constructs, they capture different domains. Finally, the scales can discriminate by groups that are likely to have different experiences, despite a relatively homogenous sample.

The findings of this study regarding validity and reliability of the scales in a sample of low-income predominantly Latinx-identifying participants are generally consistent with the original validation studies, which consisted of predominantly high-SES Black women [33, 34]. Specifically, the performance of the overall scales is similar to the original study in terms of high validity and reliability. There were, however, some differences in the item loadings: although all scale items loaded well in the in the initial validation analysis, a few items loaded poorly in this sample of Latinx-identifying participants. This is likely due to difference in the distribution of these items, potentially due to differences in the interpretation of the questions or differences in the actual care received.

The average scores across the studies are also similar: the PCPC score in the prior study was 91.8 (SD = 11.1) [33], compared to 92.01 (SD = 11.3) in the current study, while the PCMC score was 89.1 (SD = 14.0) [34], compared to 90.15 (SD = 12.93) in this study. The high scores are, however, surprising given that the study sample is a low-income minoritized population, whose experiences may be influenced by both racism and classism [16, 20, 51]. One possible reason for the high scores is that all participants were participating in an enhanced prenatal care program in which person-centered care is prioritized.

The analysis on factors associated with receipt of PCPC-US and PCMC-US are also consistent with prior studies with minoritized groups [17, 21, 22, 24, 52]. Even in this relatively homogenous group, the intersection of racism and classism is still reflected in the differences in scores based on immigration status (captured by birth country), SES (captured by unstable housing), and in worry about experiences of discrimination. The non-significant differences in other measures of SES such as having public or no insurance, employment, education, and income, in the final models, is likely due to lower variability in these measures in the low-income population, with ever having been homeless being the most discriminating factor. Similarly, little variability in racial and ethnic identities limits inferences on these, but the finding on discrimination reflects differences based on interpersonal experiences of racism [16, 53, 54].

The challenges of providing person-centered care during the early phase of the COVID-19 pandemic when several restrictions were put in place, including limiting birth companions, are also captured by the scales, reflected in the lower PCMC scores in the earlier phases of the pandemic [55, 56]. Other variables found to be associated with person-centered care in prior studies such as late start of prenatal care and lack of continuity of care and racial discordance with provider were not measured in the current study [52, 57,58,59].

Overall, the findings of our current study are consistent with existing literature regarding the characteristics of person-centered care for Latinx-identifying patients [60,61,62,63]. In a qualitative study of Latinx-identifying prenatal patients, Bergman and Connaughton reported 5 key themes of patient-centered care, including: a friendly relationship, effective medical care, spoken Spanish language, understanding of medical information, and elimination of racism [from the healthcare setting] [60]. They emphasized that training health care staff on the importance of displaying friendly communicative behaviors engenders greater trust in the healthcare team. Three additional studies had similar findings, reporting that cultural and linguistic competence were the most important factors informing person-centered care of Latinx-identifying patients [61,62,63]. To address the need for cultural competence during prenatal care, these authors conducted a follow up study of group prenatal care for Latinx-identifying pregnant people and found that patients participating in culturally competent group prenatal care experienced greater satisfaction and engagement with their care (e.g., more likely to establish a medical home for their child, and attend their postpartum appointments) [64]. Furthermore, multiple studies have found that addressing the need for linguistic competence includes the use of professional interpretation rather than ad hoc interpretation [21, 61, 65, 66]. Doing so improves patient trust and satisfaction [65]. Interestingly, in this study, completing the survey in Spanish was associated with higher PCMC scores in bivariate analysis. This may suggest that the more positive PCMC scores of Spanish-speaking participants reflect linguistically and culturally competent care. Feedback from our field staff also suggest that although language barrier influenced experiences of participants, the Spanish speaking foreign born participants tended to feel so grateful to have care that they were not bothered much by how staff treated them, which might explain the higher scores.

Strengths and limitations

This is the first study to assess the psychometric properties of the PCMC-US and PCPC-US scales in a Latinx population. Theses scales were developed using a community-engaged approach embedded in standard instrument development methods. Starting with validated tools provided a rigorous, evidence and theory-based foundation. Expert reviews and cognitive interviews with people from racial and ethnic minoritized groups ensured content validity as well as relevance to the experiences of people in those groups. The current study provides additional evidence of construct, criterion, and discriminant validity in a Latinx population. A potential limitation is generalizability to other racial and ethnic groups and people of other SES, given this was a low-income Latinx population. However, given the scale performed similarly well in the previous validation in a high-income Black population, the findings suggest the scale will likely have similar levels of validity and reliability in diverse populations. Given the major need for cultural and linguistic adaptations in the care of Latinx patients, use of validated instruments such as the PCPC-US and PCMC-US scales provide valuable measures for interventions aimed at improving person-centered care and addressing existing inequities in obstetric and perinatal outcomes. Validations in other populations are however needed to ensure their appropriateness in other populations besides Black and Latinx individuals. Further, although the scales performed well in self-administered surveys in the previous validation in a predominantly high-SES Black population, the data collection in this study was only via interviewer-administered surveys. Thus, we are unable to speak directly to how well the scales will work as self-administered surveys in the low-SES predominantly Latinx population. Future studies should examine this.

Respondent burden due to the length of the scale also may be a limitation. Several items are needed given the multidimensional nature of person-centered care and the assessment of the relevance of the items included during the initial validation activities [33, 34]. However, given the high correlation between the 20- and 26-item versions of the PCPC-US and the 30- and 35-item versions of the PCMC-US scales, these shorter versions can be used where abbreviated scales are desired. While the longer scales may be more helpful for quality improvement efforts where the goal is to identify specific behaviors for improvement, the shorter scales can be used where only the summative score is needed. The sub-scales can also be used individually where necessary, although we recommend measuring all three domains to assess PCPC and PCMC in a holistic manner.

Finally, in our data the scale scores were highly left skewed, requiring the use of a non-parametric method (bootstrapping) in the multivariate analysis with the scores as the outcomes. Thus, when the scores are used as outcome variables in statistical analyses, the distribution of the scores should be examined. In situations where the distribution is highly non-normal appropriate statistical methods, such as nonparametric methods (e.g., bootstrapping or rank-based methods), should be used. Most statistical analyses, however, make no distributional assumptions about covariates. So, the left-skew should not be an issue if the scales are being used as covariates.