Trends in Visual Acuity Impairment in US Adults: The 1986-1995 National Health Interview Survey

Objective  To assess 10-year trends in reported visual impairment.

Methods  The National Health Interview Survey is a continuous multistage areaprobability survey of the US civilian noninstitutionalized population livingat addressed dwellings. Adults within randomly selected households were administereda chronic conditions list that included questions about visual impairment.Proxy information on these conditions was obtained when household memberswere unavailable for interview. Complete data were available on 132 860adults 18 years or older in survey years 1986 to 1995. Prevalence rates wereadjusted for age and sample survey design.

Results  Annual age-adjusted rates of some visual impairment ranged from 3.6%to 4.6%. Rates of severe bilateral visual impairment ranged from 0.2% to 0.4%.There was some evidence for increasing rates of visual impairment among youngeradults 18 to 39 years of age (annual increase, 0.03%; P = .03). However, there were no significant changes in reported visualimpairment rates in older adults stratified into 10-year age groups.

Conclusions  Data from the National Health Interview Survey provide no evidence thatreported visual impairment rates are declining in the US noninstitutionalizedpopulation from 1986 to 1995. Additional treatment advances, greater use ofexisting treatments, including correcting refractive errors, and further reductionsin risk factors for disabling eye diseases may be necessary before population-levelreductions in visual impairment rates can be achieved.

There has been a vigorous debate in the literature about the possibilitythat morbidity rates in older populations will be reduced because of lifestylemodifications and advances in the treatment of chronic disease.^1,2 Evidencefor this "compression of morbidity" has emerged in recent years with respectto reduced rates of disability in older adults (eg, activity limitation).^3-6 It isunknown, however, if such a compression of morbidity extends to sensory impairmentssuch as visual impairment, despite the notable improvements in the treatmentof major disabling eye diseases in the last 2 decades.^7-9 Ifsuch a trend exists, visual impairment rates should be declining in adultsin general and in older persons in particular.

Clinical data to determine if visual impairment rates are decliningare not available because there are no ongoing clinical annual studies ofvisual impairment in the United States. However, the National Center for HealthStatistics has included questions about visual impairment in its annual NationalHealth Interview Survey (NHIS). Survey modifications after 1995 limited datacomparability with earlier survey years. Therefore, NHIS survey years 1986to 1995 represent the best, most recently available data on 10-year trendsin self- or proxy-reported visual impairment in the United States.

The NHIS is an annual cross-sectional, multipurpose, and multistageprobability area survey of the US civilian noninstitutionalized populationliving at addressed dwellings.¹⁰ Each year,approximately 50 000 households are selected to participate in the NHIS.Once selected, attempts are made to interview all adults within each household.Proxy information is collected on any adult family member not available forthe interview. Although the sampling unit for the NHIS is the household, surveysample weights are calculated for each member of the household to allow forperson-level analyses.

Each year, a subsample of 1 in 6 sampled households was administereda chronic condition list that included 2 visual impairment questions. Participantswere asked to indicate if they or any of their family members had blindnessin 1 or both eyes and any other trouble seeing with 1 or both eyes even whenwearing glasses. In most cases (62%), the participants themselves answeredall the questions, and for the adults not available, responses to the visualimpairment questions were obtained from an adult informant within the household.For simplicity, in the present study, self- or proxy-reported data are referredto as "reported." Participants were also asked to name conditions and impairmentsthat were related to the following: (1) activity limitation in the previous2 weeks and the previous 12 months, (2) a health care visit in the previous2 weeks, and (3) hospital stays in the previous 12 months. A series of standardizedquestions was used to detail the name, characteristics, cause, onset, andeffects of each reported condition and impairment.¹¹ Trainedmedical coders used this information and the InternationalClassification of Diseases, Ninth Revision (ICD-9) to generate a codefor each condition.^12,13 Speciallydesigned impairment categories were used by the National Center for HealthStatistics to classify visual impairment as follows: (1) blind in both eyes,(2) visual impairment in both eyes, (3) blind in 1 eye and visually impairedin the other eye, and (4) blind or visually impaired in 1 eye only (othereye, good vision or not mentioned). Adults classified as blind in both eyeswere considered to have "severe bilateral visual impairment." Consistent withthe standard reporting approach used by the National Center for Health Statistics,participants falling into categories 2 through 4 were grouped into a singlecategory of "some visual impairment." For the present study, complete visualimpairment data were available for 132 860 participants of the 1986-1995NHIS, ages 18 years or older.

Because of the complex sample survey design, all analyses were completedusing the Software for the Statistical Analysis of Correlated Data (SUDAAN)¹⁴ package to take into account sample weights and designeffects. To allow for comparisons of visual impairment over time, we adjustedrates for age using the direct adjustment method and the 2000 population distributionas the standard.¹⁵ To determine if there wereany statistically significant changes in visual impairment during the 10 years,a weighted linear regression model was fitted to the annual adjusted rates.The weight used for each rate was the inverse of its variance. Linear- andhigher-order polynomial models were examined.

Figure 1 shows the surveydesign and age-adjusted rates of some visual impairment and severe bilateralvisual impairment from 1986 to 1995. Age-adjusted rates of some visual impairmentranged from 3.6% to 4.6%, while rates of severe bilateral visual impairmentranged from 0.2% to 0.4%. There were no statistically significant upward ordownward trends in reported impairment rates over time. The estimates of theslopes of the linear regression models for some visual impairment and forsevere bilateral visual impairment were 0 (P = .94and P = .75, respectively).

Figure 2 shows the rates ofany visual impairment by age group. Prevalence rates were generally higherin older age groups, with the highest rates found in those 80 years or older.There were no statistically significant changes over time in reported impairmentrates for any of the age groups, except for a slight increase in rates foryoung adults 18 to 39 years of age (P = .03). Theannual increase in this age group was 0.03%.

Results of the present study indicate that rates of reported severebilateral visual impairment and milder visual impairment remained stable inthe noninstitutionalized US population from 1986 to 1995. Although there wassome evidence of a slight increase in visual impairment among young adults18 to 39 years of age, rates of overall impairment and impairment within older10-year age groups remained stable. These findings are in contrast to thosebased on analysis of data from the 1984 and 1993 Surveys of Income and ProgramParticipation (SIPP) conducted by the US Bureau of the Census.¹⁶ TheSIPP participants (or their designated household proxy respondents) were askedto indicate if they had any difficulty seeing the words and letters in ordinarynewspaper print, even when wearing glasses or contact lenses (if usually worn).Prevalence rates of reported near visual impairment declined from 15.3% in1984 to 11.6% in 1993. A statistically significant decline in rates remainedin younger and older adults after controlling for sociodemographic status.The reasons for the discrepancy between the findings of the SIPP study andthose of the NHIS study reported herein are unclear. A single visual impairmentquestion was administered in the SIPP, while in the NHIS, a series of standardizedquestions was used to detail the characteristics, cause, onset, and effectsfor each reported episode of visual impairment. This information was thenused to code the severity of visual impairment in a standardized manner.¹² The visual impairment question used in the SIPP waslimited to difficulties in reading newsprint only, while the impairment questionsused in the NHIS asked about visual impairment in general, without referenceto the ability to see near or distant objects. Questions assessing functionalaspects of vision, including the ability to read newsprint, may be more sensitiveindicators of visual impairment than more general questions such as thoseused in the NHIS.

There are several advantages to the use of the NHIS to examine trendsin reported visual impairment. The NHIS includes annual large samples andis designed to be representative of the US population; only institutionalizedand military groups have been omitted from direct sampling. Survey responserates have been excellent (ie, 95%-98%). However, there is likely some misclassificationof reported visual impairment because of the self- or proxy-reported natureof ascertainment of chronic conditions in the NHIS. Validation studies^17,18 conducted by the National Centerfor Health Statistics suggest that proxy reports lead to slightly lower prevalenceestimates of chronic conditions compared with reports obtained directly fromrespondents. Furthermore, we are not aware of any published reports validatingthe ability of study participants to correctly identify the presence and degreeof visual impairment in family members. To address this potential limitationof our study, we repeated the analyses with only the 62% of NHIS participantswho were interviewed directly. The rates of visual impairment were largelyunchanged. For example, among those directly interviewed, the rate of somevisual impairment, averaged over the 10-year period, was 4.2%, while the correspondingrate for proxy and nonproxy participants combined was 4.1%.

Smoking, diabetes mellitus, and, perhaps, uncontrolled high blood pressureare each associated with 1 or more disabling eye conditions.^19-21 Exceptfor diabetes mellitus,²² prevalence rates forthese risk factors have been decreasing during the previous 30 years.^23,24 There have also been marked increasesin the number of US adults undergoing cataract surgery or other surgical proceduresdesigned to restore vision or slow disease progression. For example, in theUnited States, the number of lens extractions completed in short-stay hospitalspeaked in 1983 at 630 000.²⁵ These numbershave since declined as lens extraction increasingly moved into an outpatientsurgical setting. Virtually all (ie, >97%) of the estimated 2.3 million lensextractions that took place in the United States in 1995 were performed onan outpatient basis.²⁶ Similar data on inpatientand outpatient surgical trends for the treatment of other major eye diseaseshave not been published. However, the number of glaucoma-related office visitsin the United States increased by 43% from 1985 to 1991, suggesting growingaccess to medical and surgical treatment options for this condition.^27-29 Despite these treatmenttrends, the present findings are not suggestive that population-level reductionsin visual impairment rates occurred from 1986 to 1995. Additional advancesin the treatment of disabling eye diseases such as macular degeneration maybe needed to measurably affect these rates.³⁰ Strategiesto reduce medical and surgical undertreatment of disabling eye diseases arealso needed.^31-33 Finally,population-based clinical studies^34-36 ofeye disease in the United States indicate that an important cause of visualimpairment is uncorrected refractive error. Therefore, population-level improvementsin visual impairment due to treatment of disabling eye diseases may be maskedin part because of continued high levels of uncorrected refractive error.

The present findings do not support the notion of a compression of morbidityin older populations with respect to visual impairment, as there was no evidenceof decline in visual impairment rates in any of the older age groups. Sucha compression of morbidity may exist only for certain organ systems (eg, musculoskeletal);alternatively, it may take longer for reductions in visual impairment ratesto become apparent at the population level. Furthermore, as already noted,a large percentage of visual impairment is due to uncorrected refractive error.Therefore, further evaluation of the hypothesis that compression of morbiditycan be applied to visual impairment may require the collection of clinicalophthalmic data in representative samples over longer periods. Comparisonof clinical data from existing studies is problematic given protocol differencesand advances in the assessment of visual impairment and disabling eye diseases.Nevertheless, comparative analyses would enable the exclusion of participantswith uncorrected refractive error.

To summarize, analyses of the NHIS data show that annual age-adjustedrates of reported visual impairment varied only slightly from 1986 to 1995.Also, there were no significant reductions in reported visual impairment ratesin any of the 10-year age groups. These findings suggest that increased useof and additional improvements in treatment of disabling eye disease and refractiveerror, in combination with population-level reductions in risk factors forthese eye diseases, are needed before significant reductions in visual impairmentcan be achieved.

Corresponding author: David J. Lee, PhD, Department ofEpidemiology and Public Health, University of Miami School of Medicine, HighlandProfessional Building, 1801 NW Ninth Ave, Room 212-I, Miami, FL 33136 (e-mail: dlee@med.miami.edu).

Submitted for publication June 23, 2003; final revision received October16, 2003; accepted November 21, 2003.

This study was supported by grant 1R03EY13241 from the National EyeInstitute, Bethesda, Md.