Survey Context May Influence Estimates of US Vision Problem Prevalence

Matthew W Brault, MPP

Credit: NORC at the University of Chicago

A recent study in JAMA Ophthalmology investigated the impact of survey design on estimates of self-reported vision problems in the American Community Survey (ACS) and Behavioral Risk Factor Surveillance System (BRFSS).¹

Disability prevalence estimates in the BRFSS were higher than similar estimates in the ACS, confusing the true size of the US population with vision issues.^2,3 However, cross-survey comparisons have not directly tested for the cause of these gaps.

In this survey of nearly 3 million individuals, subgroup analysis revealed account sample frame coverage, non-response follow-up, imputation, and proxy reporting had little effect on closing the gap between survey estimates of vision problems in the United States.¹

Instead, the focus on health with the BRFSS may impact the decision of survey participation and how respondents understand their own difficulties with vision.

"People with vision problems may find the BRFSS, a survey about health topics, to be particularly salient and consequently be upwardly biasing prevalence estimates," wrote the investigative team, led by Matthew W Brault, MPP, NORC at the University of Chicago.

Self-reported blindness and vision problems from survey data are crucial for surveillance and public health measures.⁴ However, a range of estimates could challenge the interpretation of the size of the population with vision problems.

The ACS and BRFSS pose the same question on blindness and serious difficulty with vision but have yielded different estimates on self-reported vision problems.^2,3 These differences in estimates are not well understood—however, the national surveys differ on sampling methods, mode of collection, and requirements for participation.¹

Brault and colleagues indicated survey questions and how they are ordered on the questionnaire could influence saliency and prime response. In this analysis, they sought to explain the two surveys’ different prevalence estimates for vision problems, despite the same question wording.

The cross-sectional analysis investigated the 2021 ACS and BRFSS using subgroup analysis and decomposition. Data sets were constructed for civilian adult responders ≥18 years old from 49 states and the District of Columbia. Analysis occurred from August 2022 to October 2023.

Overall, a weighted sample of 2.8 million individuals was included in the analysis. Participants had a median age of 47.7 years and were 51% male.

Brault and colleagues found the estimate of self-reported vision problems prevalence from the BRFSS (4.89%; 95% CI, 4.73 - 5.04) was approximately 1.7 times as high as the estimate from the ACS (2.95%; 95% CI, 2.92 - 2.97) for similar sample populations.

Linear probability models were used to predict vision problems in each survey as a function of common characteristics. Model 1 included covariates for age, sex, race and ethnicity, marital status, education, employment, and health insurance coverage. Model 2 further included indicators for hearing, ambulatory, and cognitive difficulties.

In decomposition analysis, investigators noted if the BRFSS had a sample composition like ACS using Model 1, the prevalence of vision problems would be 4.71% (95% CI, 4.56 - 4.86). Using the Model 2 specifications, aligning with other disability types in the ACS, the BRFSS prevalence would be 3.67% (95% CI, 3.53 - 3.80), closing nearly 63% of the gap between survey estimates.

“Weighting adjustments that address this source of nonresponse bias may help improve estimation in the BRFSS; however additional research is needed to confirm these findings,” Brault and colleagues added.

References

1. _{Brault MW, Wittenborn JS, Rein DB. Behavioral Risk Factor Surveillance System and American Community Survey Estimates of Vision Difficulty Prevalence. JAMA Ophthalmol. Published online June 13, 2024. doi:10.1001/jamaophthalmol.2024.1993}
2. _{Altman BM, Madans J, Weeks JD. An evaluation of the American Community Survey indicators of disability. Disabil Health J. 2017;10(4):485-491. doi:10.1016/j.dhjo.2017.03.002}
3. _{Lauer EA, Houtenville AJ. Estimates of prevalence, demographic characteristics and social factors among people with disabilities in the USA: a cross-survey comparison. BMJ Open. 2018;8(2):e017828. Published 2018 Feb 14. doi:10.1136/bmjopen-2017-017828}
4. _{Lee PP, West SK, Block SS, et al. Surveillance of disparities in vision and eye health in the United States: an expert panel's opinions. Am J Ophthalmol. 2012;154(6 Suppl):S3-S7. doi:10.1016/j.ajo.2012.09.006}