Psychometric properties of the Barkin Index of Maternal Functioning with Ethiopian mothers

21

22 The functional status of mothers after childbirth has implications for maternal, child, and family 23 health. There is a lack of adequate reliable and valid instruments in Ethiopia for assessing 24 women’s postpartum functional status. 25

26 This study was intended to reveal the psychometric properties of the Barkin Index of Maternal 27 Functioning (BIMF) for assessing Ethiopian mothers’ functional status. 28

29 Structured interviews were used to obtain BIMF data from 202 women who had a child less than 30 1 year of age. Descriptive statistics were calculated for the BIMF items; internal consistency was 31 assessed with interitem correlations and coefficient alphas; construct validity was examined 32 through exploratory factor analyses (EFAs) after face and content validities had been confirmed; 33 and test–retest reliability was assessed with intraclass correlation coefficients (ICCs). 34

35 Narrow standard deviations and significant skewness and kurtosis characterized most of the 36 individual BIMF items. Most interitem correlations were < |.15| and 13 of the 20 BIMF items 37 did not load satisfactorily on any factor in exploratory factor analyses. Two factors emerged 38 from the remaining items, one with three items and the other with four. Coefficient alphas were 39 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 3 .54 for the first of these factors, .48 for the second, and .58 for all 20 items. The ICCs for test–40 retest reliability were < .40. 41

42 BIMF data from the sample of Ethiopian women in this study exhibited unusually low levels of 43 variability and high levels of skewness and kurtosis. Furthermore, in the context of this study, the 44 BIMF could not be regarded as reliable either in terms of internal (interitem) consistency or 45 temporal (test–retest) consistency. Researchers using the BIMF in Ethiopian contexts are advised 46 to examine the nature of their data carefully, identify the factor structure of their samples’ data, 47 and consider ways in which the index might be improved, or replaced. 48 49

Barkin Index of Maternal Functioning, BIMF, maternal functioning, Ethiopia 50 51 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 4

52 Maternal functioning in the postpartum period is a reflection of the physical and psychosocial 53 changes associated with pregnancy and childbirth. Maternal activities, in addition to 54 physiological changes as well as individual and social activities, can make this period difficult 55 for mothers. As a result, assessing maternal functioning following childbirth can be important for 56 identifying maternal performance, productivity, and health, and can provide valuable information 57 for healthcare providers, particularly with regard to identifying mothers who might benefit from 58 supportive advice or interventions [1–3]. 59 60 Almost all of the postpartum functional status assessment tools available today were 61 developed for populations in high-income countries [4]. To date, the Inventory of Functional 62 Status After Childbirth (IFSAC) is the most widely used tool specifically designed to assess 63 several domains of maternal functioning [5]. The IFSAC contains 36 items arranged in subscales 64 encompassing five dimensions: infant care responsibilities, self-care activities, household 65 activities, social and community activities, and occupational activities. The inventory is designed 66 to assess maternal functioning from 6 to 10 weeks after childbirth with a focus on “recovery 67 from childbirth” [6]. Despite its popularity, the IFSAC has been criticized because it does not 68 account for women’s feelings or levels of satisfaction with the changes in their lives since 69 childbirth, and, as an alternative, the Barkin Index of Maternal Functioning (BIMF) was 70 recommended by its developers as a more modern, patient-centered method for assessing 71 maternal functioning [7, 8]. It is therefore the focus of this study. 72 73 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 5 The BIMF is a 20-item Likert-type scale designed to capture mothers’ perceptions of their 74 functional status across the previous 2-week period at some point within the year since the birth 75 of their most-recent, or only, child. For each item, participants are offered six response options: 0 76 corresponding to “strongly disagree”, 1 to “disagree”, 2 to “slightly disagree”, 3 to “neutral”, 4 to 77 “slightly agree”, 5 to “agree”, and 6 to “strongly agree”. Items 16 and 18 are negatively worded, 78 so responses on those items are reverse coded prior to analyzing the data to ensure that higher 79 scores on the index consistently indicate higher levels of functoning. 80 81 Because of the lack of a valid means for assessing maternal functional status in sub-Saharan 82 countries, including Ethiopia, in this article we examine the validity and reliability of the BIMF 83 when used with Ethiopian mothers to identify whether the index is likely to perform 84 satisfactorily within that population. 85 86

87 Study design and setting 88 A descriptive community-based study was implemented with participants from Lante District 89 within Gamo Zone, Ethiopia, situated 430 km south of the country’s capital, Addis Ababa. In 90 2020 G.C. (2012 E.C.), the district health office reported a total population of 236,600, with 436 91 mothers having a child under 1 year of age. 92 93 Participants 94 Eligible participants were women who had a child less than 1 year of age resulting from a single-95 child pregnancy; who had vaginal delivery, birth through caesarean section, or instrumental 96 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 6 delivery; who had self-declared physical health; and who had self-declared absence of depression 97 or any other mental disorder during the pre-pregnancy, prepartum, and postpartum periods. 98 99 Sample size determination and sampling method 100 Sample size required to determine the factor structure of the BIMF was based on there being at 101 least 10 participants per item of the index [9]. Therefore, a sample of at least 200 mothers was 102 required. In order to attain that sample, a list of the 436 mothers who had a child less than 1 year 103 of age was obtained from the district health office. Subsequently, a sample of 202 mothers was 104 identified by using computer-generated random numbers with Microsoft Excel for Windows 10. 105 Forty of the 202 mothers were selected by the same method to comprise a retest sample for the 106 BIMF, 2 weeks after the index had initially been administered. 107 108 Questionnaire 109 A questionnaire was created that comprised questions to obtain information about the mothers’ 110 sociodemographic background (the respondent and her most-recent child’s ages, the 111 respondent’s marital status, and size of her family), obstetric characteristics (parity, number of 112 living children, mode of delivery of the most recent child, and whether or not the mother had 113 intended to become pregnant), and the BIMF. 114 115 Translation process 116 Translation and cultural adaptation of the BIMF (for example, the word “diaper” is not familiar 117 to most people in rural communities, and in Amharic is referred to as “ye shint cherq”) was 118 performed according to the minimal translation criteria of the Medical Outcomes Trust [10]. Two 119 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 7 independent bilingual translators (one a psychologist and the other an English-language lecturer) 120 with advanced levels of English language and native Amharic-language skills translated the 121 BIMF into Amharic. In doing so, they made no changes to the items’ meanings. An English-122 language lecturer who is a native Amharic speaker and who had been blinded to the original 123 version, provided a back translation. There were no major difficulties in reconciling the back-124 translated version. Subsequently, Dr Barkin, the main researcher associated with development of 125 the BIMF, gave permission for the Amharic version of the index to be used in this research. 126 127 Face and content validity 128 After translation, the BIMF was reviewed by experts in maternity and child health and 129 reproductive health units who were not part of the study, and they identified no problems. The 130 tool was regarded as likely to be generally well understood, acceptable, and culturally 131 appropriate for all prospective respondents. 132 133 Data collection 134 Names and addresses of the 202 mothers identified as prospective participants were obtained 135 from the district health office. Five data collectors and three field supervisors who had 136 experience in conducting interviewer-administered data-collection procedures and the ability to 137 speak Amharic were recruited for data collection. The data collectors went to the mothers’ 138 homes and conducted structured interviews by using the smart-phone based Open Data Kit 139 (ODK) app. Prospective participants were provided verbal and written information about the 140 research, and were invited to ask any questions they might have. All mothers who were invited to 141 participate in the research agreed to do so. 142 143 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 8 Data processing and analysis 144 Data were analyzed using SPSS for Windows version 25® (IBM Corp., Armonk, NY, USA). 145 Descriptive statistics were calculated for the sociodemographic variables and BIMF items. 146 Reliability associated with internal consistency was measured by interitem correlations and 147 coefficient alphas. 148 Because previous research with the BIMF had revealed marked inconsistency regarding number 149 and composition of factors in the scale [11–14], we did not conduct confirmatory factor analysis 150 as a means of identifying the structure of the data. Instead, we conducted a series of exploratory 151 factor analyses (EFAs). In doing so, we identified the number of factors with the scree test in 152 conjunction with parallel analysis based on principal components and 1,000 randomly generated 153 matrices; we used principal axis factoring as the method of extraction in conjunction with 154 promax rotations; and we regarded item loadings > |0.40| as definitely acceptable, and loadings 155 in the region of |.35| to |.40| as possibly acceptable. 156 157 In order to assess test–retest reliability, ICCs (3,1 [two-way mixed effects, single measures] 158 absolute agreement) were calculated. 159 160 Ethical considerations 161 Ethics approval was obtained from Arba Minch University’s College of Medicine and Health 162 Science Institutional Review Board (IRB/157/2020). Verbal consent was obtained from all 163 participants and recorded in conjunction with the starting time of each interview, and all 164 interviews were conducted in private. 165 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 9

166 Participant characteristics 167 All 202 mothers reported their physical health to be good, and none indicated they had a history 168 of depression or any other mental disorder during the pre-pregnancy, prepartum, and postpartum 169 periods. Their ages ranged from 15 to 38 years (M = 26.1 years, mode = 30, SD = 4.7), their 170 children’s ages ranged from 1 to 12 months (M = 6.7 months, mode = 11, SD = 3.7), their 171 number of pregnancies ranged from 1 to 9 (M = 2.4, mode = 1, SD = 1.5), their number of 172 children ranged from 1 to 8 (M = 2.31, mode = 1, SD = 1.40), and the size of their households 173 ranged from 2 to 12 (M = 4.6, mode = 4, SD = 1.7). Other participant characteristics are shown in 174 Table 1. 175 Table 1. Participant characteristics. 176 Variable Category Number Percent Age of the mother/ caregiver (in years) ≤ 24 73 36.1 25 – 30 103 51.0 ≥ 31 26 12.9 Age of most recent or only child < 2 months 20 9.9 2 – 6 months 76 37.6 ≥ 6 months 106 52.6 Marital status Single 9 4.5 Married 191 94.6 Divorced or widowed 2 1 Family size ≤ 3 58 28.7 4 – 6 115 56.9 ≥ 7 29 14.4 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 10 Parity Primipara 69 34.2 Multipara 133 65.8 Number of living children ≤ 3 165 81.7 ≥ 4 37 18.3 Mode of delivery Vaginal 164 81.2 Caesarean 25 12.4 Instrument 13 6.4 Pregnancy intention Intended 175 86.6 Unintended 27 13.4 177 178 Item characteristics 179 Means, SDs, and indicators of skewness and kurtosis on the 20 BIMF items at Time 1 are 180 provided in Table 2. On all but one of the items, the means were above the midpoint of the 181 option range, indicating a tendency among participants to believe they were functioning 182 effectively as mothers. The exception was Item 18, on which, prior to reverse coding, the 183 participants indicated they were often anxious or worried about fulfilling their maternal 184 responsibilities. Twelve of the items had SDs < 1.0; 18 items were significantly skewed (two 185 positively, 16 negatively); and 15 items were significantly kurtotic (12 leptokurtic, three 186 platykurtic). Twelve items were significantly negatively skewed as well as being significantly 187 leptokurtic. 188 189 Table 2. Descriptive statistics for items on the Barkin Index of Maternal Functioning. 190 Itema Mean SD Skew- ness zb Kurto- sis zb All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 11 1 I am a good mother. 5.2 .46 4.3 0.2 2 I feel rested. 5.0 .91 −16.2 33.7 3 I am comfortable with the way I’ve chosen to feed my baby (either bottle or breast or both). 5.0 .55 −15.1 54.4 4 My baby and I understand each other. 5.2 .58 −0.2 −0.7 5 I am able to relax and enjoy time with my baby. 5.4 .69 −5.6 4.7 6 There are people in my life that I can trust to take care for my baby when I need a break. 4.6 1.25 −12.0 10.6 7 I am comfortable allowing a trusted friend or relative to care for my child (can include baby’s father or partner). 4.7 1.29 −11.9 10.1 8 I am getting enough adult interaction. 4.9 .61 −7.4 22.4 9 I am getting enough encouragement from other people. 4.7 .80 −13.4 26.4 10 I trust my own feelings (instincts) when it comes to taking care of my baby. 3.5 2.00 −3.9 −3.8 11 I take a little time each week to do something for myself. 3.1 1.84 −1.3 −4.9 12 I am taking good care of my baby’s physical needs (feeding, “ye shint cherq”, and keeping doctor’s appointments). 5.2 .80 −15.4 40.7 13 I am taking good care of my physical needs (eating, showering, etc.). 5.0 .79 −7.8 13.3 14 I make good decisions about my baby's health and well being. 4.9 .71 −11.1 26.7 15 My baby and I are getting into a routine. 3.6 1.83 −3.4 −4.1 16 I [do not] worry about how other people judge me (as a mother). 4.2 1.70 −7.2 −0.2 17 I am able to take care of my baby and my other responsibilities. 4.8 .96 −14.8 25.7 18 Anxiety or worry [seldom interfere] with my mothering ability. 2.1 1.73 6.3 −1.4 19 As time goes on, I am getting better at taking care of my baby. 4.8 1.03 −11.7 19.7 20 I am satisfied with the job I am doing as a new mother. 5.5 .63 −4.3 −1.3 a Responses to Items 16 and 18 have been reverse coded to attain the meaning created by bracketed text. 191 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 12 b The z values for skewness and kurtosis are based on original skewness and kurtosis values divided by 192 their respective standard errors. If these z values exceed |3.29| when N = 50 to 300, they indicate 193 statistically significant skewness or kurtosis [15]. 194 195 Interitem correlations and factor structure 196 None of the 190 interitem correlations exceeded .50, and only 37% of those correlations lay in 197 the .15 to .50 range recommended by Clark and Watson [16]. Among the remaining 63% of the 198 interitem correlations, 17% were negative, with the strongest being −.29. Despite this generally 199 low, and sometimes inverse, level of association between the items, the Kaiser–Meyer–Olkin 200 (KMO) index was .72 and Bartlett’s test of sphericity was significant (p 1, but 202 the scree plot indicated there were four factors in the data (see Fig 1), as did parallel analysis. We 203 therefore conducted an initial EFA constrained to four factors. In this solution, nine items loaded 204 uniquely on one of the factors at > |.40|, three items had cross loadings on two of the factors, and 205 the remaining eight items failed to load on any factor at > |.37|. Only 42.56% of the variance was 206 accounted for by the four factors. 207 208 [Fig1. Scree plot based on all 20 BIMF items.] 209 210 Follow-up EFAs were conducted in order to attain the most satisfactory and interpretable 211 outcome. When nonloading items, cross-loading items, and items that loaded only on single-item 212 factors were successively removed, seven items were retained, among which 53.09% of the 213 variance was accounted for, but the extraction communalities were generally low (maximum = 214 .55, mean = .35, with Items 20 and 18 exceptionally low at .23 and .13, respectively). All seven 215 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 13 items loaded at > .39 on one of two factors, although two items cross loaded with a difference ≤ 216 .20. See Table 3. Items 13, 14, and 17 loaded on a factor that we labeled Decision Making and 217 Practicalities, and Items 5, 16, 18, and 20 loaded on the other factor, which we labeled 218 Confidence, Enjoyment, and Satisfaction. 219 220 Table 3. Item loadings on the two-factor solution.a 221 Item # Item b Factor 1 Factor 2 14 I make good decisions about my baby’s health and well-being. .74 17 I am able to take care of my baby and my other responsibilities. .56 −.40 13 I am taking good care of my physical needs (eating, showering, etc.). .52 16 I [do not] worry about how other people judge me (as a mother). .64 5 I am able to relax and enjoy time with my baby. .32 .52 20 I am satisfied with the job I am doing as a new mother. .42 18 Anxiety or worry [seldom interfere] with my mothering ability. .39 222 a Loadings < |.30| are not shown. 223 b Wording of Items 16 and 18 was altered, as shown in bracketed text, to reflect prior reverse coding. 224 225 Consonant with the practice of several other researchers [11, 13, 14], we calculated a score for 226 each subscale by adding responses on their respective items, and we calculated a composite total 227 by adding responses to all 20 items. In order to confer meaning to participants’ responses in 228 relation to the range of response options, we also divided these totals by the number of items on 229 which each score was based. The results are shown in Table 4. 230 231 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 14 Table 4. Means, standard deviations, and obtainable ranges on composite factor and total 232 scores.a 233 Unadjusted scores Adjusted scores BIMFb component and items Mean ( SD) Obtainable range Mean (SD) Obtainable range Decision Making and Practicalities (Items 13, 14, 17) 14.74 (1.79) 0–18 4.91 (0.60) 0–6 Confidence, Enjoyment, and Satisfaction (Items 5, 16, 18, 20) 17.08 (1.79) 0–24 4.27 (0.81) 0–6 Total (Items 1 to 20) 91.11 (7.76) 0–120 4.56 (0.39) 0–6 234 a Unadjusted scores are based on addition of mothers’ responses; adjusted scores were obtained by 235 dividing the unadjusted scores by the relevant number of items. 236 b Barkin Index of Maternal Functioning 237 238 The means of the adjusted scores on both factors as well as the composite total indicate that 239 the mothers tended to respond with options of somewhat agree (4) or agree (5). However, 240 relative to the first factor (Decision Making and Practicalities), the means and SDs indicated that 241 self-perceptions of maternal functioning were lower and more varied on the second factor 242 (Confidence, Enjoyment, and Satisfaction). Scores on the first factor were significantly skewed 243 (−11.76) and kurtotic (21.9), and the total score was significantly skewed (−3.4). 244 245 Coefficient alphas were .54 for Decision Making and Practicalities, .48 for Confidence, 246 Enjoyment, and Satisfaction, and .58 for all 20 items. 247 248 Test–retest results 249 The ICC for Confidence, Enjoyment, and Satisfaction could not be calculated satisfactorily 250 because of negative average covariance among the items (Pearson’s r of −.04 was calculated as a 251 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 15 substitute for the ICC). The ICC for Decision Making and Practicalities was .14 [95% CI −.18–252 .43], and the ICC for the total score across all 20 items was .36 [95% CI .06–61]. 253 254

255 To the authors’ best knowledge, this is the first examination of the BIMF in an Ethiopian 256 context. The research has several desirable foundations. Face and content validity of the BIMF 257 were assessed and regarded as satisfactory; random sampling was used to obtain participants; the 258 sample included women who gave birth by vaginal delivery, instrument, and caesarean section, 259 and who were primiparous as well as multiparous, suggesting that the results were not likely to 260 be unique to a specific group of mothers; the sample was sufficiently large for identifying the 261 factor structure of the data; and an attempt was made to identify test–retest reliability. 262 263 Three aspects of the raw data are noteworthy and might carry implications for assessing the 264 suitability of the BIMF in Ethiopian settings. First, with the exception of Item 18 (referring to 265 mothers’ worry or anxiety), the participants’ responses on individual items suggest a moderate to 266 high level of maternal functioning. Whether that is an accurate indication would require 267 independent assessment, which was not conducted in this research. However, some reassurance 268 might be taken from the mean of 91 on the 20 BIMF items in this research being similar to the 269 mean of the 20 items in research where samples were composed solely of mothers who were not 270 depressed [13, 17]. Furthermore, in samples that include a large proportion of mothers diagnosed 271 with depression and who could therefore be expected to have lower 20-item BIMF scores, the 272 mean based on those items has indeed been lower, for example, ranging from 81 down to 63 [7, 273 11, 17, 18]. However, in a sample of mothers, 46% of whom had been assessed as experiencing 274 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 16 depression, the 20-item BIMF score was considerably higher, at 104 [19]. Unfortunately, this 275 range of 40 points on the total BIMF score in samples with mothers who experience depression 276 makes it difficult to place confidence in the meaning of the 20-item BIMF score when seeking to 277 gain a definitive overview of mothers’ functioning, and the composite score of 91 in this research 278 therefore provides little reassurance in relation to validity of the BIMF in an Ethiopian setting. 279 280 The second noteworthy aspect of the data is the lack of variability among the mothers’ 281 responses. This was evident in 60% of the SDs on individual items being < 1.00 despite the 282 availability of seven response options. In contrast, only two of the SDs were < 1 on the 20 items 283 in the foundational psychometric research with the BIMF [11]. Furthermore, the SD of the 284 composite score based on all 20 items in this research was 7.76 compared with noticeably wider 285 SDs, ranging from 10.8 in some research [20] up to 17.1 in other research [7]. 286 287 This smaller amount of variation in the mothers’ responses in the present study presents 288 interpretative and empirical difficulties for three reasons. First, if there is genuine lack of 289 variability, statistics such as correlations, factor analysis, and coefficient alpha—all of which rely 290 on variability—are likely to yield misleading results. Second, if the lack of variability has a 291 foundation in some kind of distorting BIMF-related artifact within Ethiopian samples (e.g., the 292 cognitive burden on the mothers having to remember what occurred during the previous 2 weeks 293 might have created a degree of bluntness in their responses), the nature of that artifact would 294 need to be identified and addressed. Third, if the lack of variability is likely to be a chance 295 occurrence, other studies need to be conducted with the prospect that their data would contain a 296 greater spread of maternal self-impressions. 297 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 17 298 The third noteworthy aspect of the data is most items having distributions that were 299 significantly skewed or kurtotic, or both, which was accompanied by one of the factors and the 300 20-item composite score having negatively skewed distributions. These results could be genuine 301 reflections of the mothers’ self-perceptions, and therefore are interesting findings in themselves. 302 However, in combination with the minimal variation in responses to the 20 items, the skewed 303 and kurtotic distributions might also have influenced the EFA outcome, coefficient alpha values, 304 and the test–retest results. 305 306 We are aware of no other research in which a similar BIMF factor structure emerged. 307 Although two factors have been found in some other studies [11, 14], the factors in those studies 308 comprised different items and were therefore conceived of as being different and were labeled 309 differently. Our results are also distinctive in that 13 of the 20 BIMF items failed to load on any 310 factor, probably because most interitem correlations were < |.15|. An additional feature of our 311

is Items 16 and 18 having loaded on one of our factors (Confidence, Enjoyment, and 312 Satisfaction), but having loaded on none of the factors in some other research [11, 12, 14]. 313 314 Although inconsistencies in factor structure could result from differences in the nature of 315 samples, particularly when cultural differences exist, some inconsistency between our findings 316 and the findings in other research is likely to result from principal components analysis versus 317 EFA being employed as a data-reduction strategy; different procedures (the Kaiser criterion, 318 scree plots, or parallel analysis) being used for identifying the number of factors in the data; the 319

used for extraction (e.g., maximum likelihood or principal axis factoring); whether or not 320 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 18 the factors were rotated; the method employed for rotation (orthogonal versus oblique)—if factor 321 rotation occurred at all; and whether or not a sequence of EFAs was used to attain greater factor 322 clarity by successively deleting poorly performing items. Publications about the BIMF provide 323 little or no information about these procedures when data-reduction analyses have been 324 conducted despite some procedures having substantial implications for the outcomes. 325 326 In this research, we believe we engaged in best-practice data-reduction procedures [21–26], so 327 the results are likely to have been dependent on the nature of the data, not deficiencies in the 328

of analysis. That aside, the EFA results in this research are not impressive. Not only 329 were loadings on 13 of the 20 BIMF items so low that most of the index’s items were discarded, 330 but also the remaining seven items had low communalities despite accounting for 53.1% of the 331 variance. Furthermore, two of the seven items cross loaded on the two factors that emerged. 332 Under these circumstances, one solution involves focusing on the composite score based on all 333 20 items, but mothers with similar scores could have noticeably different profiles and therefore 334 noticeably different needs in relation to healthcare as well as correlates in research contexts, so 335 the value of that score is limited. 336 337 In addition to differences in factor structure, our results differ in several ways from the results 338 obtained in other studies. The coefficient alpha across all 20 items in our data was unusually low 339 at .58 compared with alphas of .87 and .83 in samples of American mothers [7, 11, 19] and, at 340 .36, the test–retest ICC for the 20-item composite score in this research indicated much poorer 341 test–retest reliability than was indicated by the ICC of .80 obtained in a study conducted with 342 Iranian mothers [14]. The above differences, in which the psychometric features of the BIMF in 343 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 19 this research are not only less impressive than are results reported in other BIMF research but are 344 also unimpressive according to conventional criteria, are almost definitely attributable to limited 345 data variability in this study [27, 28]. 346 347 Although not directly associated with our examination of the BIMF, the absence of self-348 reported depression among the mothers in this research might be important. Other research has 349 indicated that perinatal depression is likely to be approximately 13% among mothers in middle- 350 and low-income countries in general [29] and in Ethiopia, specifically [30], and as high as 33% 351 in rural parts of Ethiopia [31]. The discrepancy between those statistics and the mothers’ self-352 reports in this research suggests that responses on the BIMF in this research might be regarded 353 with a degree of reservation despite the mothers having been assured by the data collectors of 354 confidentiality concerning their responses in order to discourage them from creating an 355 inaccurately favorable impression. 356 357

358 This study provides ambiguous evidence concerning the reliability and validity of the BIMF in 359 an Ethiopian setting, largely because it is difficult to identify whether the generally high and 360 similar self-perceptions of mothers were genuine and might also have been idiosyncratic to our 361 specific sample. 362 363 The ambiguity could lead to incorrect conclusions that, in Ethiopian samples, the BIMF’s 364 items carry little association with each other; that there are only two factors within the BIMF, 365 each with only a small number of items; and that the index’s test–retest reliability is poor. 366 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 20 Because these conclusions might be proved untenable if data were obtained from other samples, 367 including samples with a greater dispersion among responses, the BIMF may be appropriate in 368 other Ethiopian contexts. Furthermore, with more variation in the data there is a prospect that 369 additional, or more substantial, factors would emerge. 370 371 Because the two factors that emerged from the EFAs were dissimilar to factors found in other 372 research, and because of evident volatility in the factor structure of the BIMF revealed by other 373 studies, further use of the BIMF in an Ethiopian context should involve examination of the factor 374 structure for the particular sample(s) on which that research is conducted. Furthermore, because 375 our research indicates that the index could not be regarded as reliable either in terms of internal 376 (interitem) consistency or temporal (test–retest) consistency, prospects for improving the index 377 might be considered—although with the proviso that permission to modify the index had been 378 granted [32]. 379 380 The seven BIMF items that loaded on two factors in this research might be used to explore 381 aspects of decision making and practicalities, as well as aspects of enjoyment and satisfaction, 382 among Ethiopian women concerning their maternal role in the postpartum period. However, 383 given the volatility of the BIMF factor structure, those two factors might not feature in data from 384 other samples. Furthermore, if only those seven items were used, important aspects of 385 motherhood such as mother–infant interaction, self-care, and social support would not be 386 assessed, and this raises the possibility that, when face and content validity were assessed at the 387 start of this research, too much consideration was given to endorsing the items within the BIMF 388 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 21 without considering aspects of motherhood that were underrepresented, inappropriately 389 represented, or not represented at all — at least for an Ethiopian context. 390 391 In summary, this research demonstrates that the BIMF should be used with a degree of 392 caution in Ethiopian contexts and that, if researchers want to use it, they should anticipate 393 obtaining sufficiently large samples to be able to conduct EFAs to identify the factor structure of 394 the data produced by those samples and also consider ways in which the index might be modified 395 if doing so is likely to result in more varied responses — with the underlying proviso that the 396 greater variability was associated with genuine differences in self-perception. An alternative 397 course of action would be to seek, or create, instruments that provide greater applicability and 398 usefulness for assessing maternal functioning in Ethiopian contexts. 399 400

401 The authors thank the participants for their valuable information and the data collectors and 402 supervisors for their commitment to this research. We also extend our gratitude to Arba Minch 403 University for support and opportunities provided for us to conduct this study. 404 405 Abbreviations and acronyms 406 BIMF: Barkin Index of Maternal Functioning 407 EFA: exploratory factor analysis 408 ICC: intraclass correlation coefficient 409 IFSAC: Inventory of Functional Status After Childbirth 410 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 22

411 1. Aydin J, Barkin JL, Kukulu K. Attempts to strengthen maternal functioning in the 412 postpartum period: a literature review. J Hum Sci. 2016; 13: 5751–5759. 413 https://doi.org/10.14687/jhs.v13i3.4170 414 2. Ahn S, Youngblut JM. Predictors of women’s postpartum health status in the first 3 415 months. Asian Nurs Res. 2007; 1: 136–146. https://doi.org/10.1016/S1976-416 1317(08)60016-X 417 3. Dulaney KJ, Hallwood DM, Keeter KZ. Post-pregnancy functional ability. Master’s 418 Thesis, Grand Valley State University Allendale, Michigan. 1998. Available from: 419 https://scholarworks.gvsu.edu/theses/415 420 4. Tsang S, Royse CF, Terkawi AS. Guidelines for developing, translating, and validating a 421 questionnaire in perioperative and pain medicine. Saudi J Anaesth. 2017; 11: S80–S89. 422 https://doi.org/10.4103/sja.SJA_203_17 423 5. Sultan P, Sadana N, Sharawi N, Blake L, El-Boghdadly K, Falvo A, et al. Evaluation of 424 domains of patient-reported outcome measures for recovery after childbirth: a scoping and 425 systematic review. JAMA Netw Open. 2020; 3: 1–12. 426 https://doi.org/10.1001/jamanetworkopen.2020.5540 427 6. Fawcett J, Tulman L, Myers ST. Development of the Inventory of Functional Status 428 After Childbirth. J Nurse Midwifery. 1988; 33: 252–260. https://doi.org/10.1016/0091-429 2182(88)90080-8 430 7. Barkin JL, Wisner, KL, Bromberger, JT, Beach, SR, Terry, MA, Wisniewski, SR. 431 Development of the Barkin Index of Maternal Functioning. J Womens Health. 2010; 432 19(12): 2239–2246. https://doi.org/10.1089/jwh.2009.1893 433 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 23 8. Barkin JL, Wisner KL, Bromberger JT, Beach SR, Wisniewski SR. Assessment of 434 functioning in new mothers. J Womens Health. 2010; 19: 1493–1499. 435 https://doi.org/10.1089/jwh.2009.1864 436 9. Dixon, AE. Exploratory factor analysis. In Plitchta SB, Kelvin EA, editors. Munro’s 437 statistical methods for health care research (6th ed.). Philadelphia, PA. Wolters Kluwer; 438 2013. pp. 371–398. 439 10. Medical Outcomes Trust. Trust introduces new translation criteria. Med Outcomes Trust 440 Bull. 1997; 5: 3–4. 441 11. Barkin JL, Wisner KL, Wisniewski SR. The psychometric properties of the Barkin Index 442 of Maternal Functioning. J Obstet Gynecol Neonatal Nurs. 2014; 43: 792–802. 443 https://doi.org/10.1111/1552-6909.12505 444 12. Aydın R, Kukulu K. Adaptation of the Barkin Scale of Maternal Functioning and 445 examination of the psychometric properties. Health Care Women Int. 2018; 39: 50–446 64. https://doi.org/10.1080/07399332.2017.1385616 447 13. Ansariniaki M, Lamyian M, Foroushani AR, Curry CL, Barkin JL. Persian version of the 448 Barkin Index of Maternal Functioning (BIMF): a cross-cultural adaptation and 449 psychometric evaluation. BMC Pregnancy Childbirth. 2020; 21: 83. 450 https://doi.org/10.1186/s12884-021-03556-4 451 14. Mirghafourvand M, Barkin JL, Jafarabadi MA, Karami F, Ghanbari-Homayi S. The 452 psychometric properties of the Barkin Index of Maternal Functioning (BIMF) for the 453 Iranian population. BMC Womens Health. 2019; 19: 166. https://doi.org/10.1186/s12905-454 019-0859-2 455 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 24 15. Kim H-Y. Statistical notes for clinical researchers: assessing normal distribution (2) using 456 skewness and kurtosis. Restor Dent Endod. 2013; 38: 52–54. 457 https://doi.org/10.5395/rde.2013.38.1.52 458 16. Clark LA, Watson D. Constructing validity: basic issues in objective scale 459 development. Psychol Assess. 1995; 7: 309–319. https://doi.org/10.1037/1040-460 3590.7.3.309 461 17. Chamgurdani FK, Barkin JL, Esmaeilpour K, Malakouti J, Buoli M, Mirghafourvand M. 462 The effect of counseling with a skills training approach on maternal functioning: a 463 randomized controlled clinical trial. BMC Womens Health. 2020; 20: 51. 464 https://doi.org/10.1186/s12905-020-00914-w 465 18. Barkin JL, Wisner KL, Bromberger JT, Beach SR, Wisniewski SR. Factors associated 466 with postpartum maternal functioning. J Womens Health. 2016; 25: 707–713. 467 https://doi.org/10.1089/jwh.2015.5296 468 19. Barkin JL, McKeever A, Lian B, Wisniewski SR. Correlates of postpartum maternal 469 functioning in a low-income obstetric population. J Am Psychiatr Nurses Assoc. 470 2017; 23: 149–158. https://doi.org/10.1177/1078390317696783 471 20. Barkin JL, Willis GB, Hawkins KC, Stanfill-Thomas T, Beals L, Bloch JR. Semantic 472 assessment of the Barkin Index of Maternal Functioning in a medically underserved 473 obstetric population. Perspect Psychiatr Care. 2017; 53: 95–103. 474 https://doi.org/10.1111/ppc.12141 475 21. Beavers AS, Lounsbury JW, Richards JK, Huck SW, Skolits GJ, Esquivel SL. Practical 476 considerations for using exploratory factor analysis in educational research. Practical 477 Assess Res Evaluation. 2013; 18: 6. https://doi.org/10.7275/qv2q-rk76 478 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 25 22. Costello AB, Osborne JW. Best practices in exploratory factor analysis: four 479 recommendations for getting the most from your analysis. Practical Assess Res Evaluation. 480 2005; 10: 7. https://doi.org/10.7275/jyj1-4868 481 23. Fabrigar LR, Wegener DT, MacCallum RC, Strahan EJ. Evaluating the use of exploratory 482 factor analysis in psychological research. Psychol Methods. 1999; 4(3): 272–299. 483 https://doi.org/10.1037/1082-989X.4.3.272 484 24. Gaskin CJ, Happell B. On exploratory factor analysis: a review of recent evidence, an 485 assessment of current practice, and recommendations for future use. Int J Nurs Stud. 2014; 486 51(3): 511–521. https://doi.org/10.1016/j.ijnurstu.2013.10.005 487 25. Matsunaga M. How to factor-analyze your data right: do’s, don’ts, and how-to’s. Int J 488 Psychol Res. 2010; 3(1): 97–110. https://doi.org/10.21500/20112084.854 489 26. Preacher KJ, MacCallum RC. Repairing Tom Swift’s electric factor analysis machine. 490 Underst Statistics. 2003; 2(1): 13–43. https://doi.org/10.1207/S15328031US0201_02 491 27. Goodwin LD, Leech NL. Understanding correlation: Factors that affect the size of r. J 492 Exp Educ. 2006; 74: 249–266. https://doi.org/10.3200/JEXE.74.3.249-266 493 28. Koo TK, Li MY. A guideline of selecting and reporting intraclass correlation 494 coefficients for reliability research. J Chiropr Med. 2016; 15: 155–163. 495 https://doi.org/10.1016/j.jcm.2016.02.012 496 29. Woody CA, Ferrari AJ, Siskind DJ, Whiteford HA, Harris MG. A systematic review and 497 meta-regression of the prevalence and incidence of perinatal depression. J Affect Disord. 498 2017; 219: 86–92. https://doi.org/10.1016/j.jad.2017.05.003 499 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint 26 30. Molenaar J, Hanlon C., Alem A, Wondimagegn D, Medhin G, Prince M, Stevenson 500 EGJ. Perinatal mental distress in a rural Ethiopian community: a critical examination of 501 psychiatric labels. BMC Psychiatry. 2020; 20: 223. https://doi.org/10.1186/s12888-020-502 02646-5 503 31. Kerie S, Menberu M, Niguse W. Prevalence and associated factors of postpartum 504 depression in Southwest, Ethiopia, 2017: a cross-sectional study. BMC Res Notes. 505 2018; 11: 623. https://doi.org/10.1186/s13104-018-3730-x 506 32. Barkin, JL. The Barkin Index of Maternal Functioning. n.d. 507 https://barkinindexmaternalfunctioning.com 508 All rights reserved. No reuse allowed without permission. (which was not certified by peer review) is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprintthis version posted December 2, 2021. ; https://doi.org/10.1101/2021.12.01.21266886doi: medRxiv preprint