OBJECTIVES: A perceived limitation of generic utility measures is lack of ability to capture change relevant to disease-specific areas or interventions.  To test whether core EQ-5D items sufficiently measure variability in patients’ self-reported quality of life scores, we aimed to identify whether the presence of a series of conditions explained residual variability in EQ-5D visual analog scale (VAS) scores, beyond EQ-5D items. METHODS: We utilized generalized linear models (GLM) with a gamma distribution and log link to predict VAS by the 5 EQ-5D items and the presence/absence of 10 conditions (cancer, diabetes, anxiety disorder, hypertension, coronary heart disease, stroke, asthma, COPD/other respiratory, depression, glaucoma), controlling for age, gender, race/ethnicity and number of chronic conditions (i.e., as a proxy for co-morbidities) using the 2000-2003 Medical Expenditure Panel Survey (MEPS) data. Coefficients for disease that were statistically significant (p-value<0.01) and showed minimally important difference (MID: coefficient ≥0.03) served as criteria to support further investigation of condition-specific “bolt-on” items that extend the content of EQ-5D.  RESULTS: Of 24,830 respondents, 45.7% were male, 77.2% were white non-Hispanic and had a mean age of 45.9 years (SD 17.1). Overall mean EQ-VAS was 79.75 at first measurement.  Diabetes, stroke and depression significantly predicted VAS scores alongside the EQ-5D items and demographic characteristics (p<0.001) and met MID criteria.  When concurrently controlling for all other conditions, cancer, CHD and COPD also met criteria. CONCLUSIONS: Findings suggest respondents with diabetes, stroke, and depression, potentially with cancer, CHD and COPD, had significant heterogeneity in their VAS valuation of their own health that was not explained alone by EQ-5D items or demographics.  This study provides one approach to identifying potential chronic conditions where disease-specific “bolt-on” items may be considered for EQ-5D.