Measurement Equivalence of Diabetes Self-Management, Distress, and Quality-of-Life Measures in Adults with Type 2 Diabetes in Vietnam

Background: Patient-reported outcome comparisons require measurement equivalence, which is seldom tested in low- and middle-income country (LMIC) diabetes research. We examined equivalence of the Diabetes Self-Management Instrument-35 (DSMI-35), Diabetes Distress Scale-17 (DDS-17), and Asian Diabetes Quality of Life (AsianDQOL) scale across sex, fasting-glucose stratum, and educational attainment in Vietnamese adults with type 2 diabetes. Methods: We conducted a secondary analysis of 374 adults (female 152, male 222; lower-FBG < 154 mg/dL, n = 212; higher-FBG n = 162; secondary-or-lower n = 202; tertiary-or-higher n = 172). Multi-group CFA (lavaan) tested configural, metric, and scalar equivalence of a parcel-level three-factor model (parcel-level equivalence does not imply item-level equivalence). Path equality was evaluated with scaled Satorra–Bentler likelihood-ratio tests; indirect effects were bootstrapped (n = 5000). Results: Scalar-equivalence change-index criteria (ΔCFI ≤ 0.010; ΔRMSEA ≤ 0.015) were met for all groupings; however, for fasting glucose the configural baseline fit was weak (RMSEA 0.117–0.119), so fasting-glucose equivalence is reported only as provisional and is not interpreted at the level of the sex and education findings. McDonald’s ω was ≥ 0.959 in every subgroup. Structural paths did not differ by sex (Δχ2(3) = 1.18, p = 0.758; not powered for equivalence) but differed by education (Δχ2(3) = 71.16, p < 0.001), with the cross-sectional association structure differing by education (distress-channelled in tertiary-or-higher and partly direct in secondary-or-lower participants); because the data are cross-sectional, these are differences in association structure, not established mediation. The fasting-glucose structural comparison was not interpretable because the lower-FBG subgroup (FBG < 154 mg/dL, n = 212) had a non-positive-definite latent covariance matrix. Conclusions: Scalar equivalence criteria were met for sex and education and only preliminarily supported for fasting-glucose stratum, where elevated configural RMSEA (0.119) cautions against firm interpretation. The self-management → distress → quality-of-life pathway showed no detected sex difference but differed by educational attainment. Measurement equivalence testing, including configural-fit assessment, should be routine in LMIC patient-reported outcome validation.