Arthritis & Rheumatism, Volume 60,
October 2009 Abstract Supplement

The 2009 ACR/ARHP Annual Scientific Meeting
Philadelphia October 16-21, 2009.

A Tobit Model for Predicting Health Utilities Index Mark 3 From Osteoarthritis Disease Duration: a Population-Based Study

Sayre1,  Eric C., Fines2,  Philippe, Flanagan2,  William M., Cibere3,  Jolanda, Rahman1,  M. Mushfiqur, Aghajanian1,  Jaafar, Kang1,  Weiqun

Arthritis Research Centre of Canada (ARC), Vancouver, BC
Statistics Canada, Ottawa, ON
University of British Columbia and ARC. Vancouver, BC
University of British Columbia, Vancouver
Univ of British Columbia, Vancouver, BC
Toronto Western Res. Institute, Toronto, ON


Osteoarthritis (OA) is usually a slowly progressive disease. With longer disease duration, health-related quality of life (HRQoL) declines, but this has not been well quantified. The purpose of this study is to understand the decline in HRQoL over time due to OA disease duration (OAD).


We used biyearly longitudinal data (1994–2002) from the National Population Health Survey (n=10,920), weighted for the 2002 Canadian population. Self-reported arthritis/OA was extracted from each cycle to determine a discrete (interval) OAD variable, which was categorized into 4 levels: no OA; <4 years; 4–7.9 years; 8+ years. Latent HRQoL (LHRQoL; unbounded above) was measured by the observed Health Utilities Index Mark 3 (HUI3; bounded above at 1). LHRQoL in 2002 was predicted in a cross-sectional Tobit regression model from OAD as of 2002, adjusted for gender, age group (decades) and BMI (<25, 25–29.9, 30+). Interactions were tested between gender and OAD (dropped at alpha=0.05), and gender and BMI (retained). Variance estimates were bootstrap adjusted. Regression coefficients represent effects on LHRQoL; predicted LHRQoL can be truncated above at 1 to convert to HUI3 scale.


Table 1 lists the regression coefficients with 95% confidence intervals (CIs). Compared to no OA, the effect (95% CI) on LHRQoL due to <4 years of OA is -0.078 (-0.101, -0.055), the effect of 4–7.9 years of OA is -0.143 (-0.198, -0.088), and the effect of 8+ years of OA is -0.162 (-0.203, -0.120). Other variables that were associated with lower HUI3 included female, older age (a nearly monotonic decline decade to decade), and heavier BMI (in males, a reduced effect). Except on covariate combinations that predict LHRQoL above 1, coefficients may be interpreted as effects on HUI3.

Table 1. Tobit model for HUI3

VariableCoefficient (95% CI)
OAD (reference=No OA)*
<4 years-0.078 (-0.101, -0.055)
4–7.9 years-0.143 (-0.198, -0.088)
8+ years-0.162 (-0.203, -0.120)
Gender (reference=Female)*
Male0.060 (0.028, 0.091)
Age group (reference=80+)*
12–190.250 (0.206, 0.295)
20–290.246 (0.206, 0.286)
30–290.247 (0.206, 0.288)
40–490.215 (0.173, 0.257)
50–590.170 (0.130, 0.210)
60–690.167 (0.125, 0.209)
70–790.113 (0.070, 0.155)
BMI (reference=30+)*
<250.070 (0.044, 0.097)
25–29.90.053 (0.024, 0.083)
Male & BMI <25-0.065 (-0.100, -0.030)
Male & BMI 25–29.9-0.022 (-0.060, 0.015)
1Scale0.217 (0.208, 0.225)
1Scale estimates the standard deviation of the normal error term.


LHRQoL is affected in a monotonic, nonlinear way by OA disease duration. Adjusted for age, gender and BMI, the average drop in HUI3 between no OA and <4 years of OA is nearly matched by the subsequent drop between <4 years and 4–7.9 years of OA. However, the reduction from there to 8+ years of OA is only a quarter the size of the previous declines. Most of the negative effects of OA on LHRQoL appear to occur in the first 8 years. This may (in part) be due to disease adaptation, or treatments such as eventual surgery.

To cite this abstract, please use the following information:
Sayre, Eric C., Fines, Philippe, Flanagan, William M., Cibere, Jolanda, Rahman, M. Mushfiqur, Aghajanian, Jaafar, et al; A Tobit Model for Predicting Health Utilities Index Mark 3 From Osteoarthritis Disease Duration: a Population-Based Study [abstract]. Arthritis Rheum 2009;60 Suppl 10 :49
DOI: 10.1002/art.25132

Abstract Supplement

Meeting Menu