ACCOUNTING FOR RISK-AVERSION IN HEALTH TECHNOLOGY ASSESSMENT: UTILITY ESTIMATION ACROSS MULTIPLE HEALTH INDEXES IN NON-SMALL CELL LUNG CANCER
Author(s)
Rahul Mudumba, MS1, Karen Mulligan, PhD2, Jorge J Nieva, MD2, Darius N. Lakdawalla, PhD2;
1University of Southern California, PhD Candidate, Los Angeles, CA, USA, 2University of Southern California, Los Angeles, CA, USA
1University of Southern California, PhD Candidate, Los Angeles, CA, USA, 2University of Southern California, Los Angeles, CA, USA
OBJECTIVES: Generalized Risk-Adjusted Cost-Effectiveness (GRACE) relaxes restrictive assumptions of traditional cost-effectiveness analysis (CEA) such as risk-neutrality over health, but practical implementation is limited by reliance on visual analog scale (VAS)-based health scores and general-population risk preference estimates. This study aims to directly estimate utility functions for patients with non-small cell lung cancer (NSCLC) and establish a mapping between time trade-off (TTO) and VAS indexes, expanding health measurement and utility function choices for GRACE practitioners.
METHODS: We developed and administered a survey to elicit risk preferences over health from patients with NSCLC recruited via online platforms and oncology clinics. Respondents completed six hypothetical health gamble questions framed using a VAS to elicit certainty equivalents (CEs) reflecting indifference between risky and certain health outcomes. Participants then completed six TTO valuation tasks at matched, within-respondent health levels derived from their CEs. Based on these CEs, utility functions were structurally estimated using nonlinear least squares under three parametric specifications: constant relative risk aversion (CRRA) and one- and two-parameter expo-power (EP). We empirically mapped TTO utilities to VAS-based CEs using linear, quadratic, and piecewise-linear specifications.
RESULTS: The interim sample (data collection is ongoing) comprised 81 respondents, yielding 486 pooled observations across tasks. Across all utility specifications, respondents exhibited nonlinear returns to health. The two-parameter EP model provided the best fit, indicating risk-seeking behavior at lower health levels and a switch to risk-aversion above health levels of 0.53. Piecewise-linear mappings captured nonlinearity at lower and higher health ranges, approximating linear and quadratic models over the range of approximately 0.3-0.8. These mappings enable convenient transformation of published TTO-based health utilities (e.g., EQ-5D) into GRACE-compatible VAS equivalents.
CONCLUSIONS: We find clear departures from risk-neutrality over health among patients with NSCLC and establish a traceable link between TTO and VAS indexes. Together, these results expand GRACE feasibility and support more comprehensive value assessments.
METHODS: We developed and administered a survey to elicit risk preferences over health from patients with NSCLC recruited via online platforms and oncology clinics. Respondents completed six hypothetical health gamble questions framed using a VAS to elicit certainty equivalents (CEs) reflecting indifference between risky and certain health outcomes. Participants then completed six TTO valuation tasks at matched, within-respondent health levels derived from their CEs. Based on these CEs, utility functions were structurally estimated using nonlinear least squares under three parametric specifications: constant relative risk aversion (CRRA) and one- and two-parameter expo-power (EP). We empirically mapped TTO utilities to VAS-based CEs using linear, quadratic, and piecewise-linear specifications.
RESULTS: The interim sample (data collection is ongoing) comprised 81 respondents, yielding 486 pooled observations across tasks. Across all utility specifications, respondents exhibited nonlinear returns to health. The two-parameter EP model provided the best fit, indicating risk-seeking behavior at lower health levels and a switch to risk-aversion above health levels of 0.53. Piecewise-linear mappings captured nonlinearity at lower and higher health ranges, approximating linear and quadratic models over the range of approximately 0.3-0.8. These mappings enable convenient transformation of published TTO-based health utilities (e.g., EQ-5D) into GRACE-compatible VAS equivalents.
CONCLUSIONS: We find clear departures from risk-neutrality over health among patients with NSCLC and establish a traceable link between TTO and VAS indexes. Together, these results expand GRACE feasibility and support more comprehensive value assessments.
Conference/Value in Health Info
2026-05, ISPOR 2026, Philadelphia, PA, USA
Value in Health, Volume 29, Issue S6
Code
PCR139
Topic
Patient-Centered Research
Topic Subcategory
Health State Utilities, Instrument Development, Validation, & Translation
Disease
SDC: Oncology