Background The relationship between poverty and SLE damage has been observed in several cross-sectional studies, but it remains unclear whether the loss of work due to SLE caused poverty or the reverse. Use of longitudinal data reduces the risk of reverse causation. The aim of the present study was to examine the effects of poverty at one point on subsequent damage, to assess whether the “dose” of poverty affects the extent of damage, and to evaluate the impact on damage of permanently exiting poverty.

Materials and methods Data are from the Lupus Outcomes Study (LOS). LOS participants were recruited from diverse sources in 2003 and followed through 2015 through annual structured surveys. In each year we characterised the respondents’ poverty status based on household income and family size. Beginning in 2007, the survey included a validated measure of disease damage, the Brief Index of Lupus Damage. We used ordinary least squares regression to estimate the impact of 1) poverty in 2009, 2) the “dose” of poverty defined as the percentage of years in poverty between 2003 and 2009, and 3) the effect of permanently leaving poverty by 2009 on change in damage between 2009 and 2015, with and without adjustment for potential confounding variables (demographics, education, SLE duration, characteristics of health care, and health behaviours). To account for attrition and missing variables, multiple imputation was used.

Results In 2009, there were 783 respondents to the LOS annual survey, of whom 94% were female, 35% non-white, and 15% were in poverty. They were 49.8 (SD12.3) years of age and had had SLE for 16.9 (SD8.3) years. BILD damage scores averaged 1.9 (SD2.0, range 0–12). Table 1 shows the effect of poverty in 2007, “dose of poverty” between 2009 and 2015, and exiting poverty on change in damage, with and without adjustment. Those in poverty had greater increases in damage as did those continuously poor vs. poor some years vs. never poor. Exiting poverty was associated with change in damage scores closer to that among those who were never poor with the passage of as little as a year and smaller than those who remained poor. In all analyses, adjustment had minimal effect on results, indicating that the effect of confounding variables was minimal.

Conclusions The present study improves the certainty that poverty is etiologically related to damage and not an artefact of study design.

Acknowledgements Robert Wood Johnson Investigator in Health Policy Award, NIAMS P60 AR-053308