Instrument Development
The development of the Chinese Parental Health Literacy Questionnaire (CPHLQ) comprised three stages as illustrated in Fig. 1.
Stage 1: conceptual framework and indicator generation
The CPHLQ was based on the conceptual framework developed by Sorenson et al. in 2012, operationalized with a 3×4 matrix, including three health domains (health care, disease prevention, and health promotion) and four factors of information processing (accessing, understanding, appraising, and applying) for each domain [14].
Indicators were generated through three steps. Firstly, 10 key topics about children’s physical development in three health domains were extracted from literature review and confirmed by a 20-expert consultation (Table 1). Secondly, several indicators were developed based on the 10 key topics and four factors of information processing. Thirdly, a two-round Delphi consultation was conducted with 14 experts to confirm content representativeness, health literacy relevance, feasibility and significance of these indicators. Finally 34 parental health literacy indicators were identified by consensus [15].
Stage 2: questionnaire development
Equivalent questions were designed based on 34 indicators. Among them, 29 indicators were directly transferred into one question; for other five indicators, one indicator was divided into two to four questions. Therefore a 41-question CPHLQ were constructed. Each question reflecting “accessing”, “appraising”, or “applying” were rated with a 4-point Likert scale, similar to the practice of Sorensen et al. [16]. Meanwhile, questions relevant to “understanding” were mainly in the form of true/false questions or multiple choices with four options, which were designed to test the knowledge level among caregivers. For true/false questions, the correct answer would score 4 points. For multiple choice questions there were 4 options in a question, each option was a true/false question, and one correct choice was worthy of 1 point. Besides, each question had an option “I don’t know” which would get a ‘zero’ score. Each question had a total score ranging from 0 to 4. Examples of the questions in the CPHLQ are showed in Table 2.
The original version of 41-question CPHLQ was reviewed by one researcher, two child care doctors, two nurses to assess whether the questions consistent to the indicators. The original version questionnaire was piloted with 10 parents to identify any ambiguous or unclear questions and to revise the wording. Minor changes were made to enhance clarity and comprehension.
Stage 3: Pretest
The adjusted original version 41-question questionnaire was administered to a convenient sample of 101 caregivers with children under 3 years old. The pretest was performed to conduct an in-depth Classical Test Theory psychometric analysis of question performance [17]. The question performance is determined by item difficulty and item discrimination. Item difficulty is the average score on that question divided by the highest points in our study it was 4, and the higher the value, the easier the question [18]. Item discrimination is examined using the question-total correlation [19]. A question is considered to be deleted, when: a) item difficulty lower than 0.2 or higher than 0.8 [20,21]; and b) the coefficient of question-total correlation lower than 0.3 [19]. The results were shown in Additional file 1. Three questions met all screening criteria, e.g. the following: “See the doctor in time when suspecting that children have pneumonia”, “Analyze possible risk factors for malnutrition in children”, and “Ensure children vaccinated according to the local immunization program”. However, considering the importance of immunization for children, the third question was remained and other two questions were deleted. The questionnaire with 39 questions across 3×4 sub-domains was finalized. The final CPHLQ was organized into three subscales: 12-question health care health literacy (HC-HL), 16-question disease prevention health literacy (DP-HL), 11-question health promotion health literacy (HP-HL).
Validation of CPHLQ
Participants and data collection
A cross-sectional survey was conducted in 24 community health centers (CHCs) from eight districts in Shanghai. The target participants were the primary caregivers (including parents, grandparents and other caregivers, like nanny) with children under three years old. The inclusion criteria were as follows: a) above grade three primary educations, b) able to communicate verbally or literally with the investigators; c) willing to participate in. In Shanghai, the routine child health care is provided by CHCs. Therefore, in each participating district, three CHCs were selected as the study sites, representing high, medium and low social economic status (based on local economic indicators and child health care management rates). Before the survey, child healthcare doctors in the selected CHCs as investigators were trained about how to recruit participants and complete the self-administered questionnaire.
The caregivers coming to the CHCs between March and April 2017 and meeting the inclusion criteria were invited to join in the survey by trained doctors. 1090 caregivers were approached, and 807 (74.0%) caregivers participated in the study. Among these participants, 101 (12.5%) caregivers were asked to complete the questionnaire again two weeks later to assess test-retest reliability. Data on demographics were also collected from the participants, including caregiver’s relationship with the child, education level, family income, child’s age, gender, and Hukou (the Chinese official residency registration by location, which is directly linked to social costs, social benefits and administration), etc.
Data Analysis
Several psychometric properties of the CPHLQ and three subscales were assessed.
Internal consistency was measured with Cronbach’s α [22]. Spilt-half reliability was measured with Spearman-Brown coefficient between odd questions and even questions [22]. Test-retest reliability was measured with the Pearson correlation coefficient between the CPHLQ results completed by 101 caregivers before and after [22]. In addition, the reliability analysis of three subscales was also performed. Values greater than 0.70 indicated acceptable reliability [23,24]. The floor or ceiling effects were assessed by the proportion of respondents who got the lowest or the highest score [25].
Given that hypothesized constructs were identified with a priori, confirmatory factor analysis (CFA) was used to verify the construct validity [26]. The analysis was conducted separately for the three subscales of HC-HL, DP-HL and HP-HL, in which questions were loaded into four factors related to four information-processing domains of accessing, understanding, appraising and applying. CFA was conducted with maximum likelihood estimation by using AMOS 21.0. The model fit was considered ‘relatively good’ if the following criteria were met: root mean square error of approximation (RMSEA) lower than 0.08; goodness-of-fit index (GFI) greater than 0.90; adjusted goodness-of-fit index (AGFI) greater than 0.90; comparative fit index (CFI) greater than 0.90; and due to the large sample, c2/df lower than 5 [27,28].
The final version of the CPHLQ consisted of 39 questions. When calculating the scores for parental health literacy, the weight of each indicator was equally allocated to the questions under it based on the significance assessed during Delphi consultation. The total score was transferred to percentage grading system with the full score of 100. A higher score indicated that the caregiver had higher health literacy. The mean and standard deviation (SD) of CPHLQ score were calculated. Additionally, descriptive statistics of the participants’ characteristics were tabulated. The relationships between scores and demographic characteristics were assessed with either a t-test or a one-way ANOVA.