A low-birth-weight risk assessment scale: development and validation through a questionnaire-based survey

The present study aimed to clarify the causal relationship between maternal daily lifestyle during pregnancy and low infant birth weight and to develop a scale for the assessment of maternity risk factors for low birth weight in infants born at term. A low birth weight risk assessment scale for use in prenatal checkups was developed, and its reliability and validity were examined.

Subjects were mothers with the youngest child below school age (≤6 years).

This was a questionnaire-based survey. A total of 3400 questionnaires were distributed to 42 nursery schools and kindergartens in A Prefecture (including those in remote islands) between February and March 2015. After requesting and obtaining consent from the heads of the Nursery School Association and Private Kindergarten Association Recommended Facilities in A Prefecture, the nursery schools and kindergartens distributed questionnaire packages (containing a description of the survey, a self-administered questionnaire, and a return envelope) to parents and guardians. Postal return of the completed questionnaire to the researchers using the return envelope was considered as the final consent for participation.

Questions regarding maternal characteristics included maternal age, height, pre-pregnancy weight, age at the time of delivery, weight gain during pregnancy, participation in classes for mothers, employment status, and the number of children in the family. Questions for pre-school children included age, sex, birth weight, and gestational age at birth. In addition, a draft scale prepared by the below mentioned method was included(Additional file 1).

The preliminary study identified variables of maternal pre-pregnancy malnutrition, low weight gain, and smoking during pregnancy as factors contributing to low birth weight [6]. These findings together with causes of low birth weight identified in previous research [4, 7, 8] were reflected in the scale items. Following basic procedures for questionnaire design, the scale was developed on the basis of the preliminary survey results and previous research [3, 7, 8], and the repeated review of the scale was conducted until an agreement was reached among the joint authors and then with public health nurses with hospital-based clinical experience. Efforts were then made to ensure the accuracy and validity of the scale, seeking repeat opinions from a researcher/midwife instructor of maternal and child health nursing. Through these steps, the scale was organized into 33 items.

A four-point rating scale with the neutral mid-point eliminated was used to accurately confirm respondents’ opinions (1, strongly disagree; 2, slightly disagree; 3, slightly agree; and 4, strongly agree). Responses were scored from 1 to 4 points to attain a score for each item. The magnitude of the score demonstrated the degree of relevance to risk factors for low birth weight.

The scale was developed on the basis of item analysis, and scale reliability and validity were then examined. The inclusion of each item was determined as follows: (i) the presence of floor or ceiling effects was confirmed on the basis of mean values and standard deviations (average value + standard deviation was calculated for the ceiling effect, and when this value exceeded the maximum value of 4 points, it was considered that there was a ceiling effect. When the average value − standard deviation was smaller than the minimum value of 1 point, it was considered that there was a floor effect); (ii) Good–Poor analysis [G–P analysis; t test was performed to assess the difference between the mean values of the upper and lower thirds of the total scores (upper group and lower group, respectively) for each item, and items with a significant difference were retained] was performed; and (iii) inter-item correlation matrix was obtained.

The reliability of the scale was determined on the basis of the Cronbach’s α coefficient and correlation analysis between the item and scale [Item–Total correlation] to investigate item inclusion. The validity of the scale was examined as follows: (i) content validity of the scale was investigated; (ii) construct validity was investigated using exploratory factor analysis; and (iii) known-groups validity was investigated. Concurrent validity could not be investigated as no adequate scale exists for investigating criterion validity; therefore, known-groups validity was tested. The following three variables according to their known-groups validity were classified and subjected to comparative analysis of the mean scale total scores: (i) birth weights classified as “< 2500 g” or “≥2500 g” (t-test); (ii) birth histories classified as “primipara” or “multipara” (t-test); and (iii) maternal smoking status classified into the following five groups: “no history of smoking,” “stopped smoking before this pregnancy,” “permanently stopped smoking during this pregnancy,” “temporarily stopped smoking during this pregnancy,” and “continued smoking throughout this pregnancy” [one-way analysis of variance (ANOVA) and Tukey–Kramer honestly significant difference (HSD) test].

The study aims and outline, subject rights (voluntary nature of participation and cooperation and right to refusal to participate), protection of privacy through questionnaire anonymity, use of ID numbers instead of names, protection of personal information, data confidentiality, restriction of data use to the present study, destruction of questionnaires immediately after finishing the present study, and the fact that returning a completed questionnaire would be considered as consent for participation were explained to all subjects.

Out of the 3400 questionnaires distributed, responses were received from 671 mothers (collection rate, 19.7%). The aim of this study was to construct a scale to assess pregnant women who may deliver a low birth weight baby because of their lifestyles, which may affect the development of the fetus lurking in their daily life. Responses from mothers of twins or sick babies, which are conditions that are considered to be associated with low birth weight, were excluded from the analysis. In total, the responses from 630 mothers of babies (valid response rate, 18.5%) and 916 mothers of pre-school children were analyzed.

Maternal characteristics were as follows: mean age, 35.7 [standard deviation (SD)], 4.9; range, 23–51) years; mean number of children in the family, 2.2 (SD, 0.9; range, 1–8); mean age at time of delivery, 31.2 (SD, 4.6, range, 20–50) years; and employment status during pregnancy, 397 non-working (43.3%) and 519 working (56.7%). Child characteristics were as follows: mean age, 3.6 (SD, 1.8; range, 0–6) years; sex, 463 boys (53.3%) and 406 girls (46.7%) (Table 1).

The response rate for all 33 items in the scale was ≥99.5%. The scores for each item ranged from 1 to 4, with mean scores of 1.1–3.8 (SD, 0.5–1.2) (Table 2). The confirmation of the data distribution revealed a floor or ceiling effelatict for items 2, 5–7, 10, 12, 22–24, 27, 30, and 32. However, these items were considered necessary to investigate factor structure and were included in the factor analysis. On G–P analysis, significant differences were observed between the upper and lower groups for all items (Table 2). Therefore, all items were judged to adequately correspond with the total scores and were included in the scale.

Eleven factors were extracted on factor analysis using principal factor analysis and promax rotation of the responses for all 33 items. An eigenvalue cutoff of ≥1 was used to determine the number of factors to retain. Cronbach’s α coefficient for the 11th factor was low (0.363). Therefore, the item “I consumed alcohol”, which had the greatest number of low coefficient values in the inter-item correlation matrix, was excluded, and factor analysis was repeated for the remaining 32 items. However, Cronbach’s α coefficient for the 11th factor remained low; therefore, the factor analysis process was repeated until Cronbach’s α coefficient increased and the results stabilized. The following items were eliminated from subsequent analyses: “I consumed alcohol,” “I often experienced abdominal bloating,” “I thought that the saying, ‘deliver them small and raise them big,’ held true for me and my child(ren),” and “I refrained from driving a car.” Factor analysis was conducted on the remaining 29 items, and the lowest Cronbach’s α coefficient increased to 0.460. At this point, nine factors were extracted. Factor analysis was conducted again excluding four items with a factor loading of under 0.35 (items 4, 21, 25, and 32). As a result, an optimal scale was obtained, which was interpreted without contradictions in the meanings of the items. The final scale comprised 25 items in 9 factors. The details and results of this analysis are shown in Fig. 1.

Kaiser–Meyer–Olkin test with the sampling adequacy of 0.640 and the screen plot with changes in factor eigenvalues indicated that 9 factors were appropriate. The cumulative contribution ratio of the sum of squared load amount after extraction was 57.0%. Therefore, items with factor loading ≥0.35 were included in the scale, and subsequent examinations of reliability and validity were conducted for 25 items in 9 factors. The factor pattern and factor correlation matrix are shown in Table 3.

The 9 factors were labeled to reflect the related content assessment as follows. (i) guidance at each checkup: maternal feelings at the time of guidance (five items related to guidance from doctors and nurses regarding prenatal checkup findings); (ii) adequate rest: the state of maternal rest during pregnancy (four related items, such as maternal rest, holidays, and sleep); (iii) support from husband: support provided by the husband and related maternal satisfaction (two items); (iv) effects on the fetus: effects of alcohol consumption and tobacco on the fetus (two items); (v) support from society: social support other than that provided by the husband and family and related maternal satisfaction (two items); (vi) support from family: support provided by parents (including in-laws) and related maternal satisfaction (two items); (vii) effects of minor troubles: the extent of and related maternal feelings regarding minor troubles, such as morning sickness (three items); (viii) good lifestyle habits: good lifestyle habits during pregnancy (three items); and (ix) fall risk and lifestyle changes: concern regarding falling during pregnancy and associated lifestyle changes (two items).

The strongest positive correlation in the factor correlation matrix was between “social support” and “family support” (0.417), followed by among “support from husband”, “support from society”, and “support from family,” between “adequate rest” and “good lifestyle habits” (0.336–0.395), and between “effects on the fetus” and “fall risk and lifestyle habits” (0.234).

These findings show a structure comprising two groups, one related to daily lifestyle during pregnancy, including “adequate rest,” “good lifestyle habits,” “effects on the fetus,” “effects of minor trouble,” “guidance at each checkup,” and “fall risk and lifestyle changes” and one related to support including “support from husband,” “support from family,” and “support from society.”

The mean scale total score was 68.5 (SD, 8.1; range, 32–93), and the normal probability–probability plot showed a normal distribution. The scale score was 25–100 points. Regarding the overall reliability of the scale, Cronbach’s α coefficient for each item ranged from 0.09–0.585 and that for the overall scale was 0.701 (Table 3). A positive Item–Total correlation was also observed.

After confirming content validity among the authors, aspects of the scale, such as standard language among items and concept equivalence, were repeatedly evaluated by two authors, and two expert researchers in the field of maternal and child nursing summarized the items while maintaining their meaning to form the scale.

Exploratory factor analysis using principal factor analysis and promax rotation was performed to confirm the structure of the assessment indicators. Regarding birth weight, the mean subscale total scores of the ≥2500 g group were significantly higher than those of the < 2500 g group (t = − 3.153, p = 0.002). Regarding birth history, the mean total scores of primipara were significantly higher than those of multipara (t = 4.317, p < 0.00). One-way ANOVA revealed a significant difference in scores for maternal smoking status, whereas multiple comparison (Tukey–Kramer HSD test) demonstrated that the “no history of smoking” group had significantly higher mean total scores than the “continued smoking throughout this pregnancy” group (p < 0.00). The “stopped smoking before this pregnancy” group had significantly higher mean total scores than the “continued smoking throughout this pregnancy” group (p < 0.00), while the “permanently stopped smoking during this pregnancy” group also had significantly higher mean total scores than the “continued smoking throughout this pregnancy” group (p < 0.00) (Table 4).

Regarding subject characteristics, the number of children in a family in the present study was slightly higher than that reported in the birth rates by maternal age group, number of live births by birth order, and the sex distribution of infants in the population for 2013 [2], while maternal age at the time of birth and the sex of the child were similar to the reported values. Valid responses were obtained from 630 mothers and 916 young children, and the data were normally distributed; thus, the data volume was sufficient to examine reliability and validity.

The reliability and validity of this maternity low birth weight risk assessment scale were verified; thus, it can be used for the screening of high-risk pregnancies during health guidance. This scale would be particularly useful for identifying high-risk pregnancies during obtaining the history of mothers in prenatal checkups or during consultations with pregnant employees in the labor management section in their company regardless of the ability of the consultant.

There are some limitations in the present study. This was a cohort study regarding the daily lifestyles of pregnant women. The survey method may have introduced recall bias as responses were obtained from women with children of nursery school or kindergarten age looking back at and recalling their pregnancy lifestyles. Ideally, the survey should have been conducted either during pregnancy or soon after birth. However, the present study used a systematic oral history method focusing on mothers with children of nursery school and kindergarten age who were able to reflect and recall their lifestyle and sense of values during pregnancy [15]. The scale also included items with floor or ceiling effects, and there were factors with low Cronbach’s α coefficients. Therefore, caution should be exercised when interpreting the results. Also, this survey was conducted in a single prefecture, and conducting further studies using the present scale in other areas will increase its accuracy, making it more useful. However, to the best of our knowledge, this is the first study to develop a reliable and valid assessment scale focused on maternal daily lifestyle and psychology.