The stakes of this kind of psychometric scrutiny are higher than they might appear. Students training at universities of medical sciences represent a population with a double stake in healthy living: their own well-being during years of intense academic pressure, and their future professional roles as ambassadors of health-promoting behavior in clinics and communities. Non-communicable diseases, driven largely by modifiable lifestyle factors such as nutrition, physical activity, stress management, and interpersonal relationships, remain the leading burden of disease globally. Instruments like the HPLP-II are the measuring sticks by which interventions aimed at these behaviors are evaluated. If the measuring stick bends differently in different populations, comparisons across studies and countries become shakier than the sheer volume of published research would suggest.

The research team, led by Maryam Jalali of the Department of E-Learning in Medical Sciences, together with Zahra Karimian and Faride Ensafdaran, recruited 501 students at Shiraz University of Medical Sciences for a methodological cross-sectional validation study. Their primary task was to test the theory-specified model that underlies the HPLP-II: a correlated six-factor structure in which each of the 52 items loads onto one of six domains covering health responsibility, physical activity, nutrition, spiritual growth, interpersonal relations, and stress management. Crucially, the researchers did not treat the questionnaire’s items as if they were continuous measurements. Because each item is answered on four ordered response categories, they employed weighted least squares mean- and variance-adjusted estimation, known in the structural equation modeling literature as WLSMV, a technique specifically designed for ordinal data and widely regarded as the appropriate choice when Likert-type items are analyzed with confirmatory factor analysis.

The results of the primary confirmatory factor analysis were, in a word, mixed. The scaled fit indices told a reasonably encouraging story: a chi-square value of 3346.73 with 1259 degrees of freedom, a comparative fit index of 0.929, a Tucker-Lewis index of 0.925, a root mean square error of approximation of 0.058, and a standardized root mean square residual of 0.065. By conventional benchmarks, most of these values fall within or near the ranges typically interpreted as acceptable fit. But the robust indices, which are more sensitive to the non-normality and ordinal nature of the response data, were considerably less flattering, with a comparative fit index of 0.800, a Tucker-Lewis index of 0.790, and a root mean square error of approximation of 0.083. The discrepancy between scaled and robust statistics is itself informative, hinting that the distributional assumptions underlying the more optimistic numbers were strained, and that the six-factor model captures only part of the structure hiding in the responses.

Standardized factor loadings across the model ranged from 0.501 to 0.898, indicating that the individual items were generally meaningfully related to their assigned domains, even if the global architecture was under strain. To probe beyond the confirmatory analysis, the team conducted a post-hoc exploratory factor analysis built on a more data-driven foundation. They computed polychoric correlations, which estimate the association between ordinal variables more faithfully than ordinary Pearson correlations, then applied parallel analysis to determine how many factors the data genuinely support, extracted factors through principal axis factoring, and rotated the solution with Oblimin rotation to allow the factors to correlate. Parallel analysis pointed to an eight-factor empirical solution, a notable departure from the six domains specified by the instrument’s original theory.

Yet the story was not simply that the questionnaire’s structure is wrong. When the researchers examined the theory-guided six-factor principal axis solution, they found substantial correspondence with the original domains: most items clustered where the theory said they should. The complication lay at the item level, where several items showed cross-loadings, meaning they related meaningfully to more than one domain simultaneously. Cross-loadings are a classic signal of structural complexity, suggesting that some questionnaire items blur the conceptual boundaries between, for example, spiritual growth and stress management, or between health responsibility and interpersonal relations. In a population of medical students, whose daily lives intertwine study habits, social networks, and health behaviors in distinctive ways, such blurring may reflect genuine features of how health-promoting behavior is organized in this group rather than a defect in the translation itself.

The researchers also examined models with residual covariances, which allow pairs of items to share variance beyond what their factors explain, but they were careful to frame these as sensitivity analyses rather than as a final model. This methodological restraint matters. Adding residual covariances can improve fit statistics almost mechanically, but doing so without theoretical justification risks overfitting the model to the idiosyncrasies of a single sample. By keeping such modifications in the sensitivity column, the authors preserved the interpretability of their primary findings and avoided the common trap of presenting a statistically convenient but scientifically hollow model as the definitive structure of the instrument.

Where the questionnaire did shine was in reliability. Internal consistency was high across all six subscales, with Cronbach’s alpha ranging from 0.836 to 0.917 and model-based omega coefficients ranging from 0.886 to 0.947. For the full 52-item scale, the numbers were extraordinary: an alpha of 0.956 and a model-based omega of 0.979. These figures indicate that the items hang together coherently and that scores derived from the Persian HPLP-II are highly dependable in this population. Reliability, however, is not the same as structural validity. A scale can be internally consistent while its items measure a slightly different configuration of constructs than the theory assumes, which is precisely the pattern the Shiraz data reveal. High alpha values coexisting with an eight-factor exploratory solution and item-level cross-loadings paint a picture of an instrument that measures something real and coherent, but perhaps not exactly the six cleanly separable domains its scoring assumes.

The authors’ conclusion is deliberately measured: the findings provide partial, rather than definitive, support for the theory-specified six-domain structure of the Persian HPLP-II among students at Shiraz University of Medical Sciences. They call for independent cross-validation and further item-level work before stronger conclusions about structural validity or population-specific adaptation are drawn. For the broader research community, the study is a reminder that even the most established instruments deserve periodic, population-specific interrogation, particularly when deployed across languages and cultures. The study was approved by the Ethics Committee of Shiraz University of Medical Sciences, participants provided electronic informed consent, and the work was supported by the university’s Vice-Chancellor for Research and Technology. As health promotion research expands globally, studies like this one demonstrate that the honest answer to whether a questionnaire fits is often neither a triumphant yes nor a dismissive no, but a carefully quantified maybe that points the way toward better measurement.

Subject of Research: Psychometric validation and factor structure of the Persian Health-Promoting Lifestyle Profile II among university students

Article Title: Psychometric properties and factor structure of the persian health-promoting lifestyle profile II (HPLP-II) among students at Shiraz University of Medical Sciences: a validation study

Article References: Psychometric properties and factor structure of the persian health-promoting lifestyle profile II (HPLP-II) among students at Shiraz University of Medical Sciences: a validation study. (n.d.). https://doi.org/10.1186/s40359-026-05744-3

Image Credits: AI Generated

DOI: 10.1186/s40359-026-05744-3

Keywords: HPLP-II, psychometrics, confirmatory factor analysis, exploratory factor analysis, health-promoting lifestyle, university students, Iran, validation study, health promotion, questionnaire reliability, WLSMV estimation, BMC Psychology