National Academies Press: OpenBook

Biosocial Surveys (2008)

Chapter: 4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir

« Previous: 3 The Taiwan Biomarker Project--Ming-Cheng Chang, Dana A. Glei, Noreen Goldman, and Maxine Weinstein
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 78
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 79
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 80
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 81
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 82
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 83
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 84
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 85
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 86
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 87
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 88
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 89
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 90
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 91
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 92
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 93
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 94
Suggested Citation:"4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study--David Weir." National Research Council. 2008. Biosocial Surveys. Washington, DC: The National Academies Press. doi: 10.17226/11939.
×
Page 95

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

4 Elastic Powers: The Integration of Biomarkers into the Health and Retirement Study David Weir T he title of this chapter, “Elastic Powers,” appeared as the last words of Cells and Surveys: Should Biological Measures Be Included in Social Science Research?, the influential 2001 volume from the Committee on Population (National Research Council, 2001). In his closing reflections on the biosocial opportunities for surveys, Kenneth Wachter quoted from a Matthew Arnold poem that described the aging process as exhaustion from the cumulation of shocks and change—a literary anticipation of the concept of allostatic load that Wachter in turn stretched from a model of individual physiology in a social context into a model of paradigm shifts in the history of science (Wachter, 2001). He warned of the challenges to established modes of thinking that the integration of biology and social surveys would pose, calling for a renewal of what Arnold had termed “the elastic powers.” Beginning the story of the integration of biomarkers into the Health and Retirement Study (HRS) with the closing words of Cells and Surveys is more than mere symbolism. Several contributors to that volume went on to be contributors to the HRS effort: Robert Wallace and Robert Willis as investigators, Eileen Crimmins and Douglas Ewbank as members of advisory groups, Teresa Seeman as an adviser to the study’s sponsor, the National Institute on Aging (NIA), and Jeffrey Halter as a consultant. Richard Suzman’s considerable role in organizing the first volume was acknowledged by Jane Menken in her preface, and his leadership on the NIA side of the HRS cooperative agreement is unrelenting. Together, their own elastic powers in picking up where Cells and Surveys left off have 78

DAVID WEIR 79 been considerable, and through their influence the elasticity of the entire field of population research has been renewed. The scientific rationale for including biomarkers in HRS is not fun- damentally different from the rationale for including them in any popu- lation survey concerned with health. They validate and add nuance to self-reports of health, they allow richer modeling of pathways of influ- ence between the socioeconomic and the physical, and they may capture aspects of health unknown to survey participants. This chapter gives examples of how each of these are realized in the HRS. The development of biomarker data in other studies of older populations in the United States, such as the National Survey of Midlife Development in the United States (MIDUS) and the National Social Life, Health, and Aging Project (NSHAP), and outside the United States in the English Longitudinal Study of Ageing (ELSA) and the Mexican Health and Aging Study (MHAS), has both provided models of what can be done and created great potential for comparative work with the addition of such data to the HRS. Because of the unique place of the HRS in population surveys of aging, however, the decision to add biology to the HRS involved a num- ber of other considerations, several of which were clearly anticipated by Weinstein and Willis in their chapter of Cells and Surveys (Weinstein and Willis, 2001). The HRS is a large longitudinal study that serves a large constituency of researchers from many different disciplines. At last count, there were over 6,000 registered users of the data, and over 1,000 unique authors of written research using the data. Putting its traditional aims at risk through attrition of respondents or elimination of critical established content would have been unacceptable. Similarly, the confidentiality of respondents had to be protected, as well as the integrity of a longitudinal observation study not be transformed into an intervention study. The ethical issues were considered carefully by the HRS investigators as well as the institutional review board (IRB) governing the study. Noti- fying respondents of the results of well-established and commonly avail- able diagnostic tests was deemed an ethical responsibility that overrides any concern that the information might alter future behavior. Because the tests contemplated by HRS assess familiar risk factors and do not identify, for example, life-threatening cancers, the ethical conflict is not particularly difficult at this time. Biological material stored in repository for future use is governed by a separate IRB review. Respondents were asked to consent to having this material stored anonymously for future research. Ethical issues arising from any particular future test will need to be addressed at that time. For example, it is conceivable that some test of scientific value might not be permitted if it carried with it the ethical obligation to notify children or other nonparticipants of the possibility of an inherited disease,

80 BIOSOCIAL SURVEYS and if that notification were considered detrimental to the confidentiality of respondents. Through the use of supplemental studies, some funded through peer review as competing supplements to the HRS, most of the elements of the HRS biomarker expansion were piloted on subsamples well before their introduction to the main survey. Those pilot efforts from 2001 through 2004 are thus a crucial part of the story. Adding biomarkers without subtracting other things from an ongoing panel study leads to the question of cost. After the baseline interviews, the primary mode of interview in the HRS has been telephone. Although some biological material can be collected by mail, and the HRS had some success with this in a study of diabetes described at some length below, it was clear that, for a thorough integration of biology, some form of per- sonal contact would be needed: clinic visits, nurse visits to the home, or in-person interviewing. Because of the high cost of clinics or nurses, and because of technical innovations that have expanded what can be done by interviewers in the home, the HRS has designed its biomarker effort around conventional in-person interviewing. First described in its 2005 renewal proposal and now fully imple- mented in the ongoing 2006 data collection, the HRS has developed an integrated package of new content for a new model of in-home inter- view we describe as “enhanced” face-to-face. It includes anthropometrics, physical measures, blood spots, salivary DNA, and a self-administered psychosocial questionnaire. Most of the elements of this dramatically new development for the HRS were piloted in one way or another in smaller supplemental studies. AGING, DEMOGRAPHICS, AND MEMORY STUDY The first effort at collecting biological specimens from HRS respon- dents came in a supplemental study of dementia, known as the Aging, Demographics, and Memory Study (ADAMS). The primary aim of the ADAMS study was to establish the prevalence of dementia in the popula- tion over 70 years of age from a nationally representative sample (Langa et al., 2005). Because dementia is not a common condition even at that age, a simple random sample of the population would need to be fairly large to derive reliable estimates. The great virtue of using the HRS as a sampling frame for the ADAMS was the ability to sample at higher rates from persons with higher likelihoods of dementia based on the cognitive assessments conducted by the HRS. The stratification of sample selection for ADAMS was based on five cognitive categories from the HRS interview. These had to be established separately for persons who did their own cognitive assessments and

DAVID WEIR 81 respondents for whom interviews were taken by proxy. In the case of proxies, the proxy reporter provides to the HRS interviewer an assess- ment of the cognitive function of the proxied respondent from the Jorm Informant Questionnaire on Cognitive Decline in the Elderly (IQCODE). The cognitively normal group was further stratified by age (ages 70-79 versus 80 or older) and sex in order to ensure adequate numbers in each of these subgroups. Finally, because of the long anticipated field period for the in-home ADAMS assessments, the recruitment was split between the 2000 wave of HRS and the 2002 wave, with some geographic areas drawn in one and some in the other, on a randomly sampled basis. Because the rigorous in-home dementia assessments were conducted by collaborators at Duke University Medical School, the consent process for ADAMS was in two parts. First, an interviewer from the Institute for Social Research (ISR) at the University of Michigan contacted the respon- dent and an informed caregiver to obtain consent for Duke to contact them about the study. Duke then established contact and obtained consent for the home visit. This two-stage process worked against response rates in two ways. It extended the time interval between the HRS interview and the ADAMS visit, which resulted in loss due to mortality, and it provided respondents two opportunities to say no. Overall, the response rate among survivors was 56 percent, yielding a final sample size of 856. Mortality rates (14 percent overall) were higher among the more cognitively impaired, but response rates conditional on survival were slightly higher for those groups. There were few significant predictors of nonresponse to the ADAMS study. Racial minorities partici- pated at slightly higher rates than whites. There are no established blood tests for dementia or cognitive impair- ment. Consequently, the ADAMS protocol relied primarily on extensive neuropsychological testing and did not include any blood sampling. There is one well-established genetic risk factor for dementia, and that is the E-4 allele of the ApoE gene. The ADAMS protocol did include collect- ing a cheek swab. Nearly everyone who participated in ADAMS agreed to provide this sample; only 11 of the 856 refused (1.3 percent). The samples were sent to the pathology lab of the University of Michigan for extracting and typing of ApoE. Only three of the samples could not be genotyped. The (weighted) distribution of genotypes found in the ADAMS respondents replicates fairly closely the expected population distribu- tions (Table 4-1). As has been found elsewhere, the presence of any ApoE- 4 allele is a risk factor for dementia, but not a particularly powerful one (Hyman et al., 1996). Preliminary analyses show that presence of the ApoE-4 allele (either homozygous 4/4 or heterozygous 3/4) was associ- ated with approximately twice the odds of dementia compared with those with the 3/3 genotype, which is in the range of previously reported values

82 BIOSOCIAL SURVEYS TABLE 4-1  Distribution of ApoE Genotypes in ADAMS and Other Studies ADAMSa IOWA 65+ Framingham Genotype % Populationb Populationb ApoE e2/2 1 1 1 ApoE e2/3 12 15 12 ApoE e2/4 2 2 2 ApoE e3/3 60 58 63 ApoE e3/4 22 22 19 ApoE e4/4 2 2 3 aADAMS percentages are weighted. bData from Hyman et al. (1996). (Breitner, Jarvik, Plassman, Saunders, and Welsh, 1998; Skoog et al., 1998). In the ADAMS data, the odds ratio for dementia associated with residing in a rural area compared with urban or suburban areas is nearly as high. The ADAMS study demonstrated that HRS respondents would be willing to provide samples of DNA for research purposes. In the group of respondents willing to participate in a three-hour home interview to assess dementia, cooperation with the DNA request was nearly universal. Combined with the 56 percent response rate to the ADAMS study overall, however, only 55 percent of the HRS respondents approached for ADAMS ultimately provided a DNA sample. Diabetes Study In contrast to the ADAMS study, for which biomarker collection was a relatively small part of the overall assessment, the diabetes study was motivated in large part by the idea of collecting a clinically meaningful biomarker of the disease. Diabetes can result from a variety of underly- ing conditions (pancreatic failure to produce insulin, cellular resistance to absorbing insulin), but it is always characterized by excessive levels of glucose in the blood. High levels of blood glucose cause damage to both large and small blood vessels and to nerves, potentially leading to many severe consequences (including cardiovascular disease). Among persons with diagnosed diabetes, the management of the disease targets the main- tenance of lowered glucose levels (and, increasingly, the management of other cardiovascular disease risk factors, especially hypertension). Thus, while there is considerable interest in understanding how people manage

DAVID WEIR 83 and cope with the disease, further study without a clinical marker for glucose levels seemed of relatively low priority. A study on the scale of ADAMS, with an in-home assessment and blood draw, would have been quite expensive. Much of the nonbiological information about diabetes could easily be collected by a self-administered mail survey, which is far less expensive than even telephone interviews. The innovative aspect of the diabetes study was the attempt to gather dried blood spots (DBS) through the mail for the analysis of glycoslylated hemoglobin (HbA1c). A1c is an ideal measure for this study, as it is for the medical management of diabetes, because it summarizes the average levels of blood glucose over a two- or three-month period. It also does not require fasting and can be done from blood collected at any time of the day. Glucose levels vary widely over the course of a day and in response to the intake of food, making standard point-in-time readings very dif- ficult to interpret in isolation. Because of the reliance on A1c measures in the medical management of diabetes, commercial laboratories have developed assays for A1c that can be done in DBS. This allows patients to take their own samples and mail them to a lab from which the results can be reported to their doctors, saving time and money. DBS assays for A1c require special proprietary pretreatment of filter paper and utilize proprietary laboratory methods for analysis. Working with a commercial partner that has developed a DBS assay is therefore essential. Flexsite Diagnostics was the laboratory that did the HRS diabetes study, and their support and cooperation were outstanding. They designed a specimen collection card specifically for the study. This allowed the respondents to mail their specimens directly to the lab with only an arbitrary numeric identifier, so that the laboratory would not know the name or address of the respondent. Results were reported to HRS by numeric identifiers and then merged with the questionnaire data and the usual HRS identifications. The diabetes sample was selected from respondents to the 2002 wave of HRS. Only persons reporting a doctor diagnosis of diabetes were eli- gible. About 20 percent of the eligible sample was excluded because of their participation in another HRS mail survey (the Consumption and Activities Mail Survey). The eligible sample numbered 2,518. Of that group, 133 (5.3 percent) died prior to the beginning of the diabetes study in late 2003. The diabetes study proceeded in two stages. First, respondents were sent a self-administered questionnaire, along with a check for $40 and an explanation that they would be receiving a second request to send a blood sample later (the usual HRS incentive for a mail survey is $20). Blood test kits were sent out to respondents when questionnaires were returned. A standard protocol of reminders was followed. After about six weeks,

84 BIOSOCIAL SURVEYS duplicate questionnaires and blood kits were mailed to persons who had not responded. After about eight weeks, follow-up telephone calls were placed to some respondents. Questionnaires were returned by 1,897 sample members (79.7 per- cent). In contrast to both the core HRS interviews and the ADAMS study, but quite consistent with other HRS mail surveys, there were substantial racial and ethnic differences in participation. Blacks and Hispanics had response rates about 10 percent lower than those of whites. Blood kits were returned by 1,233 respondents, which is 65 percent of those who returned the questionnaire. There was not much difference between His- panic and other respondents on the blood test response rate conditional on participation in the questionnaire, but there was again a lower response rate among blacks. Combined with the questionnaire response rate, the net biomarker rate was 52 percent of the eligible surviving sample. The quality of the A1c data collected seems to be quite satisfactory. Figure 4-1 shows the level of A1c according to the type of treatment regime: 7.9 for those on insulin, 7.2 for those taking oral medication only, and 6.5 for those not taking medication (F-statistic = 53.8, p < .0001). The 8.5 8.0 7.9 7.5 7.2 7.0 6.5 6.5 6.0 5.5 5.0 Insulin Oral Medication None FIGURE 4-1  Mean HbA1c score by type of medication regime. SOURCE: HRS Diabetes Study.

DAVID WEIR 85 9.0 8.5 8.0 7.8 7.7 7.5 7.1 7.0 6.5 6.0 5.5 5.0 White Black Hispanic FIGURE 4-2  Mean HbA1c score by race and ethnicity. SOURCE: HRS Diabetes Study. corresponding numbers for the population ages 50 and older from the National Health and Nutrition Examination Survey (NHANES) study for 2003-2004 are 7.8, 7.0, and 6.2. Figure 4-2 shows the differentials by race and ethnicity: 7.1 for white non-Hispanics, 7.7 for blacks, and 7.8 for Hispanics (F-statistic = 35.1, p < .0001). That again differs only slightly Figure 4-2 from the comparable NHANES figures of 7.1, 7.9, and 7.8. Finally, Figure 4-3 shows that A1c also varies according to the respondent’s self-assessed performance at managing the disease. Respondents giving themselves an “A” had A1c scores of 7.0, compared with 7.3 for “B” grades, and 7.8 for “C” or lower (F-statistic = 15.1, p < .0001). Thus, in comparison with ADAMS, the diabetes study had a much higher overall participation rate but a fairly comparable net completion rate on the biomarker. Taken together, these two experiments suggested several important guidelines for future work on biomarkers in the HRS. First, multistage requests, in which the biomarker request is conditional on agreeing to one or more prior request, are bad for response rates. Sec- ond, self-administration and mailback of blood spots, while inexpensive, is unlikely to yield high response rates and seems particularly ill-suited to maintaining high response rates of minorities. In-home requests, with a trained person present to take the sample, seemed to provide the best basis for administering biomarkers.

86 BIOSOCIAL SURVEYS 9.0 8.5 8.0 7.8 7.5 7.3 7.0 7.0 6.5 6.0 5.5 5.0 A B C FIGURE 4-3  Mean HbA1c score by self-rated assessment (letter grade) of self- management of diabetes. SOURCE: HRS Diabetes Study. Face-to-Face Interviews In 2004, the HRS was given additional funding from the Social Secu- Figure 4-3 rity Administration to use in-person interviewing to improve consent rates for linkage to Social Security records for two groups: all of the origi- nal 1992 HRS cohort (born 1931-1941, plus spouses) and members of the 1998 war baby cohort (born 1942-1947, plus younger spouses) who had not yet given consent. This effort was successful. Seizing the opportunity created by in-person interviewing to pilot some other measures, the HRS obtained administrative supplements from NIA to conduct in-person interviews with samples of the other cohorts to create a representative sample of the whole. From the combined set of in-person interviews, samples of about 100 persons from each single year of birth were assigned to do physical performance measures, with a subset getting height and weight measures. Although the 2004 interviews did not include any blood or DNA work or blood pressure testing, they were an important step in develop- ing the 2006 strategy for biomarkers. We observed that there was a fairly high loss of sample due to respondents declining the in-person interview in favor of telephone—about 10 percent of those assigned. Thus, address- ing “mode switches” is important for the HRS, given its history as a

DAVID WEIR 87 telephone survey. We also observed that failure to complete the physical performance measures (timed walk, puff test, grip strength) was related to self-reported physical limitations. Having good self-report indicators of those abilities would aid in understanding that censoring. 2006 Enhanced Face-to-Face Interview All the work of the various supplemental studies and pilot projects were brought together in the design of the enhanced face-to-face inter- view for 2006. The key elements of the 2006 HRS enhanced face-to-face interview are • measured height and weight and waist circumference, • blood pressure, • timed walk, grip strength, puff test, balance test, • dried blood spots for HbA1c, total cholesterol, high-density lipo- protein cholesterol, C-reactive protein and repository, • salivary DNA for repository, and • self-administered mailback psychosocial questionnaire. Selection of Measures As a multidisciplinary population survey serving a wide community of researchers, the decision process in HRS about any survey content, including biomarkers, must consider a wide range of potential uses and not focus narrowly on specific hypotheses or interests. Input was sought from a large number of experts. The choice of measures attempted to bal- ance scientific value against cost and respondent and interviewer burden. There are two primary foci of the measures: the first is obesity and meta- bolic syndrome, for which the main goal is obtaining assessments now to model risks of future events, and the second is frailty, for which the main goal is improving our characterization of the dynamics of disability and care needs of the elderly. As an example of this selection process, the Quetelet body mass index (BMI) is obviously a critical measure for understanding obesity, and direct measures of height and weight should help to resolve any doubts about the accuracy of self-reports. But BMI is far from a perfect measure because of variability in muscle mass and other nonfat components of body weight. Waist circumference adds valuable complementary informa- tion about fatness and in particular about central adiposity. Waist-hip ratio was considered to add relatively little to waist alone, and hip measure- ment is more intrusive for respondents and difficult for interviewers. Grip strength is a somewhat expensive measure because of the high price of the

88 BIOSOCIAL SURVEYS dynamometer devices, but when this is factored over an average of 50-60 enhanced face-to-face (EFTF) interviews per interviewer in this wave, and the potential for reuse in other waves, its value in assessing loss of muscle strength clearly outweighs its cost. A more difficult set of choices had to be made regarding the physical performance measures. There are a number of well-known assessments of lower body mobility, such as chair stands and “get-up-and-go.” We determined in 2004 that timed walk was the best single measure for our needs. For 2006, after consultation with other experts, we decided that what was needed to complement timed walk was not another measure focused on lower body strength, but rather mea- sures that directly assessed balance, because that can also be useful not only for modeling falls, but also in understanding cognition. We therefore added a 10-second semitandem stand, followed by a side-by-side stand, or a full tandem stand, depending on performance. The most significant restriction imposed by cost constraints was on blood testing. Drawing of whole blood in the home or in a clinic would be extremely expensive in a dispersed national sample like the HRS. At the same time, the laboratory technologies for using dried blood spots are advancing rapidly, making the scientific potential of this relatively inex- pensive field collection protocol extremely attractive. At present, the HRS blood spots will be used to assay for HbA1c, total cholesterol, HDL choles- terol, and C-reactive protein. All these measures are of course important in metabolic syndrome and cardiovascular risk. Having established the protocol for DBS collection, other assays can be added as the technology improves and as scientific interest and funding develop. Despite the restrictions imposed by both cost concerns and scientific focus, the new measures added to HRS cover a lot of ground. In their paper in Cells and Surveys, Eileen Crimmins and Teresa Seeman outlined a table of 17 measures on 8 different physiological systems that were related to social and behavioral influences and health outcomes (Crimmins and Seeman, 2001, p. 20). Of these, the HRS covers eight measures in four of the systems. The two most significant systems not covered are the sympa- thetic nervous system and the hypothalamic pituitary adrenal axis. Good measures of functioning in these systems require either whole blood, urine, or multiple measures during the course of a day or over several days (e.g., cortisol). The HRS will continue to follow research using these measures and technologies for assessing them. Sample Design The 2005 renewal proposal called for spreading the EFTF interviews over the next three waves of the HRS—randomly assigning one-third of the sample to each. Following the successful review of the proposal, NIA

DAVID WEIR 89 recommended that this be accelerated to assign one-half of the sample in 2006 and the other half in 2008, creating the possibility of a four-year interval between biomarker collections rather than six. That recommenda- tion was adopted. The assignments were made randomly at the house- hold level. That means that both persons in a two-person household will get the biomarker interview in the same year. It also means that all sample clusters will get a mix of conventional telephone follow-ups and enhanced face-to-face interviews, and therefore that all interviewers must be trained for both types of interviews. While this increases training costs slightly, it allows for operational efficiencies when interviewers can mix the two types of activities, and it allows for complete geographical repre- sentation in each wave. Interviewer Training The HRS was fortunate to follow, with about a year’s lag, the devel- opment of in-home biomarker interviews in the National Social Life, Health, and Aging Project at the National Opinion Research Center and the University of Chicago. Their success set a high standard. Critical to that success is the successful training of interviewers in both persuading respondents to participate and in conducting the various measures suc- cessfully. The HRS had already developed protocols for height and weight measurement and physical performance measures. The ADAMS study had used cheek swabs to collect DNA. The switch to mouthwash samples offered both better quantities of genetic material and an easier mode of administration. The two main areas for which new training materials had to be developed were blood pressure and blood spots. For the latter we were advised by Professor Thomas McDade of Northwestern University, as well as the commercial laboratory conducting the assays. For both we had input from Robert Wallace and Kenneth Langa, the two medical doc- tors on the HRS investigative team. The HRS survey operations group developed a DVD that demon- strated the protocols for all aspects of the new content. This video was sent in advance to prospective interviewers interested in working on HRS (most of whom had worked for the study in previous years). It helped to screen out interviewers who were too uncomfortable with the methods to do the work. It also served as a training vehicle and continues to serve as a refresher for interviewers in the field. Consent and Reporting The HRS developed a booklet for the administration of physical mea- sures and biomarkers. Respondents of course consent to participate in the

90 BIOSOCIAL SURVEYS HRS itself before interviewing begins. The biomarker assessments occur around the middle of the interview. For each of three sections—physical measures, blood spots, and DNA sample—the respondents are shown a printed information form and asked to read it and sign a consent before proceeding. In addition, respondents are asked after signing the consent whether they feel it is safe for them to perform each measure immedi- ately before doing it. Blood pressure results can be reported during the interview, and any respondent exceeding a specified threshold is given a card recommending that they see a doctor about their blood pressure. Respondents are also told that their blood test results for HbA1c and cholesterol will be reported to them by mail. Both the blood test and the DNA consents permit future analysis to be done for HRS-related research purposes without reporting back to the respondent. Early Results At this writing, the HRS is about 10 weeks from the end of its 2006 field period, and about 90 percent of the expected interviews are com- pleted. Early indications are that cooperation with the new EFTF inter- view is going well. Relatively few respondents have refused the face-to- face mode. Of those who have been interviewed, consent rates are over 94 percent for the physical measures, 82 percent for the DNA sample, and 81 percent for the blood spots. At present, older respondents are somewhat more likely to give DNA and less likely to give blood spots than younger respondents. In the 2004 pilot work with physical performance measures, we noted a significant correlation between noncompletion of the measures and self- reported physical limitations. The distribution of measured scores thus does not represent the true population distribution of abilities. In the case of timed walk, we have several good self-reports of lower body function that allow one to assess the function level of persons who decline to do the task and potentially to impute a physical performance score. For the other measures, we do not. In 2006 we therefore added self-rating ques- tions on hand strength and on lung function to aid in understanding the functional abilities of those who do not do the measures. Table 4-2 shows that self-rated hand strength correlates very strongly with measured grip strength, and that persons reporting weakness in the hand had substantially lower completion rates on the grip strength test. For lung function, shown in Table 4-3, the question on frequency of breathlessness is not quite as good at predicting response rates to the puff test. This may be due to the fact that (unwarranted) fear of infection from the device leads some well-functioning respondents to decline this test. Self-rating does correlate well with performance on the test.

DAVID WEIR 91 TABLE 4-2  Self-Rated Grip Strength, Response Rate to Grip Test, and Measured Grip Strength Self-Rated Hand Measured Grip Response Strength Strength Rate (%) N Measured Very strong 36.7 86.5 1,440 Somewhat strong 31.9 88.1 4,149 Somewhat weak 24.7 78.8 1,213 Very weak 20.6 48.9 174 F-statistic 351.7 151.3 p-value < .0001 < .0001 SOURCE: Preliminary HRS 2006 production data, unweighted. TABLE 4-3  Self-Rated Shortness of Breath, Response Rate to Lung Test, and Measured Lung Function Self-Rated Expiratory Response Shortness of Breath Force Rate (%) N Measured Often 272.2 72.3 391 Sometimes 313.2 85.3 1,221 Rarely 367.2 87.7 2,134 Never 378.0 87.6 3,411 F-statistic 135.6 33.7 p-value < .0001 < .0001 SOURCE: Preliminary HRS 2006 production data, unweighted. There is some controversy about the quality of self-reports of height and weight, although the general finding seems to be that there is a gen- eral tendency to overstate height among the elderly (Ezzati et al., 2006; Gunnell et al., 2000). The most plausible explanation for this is that older people report their maximum adult height, not their current height after shrinkage due to age-related compression. Weight tends to be underre- ported by the overweight, and overreported by the underweight, leaving a relatively small bias on average. In preliminary results from 2006, as well as in a very small sample from 2004, the HRS tends also to find rather small errors in reported weight, and systematic overreporting of height. The self-reports of height and weight are obtained before respondents are told that they will be measured. Figure 4-4 shows the pattern of heights found in 2006, graphing the mean measured height and mean self-reported height against the self-

92 BIOSOCIAL SURVEYS 72 70 68 Measured Height 66 64 62 60 60 61 62 63 64 65 66 67 68 69 70 71 72 Self-Reported Height Measured height Self-reported height FIGURE 4-4  Measured versus self-reported height. SOURCE: Preliminary HRS 2006 production data. reported height. If self-reports were unbiased (equal to measured) on average, the graph should show a perfect 45-degree line. Instead, mea- sured heights are lower than self-reports at every level of self-report, and more so at taller self-reported heights. The average differential is just Figure 4-4 under one inch. Measured heights are recorded to the nearest quarter- inch, and self-reports are in round inches. In addition to the bias, there is some random error, as shown by the correlation coefficient of .89 between measured and self-reported height. Figure 4-5 shows a similar graph for weights, grouping self-reports in 10-pound ranges on the horizontal axis and graphing on the vertical axis the mean of measured and self-reported weight for each of those groups. The average error is about three pounds (self-reports below measured weight). The correlation is also impressively high at .97. To put this in perspective, the average HRS respondent has a BMI of 29.1 using the measured data. Using instead the self-reported height lowers this to 28.2, while using instead the self-reported weight lowers it much less, to 28.6. Both together lower it to 27.8. Based on this evidence,

DAVID WEIR 93 255 235 215 Mean Weight 195 175 155 135 115 95 105 115 125 135 145 155 165 175 185 195 205 215 225 235 245 255 Midpoint of Self-Reported Weight Measured mean Self-reported mean FIGURE 4-5  Measured versus self-reported weight. SOURCE: Preliminary HRS 2006 production data. the real scientific value from measuring weight in an older population, as opposed to relying on self-report, does not appear to be as great as the Figure 4-5 gain from measuring height. Height measurements are also less costly and less burdensome than using scales to measure weight. Blood pressure data appear to be of good quality. The HRS protocol calls for three repeated measures, using an upper arm cuff with an auto- matic inflation device. These three measures are correlated at about .95, TABLE 4-4  Biomarkers and Self-Reports: Measured Blood Pressure by Self-Reported High Blood Pressure Diagnosis and Control (mean of 3 measurements) Systolic Diastolic No high blood pressure 126.4 78.1 Under control no meds 137.4 84.5 Under control using meds 134.2 79.9 Not under control 148.4 87.6 F-statistic 138.7 71.2 p-value < .0001 < .0001 SOURCE: Preliminary HRS 2006 production data, unweighted.

94 BIOSOCIAL SURVEYS indicating good reliability. Table 4-4 shows that the mean of the three measures is reasonably well correlated with self-reported status. Interest- ingly, the mean blood pressure of persons who report a diagnosis and say it is under control is not much higher than those who say they do not have hypertension. Those who report their blood pressure is not under control do indeed have substantially higher measured levels. Conclusion The integration of biomarkers into the HRS is very much a work in progress. The first big steps have been taken to transform a primarily telephone study into one using in-person interviewing to obtain direct physical measurements and collect biological samples in the home, chal- lenging the “elastic powers” of the survey’s designers and its funders. HRS respondents have shown themselves willing to participate in this new survey experience, and the data they have provided appears to be of high quality. The HRS investigators hope to continue to expand and innovate in the inclusion of biomarkers as appropriate to the overall aims of the HRS. While this provides valuable new content to the HRS and new points of contact with clinical and lab-based studies, a large population survey like HRS cannot replace the vastly greater biological detail attain- able in small clinical studies. Soon the challenge to the elastic powers will shift from the design and implementation of the measures to their integration into longitudinal analyses using the data. It is this crucial intellectual transformation of how researchers conceive of problems that Wachter saw as the real challenge. The effort to collect such measures in population surveys will be war- ranted only by the new research insights they support. To support that challenge, we will need to seek ways to encourage researchers to develop models to make use of them. And that in turn will stimulate new ideas and new measures for future waves of data collection. REFERENCES Breitner, J.C.S., Jarvik, G.P., Plassman, B.L., Saunders, A.M., and Welsh, K.A. (1998). Risk of Alzheimer disease with the epsilon-4 allele for apolipoprotein E in a population-based study of men aged 62-73 years. Alzheimer Disease and Associated Disorders, 12(1), 40-44. Crimmins, E., and Seeman, T. (2001). Integrating biology into demographic research on health and aging (with a focus on the MacArthur Study of Successful Aging). In National Research Council, Cells and surveys: Should biological measures be included in social science research? (pp.9-41). Committee on Population, C.E. Finch, J. Vaupel, and K. Kinsella, Eds. Commission on Behavioral and Social Sciences and Education. Wash- ington, DC: National Academy Press.

DAVID WEIR 95 Ezzati, M., Martin, H., Skhold, S., VanderHoorn, S., and Murray, C. (2006). Trends in national and state-level obesity in the USA after correction for self-report bias: Evidence from health surveys. Journal of the Royal Society of Medicine, 99, 250-257. Gunnell, D., Berney, L., Holland, P., Maynard, M., Blane, D., Frankel, S., and Davey Smith, G. (2000). How accurately are height, weight, and leg length reported by the elderly and how closely are they related to measurements recorded in childhood? International Journal of Epidemiology, 29(3), 456-464. Hyman, B.T., Gomez-Isla, T., Briggs, M., Chung, H., Nichols, S., Kohout, F., and Wallace, R. (1996). Apolipoprotein E and cognitive change in and elderly population. Annals of Neurology, 40(1), 55-66. Langa, K.M., Plassman, B.L., Wallace, R.B., Herzog, A.R., Heeringa, S.G., Ofstedal, M.B., Burke, J.R., Fisher, G.G., Fultz, N.H., Hurd, M.D., Potter, G.G., Rodgers, W.R., Steffens, D.C., Weir, D.R., and Willis, R.J. (2005). The aging, demographics, and memory study: Design and methods. Neuroepidemiology, 25, 181-191. National Research Council. (2001). Cells and surveys: Should biological measures be included in social science research? Committee on Population, C.E. Finch, J.W. Vaupel, and K. Kinsella, Eds. Commission on Behavioral and Social Sciences and Education. Wash- ington, DC: National Academy Press. Skoog, I., Hesse, C., Aevarsson, O., Landahl, S., Wahlstrom, J., Fredman, P., and Blennow, K. (����������������������������������������������������������������������������������� 1998). A population study of ApoE genotype at the age of 85: Relation to dementia, cerebrovascular disease, and mortality. Journal of Neurology, Neurosurgery, and Psychia� try, 64(1), 37-43. Wachter, K. (2001). Biosocial opportunities for surveys. In National Research Council, Cells and surveys: Should biological measures be included in social science research? (pp. 329-336). Committee on Population, C.E. Finch, J. Vaupel, and K. Kinsella, Eds. Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press. Weinstein, M., and Willis, R. (2001). Stretching social surveys to include bioindicators: Pos- sibilities for the health and retirement study, experience from the Taiwan study of the elderly. In National Research Council, Cells and surveys: Should biological measures be included in social science research? (pp. 250-276). Committee on Population, C.E. Finch, J.W. Vaupel, and K. Kinsella, Eds. Commission on Behavioral and Social Sciences and Education. Washington, DC: National Academy Press.

Next: 5 An Overview of Biomarker Research from Community and Population-Based Studies on Aging--Jennifer R. Harris, Tara L. Gruenewald, and Teresa Seeman »
Biosocial Surveys Get This Book
×
Buy Paperback | $65.00 Buy Ebook | $49.99
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Biosocial Surveys analyzes the latest research on the increasing number of multipurpose household surveys that collect biological data along with the more familiar interviewer–respondent information. This book serves as a follow-up to the 2003 volume, Cells and Surveys: Should Biological Measures Be Included in Social Science Research? and asks these questions: What have the social sciences, especially demography, learned from those efforts and the greater interdisciplinary communication that has resulted from them? Which biological or genetic information has proven most useful to researchers? How can better models be developed to help integrate biological and social science information in ways that can broaden scientific understanding? This volume contains a collection of 17 papers by distinguished experts in demography, biology, economics, epidemiology, and survey methodology. It is an invaluable sourcebook for social and behavioral science researchers who are working with biosocial data.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!