Integration of job satisfaction into the Personal Wellbeing Index: analysis of psychometric properties and sociodemographic influences among employees of the Croatian Health Insurance Fund

Between October and November 2018, we recruited via email a convenience sample of 1,067 workers of the Croatian Health Insurance Fund – the central national insurance organisation overseeing the implementation of universal healthcare coverage – who were based in the nation’s capital Zagreb. Following the exclusion of 16 respondents who provided consistently extreme scores (0 or 10) across all domains – a step taken to ensure response validity as recommended in PWI validation studies (16) – the final sample included 1,051 participants. These individuals held a variety of professional roles within the health insurance system, including medical and legal professionals, administration and finances. We believe that data collected in 2018 provide a pre-COVID baseline for understanding wellbeing trends, which are particularly relevant given the subsequent pandemic’s impact on healthcare-related administrative workers. Such a baseline is essential for a comparative longitudinal research, allowing future studies to track changes in subjective wellbeing over time and to assess the long-term effects of the pandemic on healthcare-related administrative workers in transitional economies like Croatia.

The sample comprised 853 women (81.2 %) and 198 men (18.8 %), reflecting the healthcare sector’s gender distribution, though this imbalance may limit the generalisability of gender-based analyses. In terms of education 532 (50.6 %) had a secondary level, 129 (12.3 %) a bachelor’s or higher vocational degree, and 390 (37.1 %) held a master’s degree.

The sample size was sufficient to detect medium effect sizes (η²=0.06) with 80 % power at α=0.05 for the analysis of variance (ANOVA), multivariate analysis of variance (MANOVA), and factor analyses, as reported by earlier PWI-A studies (17, 18).

Subjective wellbeing (SWB) was assessed with the newly developed PWI-J instrument combining the Croatian version of the PWI-A, standardised by Kaliterna Lipovčan and Prizmić-Larsen (18) and validated by Kaliterna Lipovčan et al. (5), and the newly added job satisfaction domain to the original seven domains: standard of living, health, achievements, relationships, safety, community connectedness, and future security. The eighth domain, job satisfaction, was tailored to assess psychosocial factors relevant to the quality of life of public sector workers. Each domain was rated on an 11-point Likert scale (0 – completely dissatisfied to 10 – completely satisfied) converted to a 0–100 scale.

The survey was completely anonymous and confidential and was approved by their employer, the Croatian Health Insurance Fund in accordance with Croatian regulations governing research in public-sector organisations that lack a dedicated ethics committee. All procedures were performed in full compliance with the Declaration of Helsinki.

Participants were informed that there were no correct or incorrect answers, and data were analysed at an aggregate level, preventing individual identification.

Normality of distribution was assessed as the first step in evaluating assumptions for parametric testing. Given the large sample size (N=1,051), Kolmogorov-Smirnov tests were significant for all items (P<0.001), which was attributed to high statistical power rather than meaningful deviations from normality (19). Normality was therefore judged primarily using skewness and kurtosis statistics. All values (|skewness|≤1.33; |kurtosis|≤1.64) fell well within the recommended thresholds (<2 and <7, respectively) for robust parametric analyses in large samples (20). Based on this evidence, we used one-way ANOVA (with education level, gender, or age group as the independent variable, depending on the hypothesis tested) for between-group comparisons of PWI-J domain and composite scores. Post-hoc comparisons were done using the Tukey’s honestly significant difference (HSD) test to control for the family-wise error rate.

The internal consistency of the PWI-J was high, with Cronbach’s α=0.896 indicating high reliability. Corrected single domain to total correlations were satisfactory, ranging from 0.492 (relationships) to 0.729 (safety). Although the correlation for the relationships domain was relatively low, it was above the conventional threshold of >0.40. The reliability analysis further supported the inclusion of the job satisfaction domain, as its deletion would lower the scale’s α to 0.848 (Table 3).

Pearson correlation analysis shows that all inter-domain correlations, as well as the correlations of each domain with the total PWI-J score are positive and statistically significant (P<0.001) (Table 4). The strongest correlations with the PWI-J total score are those for safety (r=0.811), achievements (r=0.777), and future security (r=0.774). The highest inter-domain correlations are those between safety and future security (r=0.685) and between community and relationships (r=0.578), whereas the lowest are between relationships and standard of living (r=0.217).

Job satisfaction and the overall PWI-J score show a strong Spearman’s correlation (ρ=0.633, P<0.001; applied due to non-normal data distribution). Given the cross-sectional design, causal direction cannot be inferred. However, this association is consistent with the spillover hypothesis, which suggests that experiences in one central life domain, such as work, can influence perceptions of overall life satisfaction (7, 8, 21).

We tested our third hypothesis of the association between sociodemographic factors and PWI-J domains with a series of ANOVAs (Table 5). Education level shows significant effects on the standard of living, health, achievements, and future security, with the post-hoc Tukey’s tests indicating higher scores among higher-educated participants. Gender showed significant effects on achievements, relationships, safety, community, and job satisfaction, with women reporting higher satisfaction. Age showed significant effects on health, relationships, safety, community, and job satisfaction, with younger participants (19–34-year-olds) generally reporting higher satisfaction than older groups.

The data proved suitable for the exploratory factor analysis (EFA), as indicated by the Kaiser-Meyer-Olkin measure of sampling adequacy of 0.861 and the significant Bartlett’s test of sphericity [χ²(28)=3593.39, P<0.001]. We applied principal component analysis with Promax rotation, given the expected correlation of factors, which yielded two components with eigenvalues exceeding 1. These account for 65 % of the total variance (Table 6).

The pattern matrix (Table 7) highlights a clear two-factor structure: Factor 1, explaining 52.2 % of the variance, shows strong loadings from the standard of living (0.945), achievements (0.625), safety (0.747), future security (0.844), and job satisfaction (0.664) domains. These domains capture material and practical wellbeing, reflecting the aspects tied to economic stability and security.

Factor 2, accounting for 12.8 % of the variance, shows high loadings from the health (0.523), relationships (0.965), and community (0.799) domains and, being focused on interpersonal and health-related dimensions, represents socioemotional wellbeing.

In line with the EFA, the PWI-J domains were grouped into two composite scores: material security (averaging standard of living, achievements, safety, future security, and job satisfaction) and socioemotional wellbeing (averaging health, relationships, and community) to establish overall wellbeing patterns. The effects of gender, education level, and age on these two dimensions were tested with the MANOVA.

Box’s M test indicated a violation of the homogeneity of covariance assumption (M=126.34, P<0.001), but given the large sample size (N=1,051), the results are considered robust (22). Levene’s test indicated heterogeneity of variance for material security (P=0.020) but not for socioemotional wellbeing (P=0.058).

The effects of all three sociodemographic factors were significant, as follows: gender [Wilks’ Λ=0.991, F(2, 1026)=4.42, P=0.012, partial η²=0.009], education level [Wilks’ Λ=0.975, F(4, 2052)=6.40, P<0.001, partial η²=0.012], and age [Wilks’ Λ=0.969, F(6, 2052)=5.40, P<0.001, partial η²=0.015]. All effect sizes were small, and we observed no interaction effects (all P>0.05).

Follow-up univariate analyses of variance indicated that gender had an effect on both material security [F(1, 1027)=7.14, P=0.008, partial η²=0.007] and socioemotional wellbeing [F(1, 1027)=6.96, P=0.008, partial η²=0.007], as women reported higher scores on both: (59.25 vs 56.89 for material security and 75.63 vs 74.06 for socioemotional wellbeing).

Education level affected material security [F(2, 1027)=12.40, P<0.001, partial η²=0.024] and participants with graduate or higher education scored higher than those with secondary education (63.95 vs 55.42, P<0.001) or undergraduate education (57.26, P=0.003).

Age had an effect on socioemotional wellbeing [F(3, 1027)=6.13, P<0.001, partial η²=0.018], and the youngest age group of 19–34 years reported higher scores than the 45–54 group (80.14 vs 73.39, P<0.001) or 55 and over (72.72, P<0.001).

Compared with the Australian wellbeing triage ranges (4), the Croatian sample showed a markedly smaller proportion of participants in the “Well” category (70–100 points: normal homeostatic control) and a substantially higher proportion in the “UnderWell” (50–69 points: challenged homeostatic control) and “NoWell” (0–49 points: defeated homeostatic control) categories (Table 8). The higher prevalence of “UnderWell” and especially of “NoWell” scores indicates significant psychosocial challenges and homeostatic defeat in this population.

Our PWI-J shows excellent internal consistency (α=0.896), well above the conventional threshold, and falls in line with reliability estimates reported internationally (4, 16, 23, 24). All domain-to-total-score correlations are satisfactory, supporting the inclusion of job satisfaction, particularly in occupational contexts where employment permeates multiple life domains (25).

The two dimensions accounting for 65 % of the variance, namely, material-practical and socioemotional wellbeing, extend the theoretical unidimensional model of the PWI-A often supported in homogeneous Western samples like Australia (4, 26). Cross-cultural studies, however, frequently report more complex structures, particularly in transitional contexts, where economic and security concerns may form a distinct dimension (23, 27). The moderate correlation between the two factors supports their interpretation as related yet distinct components of a higher-order SWB construct.

The overall mean score for the first seven domains (job satisfaction excluded) of ~65 places our sample in the “UnderWell” group (50–69 points), indicating a state of homeostatic challenge where external pressures have lowered SWB below its normal set-point (70–100 points) but has not resulted in full homeostatic defeat (“NoWell”: 0–49 points) (4). The notably low satisfaction scores with the standard of living and future security likely act as primary “homeostatic attackers”, which is consistent with the reports on populations experiencing economic strain or transitional stressors (60–70 points) (18). In contrast, the relatively high scores in the relationships and community domains may function as key internal and external buffers, helping to mitigate these negative pressures and maintain a degree of stability (29, 30). Furthermore, as stated in the PWI-A manual, providing domain-relevant services to individuals in the “UnderWell” ranges – especially to those scoring below 50 points – has a high probability of raising SWB (4, p. 3–9).

Inter-domain correlations were all positive and significant, with the strongest associations between safety and future security and between community and relationships, whereas the weakest correlation was between relationships and the standard of living, suggesting that social and economic domains operate somewhat independently in our sample.

The spillover effect between general wellbeing (PWI-A) and job satisfaction was also confirmed, supporting the notion that work experiences significantly influence but do not fully determine overall personal wellbeing (7, 9, 21). According to Diener et al. (1), SWB encompasses satisfaction with key domains such as work, which has been confirmed by a meta-analysis on the bidirectional relationship between job satisfaction and SWB (8). Iqbal et al. (31) emphasise that work-life balance acts as a mediator, with intrinsic motivation as a moderator. In the context of contemporary challenges such as hybrid work and automation (32), incorporating job satisfaction into the PWI-A enhances its applicability to the working population, as evidenced by our results and supported by Kuykendall and Tay (33).

Compared to Australian triage ranges (4), the Croatian sample shows a substantially smaller proportion of participants in the “Well” range. This divergence underscores the influence of macroeconomic and socio-political context on population wellbeing and cautions against an uncritical application of norms derived from stable, high-income countries (24).

Sociodemographic factors exerted statistically significant yet modest effects, suggesting that they can explain a limited portion of variance in SWB compared to other contextual factors.

Higher education was associated with greater satisfaction in material and security-related domains, reinforcing its role as a resource buffer (34).

Women reported higher satisfaction than men across several domains, particularly those related to social connectedness and safety. Although this gender pattern is frequently observed in Western European samples (35), studies in Croatia and other post-Yugoslav countries have yielded mixed or context-specific findings. Our finding of higher satisfaction of women in social domains may reflect the enduring influence of collectivist cultural norms and stronger family/community ties in transitional societies recovering from the conflicts in the 1990s. These are the factors that appear to buffer women more effectively against economic stressors than men (6, 23, 36).

Younger participants generally reported higher socioemotional wellbeing, possibly reflecting cohort effects or life-stage advantages in social engagement and health (37, 38).

The cross-sectional design precludes causal inferences, as it captures wellbeing at a single point in time without tracking changes over the years. This is particularly relevant, given that the data were collected in 2018, prior to the COVID-19 pandemic, which has since profoundly impacted workers’ stress levels and subjective wellbeing globally. Longitudinal studies would therefore be essential to examine how these patterns evolve.

Another design limitation is that the measures are self-reported and may introduce biases, such as social desirability or recall inaccuracies, which could inflate correlations between domains like job satisfaction and overall SWB.

The third is convenience sampling, as it may have led to selection bias. This resulted in a sample skewed toward women (81.2 %), residents of the Croatian capital, and public sector, limiting generalisability to other regions, private sector workers, or more diverse demographic groups in Croatia.

Finally, although sociodemographic factors were analysed, we did not control for additional confounders like work tenure or shift patterns, which may have influenced the observed effects.

Despite these limitations, the study’s strengths in psychometric validation and contextual relevance provide a solid foundation for future research in transitional economies.