Title

Convergent validation of self-reported and biometric indicators of employee health

Date of Completion

January 2003

Keywords

Health Sciences, Occupational Health and Safety|Health Sciences, Public Health|Psychology, Clinical

Degree

Ph.D.

Abstract

The validity of self-reported hypertension, hypercholesterolemia and diabetes was examined in a representative sample of Connecticut State employees, utilizing Johnson & Johnson's INSIGHT+ Health Risk Appraisal. Self-reported and biometrically measured health was assessed concomitantly. Sensitivity, specificity, positive predictive values (PPV) and negative predictive values (NPV) of self-reported health-risk were calculated against the most current definitions of risk for each health indicator including: (a) blood pressure ≥140/90 mm Hg and/or taking antihypertensive medication, as recommended by the Third Joint National Committee on Hypertension, (b) TC > 200 and/or HDL < 40mg/dL and/or on hyperlipidemic medication, in accordance with newly established ATP III recommendations, and (c) fasting blood serum levels ≥110 mg/dL and/or on medication, as defined by the American Diabetes Association. Sensitivities for self-reported hypertension, hypercholesterolemia and diabetes were 48%, 35%, and 13%, respectively. Hypertension differed by gender (men: 44%, women: 78%). Diabetes incidence was restricted to men. Specificity was >92% for all risk factors except hypercholesterolemia in men (87%). PPV was lower for hypertension (77%) than for hyperlipidemia (91%) and did not differ by gender. NPV was higher for hypertension (men: 81%, women: 97%) then for hyperlipidemia (men: 29%, women: 57%). Self-report failed to identify 52% of hypertensives, 65% of hyperlipidemics, and 88% of diabetics. Across all conditions, more men than women underreported risk. Among hyperlipidemics, risk awareness was associated with higher TC and LDL levels. This may reflect physicians' use of higher diagnostic thresholds to identify and communicate risk. When analyses were repeated for high-threshold hyperlipidemia (≥240 mg/dL), as expected, awareness increased as prevalence decreased. Overall, results suggest that self-report alone does not adequately reflect health status. For comprehensive disease detection, physiological measurement should be employed. Finally, the Johnson & Johnson INSIGHT+ Health Risk Appraisal was examined to ascertain the construct validity of the computer generated Summary Scores related to Nutrition, Tobacco Use, Activity, Motor Vehicle Safety, and Stress. All score items demonstrated good construct validity and discriminability, with some overlap among 2 of the 3 Nutrition Scores. No strong associations emerged between the INSIGHT+ Score variables and health outcome, measured by 10-year risk of CHD and biometric health indicators. ^