Background: Maternal mortality continues to be the biggest challenge facing Ethiopia and other developing countries. Although progress has been made in making maternity services available closer to the community, the rate of deliveries attended by skilled birth attendants has remained very low. Absence of respectful maternity care (RMC) is believed to have contributed to low utilization of facility delivery services. This study outlines steps undertaken to construct and validate a scale that measures women’s perception of respectful maternity care provided in health facilities. Methods: An inductive item generation process that included a literature review and in-depth interviews with labor and delivery clients, followed by an expert review, assured face validity and content validity of the tool. A draft RMC scale with 37 items and two additional measures of global satisfaction items, measured on a fivepoint Likert scale, were administered to a developmental sample of 509 postnatal care clients visiting facilities immediately after childbirth to 7 weeks postpartum. IBM SPSS 20 was used to perform exploratory factor analysis (EFA) using principal component analysis (PCA) with oblique rotation method. Results: The final RMC scale with 15 items was loaded on four components. The extracted components were labeled as friendly care, abuse-free care, timely care, and discrimination-free care. The final RMC scale correlated strongly with the global satisfaction measures, indicating criterion-related validity of the scale. Content-related validity was assured by the process of item generation. Construct validity of the RMC scale was confirmed by high average factor loading of the four components ranging from 0.76 to 0.82 and low correlation between the components. Stability of the scale was confirmed by running PCA in a randomly selected split sample of 320 samples from the validation sample. The final 15-item scale showed an adequate reliability with a = 0.845. Conclusion: The 15-item RMC scale is a valid and reliable measure of women’s perception of RMC received in health facilities. We recommend that health facilities use the RMC scale in urban public health facilities and that other researchers conduct further exploratory and confirmatory factor analysis in different geographic areas.
The study was conducted in 11 urban-based public health facilities (three hospitals and three health centers in Addis Ababa, one hospital and one health center in Bishoftu, and one hospital and two heath centers in Adama town). This population, which was used for developing and validating the scale, is referred to as the developmental group. The target population for this study consisted of postpartum women who delivered in public health facilities within seven weeks prior to data collection. The study utilized a mixed approach of qualitative and quantitative methods. The qualitative approach used in-depth interviews with postpartum women. In the quantitative approach, expert review was undertaken by trained data collectors using email and interviews with postpartum women. The study was conducted in three phases. First, a formative phase was carried out to determine potential items that could be included in the tool. This initial phase included a comprehensive literature review followed by in-depth interviews with eight postpartum women in two health facilities. In the second phase, the draft items were pilot tested among 40 postpartum women in five health facilities. In the third phase, a quantitative assessment was conducted in a private area within the selected health facilities. Postpartum women interviewed were those who received labor and delivery services in public health facilities within 7 weeks prior to the date of the interview, consented to participate in the study, and visited health facilities during the data collection period. A consecutive sampling approach was utilized. In-depth interviews with eight postpartum women helped to saturate RMC dimensions. For piloting the draft tool, interviews with 40 postpartum women were conducted. The probability that a factor structure can be replicated in another study depends partially on the sample size used in the initial analysis [13]. For the final administration of the RMC tool, sampling recommendations, stated in terms of the ratio of a minimum sample size (N) for a particular analysis to the number of variables (p), were used [14, 15]. Tinsley and Tinsley (1987), cited by DeVellis (2003), suggest that proportions of 5 to 10 subjects to one variable is sufficient [16]. An empirical test conducted by Costello and Osborne on the effect of sample size on the results of factor analysis reported that larger samples tend to produce more accurate solutions [13]. In this study, the number of variables (p) was 37 items, and a total of 509 women (N) were interviewed, which resulted in nearly 14 subjects to one variable. Data were collected from postpartum women during March 2014 in the 11 health facilities across three cities. The interviews were conducted at intervals ranging from 6 h to 7 weeks after delivery. Forty-two percent of mothers were interviewed within 2 days of delivery, 14 % were interviewed from 3 to 42 days after delivery, and the remaining 44 % were interviewed from 43 to 49 days after delivery. Inclusion criteria for women were as follows: use of delivery services in public health facilities from 6 h to 49 days before data collection, ability to speak Amharic, and willingness to participate in the study. To avoid professional bias during data collection, we selected experienced non-health professional (information technology and social science background) data collectors. The principal investigator and co-investigator supervised the data collection process. RMC tool development was conducted using psychometric procedures recommended by DeVellis (2003) on procedures for new scale development [16]. This included initial item generation, expert review, pilot testing, and final administration of the draft tool to the developmental group. Each step of the development process is described in Fig. 1 and in the next section. RMC tool development process The literature review identified seven a priori dimensions. In each of these, 5–12 items were selected from the pool of items generated by in-depth interviews conducted to understand the perception of care received by eight postpartum women during the delivery and postnatal periods. This resulted in a draft tool, or scale, with 60 items. A five-point Likert scale (with 5–strongly agree, 4–agree, 3–I don’t know, 2–do not agree, and 1–strongly do not agree) was used. The 60-item draft scale was reviewed by five maternal and newborn health experts in Ethiopia. The experts were all public health practitioners with masters’ degrees in public health as well as a bachelor’s degree in midwifery, public health, or medicine. These experts had from 10 to 35 years of experience in teaching, program management, and clinical work related to maternity care. All experts who participated were executive board members in their respective associations. These associations included the Ethiopian Public Health Association (EPHA), Ethiopian Public Health Officers association, Ethiopian Midwifery association, and Ethiopian Evaluation association. Using the comments of experts, four items were excluded, five items rephrased, and three new items added. After incorporating experts’ feedback, the 59-item draft scale was random ordered and formatted to use in a pilot test conducted among a sample of 40 postpartum women in two hospitals and three health centers in Adama and Addis Ababa. The findings guided several changes to the tool: three items were merged into one and 20 items (those that were not clear to respondents or were redundant) were excluded, resulting in a tool with 37 items for final administration to the developmental group. Data were collected in March 2014. All data collectors received a half-day orientation on administration of the scale, the informed consent process, confidentiality of data, and the role of the data collector during the survey. Data collectors presented a support letter, obtained from the regional health bureaus of Addis Ababa and Oromia for the 11 facilities, to health facility managers and maternity unit coordinators to inform them of the study objectives. After this, the data collectors began their work. All women who used labor and delivery services in public health facilities within 49 days preceding the survey were invited to participate. Informed consent was requested and obtained for all women. To maintain the women’s privacy, all interviews were conducted in a private area inside the health facility. Interviews with immediate postnatal clients were conducted in the postnatal room when health providers were not around and no other mothers were in the room. After the data collected were reviewed for completeness, data entry was conducted. Data entry was managed using a data entry template prepared in Microsoft Access 2010. Double data entry was used for 25 % of cases and validated with the original and no discrepancy was obtained to proceed to 100 % double entry. Data analysis was performed using the IBM SPSS 20 statistical package. Data analysis followed steps for new scale construction outlined by Worthington and Whittaker (2006) [15] and DeVellis (2003) [16]. The steps are outlined in Fig. 2. Five-step data analysis process Exploratory factor analysis (EFA) using a principal component analysis (PCA) was used to identify a parsimonious list of factors that describe women’s perception of RMC and consolidate variables and generate hypotheses about underlying processes [13]. The Kaiser-Meyer-Olkin (KMO) measure of sampling adequacy and Bartlett’s test of sphericity were used to check the suitability of data for factor analysis. The reliability of each component was assessed using Cronbach’s alpha. Correlation analysis and independent samples t-tests were used to assess validity of the tool with criterion satisfaction items and other background information of women and type of delivery. Establishing dimensionality of a construct is an important step in the scale development process [17]. In this analysis a KMO value of 0.6 was used as the criterion for sampling adequacy. To produce scale uni-dimensionality and simplify the factor solutions, scree plot and parallel tests were used as criteria for factor extraction. Rotation is a statistical technique used to simplify interoperability of factor solution [13]. Oblique rotation was used as a method of rotation. Use of oblique rotation was justified because RMC components are closely correlated. The rotation was conducted in a series of seven iterative processes, deleting one or more items at a time and examining the remaining items. Item loading (which refers to the degrees to which the original item scores correlate with the components), cross loading, and communalities were used as criteria for item deletion. If factors shared items that cross-loaded too highly on more than one factor (e.g., > 0.32) or if factors shared items that cross-loaded and the difference in item loading from the highest was less than 0.15, it was rejected. Communalities (the amount of variance of a measure that is accounted for by a component or group of components derived from factor analysis before rotation) was the third criterion, where item communalities of less than 0.6 after rotation were used as the lowest limit for item deletion. Cross loading was not used as a pragmatic statistical criterion for item deletion; instead, the judgment of the researcher and study team members was used to delete or retain items to relevant factors based on their theoretical significance. Evidence about different forms of validity of RMC components was obtained using PCA. This section describes how the evidence for different forms of validity was inferred. Content-related validation involved assessing the degree to which the sample of items, tasks, or questions on a test is representative of some defined universe or domain of content, based on expert judgment. Face validity, which is closely related to content validity and refers to whether a measure appears to be measuring what it is supposed to measure, was also assessed [10, 18]. Criterion-related validation consisted of verifying whether a test score on the scale was correlated with criteria measured at the same time. This is usually based on comparison between an existing scale and the one under development, but in our case, no appropriate scales existed for the construct. Therefore, we selected two criteria using the experiences of other researchers on a closely related variable: satisfaction with overall service and recommendation to others [10]. Construct validation relates to how well the items on a questionnaire represent the underlying conceptual structure. Construct validation was assessed by examining the Pearson correlation coefficient between components identified by factor analysis. Known-groups validity (also a form of construct validity) was ensured by assessing the scale’s ability to differentiate the level of RMC reported for normal and complicated deliveries. Reliability analysis was used to assess the internal consistency of the scale. The internal consistency of each component of the RMC scale was assessed using Cronbach’s alpha. To be considered consistent, the minimal coefficient for a component had to be above 0.70 [10]. The proposal for this study was reviewed and approved by the Addis Ababa University Faculty of Education and Behavioral Sciences ad hoc research ethics committee, the Addis Ababa Regional Health Bureau institutional review board, and the Oromia Regional Health Bureau institutional review boards. All women interviewed were asked for their informed consent to participate.
N/A
DIMA AI Care
