Study design and data collection. Researchers conducted a cross-sectional survey from April 2023 to June 2023, involving 730 undergraduate students from the College of Medicine and Allied Medical Sciences at Federal University Dutse, Nigeria (FUD). Convenience sampling is recognized as a more stringent approach for participant selection in research, but it may be appropriate for exploratory or preliminary investigations aimed at generating hypotheses or insights (Etikan et al., 2016, Hills et al., 2015). This method allows researchers to swiftly gather data, laying the groundwork for subsequent research activities (Etikan et al., 2016). Convenience sampling applied for participant selection in this study, is a straightforward and easily accessible approach that accurately represents the population from which the sample was drawn (Winton and Sabol, 2022), thereby establishing a strong basis for generalizability.
Students meeting the study's inclusion criteria received the Google Form link for data collection due to its practicality, versatility, and cost-effectiveness. Researchers and surveyors commonly use Google Forms to effectively mitigate response bias (Nayak and Narayan, 2019). The inclusion criteria included enrolment in Federal University Dutse (FUD) College of Medicine and Allied Medical Sciences; undergraduate status ranging from first to final year; registered students during the data collection period; and expressed consent to participate. We specifically selected participants from the College of Medicine and Allied Medical Sciences because they are presumed to be familiar with and comprehend the underlying concepts and constructs being assessed. Participants who have a certain level of understanding of a specific scale contribute to improving the scale's construct validity, ensuring that it accurately measures the intended concepts (Jones et al., 2019).
Ethical approval. The Human Research Ethics Committee, Ministry of Health, Jigawa State, Nigeria [JGHREC/2023/151], and Universiti Sains Malaysia’s Human Research Ethics Committee [USM/JEPeM/22110695], granted ethical approval for the study. The participants were informed about the research aim and methods before signing the informed consent form. Written informed consent was obtained from each participant. The study conforms to the principles outlined in the Declaration of Helsinki.
Scales' item generation. Based on the Meikirch model (Bircher, 2020, Bircher and Kuruvilla, 2014), the present study creates the IP-Q, which posits two hypothetical constructs: biologically given potential and personally acquired potential. Expert input was sought from professionals in public health, psychometrics, health psychology, and questionnaire validation to refine the generated items. Additionally, in-depth interviews were conducted with twelve undergraduate students to gather further insights. Initially comprising 14 items, with six under biologically given potential and eight under personally acquired potential, the IP-Q was expanded to 56 items (14x4) by providing four alternative options for each original item. Subsequently, experts in pertinent fields evaluated these items to identify the most suitable 14, selecting one item from each set of four. The authors formulated the final set of 14 items through consensus. These items were evaluated using a four-point rating scale, ranging from 1 (none) to 4 (severe) for biologically given potential construct and from 1 (not at all) to 4 (very often) for personally acquired potential construct.
For the biologically given potential, the items were generated based on the health belief model (HBM) construct of perceived severity (Champion and Skinner, 2008, Green et al., 2020). According to this model (Champion and Skinner, 2008, Green et al., 2020, Sulat et al., 2018), evaluations of potential medical and clinical outcomes (such as mortality, impairment, and discomfort), as well as potential social repercussions (such as effects on employment, family dynamics, and social connections), are determinants that shape perceptions of the severity of illness or the consequences of not treating it. Perceived threat encompasses both susceptibility and severity (Ritchie et al., 2021, Abraham and Sheeran, 2015). Further insights were obtained from longitudinal studies, which indicated that self-perceived health serves as a predictor for the onset of chronic diseases (Allen et al., 2016, Froehlich-Grobe et al., 2016, Shields and Shooshtari, 2001), recovery from illnesses (Latham and Peek, 2013), and deterioration in functional abilities (Assari et al., 2016, Barry et al., 2020, Rnic et al., 2023).
In terms of personally acquired potential, the items were formulated based on Antonovsky's salutogenic model (Antonovsky, 1987). This model emphasizes coping abilities rather than stressors (Antonovsky, 1987). This theory posits that the sense of coherence elucidates effective coping strategies for managing stress. It denotes a universal outlook, signifying individuals' capability to believe that the stimuli arising from both internal and external environments in their lives are organized, predictable, and understandable (comprehensibility); that resources are accessible to address the demands posed by these stimuli (manageability); and that these demands are worthwhile and deserving of investment and engagement (meaningfulness) (Dantas, 2007, Eriksson and Lindström, 2005).
Content validity. After item generation, six experts in health psychology, psychometrics, public health, and questionnaire development determined the content validity index (CVI). Through the use of a Google Form link, each expert assessed each item's relevance to their specific construct. Next, we calculated the item content validity index (I-CVI) and scale content validity index (S-CVI) based on previously established standards (Polit and Beck, 2006, Lynn, 1986, DeVon et al., 2007, Polit et al., 2007) to evaluate the CVI. For every item, the relevance rating was transformed to 0 (the item is not relevant or is somewhat relevant) or 1 (the item is quite relevant or highly relevant). We estimated the I-CVIs by calculating the propositions of the experts, giving the items a relevance rating of 1, and the I-CVIs for each construct were averaged to determine the S-CVI/Ave In addition, we calculated the S-CVI/UA by averaging the proportion of scale items that achieved a 1 for relevance from all experts (Polit et al., 2007). The I-CVIs for all 14 items ranged from 0.83 to 1. The S-CVIs/Ave were 1 for biologically given potential and 0.98 for personally acquired potential. The S-CVIs/UA were 1 for biologically given potential and 0.88 for personally acquired potential. As a result, these CVI values satisfied the required cut-up value of 0.83 (for six experts) (Polit et al., 2007).
Face validity. We also determined the face validity index (FVI) to assess the clarity and comprehension of the items. Ten undergraduate students from the targeted population evaluated each item for clarity and comprehension via a Google Form link. We estimated the item face validity index (I-FVI) and scale face validity index (S-FVI) based on the standard recommendations (Yusoff, 2019, Marzuki et al., 2018). Each item was rated as either 1 (clear and understandable, or very clear and understandable) or 0 (not clear and understandable, or somewhat clear and understandable) based on relevance. We calculated the I-FVI by determining the proportion of students who assigned a relevance rating of 1 to the items. We computed the S-FVIs/Ave by averaging the I-FVIs for each construct on the IP-Q. Finally, we estimated the S-FVIs/UA by averaging the proportion of scale items that all students rated as 1 for relevance (Marzuki et al., 2018). The I-FVIs for all 14 items are equal to 1. For the constructs, the S-FVIs/Ave and S-FVIs/UA were each rated as 1. As a result, these FVI values satisfied the required cut-off value of 0.83 (for 10 raters) (Yusoff, 2019).
Sample size estimation. The recommended minimum sample size for exploratory factor analysis (EFA) falls within the range of 100 to 250 individuals (Kyriazos, 2018). To account for potential missing values, we set the adjusted sample size at 286 by adding 30%. As a result, the EFA sample size is rounded to 300. Additionally, according to Tabachnick et al. (2013), a sample size of 300 is considered reasonable for EFA. Studies involving seven or fewer constructs in confirmatory factor analysis (CFA) should aim for a minimum sample size of 300 (Black and Babin, 2019). Following this guideline, we maintained a sample size of 300 for the CFA phase. After incorporating a 30% correction for missing values, the final corrected sample size for CFA was 430.
Statistical analysis. The data underwent pre-screening to identify erroneous data entries and missing values. Subsequently, EFA was carried out using Statistical Product and Service Solutions (SPSS) version 27 (IBM, Armonk, NY, USA). Following this, CFA was conducted usingMplus 8 to validate the EFA model. The researchers employed the MLR estimator during the CFA due to its robustness to non-normal data distributions (Muthén and Muthén, 1998).
The EFA sample comprised 300 participants. To identify the principal contributing factors, the 14 items on the IP-Q underwent testing using principal axis factoring with Promax rotation selected. Researchers often select Promax rotation for EFA when they anticipate or have a theoretical rationale for correlated factors, as well as when seeking a more pragmatic and interpretable factor structure (Tabachnick et al., 2013). Furthermore, Promax rotation aids in better aligning the hypothesized model with established theories or expectations (Tabachnick et al., 2013). Upon identifying factors with eigenvalues surpassing one, those displaying factor loadings exceeding 0.40 were considered satisfactory and retained for subsequent CFA (DeVon et al., 2007, Brown, 2015). DeVon et al. (2007) also suggest that a Cronbach's alpha value of 0.60 or higher indicates acceptable reliability for each factor.
CFA was applied to validate the EFA model with a sample size of 430 respondents. A standardized factor loading equal to or greater than 0.40 served as the criterion for retaining or removing an item in the current study (Kline, 2023, Fornell and Larcker, 1981). According to Hair (2009), for sample sizes exceeding 250 and consisting of 12 items or more, the acceptable fit indices included: root mean square error of approximation (RMSEA) below 0.07; standardized root mean square residual (SRMR) below 0.08; and comparative fit index (CFI) or Tucker and Lewis index (TLI) exceeding 0.94. After considering adequate theoretical support, the model was revised based on the CFA modification index to enhance the model fit indices.
By computing composite reliability (CR) and average variance extracted (AVE), we were able to delve deeper into assessing the convergent validity of the IP-Q. Acceptable threshold values for CR and AVE were set at 0.70 and 0.50, respectively, or higher (DeVon et al., 2007, Brown, 2015, Byrne, 2013). To evaluate discriminant validity, which gauges the extent to which one factor differs from another, correlations between factors were examined (Brown, 2015). For discriminant validity, a correlation coefficient of 0.85 or less between two factors is considered adequate (Brown, 2015). According to Fornell and Larcker (1981), confirmation of discriminant validity necessitates that the AVE of constructs exceeds the squared correlation coefficient, representing shared variance among variables. To assess test-retest reliability, a subset of 70 respondents completed the IP-Q twice within a seven-day period. A value exceeding 0.70 for the intra-class correlation coefficient (ICC) indicates satisfactory stability (Koo and Li, 2016).