ASSESSING THE VALIDITY AND RELIABILITY OF LATENT CLASS ANALYSES ON A POLICY-RELEVANT CHOICE EXPERIMENT ON HIV MOLECULAR EPIDEMIOLOGY
Author(s)
John F. Bridges, PhD1, Anne L. Schuster, PhD1, Juli Bollinger, MS2, Jeremy Sugarman, MD2, Nicola Campoamor, BA1, Norah L. Crossnohere, PhD1;
1The Ohio State University, Columbus, OH, USA, 2Johns Hopkins University, Baltimore, MD, USA
1The Ohio State University, Columbus, OH, USA, 2Johns Hopkins University, Baltimore, MD, USA
OBJECTIVES: Latent class analysis (LCA) is widely used in choice experiments to characterize preference heterogeneity, yet there is a paucity of research demonstrating validity and reliability of LCA. We developed and demonstrated a novel taxonomy of tests to assess the validity and reliability of LCA on a policy-relevant choice experiment.
METHODS: We developed a taxonomy of tests to assess LCA across four domains: reliability, internal validity, cross-method triangulation, and external validation. Reliability was assessed using concordance between class assignment and alternative scoring approaches and through split-sample replication. Internal validity was evaluated by testing whether class membership was associated with conceptually relevant attitudes while remaining independent of demographic characteristics or HIV status. Cross-method triangulation examined alignment between class membership and themes from open-ended responses. External validity was assessed by testing for theoretically consistent differences in personality traits. These tests were applied to LCA of a national survey using best-worst scaling (BWS) that assessed community priorities across 11 implications of HIV molecular epidemiology (HIV-ME).
RESULTS: LCA identified two classes with distinct priorities: “Use the Data” (63%) and “Protect the Data” (37%) groups. Reliability tests showed strong concordance between LCA results and simple scores (ρ=0.98), 3-point importance scores (ρ=0.78), and split sample replication (ρ=0.99). Internal validity was supported by group differences in HIV-ME related attitudes (e.g., greater comfort with HIV-ME in “Use the Data” group, p<0.01) and no differences by demographic characteristics or HIV status. Cross‑method triangulation demonstrated alignment between LCA results and open‑ended responses. External validity was demonstrated by theoretically consistent differences in personality traits (e.g., higher optimism in “Use the Data” group, p=0.02).
CONCLUSIONS: We demonstrate the validity and reliably of LCA applied to a BWS studying community priorities for HIV-ME. The proposed taxonomy of tests offers a framework that is applicable for evaluating latent classes across choice experiments and policy contexts.
METHODS: We developed a taxonomy of tests to assess LCA across four domains: reliability, internal validity, cross-method triangulation, and external validation. Reliability was assessed using concordance between class assignment and alternative scoring approaches and through split-sample replication. Internal validity was evaluated by testing whether class membership was associated with conceptually relevant attitudes while remaining independent of demographic characteristics or HIV status. Cross-method triangulation examined alignment between class membership and themes from open-ended responses. External validity was assessed by testing for theoretically consistent differences in personality traits. These tests were applied to LCA of a national survey using best-worst scaling (BWS) that assessed community priorities across 11 implications of HIV molecular epidemiology (HIV-ME).
RESULTS: LCA identified two classes with distinct priorities: “Use the Data” (63%) and “Protect the Data” (37%) groups. Reliability tests showed strong concordance between LCA results and simple scores (ρ=0.98), 3-point importance scores (ρ=0.78), and split sample replication (ρ=0.99). Internal validity was supported by group differences in HIV-ME related attitudes (e.g., greater comfort with HIV-ME in “Use the Data” group, p<0.01) and no differences by demographic characteristics or HIV status. Cross‑method triangulation demonstrated alignment between LCA results and open‑ended responses. External validity was demonstrated by theoretically consistent differences in personality traits (e.g., higher optimism in “Use the Data” group, p=0.02).
CONCLUSIONS: We demonstrate the validity and reliably of LCA applied to a BWS studying community priorities for HIV-ME. The proposed taxonomy of tests offers a framework that is applicable for evaluating latent classes across choice experiments and policy contexts.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
PCR171
Topic
Patient-Centered Research
Disease
No Additional Disease & Conditions/Specialized Treatment Areas, SDC: Infectious Disease (non-vaccine), SDC: Reproductive & Sexual Health