Scientific Papers

Psychometric evaluation of the Protection Motivation Theory scale in assessing fall protection motivation among older adults to reduce fall risk | BMC Geriatrics


Aims

The study aims to evaluate the reliability and validity of the culturally adapted Protection Motivation Theory scale for older adults’ fall protection motivation or protective behaviours to reduce fall risk.

Study design

A cross-sectional study was conducted in Sarawak, Malaysia, from November 2021 to January 2022. The study consisted of two phases, (i) translation of the PMT Scale, cross-cultural adaptation and face validation, (ii) pre-testing of the PMT Scale.

Sample

Three hundred eighty-nine older adults aged 55 years and above were identified and included from a local primary healthcare clinic in Sarawak. Multistage random sampling was adopted to identify and select a primary healthcare clinic and participants from the clinic. This sampling was appropriate to be adopted for large or dispersed populations [16]. There was an estimation of 12 divisions governed by the Sarawak state’s administration. A primary healthcare clinic was randomly selected from a division after a division was randomly identified from the 12 divisions registered within the Sarawak State Health Department. Next, 10 to 30 participants registered with the clinic were randomly chosen in each community within 18 settlements receiving healthcare services from this primary healthcare clinic.

The sample size was determined based on the minimum ratio between items per response: one item to five responses for exploratory factor analysis (EFA) [17]. Several authors justified that a ratio of one item to five responses was able to approximate about 40% of samples with correct structures [18]. They also recommended a larger sample of 300 or applying a higher ratio of one item to ten or twenty responses for better sample size estimation [18]. On the contrary, other authors suggested that samples in the range of 100–200 were appropriate with well-determined factors, such as the main factors defined by many indicators or indicator variables with loadings > 0.80 and communalities within the range of 0.5 or above [19]. They also stated that if the communalities fall into the range of 0.40 to 0.70, then the sample size should be at least 200 [19]. The instrument tool used in this study was adapted from a previous study with constructs or factors derived from an established theory. Therefore, it was considered to have well-determined factors. This study’s communalities of factor analysis ranged between 0.5 and 0.84, with each factor comprising several indicators. Hence, 184 participants were considered as an acceptable sample size to be included for EFA, in which a total of 389 participants were randomly split into the first half for EFA and the second half, with a total of 195 participants were included in the confirmatory factor analysis (CFA) using the partial least square structural equation modelling (PLS-SEM) [17]. In addition, the total number of participant has met the minimum sample size of 160 for PLS-SEM analysis [20].

The inclusion criteria included community-dwelling older adults at the primary care clinic of Kota Samarahan aged 55 years old and above who could read, write or understand Malay or English. Those suffering from mental health problems were excluded from the study.

Validity and reliability of the instrument

Validity was assessed on an instrument tool that correctly measures what it intends to measure [21]. This assessment includes content and construct validity [22]. Both validities refer to the degree to which instrument content sufficiently reflects the construct being measured and to which a set of variables represents the construct to be measured [17]. Therefore, content validity was assessed using the panel committees’ expert opinions on the items within the instrument and rated according to their equivalent in the content validity index (CVI). Next, the construct validity was inquired using the EFA and CFA.

Reliability was tested to ensure that the instrument tool has the ability to reproduce a consistent result or refer to stability, internal consistency and equivalence of a measure [22]. Several examples of the tests were used, such as the intraclass correlation coefficient (ICC), mainly used to assess continuous variables stability while also considering the measurement errors [23]. Secondly, another assessment was test–retest reliability, which was performed to estimate the consistency of measurement repetition [22]. Thirdly, Cronbach’s alpha coefficient was a commonly used assessment to determine the instrument tool’s internal consistency [17].

The questionnaire consists of two parts and was approved for use by the original authors [9, 24]. Part I is the adapted and modified version of the participants’ sociodemographics: age, gender, educational level, ethnicity and fall history. Meanwhile, the adapted Part II questionnaire consists of 35 items with eight constructs of the PMT scale. The PMT scale was scored using a 5-point Likert Scale, from strongly disagree to strongly agree, ranging from one to five for perceived sensitivity/severity, self-efficacy, response efficacy and perceived rewards. Meanwhile, another five-point Likert Scale ranging from not at all (1), a little (2), somehow (3), much (4), to too much (5) rated for fear, perceived costs and protection motivation.

Phase I: Translation of the PMT scale, cross-cultural adaptation and face validation

The translation and adaptation process were referred to guidelines of WHO [25], Gjersing, Caplehorn, and Clausen [26] and Beaton, Bombardier [27] (as illustrated in Fig. 2). First, five professionals consisting of three lecturers in health sciences, a physician and a geriatrician reviewed the original PMT scale for content suitability according to the local setting. Next, three panel committees assessed and rated the instrument’s CVI.

Fig. 2
figure 2

Flow chart of the adaptation, translation, face validation and pre-testing process of the PMT Scale

Second, the questionnaire was translated into the Malay language by an independent certified translator who is bilingual (Malay or English). In addition, another translator with a health sciences background has independently translated the scale into Malay. The study members then reviewed and clarified both versions of the translated Malay questionnaire. Fall is defined during the harmonisation process based on the WHO [1] and WHO [28] for the Malay translation. The contents were identical to the original version and apart from that, a slight sentence enhancement was made. Another independent translator then translated it back into English. In the final phase of this process, the committee members reviewed and assessed the original instrument by comparing the translated and back-translated questionnaires for accuracy.

Thirteen older adults aged 55 years old and above from various educational backgrounds and ethnicities were randomly selected for face validation to validate the cultural appropriateness [25]. Participants were interviewed individually based on the PMT scale. The principal investigator requested the participants to advise of any unclear, confusing statements or scoring methods in the questionnaire [25], such as their thoughts about a particular question that was being asked, what were their thoughts or understanding of a question, phrase or term used when it was read out to them or alternative words to conform among them [25]. A majority of them stated they could follow and understand the questions. However, they suggested including examples of older adults’ daily routines or activities in several statements, especially for response efficacy, self-efficacy, perceived costs and protection motivation items, to enhance their understanding of the questions. Older adults were also asked for the rationales of their selected answers. Their responses were compared between the first and second responses. Their second response was collected within 14 days after the first administration to ensure consistency between both responses [25]. Several subscales and questions were later improved by adding examples, including seven items from protection motivation, self-efficacy, response efficacy, perceived costs, and perceived severity, which were enhanced with sentence adjustment or additional examples.

Phase II: Pre-testing of the PMT Scale

Older adults aged 55 years and above have participated in this cross-sectional survey. The final version of the PMT Scale was tested among 389 participants who had fulfilled the same inclusion criteria. Furthermore, about one hundred fifty participants from the same population were included for a test–retest of the PMT scale. They were requested to answer a similar questionnaire for the second time, ranging from seven to 14 days later [22].

Data collection

Data collection was carried out when the participants were at the clinic, followed by a test–retest assessment performed at their homes. The participants were informed about the purpose of the study and their consent was obtained after fulfilling the inclusion or exclusion criteria. It was conducted through (i) face-to-face interviews and (ii) a self-administered questionnaire for those willing to answer independently. Both interviews and self-administered questionnaires used a similar questionnaire. They were also informed and allowed to withdraw from participating in the study without any penalties.

The principal investigator had briefed assistant investigators on the content of the questionnaires before the data collection started and investigators calibration was also conducted two weeks later to ensure the consistency of the data collected between the assistant investigators. Next, participants were invited to participate in the study and offered to self-administer the questionnaire. Most participants requested or preferred to be interviewed and some were being assisted during the self-administered questionnaires. The assistant investigators selected and interviewed participants whilst monitored by the principal investigator. Participants were also informed that there would be a second visit from the investigators at their home. Therefore, their contact numbers were obtained to schedule a second test–retest assessment visit. The investigators also checked for any incomplete information at the end of the interviews or during the collection of the questionnaire. The interview session lasted about 30 to 35 min.

Ethical considerations

The ethical approval to conduct the study was obtained from the National Medical Research Register, Medical Research and Ethics Committee, Ministry of Health Malaysia (NMRR-21–1680-61095) and the Sarawak State Health Department. All participants were provided with oral and written informed consent. Their identity was kept confidential and not entered into the database.

Data analysis

Data entry and statistical analysis were performed using IBM SPSS version 26 for EFA and SmartPLS version 3.3.7 for PLS-SEM. The Mahalanobis’s Distance was also performed to identify extreme outliers and ten participants were deleted from the analysis with the Mahalanobis multivariate outlier test result with significant observations (p < 0.001) [29]. EFA was tested for the adapted version of the instrument tool from another language to the local language and aimed to explore a new measure in determining the factors within an unfactorised measure [17, 30]. Meanwhile, CFA using PLS-SEM was tested to confirm a pre-existing factor structure that has already been determined [17, 30].

PLS-SEM was considered an appropriate analytical method with models that consisted of many constructs and indicators, in which the study has eight constructs with thirty-five indicators in total [31]. Additionally, this approach was performed to predict and describe the essential target constructs, identify the important driver constructs, then allow one to form a higher-order construct to explain a relationship between a newly formed indicators and constructs [31]. Furthermore, PLS-SEM was considered as soft-modeling due to its high flexibility in adjusting assumptions of data distributions [32]. Hence, Smart PLS is considered as an appropriate software for analysing structural equation modelling (SEM) [33].

Exploratory factor analysis

EFA using principal component analysis (PCA) with oblique rotation (Promax) and Eigenvalues of > 1 was performed to identify the constructs and their dimensions. The value for significance was only limited to 0.40 and above to be accepted. The item loading selection was determined when the i) primary loading was > 0.40, ii) cross-loadings were > 0.2 between the primary and secondary loading, iii) the minimum of two items was necessary to be loaded in a factor, iv) the relevance of items in a factor loading [17].

Confirmatory factor analysis: PLS-SEM

The analysis aims to examine the convergent validity using the average variance extracted (AVE: > 0.50) and composite reliability (CR: > 0.70) [32, 34]. The discriminant validity was analysed using Henseler’s Heterotrait-Monotrait (HTMT: < 0.90) correlation ratio [32, 34]. Higher-order constructs (HOCs) were also performed to merge the self-efficacy and response efficacy for the coping appraisal construct, vulnerability and severity for the threat appraisal construct. The blindfolding assessment was performed to identify the protection motivation’s predictive relevance (Q2: > 0), followed by PLSpredict to assess the model’s out-of-sample predictive power [35].



Source link