Phase 1: Translation of the PCMC scale and pretest

First, we obtained the permission to use the PCMC scale from the developer of the original PCMC scale, for the purpose of its adaptation to Cambodian contexts. A total of 31 items were used as initial item pool, i.e., all the original 30 items validated in Kenya [16] and one additional item ‘being asked to pay bribe’ validated in India [17]. The overall Phase 1 procedures were performed, by following the WHO guideline on translation and adaptation of instruments [24]. Important considerations for cross-cultural research are conceptual and cross-cultural equivalence.

The English version of the original PCMC scale items were first translated into Khmer language, employing the team translation approach. The translation team was composed of one bilingual English-Khmer translator, one bilingual Japanese-Khmer translator and two trilingual English-Khmer-Japanese linguistic experts. It was subsequently reviewed by eight Cambodian content experts to identify unnatural expressions and to review cultural appropriateness for the Cambodian context. The Khmer version of the PCMC (Kh-PCMC) scale was then pre-tested among 20 Cambodian postpartum women using cognitive interviewing from 20 January to 28 March 2021. We spent a substantial amount of time attempting to accurately capture the cultural context of Cambodia and select appropriate words and phrases. Discrepancies and nuanced translations were discussed and resolved among the translation team. The translation team and the tool developer approved the retention of the 31 items for use in a field survey. The details available in a separate paper [25].

Phase 2: A survey for psychometric assessment

Ethics

The study received ethical approval form the Ethics Committee, Faculty of Medicine, University of Tsukuba on 24 December, 2020: Reference: IRB1605, and National Ethics Committee for Health Research, Ministry of Health Cambodia on 30 December, 2020: Reference: #322 NECHR. Oral informed consents were obtained from all respondents prior to participation. Participants were informed their participation in the study was voluntary, and they had the right to refuse to participate in the study and could withdraw at any time after giving their consent without giving any reasons. Interviews were conducted in a private space of the facilities. Participants’ responses were kept confidential by de-identifying the data using a unique identifier code.

Respondent characteristics

A total of 300 postpartum Cambodian women were interviewed. Table 1 shows the demographic characteristics of respondents. The mean age of the women was about 29 years (range of 18 to 46) with the mean parity of 2.26 (range of 1 to 8 children). Almost all of the women were married (99.3%) and Buddhist (97.3%). About half (48.5%) of women had less than primary education, and 69.1% had some difficulty in reading the Khmer language or were illiterate; 6.6% were certified as the poorest to be exempt from paying medical expenses. The postpartum length for women interviewed was between one and seven days.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of 300 women.

https://doi.org/10.1371/journal.pone.0288051.t001

Data quality

All data were normally distributed. While seven items (#7,21,22,23,29,30,31) were of a particularly high mean (+1SD) of greater than 2.9 in upper limit of 3, we retained all items at this stage (S1 Table).

Score distribution of 30-item scale

The mean Kh-PCMC full score for the sample based on the sum of the original 30 items was 69.52 (SD = 9.47) with a range of 48 to 89 (where 0 is the worst score and 90 is the best score). The mean Kh-PCMC sub-scale scores for the sample were 16.01 (SD = 1.53) with a range of 8 to 18, 15.43 (SD = 3.92) with a range of 6 to 24, and 36.26 (SD = 4.38) with a range of 24 to 44, for dignity and respect, communication and autonomy, and supportive care, respectively. Standardized scores ranging from 0 to 100 are 77.02 for full score, and 88.94, 57.15, and 80.58 in three subscales, respectively. The score distribution and the distributions of 30-item Kh-PCMC scale are shown in S2 and S3 Tables, respectively.

Psychometric properties

(1) Content validity

Table 2 shows Content Validity Index (CVI) evaluation of the 31 items by eight content experts. The S-CVI/Avg (scale-level content validity index, average) was 0.96 and the S-CVI/UA (scale-level content validity index, universal agreement) was 0.74 with a total item agreement of 23 of 31 items (7 items at 0.87, and 1 item at 0.75).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. CVI evaluation on a 31-item scale by eight experts.

https://doi.org/10.1371/journal.pone.0288051.t002

(2) Structural validity

The KMO values of 0.83 and the Bartlett’s test of sphericity (Chi-squared value = 3484.092 and df 465, P < 0.001) indicated that the overall variables were satisfactory for factor analysis.

The initial exploratory factor analysis using principal factor with 31 items yielded ten factors with one dominant factor with eigenvalues of greater than one, (7.22, 2.284, 2.077, 1.75, 1.624,) respectively, accounting for 66.83% of the total variance. Because the original PCMC scale has a three-factor structure, the second exploratory factor analysis was performed using principal factor and Promax rotation assuming a three-factor structure. The second exploratory factor analysis with 31 items yielded three factors including one dominant factor, 17 items loaded on the first factor, 11 on the second factor, and three on the third factor. If we used a cut-off of 0.3, 11 items would be eliminated, leaving 20 items. While if we used a cut-off of 0.1, two items (physical abuse and verbal abuse) would be eliminated, leaving 29 items.

Another round of exploratory factor analysis using principal factor and Promax rotation with 20 items and 29 items (S4 Table) were performed. There was significant positive correlation between the first factor and the second factor for both 20 items (r = 0.56) and 29 items (r = 0.58). When we compare the scree plot, the “break” in the scree plot for 20 items after exploratory factor analysis showed steeper bend between the third factor and the fourth factor (Fig 1), indicating that a three-factor structure would be an appropriate and data-driven solution. The scree plot for the 29 items after exploratory factor analysis yielded one dominant domain, but did not show a clear three-factor solution (Fig 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Comparison of scree plot after exploratory factor analysis.

https://doi.org/10.1371/journal.pone.0288051.g001

The items and data were carefully analyzed, and the decision was made to eliminate 11 items with cut-off of less than 0.3 using a data-driven approach. The items excluded were “2. introduce themselves”, “7. record confidentiality”, “9. consent to procedures”, “21. verbal abuse”, “22. physical abuse”, “23. bribes”, “26. trust”, “28. clean”, “29. water”, “30. electricity”, and “31. safe”. The decision was made based on the following reasons: (1) all items had low factor loadings of less than 0.3; (2) item #7, 21, 22, 23, 29, 30, and 31 had particularly high mean (+1SD) greater than 2.9 in an upper limit of 3; (3) item #2 had low I-CVI (0.75); (4) items #28, 29, 30, and 31 are theoretically classified into health facility environmental dimension of quality of care, which may be conceptualized differently from experience of care (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Items for person-centered maternity care scale.

https://doi.org/10.1371/journal.pone.0288051.t003

Exploratory factor analysis of 20 items yielded three factors, 12 items loaded on the first factor, six items on the second factor, and two items on the third factor (Table 4). The three factors were named PCMC factor 1, PCMC factor 2, and PCMC factor 3 namely birth companionship, respectively. The items loading on each of the three factors did not represent clear conceptual domains, because the factors extracted included a mix of items from each of the original domains. For example, the first factor included “11. language”, “13. call by name”, “16. able to ask questions”, which conceptually should have loaded on the second factor, and “14. talk about feeling”, “19. attention when needed help”, “20. control pain”, which conceptually should have loaded on the third factor.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 4. Exploratory factor analysis result of 20 items.

https://doi.org/10.1371/journal.pone.0288051.t004

We, therefore, regrouped the retained items into three conceptual domains drawn from the “experience of care” dimension of the WHO Quality of Care Framework. However, some items loaded negatively on the theoretically derived domain and positively on the data-driven domain. The original three-factor structure was not reproduced from Cambodian data. Instead, the distribution of the items, cultural rationale, and the judgment from the tool developer was considered.

(3) Internal consistency

The 20-item Kh-PCMC scale had a Cronbach’s alpha of 0.86, suggesting good internal consistency. Cronbach’s alphas of the three subscales: PCMC factor 1, PCMC factor 2, and PCMC factor 3 were 0.85, 0.76, and 0.91, respectively (Table 5). In contrast the Cronbach’s alpha for the theoretically derived 30-item scale is 0.85, with Cronbach’s alphas 0.47, 0.77, and 0.68 for the dignity and respect, communication and autonomy, and supportive care subscales respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 5. Lists of reliability and validity.

https://doi.org/10.1371/journal.pone.0288051.t005

(4) Criterion validity

The 20-item Kh-PCMC full scale score was significantly correlated with satisfaction with care (r = 0.249, P < 0.001) and quality of care rating (r = 0.593, P < 0.001) (Table 5).

(5) Cross-cultural validity

The preliminary work for cultural translation and adaptation supported acceptable cross-cultural validity. The correlation between the 20-item and 30-item scale is 0.99 (P < 0.001).

Score distribution of 20-item Kh-PCMC scale

The 20-item Kh-PCMC scale with a four-point frequency response ranging from 0 to 3 (“0 = No, never”, “1 = Yes, a few times”, “2 = Yes, most of the time”, “3 = Yes, all the time”) was proposed. The item ratings were aggregated to scale scores by summing each item. The total possible summative score ranged from 0 to 60, with scores representing better person-centered maternity care. The mean scores of 20 item Kh-PCMC scale for this sample is 44.25 (SD = 8.68). The mean scores of the three subscales are 25.68 (SD = 5.87), 13.40 (SD = 3.79), and 5.18 (SD = 1.72), respectively. Standardized scores ranging from 0 to 100 are 73.75 for full the score, and 71.33, 74.44, and 86.33 for the three subscales, respectively (Table 6).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 6. Distribution of 20 item Kh-PCMC scale and subscales in Cambodia (n = 300).

https://doi.org/10.1371/journal.pone.0288051.t006

Factor interpretation

Due to the potential cultural and social differences, it is necessary to validate the PCMC scale in a different context. Nineteen of 30 items were common across Kenya (Africa), India (South Asia), and Cambodia (Southeast Asia), which enable meaningful international comparisons among very different settings. Our exploration found the items loading on each subscale differed from the original version, while the overall PCMC concept remained similar. For example, the first factor (dignity and respect) included “3. call by name”, “10. delivery position choice”, “11. language”, and “16. able to ask questions”, which conceptually should have loaded on the second factor (communication and autonomy), and “14. talk about feelings”, “19. attention when needed help”, and “20. control pain”, which conceptually should have loaded on the third factor (supportive care). Our finding may indicate that differing local contexts and cultures influenced the women’s experience of received care, which influenced item loading.

There are four potential explanations for the difference in item location. First, it is probably attributable to the overarching themes of the PCMC that produce meaningful interactions between the subscales. Our finding showed a significant positive correlation between the subscales of “dignity and respect” and “communication and autonomy” (r = 0.51). This is consistent with the original version in which the subscales were shown to be strongly correlated with each other, with correlation coefficients (r) ranging from 0.53 to 0.63, and with the main scale (r = 0.75, 0.86, and 0.9 for dignity and respect, communication and autonomy, and supportive care, respectively) [16]. The original PCMC scale was developed as a theory-based practical tool that can be easily administered in various contexts [16]. A recent study proposed a unidimensional 13-items PCMC short scale using a data-driven approach that could be applied to multiple settings [40]. Thus, there may be flexibility of which items fits which subscales according to the context.

Second, the difference in item location could be explained by contextual difference. This is supported by the previous validation studies which showed that the factor loading was different between urban and rural populations within Kenya [16], and the factor loading from Indian data was also different from the conceptual domains [17]. Because the total number of respondents were 1,407 in Kenya [16], and 2,018 in India [17], the difference was not due to sampling issues. Rather, the factor structure may differ across different contexts and different sub-populations. In that sense, our findings reflected local reality where the concept of person-centered maternity care was not yet familiar and not commonly practiced [41]. The original PCMC scale consists of three conceptual domains, however, there may not have been clear differences among “dignity and respect,” “community and autonomy,” and “supportive care” for our respondents under the current situation in Cambodia.

Third, language issue related to the equivalence of translation may be another potential reason. The PCMC scale was validated in Kenya and India, where English is one of the official languages and the interviews were conducted in English, Swahili, and Luo in Kenya, in Hindi in India, respectively. On the other hand, in Cambodia, the official language and interview language were in Khmer. The limited vocabulary of the Khmer language and the issue that English is not commonly used in the country may have influenced the limited nuanced translation from English into Khmer. This is consistent with a recent study from Cambodia in which the translation from English to Khmer was a big challenge due to unfamiliarity with nuanced technical jargon in the cultural and linguistic settings [42]. This is also consistent with other studies that have shown how terms can be influenced by culture and render translations conceptually different [21,43]. The language barrier is one of the limitations of any cross-cultural study.

Fourth, comprehension errors among Cambodian postpartum women may have affected the quality of data. Thus, the obtained data from the respondents may have influenced the results of factor analysis.

Another potential issue is whether the third factor holds as a factor, because there were only two items of labor companion and delivery companion in that factor. This can be justified by cultural importance. In Cambodia, the cultural values based on the mixture of Animism, Hinduism, and Buddhism are strongly reflected in the perspectives and behaviors of women during maternity such as “reincarnation” and “karma” [44]. The items included in the first factor (PCMC subscale 1) may reflect items in which women felt the medical staff did something good to/for them. In the Cambodian context, this was probably attributed to karma, as it is also considered as good karma to let the others do good deeds. This is empirically supported by a JICA project (2010–2015) in which when introducing the new concept of midwifery care, Cambodian medical staff incorporated the concept in connection with the heart of mercy [41]. Two items of labor companion and delivery companion that loaded on the third factor (PCMC subscale 3) were both related to family presence. Our result agrees with a previous report which found that family-like care was a reasonable way for Cambodian medical staff to understand the concept of person-centered maternity care [45]. Cambodian people attach great importance to the family which is reflective of its collectivistic culture [46]. This is consistent with the evidence from Nigeria and Uganda where women “desired midwives who acted as “mums” to them, who warmly received them, and who provided reassurance and encouragement to give birth well” [47]. Further, our result is consistent with a unique Cambodian contextual feature that nursing tasks are shared among doctors, nurses, and patient families. Many non-invasive nursing cares including bedside hygiene, bathing, and changing sanitary napkins, were normally provided by the patient family [48]. Cambodian women are more likely to seek emotional support and reassurance from their family. Therefore, even though it only consisted of two items, the third factor was retained as a factor that reflects a context where family support is important.

Item interpretation

The four items related to the health facility environment (clean, water, electricity, and safe) did not load well and were eliminated from the 20-item Kh-PCMC scale, which was included in the “supportive care” sub-scale in the Kenyan validation. This is consistent with previous PCMC validation studies in which three items related to the health facility environment (water, electricity, and crowding) were removed from the version in India [17] and also from the 13-item short scale due to poor factor loading [40]. In the original PCMC scale, items related to the health facility environment were retained because they are conceptually and empirically important aspects of person-centered care [49,50], and because the independent health facility environment subscale had low reliability [16]. On the contrary, the health facility environment is theoretically an independent dimension from “experience of care” within the WHO Quality of Care Framework [1]. Thus, the poor loading of items related to the health facility environment may be attributable to this theoretical difference. The facility environment is a foundational requirement in care settings that influences experience of care but may be distinct from experience of care.

Two items related to disrespect and abuse (verbal abuse and physical abuse) did not load well and were eliminated from the 20-item Kh-PCMC scale, which were included in “dignity and respect” sub-scale in the Kenyan validation. In Kenya, the item physical abuse had poor loading but was retained due to conceptual and empirical significance [16]. The poor loadings of items related to abuse were likely due to the low prevalence of verbal and physical abuse in Cambodia. Comparing to other available studies, the percentage of verbal abuse was 4% and that of physical abuse was 3% in Cambodia, 10% and 4% in rural Kenya, 18% and 1% in urban Kenya, 13% and 4% in Ghana, and 19% and 3% in India, respectively [19]. Worldwide, there have been rising reports of disrespect and abuse in maternity care in institutional settings [51–53], hence eliminating disrespect and abuse during the childbirth is an urgent problem [54]. Previous research has shown that self-reported measures were likely to underreport instances of disrespect and abuse during childbirth compared to direct observation, because it becomes internalized and normalized for both care provider and care receiver [55]. Thus, direct observation may be a more effective way to investigate the reality of disrespect and abuse.

Item related to consent to procedure did not load well and was eliminated from the 20-item Kh-PCMC scale, which was included in “communication and autonomy” sub-scale in the Kenyan validation. The results of the preliminary study for cultural translation and adaptation of PCMC scale in Cambodia suggested that some global concepts of PCMC may not resonate with local women’s views and perceptions on PCMC [25]. For example, some Cambodian women did not distinguish well between the concepts of consent to procedures and explanation of procedures. This is particularly so, where women are in a society noted for blind adherence to expert suggestion, implicit consent, poor awareness of alternative options, and gender norms and social hierarchy between medical staff and care receivers. And due to Cambodian women’s way of cognition, we often needed to add explanation to the structured questions. This may have introduced some variation in how women understood the questions if the explanations were not the same for everyone, hence, it may have influenced the quality of data.

Limitations and strengths

There are several limitations in the present study. First, the sample was not generalizable to all Cambodia. Future studies should aim to include samples from all regions and all levels of health facilities and private clinics.

Second, limited resources may have influenced quality of data. Although our translators, expert panels, and data collectors were not experienced professionals, we did our utmost to proceed with the study within the limited available resources and under COVID-19 restrictions.

Third, social desirability bias is a concern as the interviews were conducted in the maternity ward before discharge. Previous studies suggested that women were less likely to report negative experiences inside health facilities [56,57]. In addition, other research suggested that women were more likely to report their experiences positively when interviewed earlier in the postpartum period due to the joy of having just delivered a baby [58,59]. The mean postpartum length of our respondents was 2.5 days, which is similar to within 48 hours in the study from India, but shorter than the within nine weeks period used in the study from rural Kenya, within one week in urban Kenya, and within eight weeks in the Ghanaian study [19]. On the contrary, another study suggests that two to seven days of health facility stay were associated with a significantly decreased PCMC score, due to the increasing probability of experiencing poor person-centered care during the stay [60]. In this sense, the PCMC score found in this study is likely to overestimate actual levels. Relatedly, as noted in another paper [25], it was difficult to conduct an interview with a woman in private. This is because it was not culturally appropriate to keep her family member(s) away for an early postpartum woman. And it is also because a woman took it for granted that family member(s) would be present. Women may be inclined to provide socially desirable responses when family members were present.

Data collection was conducted by having an interviewer read out each question in person. This data collection mode was not only time-consuming and expensive, but also lead to biased or induced response. Nevertheless, the data collection mode is probably the most practical way to obtain accurate data from Cambodian women. This is because all 20 participants in Phase 1 expressed their preference to have an interviewer read the questions and response to them. When asking socio-culturally sensitive questions in a survey targeting women (e.g. person-centered maternity care), it should be necessary to pay the utmost consideration to the way of asking question. That will help a survey avoid ensure women’ comfort during an interview particularly during immediate postpartum period. It is also important for researchers to interpret the data assuming possible biases and limitations in women’ responses, when conducting interviews not at their home but at health facilities.

Fourth, similar to previous studies [22,61], balancing content validity and maintaining linguistic and statistical accuracy was a big challenge. The content validity is the single most important psychometric property of a questionnaire [28]. Only when there is good content validity, can the questionnaire can be considered successful, and the rest of the psychometric properties become useful [36]. In the present study, the content validity of the 31-item Kh-PCMC scale was assured by literature review, expert review, cognitive interviewing, and CVI. However, 11 items were excluded due to poor psychometric properties. We were therefore concerned about eliminating items that matter to Cambodian women during childbirth. The present study used a cut-off of 0.3 [34] for a more data-driven approach, while the validation in Kenya and India used conservative and inclusive decisions, and a relaxed cut-off of 0.1 to retain some key items such as verbal and physical abuse [16,17]. Given the potential importance of some excluded items and the limited generalizability of the validation sample, we recommend future studies include all 31 PCMC items to reassess the psychometric properties in more diverse samples. In addition, without qualitative research in Cambodia for an item generation stage, it is possible that we excluded some aspects of what matters to Cambodian women during childbirth. Future qualitative research is required to comprehensively capture PCMC for the Cambodian population.

Despite these limitations, there are strengths of the present study. First of all, to the best of our knowledge, this is the first reliable and valid instrument to quantitatively measure women’s experience of care received during childbirth in Cambodia. Since we have an assumption that potential cultural and social differences may influence the conceptualization of person-centered maternity care, the notable strength of the present study is to emphasize cultural context, language, and local practices for use in Cambodia.

Second, a recent qualitative evidence synthesis suggested that what matters to women during childbirth were consistent across many settings, albeit that the evidence to support this claim thus far has come from only one continent of the world (Africa) [62]. The previous validation studies of the PCMC scale also called for further validation in additional settings including Southeast Asian populations with a data-driven approach [40]. Therefore, we believe this is the first significant response from Cambodia using such a data-driven approach.

Third, developing the global standard to measure person-centered maternity care is an urgent priority in this area [16] and, the present study contributed to additional validation of the PCMC scale in the Asian context to facilitate meaningful international comparison.