Disaster preparedness belief in older adults: a cross-sectional psychometric evaluation of the general disaster preparedness belief scale, its refinement needs, and sociodemographic correlates

Older adults are among the groups most vulnerable to disasters, yet the General Disaster Preparedness Belief (GDPB) scale has not been psychometrically evaluated in this group. We evaluated the construct validity, reliability, and sex-based measurement invariance of the 31-item GDPB scale in adults aged 60 and over, identified components requiring refinement, and examined sociodemographic correlates. Four hundred three community-dwelling adults aged 60 and over in Kepez, Antalya, Türkiye, completed the 31-item GDPB scale in January–February 2026 in this cross-sectional study. Confirmatory factor analysis (WLSMV) compared correlated six-factor, single-factor, higher-order, and bifactor models. Convergent and discriminant validity (AVE, composite reliability, Fornell–Larcker, HTMT), reliability, and ordinal sex invariance were examined, and sensitivity analyses excluded misfitting subscales. Correlates of the total score were analysed with multiple linear regression (HC3 robust standard errors; education modelled categorically). The six-factor model outperformed the single-factor model (scaled Δχ² = 942.73, Δdf = 15, p < .001); a higher-order model was retained, with CFI and TLI below 0.90 (CFI = 0.878, TLI = 0.867, RMSEA = 0.064, SRMR = 0.091). Misfit concentrated in perceived severity (second-order loading − 0.23). A 28-item version excluding severity met the CFI, TLI, and RMSEA thresholds (CFI = 0.910, TLI = 0.901, RMSEA = 0.060; SRMR = 0.081), with higher reliability (alpha = 0.865, omega = 0.903), hierarchical omega = 0.83, and r = .98 with the full total. After merging sparse response categories, threshold and loading invariance across sex held for the 28-item version but not for the severity items. Education was the strongest correlate, following a threshold rather than a linear pattern (nonlinearity p < .001): adjusted means were 65–68 up to secondary education versus 85–106 for high school and above (R² = 0.538, adjusted 0.524), replicating with the 28-item total. In older adults, the general GDPB dimension received partial support, but subscale-level validity was limited, and the perceived severity items, which undermined the 31-item total, may require age-sensitive redevelopment. The 28-item total is proposed as a provisional scoring form pending independent confirmation, and older adults with primary education or less, who showed uniformly low belief, may warrant priority attention.