When Walking Fails

Mobility Problems of Adults with Chronic Conditions

Lisa I. Iezzoni

Technical Appendix 2: Survey Information

The primary source of numerical information is the 1994-95 National Health Interview Survey (NHIS) and four supplemental surveys:

• 1994-95 Disability supplement (NHIS-D), Phase I
• 1994-95 Disability supplement (NHIS-D), Phase II
• 1994-95 Family Resources supplement
• 1994 Healthy People 2000 supplement

Ellen McCarthy, Ph.D., epidemiologist, conducted the analyses, with input from Roger Davis, Sc.D., biostatistician. Hilary Siebens, M.D., physiatrist and geriatrician, assisted with designing our mobility measure. Details about our work with these surveys are available elsewhere (Iezzoni et al. 2000a, 2000b, 2001).

National Health Interview Survey

Since 1957, the NHIS has continuously gathered information on "the health of the civilian noninstitutionalized population" living in the U.S. (P. F. Adams and M. A. Marano, "Current Estimates From the National Health Interview Survey, 1994," Vital and Health Statistics 10, no. 193 [National Center for Health Statistics, 1995], 1). Overseen by the Division of Health Interview Statistics, National Center for Health Statistics (NCHS), this nationwide survey covers all fifty states and the District of Columbia, revising its sampling procedures after each decennial census to keep abreast of population changes.[Note 1] Because the NHIS surveys only households of civilian noninstitutionalized persons, its results do not account for homeless people, military personnel, inmates of penal and other institutions, and residents of nursing homes and other health-related facilities. Population estimates based on the NHIS therefore do not include these important populations, some of which may have high rates of mobility difficulties (e.g., nursing home residents).

Each week, the NHIS selects households for interviews based on a carefully constructed sampling scheme that identifies persons representative of the target population (Adams and Marano 1995). By reweighting the survey results based on probabilities of being selected for interviews, analyses produce nationally representative figures. Starting in 1985, the NHIS began oversampling African American households to improve the precision of estimates for black persons, and Hispanic households were oversampled in 1995 (and several prior years). All numbers in this book drawn from the NHIS and its supplemental surveys are reweighted using SUDAAN statistical software to produce national estimates.[Note 2] Because the NHIS surveys only a sample rather than the entire population, its results are subject to "sampling errors"—variations that occur by chance within the respondent population.[Note 3]

Employees of the U.S. Bureau of the Census conduct NHIS interviews year-round and in-person. Households are typically contacted by mail before the NHIS interviewer arrives, and interviewers return multiple times to homes when respondents are unavailable. Over the years, the NHIS has achieved a 94 to 98 percent response rate (Adams and Marano 1995, 132). Whenever possible, all adult family members are interviewed. Other adults give "proxy" responses for family members who are not at home and, as required, for all children and for household members who cannot respond for themselves (e.g., because of physical, emotional, or cognitive difficulties). NHIS interviewers do not independently confirm the accuracy of responses. Errors may result because respondents do not understand the question, do not know or wish to divulge the information, or interpret the question in unintended ways. In particular, proxy's responses may be problematic especially when asked to report personal attitudes (Iezzoni et al. 2000b).

The NHIS asks for basic demographic information, including income, marital status, living arrangements, and education, as well as questions about use of physician and hospital services, activity restrictions (in housework, schooling, or employment) attributable to health, self-perceived health, height and weight, injuries, and presence of specified chronic health conditions. The chronic conditions are arrayed within six lists, and one-sixth of the survey respondents are asked about conditions on each of the six lists. Therefore, consistent information about chronic health conditions is not collected across all NHIS respondents, and we do not use this information.

1994-95 NHIS-D Phase I

The Phase I disability supplement (NHIS-D) was asked of all respondents to the 1994 and 1995 NHIS—202,560 persons (self- and proxy-respondents), representing a 93 percent response rate. The NHIS-D Phase I asked detailed questions about ADLs, IADLs, and assistance with daily activities, ability to perform various physical functions, sensory limitations (vision, hearing), mental health, specific signs and symptoms, use of assistive devices, and self-perceived disability. We used responses from 145,007 persons 18 years of age and older.

Identifying Mobility Difficulties

The NHIS-D Phase I asked eight questions about physical functioning of the upper and lower body, including bending, stooping, and lifting heavy objects. Here we concentrate on the three questions related to moving independently around the community:

1. whether the person has "any difficulty walking a quarter of a mile—about 3 city blocks"
2. whether the person has "any difficulty walking up 10 steps without resting"
3. whether the person has "any difficulty standing for about 20 minutes"

Persons reporting difficulty on each question were queried about level of difficulty: "some," "a lot," or "completely unable." The NHIS-D asked only persons reporting being "completely unable" whether this limitation would persist at least 12 months, and about 77 percent said that it would.

We considered whether the three functions (walking, stair climbing, standing) deserve equal weighting in measuring mobility. Reports of any difficulty are highly (p = 0.0001) but not perfectly correlated between pairs of functions: Pearson correlation coefficients are 0.70 for walking and stair climbing; 0.63 for walking and standing; and 0.61 for stair climbing and standing. On a separate ADL question, relatively few respondents (just over 1 percent) reported "any difficulty getting around inside of the home." We viewed this ADL as representing the most impaired end of the mobility continuum. Pearson correlation coefficients for reporting difficulty on this ADL and the three mobility questions are similar: 0.35 for walking; 0.38 for stairs; and 0.37 for standing (all p = 0.0001). Since the variance in the "level of difficulty" ascribed to each of the three functions is roughly comparable (walking, 0.28; stair climbing, 0.18; and standing, 0.18), we weighted each of the three questions equally for our mobility measure. This approach gives walking (with its somewhat higher variance) more influence than the other functions, which seems appropriate.

Inquiries about physical functioning did not ask about use of mobility aids; assistive device questions appeared elsewhere in the questionnaire. Therefore, it was unclear whether people considered reliance on assistive technologies when reporting functional limitations; 0.5 percent reported "no difficulty" with walking, stairs, or standing yet said they used mobility aids.[Note 4] Following the approach of Freedman and Martin ("Understanding Trends in Functional Limitations Among Older Americans", American Journal of Public Health 88, no. 10 [1998]:1457-62), we classified persons reporting no difficulty walking but who used mobility aids as having difficulty walking. We assigned respondents to the highest (i.e., most severely impaired) level for which they qualified, eliminating 2,435 (1.7 percent) persons who failed to respond to the three mobility function, level of difficulty, and assistive device questions.

Each functional limitation question asked specifically about whether difficulties resulted from motor vehicle accidents. Determination of other causes was more complicated. Persons indicating any difficulty across the eight physical functional limitation questions, including our three mobility questions, were asked an open-ended question about the "main problem or condition" causing the difficulty. To minimize respondent burden, persons were not asked separately about causes for specific deficits or about multiple causes. We assumed that difficulty with walking, stair climbing, and standing would dominate reporting decisions if respondents had multiple physical functional problems. Because of the survey design, we cannot test this assumption or identify multiple coexisting causes. Twenty percent of respondents did not report causes for functional difficulties. In addition, for persons not reporting functional limitations, we cannot determine whether they had similar conditions (e.g., arthritis or diabetes). NCHS survey staff assigned diagnosis codes, similar to International Classification of Diseases, Ninth Revision, Clinical Modification codes, to the self-reported conditions. They then grouped these codes into condition categories. We aggregated clinically related categories to examine frequencies of conditions.

Limitations of Our Mobility Measure

The NHIS-D did not address several crucial physical functions, such as surmounting curbs, essential to full mobility in the community. In addition, as noted above, NHIS interviewers do not ascertain the accuracy of responses. People could report that they have difficulties when they do not. While self-reports provide the only authentic information about persons' own perceptions of their functioning, some respondents could exaggerate or underestimate their deficits. Different people will view the response options ("some," "a lot," "completely unable") differently, perhaps relating to their self-perceived stoicism, independence, and feistiness. Many people reporting "some" difficulty may experience only minor inconveniences that other persons would dismiss altogether. NHIS-D questions do not specify whether they are asking about "capability" (physical ability to do an activity) or "performance" (actually doing the activity). Capability typically exceeds performance (N. L. Young et al., "The Context of Measuring Disability: Does It Matter Whether Capability or Performance Is Measured?" Journal of Clinical Epidemiology 49, no. 10 [1996]:1097-1101).

Use of proxies for nonrespondents further complicates interpretation of the results. Proxies provided about one-third of the responses. The mean age of self-respondents was 46 years, compared to 41 for persons with proxies; men were less likely to respond themselves (45 percent) than women (67 percent). Self-respondents were more likely to report mobility problems (13 percent) than respondents with proxies (7 percent). One possibility is that self-respondents could have been at home explicitly because of mobility problems, while those without mobility difficulties were out and unable to respond in person. Determining the true effect of proxy responses on mobility problems rates requires further study.

1994-95 NHIS-D Phase II

The NHIS-D Phase I was administered to everyone answering the 1994-95 NHIS. Although it contained more information on sensory, emotional, cognitive, and physical functioning than the usual NHIS, many observers wanted more detailed insight about people with potentially disabling conditions. Planning for the NHIS-D had begun shortly after passage of the ADA, and an important goal was to inform the public and policymakers about the extent of disabilities in the U.S. population and their effects on people's daily lives. The Phase I supplement served as a screening survey to identify respondents who could be considered "disabled." In Phase II, these people were asked about a variety of topics, ranging from employment to recreational activities to transportation.[Note 5]

The NHIS-D defined disability as "a general term that refers to any long- or short-term reduction of a person's activity as a result of an acute or chronic condition" (Adams and Marano 1995, 137). Using questions asked in the NHIS-D Phase I, NCHS staff specified groupings of adult disabilities as follows:[Note 6]

• mental retardation/developmental disability
• sensory impairment
• mental illness
• cognitive impairment
• limitations in ADLs, IADLs, or physical functions[Note 7]
• SSDI or SSI recipient or applicant

Although this process appears straightforward, the selection criteria for Phase II participation were complex. Partially because of resource constraints, NCHS interviewed primarily those with more severe disabilities: for example, persons "unable to do" an ADL or IADL and those reporting "serious difficulty" communicating with others. For my study, however, with Phase I information, we had looked across the continuum of impaired mobility, from mild to severe difficulties. Therefore, findings taken from Phase II reflect a sample of respondents with relatively major problems, although by using appropriate sampling weights, we again produce nationally representative figures.

The NCHS tried to interview in-person those people meeting their disability criteria within eight to twelve months of the Phase I interview, even if the person had since moved into an institutional setting (such as a hospital or nursing home). The Phase II disability sample includes 25,805 adults, for an 85 percent response rate; we used responses only from the 25,470 people who were not institutionalized at the time of the Phase II interview. The questionnaire covered a range of topics, including home environment, living arrangements, long-term care, transportation, recreation and leisure activities, vocational rehabilitation, special education, assistance with key activities, health insurance, use of a variety of clinical services, and assistive technologies.

Phase II Mobility Measure

To the extent possible, we replicated the Phase I measure of mobility difficulties using the Phase II data. Unfortunately, a key question differed between the two phases: while Phase I had asked about difficulty standing for 20 minutes, Phase II asked about difficulty standing for 2 hours. The Phase II standard is therefore much more inclusive than the Phase I question: certainly more people have difficulty standing for 2 hours (Phase II) than for 20 minutes (Phase I). The Phase II assistive device question also did not separate manual from power wheelchairs, but this distinction between the devices does not affect assignment (all wheelchairs users have "major" mobility difficulties).

Just over 50 percent of the noninstitutionalized Phase II sample (8,926 persons or 52 percent) had at least some mobility difficulty measured using the Phase II definition: 16 percent with minor, 13 percent with moderate, and 22 percent with major problems.[Note 8] About 85 percent of Phase II respondents who reported at least some mobility difficulty in Phase I also had a mobility problem in Phase II. For the Phase II analyses of people with walking difficulties, we looked only at people who had also reported problems in Phase I, thus presumably focusing on persons with chronic difficulties.

Limitations of Phase II Data

The Phase II data have identical limitations to those from Phase I relating to accuracy of responses and reliance on self-perceptions of respondents. Although Phase II interviewers tried to interview individuals with disabilities, proxies supplied roughly 20 percent of Phase II responses, raising similar questions about validity as in Phase I.[Note 9] The Phase II data are significantly limited by the relatively small sample size. Many of the Phase II questions had multiple stages: persons were first asked a broad question (e.g., if they had adapted their car); then, if they answered "yes," they were asked more detailed questions (about specific types of adaptations to their car). Sample sizes sometimes became exceedingly small for these more detailed questions, making it impossible to compute reliable nationally representative estimates. I do not report figures when fewer than five persons fell into a response category.

1994-95 Family Resources Supplement

All respondents to the 1994 and 1995 NHIS were asked questions from the Family Resources supplement. This survey gathered detailed information on employment, income, public assistance, SSDI and SSI benefits, disability pensions, assets, health insurance type, coverage of specific services, access to health care, and delays in obtaining care because of cost concerns. We used household income levels calculated by NCHS, which imputed values for people who did not report income.

1994 Healthy People 2000 Supplement

The decennial "Healthy People" initiatives overseen by the U.S. Department of Health and Human Services (1995, 2000) aim to improve the overall health of the U.S. population by highlighting and monitoring risk factors for acute and chronic illnesses and implementing screening and preventive services. While all households received the Phase I Disability and Family Resources supplements, only one randomly selected adult from half of the households was asked Healthy People 2000 questions. Although NCHS also conducted this supplemental survey in 1995, Healthy People 2000 questions differed between 1994 and 1995. Here we examine 1994 questions (19,337 adult respondents). Sample sizes become relatively small for some Healthy People questions, but they are sufficient to produce reliable nationally representative estimates.

The 1994 Healthy People 2000 supplement asked about specific screening and preventive services for people reporting having had a routine physical examination within the prior 3 years. Depending on the specific service, only women (e.g., Papanicolaou smear) or persons in certain age groups (e.g., persons 65 years of age and older for flu shots) were asked the question. Questions also sometimes specified time frames for the service (e.g., tetanus shot in the past 10 years, mammogram in the past 2 years for women at least 50 years old).

Statistical Analysis

We used SAS-callable SUDAAN to conduct all analyses. Because these data come from a cross-sectional survey, we could not examine causal linkages. For example, although we found that people with mobility difficulties have higher poverty rates than other people, we could not examine whether mobility problems cause poverty. All we can say from this finding is that mobility and poverty are "associated"—they tend to occur together, but we cannot say why or that one causes the other.

Older people are more likely to report mobility problems than younger people, and women report more difficulties than men. To account for these differences by age and sex, most percentages presented in the tables are adjusted for age and sex using direct standardization. Age adjustment employed 5-year groups between ages 25 and 85, and additional groups for 18-24 and 85+ years.[Note 10] Adjusted rates used only known responses about presence of relevant characteristics in the numerators, with the entire sample (including those with unknown responses) as the denominators. Because of the large sample size with the full NHIS-D Phase I data, even small differences were highly statistically significant.

Notes for Technical Appendix 2: Survey Information

¹. Complete information about the NHIS and its supplemental surveys, including the questionnaires, is available on the NCHS Internet web site (http://www.cdc.gov/nchs/nhis.htm). Specific information about the 1994-95 NHIS-D, including the survey instruments and data collection methods, is also available (http://www.cdc.gov/nchs/about/major/nhis_dis/nhis_dis.htm). The following chart lists the data sources for tables used throughout this work. back to text

². We conducted all analyses using SAS-callable SUDAAN (version 7.5, Research Triangle Institute, Research Triangle Park, N.C.). SUDAAN facilitates the Taylor series linearizations necessary to obtain valid standard errors and statistical tests when applying sampling weights drawn from complex survey samples. Specific aspects of the sampling procedures changed between 1994 and 1995, necessitating adjustments in reweighting during analyses. back to text

³. Whenever possible, we combine results from 1994 and 1995. Analysis of more than one year of data reduces sampling errors. back to text

⁴. People probably report they have no difficulty with these mobility activities because they no longer perform them. They now use wheeled mobility instead. back to text

⁵. For the Phase II "followback" survey, the NCHS designed different questionnaires for four separate groups of respondents: children under 18 with disabilities; adults age 18+ with disabilities; adults age 18+ who had had polio; adults age 69+ without a disability. We only used results from the survey of adults with disabilities. back to text

⁶. Persons who had polio were a separate group and received their own questionnaire (see note 5). back to text

⁷. For the ADLs, people qualified as disabled if they responded: gets help; a lot of difficulty, doesn't do, unable and expected to last 12 or more months; uses special equipment; needs to be reminded; some difficulty and expected to last 12 or more months. back to text

For the IADLs, the following responses qualified as disabled: gets help; a lot of difficulty, doesn't do, unable and expected to last 12 or more months; some difficulty and expected to last 12 or more months.

For any physical function activity (lifting, walking up steps, walking, standing, bending, reaching, using fingers, holding pen/pencil), one of the following is flagged: unable to do and expected to last 12 or more months; a lot of difficulty with two or more of the functional activities.

⁸. In the NHIS-D Phase I sample, about 1 percent had missing values for the variables defining the mobility levels; we therefore excluded these respondents. In the NHIS-D Phase II sample, we eliminated 4 percent of persons without answers to the mobility questions. back to text

⁹. Proxies alone provided 12 percent of Phase II responses; 5 percent of respondents with disabilities had assistance from proxies; and proxy respondent status was unknown for 4 percent of the sample. back to text

¹⁰. Some questions, especially in the Healthy People 2000 supplement, were asked only of people in certain age ranges. In these instances, age adjustment considered only relevant categories. back to text