Development of a health literacy scale for preconception care: a study of the reproductive age population in Japan

Background Preconception care aims to improve both maternal and child health in the short as well as long term, along with providing health benefits to adolescents, women, and men, whether or not they plan to become parents. However, there is limited evidence regarding the effectiveness of interventions for improving preconception health in population-based settings. To accumulate evidence in this field, this study focused on the concept of health literacy, and aimed to develop a self-report health literacy scale in Japanese, focusing on preconception care. Methods We conducted a cross-sectional online survey. Participants were recruited from December 2019 to February 2020 from the registered members of a web-based research company. Participants were Japanese men and women aged 16–49 (n = 2000). A factor analysis was conducted to select both factors and items for health-related behavior and skills (33 initial items were generated), along with an item response theory analysis to examine how the 16 items were related to people’s knowledge of preconception care. Results We developed a 6-factor (including “appropriate medical examinations,” “appropriate diet,” “stress coping,” “healthy weight,” “safe living environment,” and “vaccinations”), 25-item behavior and skills scale, as well as a 13-item knowledge scale, to evaluate participants’ health literacy around preconception care. A shortened version, consisting of 17 items, was also prepared from the 25 items. The reliability coefficients of total scores and each factor of the behavior and skills scale were comparatively high, with weak-to-moderate correlation between behavior and skills and knowledge. Conclusions The new scale will, ideally, provide information on the current state of preconception care health literacy of the general population. In addition, this scale, which consists of both behavioral/skills and knowledge dimensions, should help support the effective implementation of risk assessment programs and interventions aimed at promoting behavioral changes using a population-based approach. Future studies using different question/administration formats for diverse populations, and considering respondents’ opinions on health literacy scales should be effective in improving this scale. Supplementary Information The online version contains supplementary material available at 10.1186/s12889-021-12081-0.


Background
Preconception care aims to improve both maternal and child health in the short and long term by improving the health statuses of women and couples before pregnancy, while reducing behavioral and environmental factors that could become future maternal and child health risks [1]. Lifestyle habits, such as drinking alcohol, smoking, drug use, high-risk sexual behavior, and the nutrition (e.g., obesity and lack of essential vitamins) of parents before/ during pregnancy, have been found to affect the health of future mothers and children. They are known to trigger gestational diabetes and hypertensive disorders during pregnancy, and low birthweight, preterm birth, birth defects, obesity and chronic disease in offspring, making it necessary to make women and men of reproductive age aware of these behaviors and of their influence on reproductive health and childbearing [2][3][4]. The greatest impact occurs in early pregnancy, often before women know that they are pregnant, and considering the substantial time needed to reach a healthy lifestyle, intervention before pregnancy is recommended [2,4]. In this study, we defined preconception care as interventions provided to women and couples of childbearing age to improve maternal and child health in the short and long term; simultaneously, preconception care improving their health statuses provides health benefits to both adolescent and adult women and men, whether or not they plan to be parents [5]. The World Health Organization (WHO) and the Centers for Disease Control and Prevention (CDC) advocate the importance of pre-pregnancy healthcare and the need for a provision of comprehensive information and professional health services [1,6].
However, there is limited evidence regarding effective interventions for preconception health in the primary care setting or in the public health and community settings [4,7,8]. To accumulate knowledge in this field, more evidence is needed regarding the effectiveness of interventions aimed at the general lifestyles and behavioral risks for the general population in terms of preconception care [9]. To achieve this, the present study focused on the concept of health literacy. Although, there is no consensus about the definition of health literacy or its conceptual dimensions [10], in this study, we defined health literacy as the knowledge and competency to access, understand, appraise, and apply health-related information; the competencies also incorporate the qualities of functional, interactive and critical health literacy as proposed by Nutbeam [10,11]. Health literacy is considered an asset for improving people's empowerment in the areas of healthcare, disease prevention, and health promotion, and through the health literacy process, people are expected to be able to take control of their own health by applying their specific health literacy skills [10]. It is important to enhance health literacy, specifically within the preconception care setting wherein risk assessment and reduction programs are implemented to reduce lifestyle-related problems, such as alcohol consumption, smoking, and under-or over-nutrition before pregnancy, in addition to promoting future maternal and child health.
To improve people's health literacy, it is necessary to properly evaluate their competencies and the effectiveness of intervention programs. Evaluations for preconception care have been conducted in terms of health knowledge(e.g., folate can prevent birth defects), behavioral changes (such as folate intake, alcohol, and smoking), and health outcomes (e.g., adverse pregnancy outcomes, such as low birth weight and preterm birth) [7,8]. However, there are no appropriate indicators to measure health literacy specific to preconception care among the general population. There are scales, such as comprehensive concept-based health literacy scale for the general population, [12] also in Japanese [13], as well as health literacy measurement tools that focus on specific aspects, such as the recognition and pronunciation of medical terms, numeracy, comprehension, and decision-making competencies [12]. There are scales that cover a comprehensive range of skills including functional, communicative/interactive, and critical health literacy in the clinical and public health contexts [14][15][16]. However, these are designed for general use or are specific to diabetic patients and have limited capacity to assess the specific skills in preconception care. Kawata et al. developed a health literacy scale for women of reproductive age comprising the following factors: women's choice for adopting health information and practice, self-care during menstruation, knowledge of the female body, and sexual discussions with partner [17]. However, no study has developed a health literacy scale focused on preconception care for men and women of reproductive age. The health literacy scale specific to preconception care is needed for assessment and reduction of risks related to short-and long-term maternal and child health outcomes, a key issue in preconception care. Therefore, this study aimed to develop a self-report health literacy scale in Japanese, to assess the knowledge and skills necessary to acquire information, focusing specifically on preconception care, understanding this information, and critically analyzing and appraising it. The use of this scale should allow measuring the health literacy of both women and men of reproductive age, along with the evaluation of interventions occurring within the primary care setting or public health and community setting.

Methods
We defined health literacy as knowledge and competencies to access, understand, appraise, and apply health-related information. We considered health knowledge as a domain of health literacy, and developed the scale as divided into items on knowledge and behavior/skills related to the health literacy competencies. We conducted a factor analysis to select the factors and items for a health literacy scale relating to people's behavior and skills. In addition, to develop more readily available screening tools, a shortened version with a reduced number of items was also examined. Items relating to people's knowledge of preconception care were analyzed using the item response theory (IRT) to ensure that the scale contained items with varying difficulties, and that items had sufficient capacity to differentiate between participants. This study follows the COSMIN reporting guideline [18].

Participants
We conducted a cross-sectional online survey. Participants were recruited from December 2019 to February 2020 from the registered members of a web-based research company (commissioned by Cross Marketing Group, Inc.). In this study, we targeted a wide range of age groups, both men and women, with the aim of developing the proposed measurement tool in universal environments, such as in schools and workplaces. Inclusion criterion was men and women aged 16-49 years. The survey included participants across Japan, with the extracted sample adjusted for their sex, place of residence, and age. We designed it so that 30% of all participants were pregnant or had delivered a child at some point in the past (aimed to include people with a broad knowledge level about pregnancy and childbirth).
Participants were involved via e-mail with a link to the response Webpage for survey cooperation. They were asked to respond only if they agreed to participate in this study. The survey was first administered to 1000 participants, then to a different set of 1000 participants, to examine if the original findings could be replicated with different samples (resulting in a total of 2000 participants).

Questionnaire design
The health literacy scale items Contents of behavior/skills and knowledge for the preconception care scale were extracted from the relevant literature to create a first items' pool, with specialists in maternal medicine developing each questionnaire item. The following topics were identified: reproductive life plans, nutrition, physical activity, folic acid supplementation, reproductive health, sexually transmitted diseases, immunizations, infectious diseases, environmental exposures, psychosocial stressors, mental health, tobacco use, alcohol and other substance use, partner violence, chronic medical conditions, medication use, family and genetic history, and regular checkups [5,19,20]. The questionnaire was designed to cover four domains of health literacy competencies (access, understanding, appraisal, and application of health information). Questions were phrased as: "Do you read …? " (examples of Access competency); "Do you know …? " (examples of Understanding competency); "Are you able to assess …? " (examples of Appraisal competency); and "Do you make it a habit …? " (examples of Application competency). Questions on respondents' behavior and skills were answerable via a 4-point Likert scale (1 = "strongly true," 2 = "somewhat true," 3 = "rarely true," and 4 = "not at all"). For the questions on respondents' knowledge around preconception care, the participants were asked to choose one correct answer from a range of multiple options.

Socioeconomic variables
Participants' current place of residence, sex, age, marital status, parity (including if currently pregnant), employment, education history, and annual household income were surveyed.

Health status
Data were collected on respondents' height, weight, sleep duration, current health status, regular medical screenings, plans to become pregnant within a year, alcohol use, and whether or not the participant smoked.
The Communicative and Critical Health Literacy (CCHL) scale is a 5-item, 5-point (theoretical range: 1-5) health literacy scale, with higher scores indicating higher health literacy. It has a Cronbach's α of 0.86 according to a previous study [14]. Additionally, the Japanese version of the Cardiff Fertility Knowledge Scale, a 13-item scale for assessing fertility knowledge was used. Correct answers are assigned one point and incorrect or "do not know" answers are assigned zero points; scores are then reported as a percentage of the highest possible score. This scale has a Cronbach's α of 0.74 based on a previous study [21]. Both of these scales were used to verify the validity of our questionnaire. We adopted these scales for criterion-related validity verification because they are both validated in Japanese and contain a relatively small number of items.

Data analysis
Factor analysis for the health literacy behavior and skills of preconception care An exploratory factor analysis was conducted to select the factors and items for the health literacy scale relating to people's behavior and skills. We used 33 items that sought respondents' answers on their behavior and skills using a 4-point Likert scale. Cronbach's alpha coefficients were calculated for the factors of each scale to determine their reliability. In addition, in consideration of practicality, a shortened version, with a reduced number of items, was also examined. We permitted correlation among factor score scales, so we used promax rotation as a method of oblique rotation.
We used a two-way analysis of variance (ANOVA) to examine sex, history of pregnancy, and childbirth related differences in the subscale obtained from the factor analysis. There were a few participants (or their partners) who were currently pregnant, so we excluded them from this ANOVA.

IRT analysis for the knowledge scale
In total, 13 questions relating to respondents' knowledge of preconception care (that asked them to choose one answer from a range of options) were summarized as dichotomous data, which marked each response to these items as either correct (1) or incorrect (0). Next, the dichotomous data were checked to see if the scale was unidimensional using factor analysis, with the results then demonstrating the unidimensionality of the scale. Then, the dichotomous data were analyzed using IRT to obtain the item difficulty index, which excluded the distribution effect of participants' knowledge levels; the scale needed to include items with varying difficulties and to have sufficient capacity to differentiate between participants. We assumed a two-parameter logistic model (2PL) to examine the discrimination parameters (slope: how well the items differentiated between participants with low and high level of knowledge of preconception care) and difficulty parameters (threshold: the item's difficulty with larger threshold estimate indicated greater difficulty) of each item. Then, we compared these with those of the Cardiff Fertility Knowledge Scale. Additionally, expecting a significant difference in the correct response rates for each item between men and women, we analyzed the data using a multiple group IRT model, with women as the reference group.

Confirmation of scale reliability and validity
Cronbach's α was calculated to assess internal consistency of scores for each scale. To confirm the validity of the scale, we conducted a correlation analysis using the variables of the behavior and skills and the knowledge scales of preconception care. Pearson's r was calculated between the subscale and total scores obtained from the factor analysis using items for respondents' behavior and skills; the scores of participants' knowledge of preconception care were scaled by the IRT analysis, the related scales of the CCHL scale scores, and Cardiff Fertility Knowledge Scale scores to confirm the criterion validity ( Table 5). As significant differences in the knowledge scores between men and women were predicted, we conducted the analysis by sex.
The significance level was set at p < 0.05. The data were analyzed using statistical packages SPSS 26.0 (factor analysis), BILOG-MG 3 (IRT), and AMOS 26.0 (multivariate regression with a multiple population simultaneous analysis).

Participant characteristics
The characteristics (n = 2000) are described in Table 1 (Additional file 1 shows the first and second group results, separately). There were 1000 men and women each, of average age of 34.8 years (standard deviation (SD) ±9.1). Although no sex differences were found in most of the items, men were more likely than women to work full-time, have an educational level of "universities and graduate schools," and drink or smoke. Fewer women had regular checkups "every year" than men (63.9% men and 49.9% women).

Factor analysis results for the health literacy behavior and skills of preconception care
The explanatory factor analysis was conducted using all 33 items relating to people's behavior and skills. Each item was evaluated on a 4-point scale (maximum likelihood estimation method with a promax rotation). To achieve a simple structure, we changed the number of factors and excluded items with low commonality (< 0.1) and those that overlapped with multiple factors with a high factor loading (> 0.3, < -0.3). As a result, six factors, comprising 25 items, were retained (Table 2; excluded items are shown in Additional files 2 and 3). Then, the final 25 items were matched in the first and second surveys. We labeled the extracted factors as "appropriate medical examinations" (seven items), "appropriate diet" (four items), "stress coping" (four items), "healthy weight" (four items), "safe living environment" (three items), and "vaccinations" (three items), respectively.
The cumulative contribution ratio of the six factors in the above factor analysis was 68.9%. The eigenvalue of Factor 6 was above 1 (1.052), but below 1 (0.810) for Factor 7.
In addition, a shortened version of this scale, with a reduced number of items, was examined (n = 2000). We examined a model of the shortened version of the scale with items selected based on a factor loading of ±0.6, preserving the six-factor structure of the original 25-item total scale. After factorial loading and validity were investigated, and the number of items was reduced, a shortened version comprising 17 items relating to people's behavior and skills was obtained (Table 3).
Participants' sex and history of pregnancy and childbirth related differences Figure 1 presents participants' sex and history of pregnancy and childbirth related differences in the subscale score of behavior and skills scale (participants who or whose partners were pregnant [n = 43] were excluded). In total, men scored higher than women and, in terms of history of pregnancy and childbirth, "No" scored higher      than "Yes," indicating a tendency for a lack of health literacy in men and among participants who had no history of pregnancy and childbirth. Regarding the subscale scores on "appropriate medical examinations" (Factor 1), "healthy weight" (Factor 4), "vaccinations" (Factor 6), sex and a history of pregnancy and childbirth related differences were consistently significant in the first and second surveys, with sex and a history of pregnancy and childbirth interacting significantly with "vaccinations" (F (1, 1953) = 16.34, p < 0.001; results for Total (n = 2000)) (Additional file 6).

IRT analysis results for knowledge of preconception care items (knowledge scale)
In total, 13 items around knowledge of preconception care, which asked respondents to choose a correct answer, were scaled using the IRT (Table 4) (Additional file 4 shows the knowledge question and answer items). Three other items about whether participants had a regular doctor, received regular health screenings, and had a dental treatment plan were excluded from the analysis, due to considerations around content validity and the low factor loading in the results of the factor analysis. Additionally, after examining only the factor loadings, two items about folic acid intake demonstrated low factor loadings. However, these questions pertained to the respondents' knowledge and were considered essential to the analysis. Therefore, these questions were retained. We examined the item parameters, including the discrimination (slope) and difficulty (threshold) of each item under the assumption of 2PL. The Cardiff Fertility Knowledge Scale score was scaled using IRT (Additional file 5), along with comparing the item parameters (discrimination/difficulty). Our 13-item knowledge scale had a lower mean for the item discrimination parameter than the Cardiff Fertility Knowledge Scale (0.693 vs. 0.897). However, the item difficulty parameter of our 13item knowledge scale varied from approximately − 4.0 to 3.9, while that for the Cardiff Fertility Knowledge Scale ranged only from − 0.8 to 1.3.
As significant difference was found in the correct response rate for each item between men and women, we analyzed the data by sex. We calculated the IRT ability parameters and found that, for our 13-item knowledge scale, the mean for women was 0 (set as reference; SD = 0.787) compared with − 0.388 (SD 0.867) for men. In all 13 of our items, men scored lower than women (higher scores indicating higher knowledge).

Scale reliability and validity Internal consistency and measurement error
We obtained relatively high reliabilities for the 25-item behavior and skills scale with the total score (0.935) and for all six factor scales (Factor 1 = 0.898, Factor 2 = 0.854, Factor 3 = 0.852, Factor 4 = 0.852, Factor 5 = 0.823, and Factor 6 = 0.686) even though there were fewer items in the lowest factor. Further, we obtained relatively high reliabilities for the shortened version with the 17-item total score (0.910) and for all six factor scales (Factor 1 = 0.900, Factor 2 = 0.846, Factor 3 = 0.823, Factor 4 = 0.848, Factor 5 = 0.823, and Factor 6 = Values of Cronbach's α for the 13-item knowledge scale was 0.661 (standard error of measurement was 1.388), and 0.831 for the Cardiff Fertility Knowledge Scale (standard error of measurement was 1.472). Although the 13-item knowledge scale had a lower reliability score, there was almost no difference in measurement error. Table 5 presents the correlation coefficients between the subscale and the total scores of each scale. In total, six factor scores and total scores of behavior and skills scale (6 factors: "appropriate medical examinations," "appropriate diet," "stress coping," "healthy weight," "safe living environment," and "vaccinations") were significantly correlated with the 13-item knowledge scale and the Cardiff Fertility Knowledge Scale scores, with an exception in the correlation between "vaccinations" and the 13-item knowledge scale scores for women (Table 5). There was almost no correlation between the CCHL (a 5-item health literacy scale) and other scales (subscale and total scores of 25-item behavior and skills scale, 13-item knowledge scale, and Cardiff Fertility Knowledge Scale).

Discussion
In this study, we developed a 6-factor ("appropriate medical examinations," "appropriate diet," "stress coping," "healthy weight," "safe living environment," and "vaccinations") 25-item behavior and skills scale, as well as a 13-item knowledge scale, to evaluate people's health literacy in terms of preconception care. The final 25 items were matched in the different populations of the first and second surveys. A short version, consisting of 17 items, was also created from the 25-item behavior and skills scale. The item difficulty parameter of our 13item knowledge scale showed wider variability than that of a previous knowledge scale [21], meaning that this would allow a more accurate measurement of the extent of participants' preconception care knowledge, especially for those with extremely low knowledge in this area and consequently, higher needs for health promotion.
The reliability coefficient of the total scores and each factor of the behavior and skills scale was  Factors 1-6: higher scores indicate a greater tendency for a lack of health literacy Overall (25-item total score): higher scores indicate a greater tendency for a lack of health literacy CCHL: Communicative and Critical Health Literacy scale, with higher scores indicating higher health literacy Our 13-item scale assessing preconception care knowledge, with higher scores indicating higher knowledge CFKS-J: the Japanese version of the Cardiff Fertility Knowledge Scale, with higher scores indicating higher knowledge comparatively high, with weak-to-moderate correlation between our behavior and skills scale and knowledge scale. In this study, there was almost no correlation between the CCHL (a 5-item health literacy scale) and our scales, including the 25-item behavior and skills scale and the 13-item knowledge scale, as well as Cardiff Fertility Knowledge Scale. We adopted CCHL for criterion-related validity verification because it was validated in Japanese and contained a relatively small number of items. However, the CCHL may not cover the whole concept of the communicative and critical health literacy [14], and further research with other validated health literacy instruments, such as HLS-EU-Q [13], is required to examine our scale's validity as a health literacy scale.
Regarding the subscales of the behavior and skills scale, men scored higher than women, and participants with a history of pregnancy and childbirth scored "No" more often than "Yes." This indicated higher needs for health promotion among men and participants with no history of pregnancy and childbirth. Although this study involved participants from a wide range of age groups including teenagers, not having a pregnancy history does not necessarily mean low health literacy. Additionally, for the knowledge scale, across all 13 items, men scored lower than women. These results are consistent with our hypothesis and suggest that we need to focus on these particular groups in future preconception care programs.
Poor nutrition and obesity among reproductive age women are common problems in both low-and highincome countries, and few interventions related to both diet and lifestyle during the pre-pregnancy period are available [2]. From a preventive medicine perspective, it is important to adequately inform both men and women before pregnancy about the potential impact of lifestyle factors, such as smoking, drinking alcohol, nutrition, infectious diseases, and exposure to pollutants, on their reproductive and future maternal and child health. In addition, more risk assessment and reduction programs need to be provided. The health literacy scale, consisting of both behavior and skills scale and knowledge items, for preconception care developed in this study, will provide information about preconception health to the respondents and support the effective implementation of risk assessment and interventions to promote healthy behavioral changes.

Conceptual issues
In this study, we developed a scale for a wide range of age groups (from adolescents to people in their 40s) for both men and women, with the aim of using it in a universal range of environments, such as schools and workplaces. It has been discovered that a populationlevel approach is important because it is necessary to make long-term efforts to improve people's appropriate body weight; it is especially ideal to establish appropriate lifestyle habits from adolescence [2]. In addition, although the WHO recommends male involvement in the health promotion dimension of preconception care [1], previous studies on preconception care have found little risk assessment, screening, or other preconception interventions for men [7,22]. Additionally, our results demonstrate that men score poorer results than women in all subscales of the behavior and skills scale and knowledge scale items, indicating the need for specific preconception health promotion among men.
Conversely, this scale could not include genderspecific reproductive health issues, such as menstrual cycle and self-care. In addition to the universal approach adopted in this study, further research is needed to measure specific risks, such as underlying diseases or detrimental socioeconomic attributes, in both clinical setting and outreach programs [23]. Depending on an individual's life stage, lifestyle, and level of health literacy, appropriate intervention programs may differ. In future, it will be necessary to identify targeted populations that specifically require preconception care, which will then facilitate the creation of an item pool and development of a behavior and knowledge scale targeting such populations.

Methodological issues
This study has the following limitations. First, participants in this study were recruited from registered members of a web-based research company, which could lead to a selection bias. Although the internet usage rate (individual) was over 90% for each age group between 13 and 69 in 2019 in Japan [24], internet surveys involving only registered members of a web-based research company are not guaranteed to be representative of the sample, and we could not know the difference between health literacy of registered members and the rest of the population. Muscat et al. suggested assessing the administration time among a diverse sample, and to use strategies to increase engagement (e.g., use of images and different question formats), as well as to consider use of alternative administration formats (e.g., paper vs online) for the development of health literacy instruments [25]; however, this was not evaluated in this study. Therefore, future research should compare the scale scores among diverse populations, and the scale's reliability and validity as a health literacy scale compared with standard health literacy measures, such as HLS-EU-Q [13], must also be examined. Furthermore, Muscat et al. suggested involving consumers in the development of health literacy instruments and examining whether consumers