Quality Life Year before Death · older from the 1986 National Mortality Followback Survey (NMFS)(n...

6
The Quality of Life in the Year before Death Harold R Lentzner, PhD, Richard Rothenberg, MD, Introduction Americans born in the late 1980s could expect to live into their 70s, and those reaching their 65th birthday at that same time could expect to live almost two additional decades.' With the expecta- tion of long life after age 65 has come the desire for accompanying good health, and, correspondingly, the fear of spend- ing one's final years in a state of severe disability and dependency. As was noted in the now decade-old Healthy People: the Surgeon General's Report on Health Promotion and Disease Prevention, "older Americans hope for a state of well-being which would allow them to perform at their highest functional capac- ity. ... Their greatest fear is of being useless, sick or unable to care for them- selves."2 Until recently, little has been known about the distribution of functionality in populations of older persons. Several re- cent studies have looked at predictors of health status among samples of the elderly3 8; however, none of these efforts has focused on the period preceding death. This period is important, in part, because those who live long lives and are vibrant until shortly before death may pro- vide the best possible example of "suc- cessful aging."9 In this study, we developed a simple multidimensional index to estimate ex- tremes in health status in the last 12 months of life among a nationally repre- sentative sample of persons who died in 1986. We also examined several demo- graphic, health, and social factors to iden- tify important correlates of being fully functional or, conversely, severely dis- abled in these final months. Elsie R Pamuk, PhD, Elaine P. Rhodenhiser, MPH, and Eve Powell-Gnner, PhD Methods National Mortality Followback Survey Sample For this analysis, we extracted all deaths of individuals aged 65 years and older from the 1986 National Mortality Followback Survey (NMFS) (n = 7586). The NMFS is a stratified random sample of approximately 1% of all deaths among US residents (excluding residents of Oregon, who were not sampled) aged 25 years and older that occurred in 1986 (n = 18 733). 10 This survey, conducted by the National Center for Health Statistics, collected detailed information on health status and health care provided in the last year of life and on aspects of lifestyle that are potential risk factors for disease and disability. Most information was collected from proxy respondents by mail question- naire, by telephone, or by personal inter- view. Information on the underlying cause of death and all associated causes came from the official death certificate. Because information for comorbid conditions (con- ditions that occurred or were prevalent at least 12 months before death and were not Harold R. Lentzner, Elsie R. Pamuk, and Richard Rothenberg are with the National Center for Chronic Disease Prevention and Health Promotion, and Elaine P. Rhodenhiser is with the National Center for Environmental Health and Injury Control, all at the Centers for Disease Control, Atlanta, Ga. Eve Powell- Griner is with the Department of Public Health Sciences, School of Public Health, University of Alabama at Birmingham, Bir- mingham, Ala. Requests for reprints should be sent to Harold R. Lentzner, PhD, 1600 Clifton Road, Mail Stop (K30), Atlanta, Ga. This paper was submitted to the Journal March 6, 1991, and accepted with revisions No- vember 12, 1991. American Journal of Public Health 1093 11" .,. .. A I mm

Transcript of Quality Life Year before Death · older from the 1986 National Mortality Followback Survey (NMFS)(n...

Page 1: Quality Life Year before Death · older from the 1986 National Mortality Followback Survey (NMFS)(n = 7586). TheNMFSis astratified randomsample ofapproximately 1%ofall deaths among

The Quality of Life in the Yearbefore Death

Harold R Lentzner, PhD,Richard Rothenberg, MD,

Introduction

Americans born in the late 1980scould expect to live into their 70s, andthose reaching their 65th birthday at thatsame time could expect to live almost twoadditional decades.' With the expecta-tion of long life after age 65 has come thedesire for accompanying good health,and, correspondingly, the fear of spend-ing one's final years in a state of severedisability and dependency. As was notedin the now decade-old Healthy People:the Surgeon General's Report on HealthPromotion and Disease Prevention,"older Americans hope for a state ofwell-being which would allow them toperform at their highest functional capac-ity. ... Their greatest fear is of beinguseless, sick or unable to care for them-selves."2

Until recently, little has been knownabout the distribution of functionality inpopulations of older persons. Several re-cent studies have looked at predictors ofhealth status among samples of theelderly3 8; however, none of these effortshas focused on the period precedingdeath. This period is important, in part,because those who live long lives and arevibrant until shortly before death may pro-vide the best possible example of "suc-cessful aging."9

In this study, we developed a simplemultidimensional index to estimate ex-tremes in health status in the last 12months of life among a nationally repre-sentative sample of persons who died in1986. We also examined several demo-graphic, health, and social factors to iden-tify important correlates of being fullyfunctional or, conversely, severely dis-abled in these final months.

Elsie R Pamuk, PhD, Elaine P. Rhodenhiser,MPH, and Eve Powell-Gnner, PhD

Methods

National Mortality FollowbackSurvey Sample

For this analysis, we extracted alldeaths of individuals aged 65 years andolder from the 1986 National MortalityFollowback Survey (NMFS) (n = 7586).The NMFS is a stratified random sampleof approximately 1% of all deaths amongUS residents (excluding residents ofOregon, who were not sampled) aged 25years and older that occurred in 1986(n = 18 733). 10 This survey, conducted bythe National Center for Health Statistics,collected detailed information on healthstatus and health care provided in the lastyear of life and on aspects of lifestyle thatare potential risk factors for disease anddisability. Most information was collectedfrom proxy respondents by mail question-naire, by telephone, or by personal inter-view. Information on the underlying causeof death and all associated causes camefrom the official death certificate. Becauseinformation for comorbid conditions (con-ditions that occurred or were prevalent atleast 12 months before death and were not

Harold R. Lentzner, Elsie R. Pamuk, andRichard Rothenberg are with the NationalCenter for Chronic Disease Prevention andHealth Promotion, and Elaine P. Rhodenhiseris with the National Center for EnvironmentalHealth and Injury Control, all at the Centersfor Disease Control, Atlanta, Ga. Eve Powell-Griner is with the Department of PublicHealth Sciences, School of Public Health,University of Alabama at Birmingham, Bir-mingham, Ala.

Requests for reprints should be sent toHarold R. Lentzner, PhD, 1600 Clifton Road,Mail Stop (K30), Atlanta, Ga.

This paper was submitted to the JournalMarch 6, 1991, and accepted with revisions No-vember 12, 1991.

American Journal of Public Health 1093

11" .,. .. A

I

mm

Page 2: Quality Life Year before Death · older from the 1986 National Mortality Followback Survey (NMFS)(n = 7586). TheNMFSis astratified randomsample ofapproximately 1%ofall deaths among

Lentzner et aL

the underlying cause of death) was un-available for a number of decedents, wehad a reduced sample (n = 5582) for thisportion of the study. However, the char-acteristics of this subsample closelymatched those of the full sample.

Creation ofan IndexWe examined a subset of responses

to questions asked on the NMFS. Proxyrespondents were asked if the decedenthad "received help or used special equip-ment" in the last year of life in five sep-arate activities of daily living (ADLs)-eating, using the toilet, dressing, walking,and bathing-and if so, over what periodof time. They were asked how often inthat same period the decedent had trou-ble understanding where he or she was,remembering what year it was, or recog-nizing family or good friends. The re-spondent was also asked how manynights the decedent spent in hospitals andnursing homes in the last year of life.

Results from all three sets ofquestionswere used to construct a simple additiveindex of health, and two groups of dece-dents with opposite profiles were identi-fied. We defined as "fully functional"those who during the last year of lifeneeded no assistance or special equipmentto carry out the five basic ADLs listedabove, who were lucid until the last fewhours or days before death, andwho spentless than a week in a hospital or nursinghome. At the other extreme were "severe-ly restricted" individuals, who needed helpor special equipment for 6 months or morein at least three of five ADLs and more

limited assistance in the remaining two,who exhibited at leastsome decline inmen-tal functioning, and who spent at least halfof their last year in a hospital or nursinghome.

Statistical Methods

The number and proportion ofdecedents who were fully functional intheir last year and of those who were se-verely restricted were estimated for threeage groups: 65 to 74 years, 75 to 84 years,and 85 years and older. Variances andconfidence intervals were computed byusing generalized parameters providedby the National Center for Health Statis-tics.11

We used separate logistic regressionmodels to estimate the likelihood of beingfully functional or severely restricted inthe last year. Decedents who fit these def-initions were compared with all other de-cedents; that is, the fully functional werecompared with all decedents experiencinggreater restrictions, and the severely re-strictedwere compared with all decedentsexperiencing fewer restrictions.

We calculated regression coeffi-cients, odds ratios, and variances by us-ing the raw sample reweighted to accountfor the stratified sampling design but notfor the effects of clustering. For both setsof models, the odds ratios reflect the rel-ative likelihood of being fully functionalor severely restricted, adjusted for singleyear of age, and in the presence of otherimportant factors.

ResultsCharacteristics ofthe Respondents

The relationship of the respondentsto the deceased varied according to thecharacteristics of the decedent. Althougha member of the immediate family wasmost often the proxy respondent, otherrelatives, friends, and acquaintances weremore likely to be proxies for female thanfor male decedents (24% vs 15%), forBlack than for White decedents (25% vs19%), and for those who died at age 85years or older (27%) than for those whodied at ages 75 to 84 years (20%o) or 65 to74 years (12%).

Components ofthe IndecThe proportion of respondents re-

quiring assistance with ADLs rose withage at death (Figure 1). For all age groups,decedents were least likely to need helpeating and most likely to need help bath-ing.

Among those who died between theages of 65 and 74, roughly 15% had hadtrouble knowing where they were, 13%remembering theyear, and about 10%o rec-ognizing family or good friends (Figure 2).These proportions increased with age; theoldest decedents were more than twice aslikely as the youngest to have cognitivelimitations.

Of those who died between the agesof 65 and 74, 7.2% spent more than half oftheir last year in a nursing home. Amongthose who died between the ages of75 and84, the corresponding figure was 17.4%.But of those who died at the age of 85years or older, 39.3% spent at least half oftheir last year in a health care facility.

Health Status and DemographicCharactenstics

Approximately 14% of all decedentswho died at age 65 years and older weredefined as fully functional in the last 12months of life. A slightly lower propor-tion, 10%, met our definition of severelyrestricted. As might be expected, the pro-portion of decedents at the definitional ex-tremes varied with age in very differentways. One fifth ofthosewho died betweenthe ages of 65 and 74 met our definition offully functional in the last 12 months, butthe percentage dropped to 6% among theoldest group; by contrast, only 3% of theyoungest group but 22% of those aged 85years and older were severely restricted.

Differences by sexwere pronounced.A smaller proportion ofwomen than menwere fully functional in the last 12 months

1094 American Journal of Public Health

_ Eating

65-74 - Using toilet

_ Dressing

WalkingcoX BathingZ)

X 75-84

85+

0 20 40 60 80Percentage

FIGURE1 a of deedents receMng help or using spedal equipment toperform afles of daillyMng In last year of ife, by age at death.

.l

August 1992, Vol. 82, No. 8

Page 3: Quality Life Year before Death · older from the 1986 National Mortality Followback Survey (NMFS)(n = 7586). TheNMFSis astratified randomsample ofapproximately 1%ofall deaths among

Quat ofmfe in the Year before Death

of their lives, and a larger proportion wereseverely restricted. This was true for de-cedents in all three age groups (Figure 3).For the oldest decedents, the differencesare substantial: roughly 10o ofthe menvs4% ofthewomen remained unrestricted inthe last year, whereas 13% of the men vs26% of the women met the definition ofseverely restricted.

We also examined health status byrace and marital status at death.We foundfew signifcant differences and noted noconsistent pattern across marital statussubgroups. Of those who died before theage of 85, Whites appeared more likelythan Blacks to be either fuly functional orseverely restricted, although the differ-ences were not always statistically signif-icant.

Health Status and Cause ofDeathIn general, a higher proportion of

those who died of acute myocardial in-farction (AMI) were fuly functional, anda lower proportion were severely re-stricted, thanwere thosewho died ofothercauses (other forms of heart disease, can-cer, cerebrovascular disease and all othercauses) (Table 1). In contrast, fewer than1 in 20 persons who died of cancer couldbe defined as fully functional, regardlessof age at death. However, a relativelysmall proportion of these decedents couldbe categorized-at the other extreme-asseverely restricted.

For most causes, a significantlyhigher proportion ofwomen than men atthe older ages were severely restricted inthe last year of life, but the sex-specificrelationships between cause and full func-tionality were less uniform. For those dy-ing of heart disease and of all causes otherthan heart disease, cancer, and stroke,women were less likely to be fully func-tional; for thosewho died of cancer, therewere no differences. Among persons dy-ing of cerebrovascular diseases, womenappeared more likely to be fully func-tional, although the differences were notlarge enough to reach statistical signifi-cance.

Health Status and PreexstingComo,bid Conditions

At all ages, decedents with a historyof stroke were less likely than most othersto be fully functional and more likely to beseverely restricted in the last year (Table2). However, decedents with conditionsother than stroke were not consistentlyless likely to be filly functional or morelikely to be severely restricted than weredecedents with none ofthese chronic con-

ditions. In most cases, lower proportionsofwomen than men were filly functionaland higher proportions were severely re-

stricted for similar types of preexistingconditions.

MultivariateAnalysisSex was highly predictive of health

status even after we controlled for age,race, marital status, cause of death, andsix preexisting conditions. In the last yearof life, women were about 40%o less likelythanmen to havebeen fully functional and70%o more likely to have been severelyrestricted (Table 3).

Decedents of races other than Whitewere less likelytobe either fully functional

or severely restricted, although the oddsratio for these races was significantly dif-ferent only for the model that examinedsevere restriction. Marital status at time ofdeath did not appear to be related to fullfunctionality. In contrast, being other thanmarried-widowed, divorced, separated,ornever married-doubled the odds ofbe-ing severely restricted in the last year oflife.

The underlying cause of death was

associated with the likelihood of beingfuly functional in the year before death.Those who died from AMI were far morelikely to have been fully functional thanthose who died from any other cause. De-cedents with a history of stroke or chronic

American Journal of Public Health 1095

50 -

* Understanding where they were9 Remembering what year it was

40 Recognizing family

e30-

0. 20-

10 i

0 rI65-74 75-84 85+

Age at Death (years)

FIGURE 2-PREnt of decedent having cognvedMlfculUes In last year of life, byage at death.

30 -

| Men S Women

25

20-

10

15

0.

65-74 75-84 85+ 65-74 75-84 85+Fully Functional Age at Death (years) Severely Restricted

FIGURE 3-Percenta of fully fn donal and esIcted dec by sexand age at death.

August 1992, Vol. 82, No. 8

Page 4: Quality Life Year before Death · older from the 1986 National Mortality Followback Survey (NMFS)(n = 7586). TheNMFSis astratified randomsample ofapproximately 1%ofall deaths among

Leutzner et al.

obstructive pulmonary disease, or with abistory of two or more preexisting condi-tions, were less likely to have been fullyfunctional than decedents with none ofthese conditions. Dying from heart dis-ease other thanAMI, dying from stroke or

"other causes," and having a history ofstroke or ofthree or more preexisting con-ditions were all associated with a greaterlikelihood of very poor functional statusthan the appropriate reference categories.

We used sex-specific models to eval-

uate the relative importance ofeach oftheremaining variables for men and women.In general, the direction and strength ofthe associations were comparable to thefirst set of models. Few differences wereapparent in the fully functional models,except that preexisting conditions seemedto be more important for men. Marital sta-tus appeared to be more strongly associ-ated with the likelihood of having beenseverely restricted inwomen than in men,whereas cause of death and disease his-torywere more important formen than forwomen. Women decedents with hyper-tension and no other preexisting conditionwere less likely to have been severely re-stricted than werewomen decedents withno preexisting conditions.

DiscussionOur results suggest that elderly dece-

dents are fully functional or severely re-stricted in the 12 months preceding theirdeath in roughly equal proportions, butthat the proportion varies with age. Usingprovisional mortality counts for 1990 (themost recent year for which data are avail-able) and applying the age- and sex-spe-cific prevalences for functionality, wewould estimate that about 216 000 elderlydecedents experienced little decline inphysical independence or mental aware-ness and few nights in health care facilitiesduring the last year of life.12 On the otherhand, we estimate that 169 000 elderly de-cedents suffered both mental and physicalimpainnent and were institutionalized forthe majority of their last 12 months of life.However, when we excluded the dece-dents who failed to reach average life ex-pectancy (whose deaths must be regardedas premature no matter what the circum-stances), we found that a much larger pro-portion of decedents required consider-able care. Of the oldest decedents-thoseaged 85 years and older at the time ofdeath-we estimate that about 100 000died with profound physical impairmentsand required assistance through much ofthe last year of life, whereas only about28 000 were fully functional.

As the oldest-aged population growsin the United States and in other agingsocieties, the number who are severelydisabled may be expected to grow, evenwith declining mortality rates, unless thepresent pattern changes. By theyear 2000,the number of persons aged 85 years andolder who will be severely restricted intheir final months may be expected to in-crease by 50%, given recent official pop-ulation projections.13 If health care costs

1096 American Journal of Public Health August 1992, Vol. 82, No. 8

Page 5: Quality Life Year before Death · older from the 1986 National Mortality Followback Survey (NMFS)(n = 7586). TheNMFSis astratified randomsample ofapproximately 1%ofall deaths among

Qualy of Ufe in the Year before Death

continue to escalate at present rates, thecosts of the care associated with their dis-abilities may be expectedto increase by amuch larger percentage.

Our results showing sex as a signifi-cant predictor of disability are consistentwith findings from recent research on pre-dictors of "healthy aging."4,5 Guralnikand Kaplan note that althoughmen tend tolive shorter lives than women, they areless likely to be functionally impaired.5Unfortunately, we cannot determinewhether other diseases are responsible fordifferences by sex in the quality of the lastyear of life without a more complete in-ventoryofprevalent chronic diseases thanis presently on the NMFS. For example,arthritis is a disease that frequentlyends insevere disability, and arthritic conditions,whether self-reported or physician-diag-nosed, are more common among olderwomen than among older men.14.15 Aftercontrolling for 13 disabling conditions (in-cluding arthritis), Verbrugge et al. foundthat women aged 55 years and older hadonly slightly higher levels ofdisability thandid men in that age group.3

In addition, our definition of"severe-ly restricted" maybe responsible forsomeof the apparent male advantage in func-tionality, because men are more likely tohave a lhving spouse than are women andare thus more likely to avoid institution-alization in the last year of life. However,when we examined the separate compo-nents of the index, we found that femaledecedents, besides having poorer resultsthan male decedents for ADLs and insti-tutional use, were also more likely to havehad cognitive impairments.

Our results also show that factorsother than sex are associated with qualityof life in the last year. Dying of anythingother than a heart attack sharply de-creased the likelihood of being fuly func-tional in the 12 months prior to death.Moreover, death from diseases or condi-tions with prolonged morbid processesthat are associated with both physical andmental deterioration, such as stroke anddiseases of the heart other than AMI, in-creased the likelihood that decedents ex-perienced restriction and dependency intheir last year of life. Our research sug-gests that, all other things being equal, theconsequence of replacing an earlier deathfromAMI with a later death from stroke isa greater likelihood of functional disabil-ity. This finding suggests that changes inthe cause-of-death structure will have animpact on the aggregate assessment ofthequality of life in the year before death.

The validity of proxy response willinfluence the results. Responses fromproxies are believed tobe mostvalidwhenthey are based on directly visiblephenomena-for example, functional ac-tivities, or habits or events such as hospi-talizations.16 The amount of time theproxy spends assisting the patient also in-fluences the proxy's assessment of func-tional impairment.17

To minimize these potential prob-lems of proxy response, we focused onthose decedents who were either sohealthy or so sick in the last year of lifethat their condition was a clear matter ofrecord. We evaluated the sensitivity oftheresults to changes in the definitional crite-ria by including adjacent categories in theanalysis. These changes in definitions re-sulted in larger numbers of decedents inboth the fully functional and the severelyrestricted groups and, as expected, some-what lessened the distinction between thetwo. However, the distinction's persis-tence suggested that definitional alterationwould produce only minor changes in theresults.

The NMFS obtained information onthe last year of life only. Although it isunlikely that personswhowere fully func-tional in their last year were partially orseverely restricted in the years preceding,the reverse may be true: restriction in thefinalyearmay follow a longperiod offunc-tionality. Even though the picture derivedfrom NMFS data is incomplete, we canbegin to develop profiles of "successfulagers"-men and women who lived intotheir ninth decade and were fully func-tional through most of their last year oflife-as well as profiles of those who arenot so fortunate. We know, for example,that only 25% of the women and 40To ofthe men who were successful agers hadany of the comorbid conditions identifiedin the NMFS. By contrast, of those whodied earlywith severe restrictions, 66% ofthewomen and more than 90% ofthe menhad at least one of the preexisting condi-tions. Moreover, personswho died beforeage 80 after a period of severe restrictionwere more likely to have died of causesother than AMI than were successful ag-ers.

American Journal of Public Health 1097August 1992, Vol. 82, No. 8

Page 6: Quality Life Year before Death · older from the 1986 National Mortality Followback Survey (NMFS)(n = 7586). TheNMFSis astratified randomsample ofapproximately 1%ofall deaths among

Ientrmr et aL

Avaluable addition to a health profileof successful agers would be the quantifi-cation of the benefits, if any, obtainedfrom a lifetime of healthy living or riskavoidance. The 1986NMFS requested in-formation from proxies on the decedent'slifetime smoking, drinking, dietary, andexercise habits and on the decedent'sweight. Concerns about data quality andresponse distribution made it difficult toevaluate the association between success-ful aging and a multidimensional healthyliving index. However,we found that suc-cessful agers (defined here as personswhodied at age 80 or older andwere fully func-tional) ofeithersexwere less likelythan allothers to have ever smoked, although theresults were significant for men only (datanot shown). There was also some evi-dence that successful agers were lesslikely than others either to have abusedalcohol or to have abstained from it, andthat women who were successful agerswere less likely than others to have beenseverely overweight or underweight.

These results are only suggestive. Amore fundamental question that cannot beanswered here is whether the life-stylechanges that improve the probability ofsurvival also improve the probability ofremaining healthy. Future researchshould be directed toward answering thisimportant question.

The concept of successful agingshould embody both longevity and vital-ity. That we are experiencing unprece-

dented gains in life expectancy is well rec-ognized. The quality of life in the 12months before death provides a poten-tially important measure of the success orfailure of those gains. The next NationalMortality Followback Survey, to be con-ducted in 1993, will provide an importantsecond point for a vector of the quality ofaging. O

References1. Healt, United States, 1989. Hyattsville,

Md: National Center for Health Statistics;1990.

2. Healthy People: 7he Surgeon General'sRepor1 on Health Prmotion and DiseasePrevention. Rockville, Md: US Dept ofHealth, Education, and Welfare, PublicHealth Service, Office ofthe Assistant Sec-retary for Health and the Surgeon General;1979, US Dept of Health, Education, andWelfare publication PHS 79-55071.

3. Verbrugge L, Lepkowsld JM, Imanaka Y.Comorbidity and its impact on disability.Mibank Q. 1990;67:450484.

4. Harris T, Kovar MG, Suzman R, Klein-manJC, Feldman JJ. Longitudinal study ofphysical ability in the oldest-old.AmJPub-lic Health 1989;79:698-702.

5. Guralnik JM, Kaplan GA. Predictors ofhealthy aging: prospective evidence fromthe Alameda County Study. Am J PublicHeakh 1989;79:703-708.

6. Guralnik JM, LaCroix AZ, Branch LG,Kasl SV, Wallace RB. Morbidity and dis-ability in older persons in the years prior todeath. Am J Public Health. 1991;81:443-447.

7. Guralnik JM. Prospects for the compres-sion of morbidity: The challenge posed byincreasing disability in the years pnor to

death. JAging Heakh 1991;3:138-154.8. Roos NP, Montgomery P, Roos LL.

Health care utilization in the years prior todeath. Milbank Q. 1987;65:231-254.

9. Fries JF. Aging, natural death, and thecompression of morbidity. NEngl JMed.1980;303:130-135.

10. Seeman I, Poe G, Powell-Griner E. Devel-opment methods and response character-istics: 1986 National Mortality FollowbackSurvey. Via Health Stat [21. In press.

11. Kapantais G, Powell-Griner E. Character-istics of Persons Dying from AIDS: Pre-liminary Data from the 1986 National Mor-tality Followback Survey. Advance datafrom vital and health stattis, no. 173.Hyattsville, Md: National Center forHealth Statistics; 1989.

12. Annual summary of births, marriages, di-vorces, and deaths: United States, 1990.Monthly vital statistics repot, voL 39, no.13. Hyattsville, Md: National Center forHealth Statistics; 1991.

13. Wade A. Social SecurityArea PopulationProjections: 1988. Baltimore, Md: SocialSecurity Administration; June 1988.

14. Verbrugge L. Arthritis and disability inolder adults. Presented at the annual meet-ing of the Population Association ofAmer-ica; April 21-23, 1988; New Orleans, La.

15. Branch LG, Meyers AR. Assessing phys-ical function in the elderly. Clin GenatrMed 1987;3:29-51.

16. Epstein AM, Hall JA, Tognetti J, Son LH,Conant L. Using proxies to evaluate qual-ity of life. Med Care. 1989;27:S91-S98.

17. Rodgers WL, Herzog AR. The conse-quences ofacceptingproxyrespondents ontotal survey error for elderly populations.In: Fowler FJ, ed. Conference Proceed-ings: Health Swvey Research Medtds.Washington, DC: National Center forHealth Statistics Research; 1990.

1098 American Journal of Public Health August 1992, Vol. 82, No. 8