rabbitt (2008) j gerontol psychol sci death dropout longitudin measures cognitive change

Death, Dropout, and Longitudinal Measurements ofCognitive Change in Old Age

Patrick Rabbitt,1 Mary Lunn,2 and Danny Wong2

1Department of Experimental Psychology, University of Oxford, England and University of W. Australia.2Department of Statistics, University of Oxford, England.

During a 20-year longitudinal study of cognitive change in old age 2,342 of 5,842 participants died and 3,204dropped out. To study cognitive change as death approaches, we grouped participants by survival, death, dropout,or dropout followed by death. Linear mixed-effects pattern-mixture models compared rates of cognitive changebefore death and dropout from four quadrennial administrations of tests of fluid intelligence, vocabulary, andverbal learning. After we took into account the significant effects of age, gender, demographics, and recruitmentcohorts, we found that approach to death and dropout caused strikingly similar reductions in mean test scoresand amounts of practice gains between successive quadrennial testing sessions. Participants who neither droppedout nor died showed significant but slight cognitive declines. These analyses illustrate how neglect of dropoutmiscalculates effects of death, of worsening health, and of all other factors affecting rates of cognitive change.

Key Words: Cognitive changeDeathDropout.

M ANY studies of changes preceding death in old age havefound that individuals who are between 18 months and11 years from death score less well on tests of mental abilitiesthan survivors do (e.g., Johansson & Berg, 1989; Lieberman,1965; Palmore & Cleveland, 1976; Rabbitt, Watson, Donlanet al. 2002; Reimanis & Green, 1971; Riegel & Riegel, 1972;Riegel, Riegel, & Meyer, 1967; Siegler & Botwinick, 1979;Small, Fratiglioni, von Strauss, & Backman, 2002). The typicalmethodology has been to assess participants only once and tocompare the scores of survivors and decedents at a later,arbitrary, census date. This underestimates the effects of death,because younger decedents are compared against youngersurvivors who will survive longer beyond the census datewhereas elderly decedents are compared against elderly sur-vivors who are likely to die soon after census (Rabbitt, Lunn, &Wong, 2005; Rabbitt, Watson, Donlan, Bent, & McInnes,1994). This explains the otherwise paradoxical findings thatdifferences in ability between deceased persons and survivorsare larger in younger than in older samples (e.g., Riegel &Riegel; Riegel et al.).

Longitudinal studies in which participants are repeatedlyassessed over many years avoid these problems but encounterother methodological difficulties. Participants improve withpractice as a result of repeated testing (Rabbitt, Diggle,Holland, McInnes, Bent, et al., 2004); recruitment cohortsmay markedly differ in ability; and participants typically dropout of studies because of deteriorating health and survivorsbecome increasingly elite and able (Lachman, Lachman, &Taylor, 1982; Rabbitt, Watson, Donlan, Bent, & McInnes,1994; Schaie, Labouvie, & Barrett, 1973). Thus, if scores ofdecedents are compared against those of all survivors, includingless able dropouts, then the effects of approaching death areunderestimated (Rabbitt et al., 2005). Demographic factorsmust be taken into consideration. Women, the socioeconom-ically advantaged, and the most able live longer, so deathsselectively alter sample composition (Hart et al., 2003). Moreand less advantaged individuals tend to die from different

causes and so experience different patterns and rates of terminaldeclines in health and cognition (Nagi & Stockwell, 1973;Pincus, Callahan, & Birkhauser, 1987; Snowden, Ostwald,Kane, & Keenan, 1989). Differences in age must be consideredbecause, independent of approach to death, age acceleratesthe rate of change in performance over time; the effects ofpathologies, and so of approaching death, may also alter as ageadvances (Rabbitt, Diggle, Holland, & McInnes, 2004).

Data from the University of Manchester LongitudinalStudy, described in detail elsewhere (Rabbitt, Diggle, Holland,McInnes, Bent, et al., 2004), allowed us to make analyses ofrates of change preceding death and dropout after the effectsof practice, age, cohort effects, and demographics had also beenconsidered.

METHODS

Participants and ProcedureA panel of 5,842 volunteers, that is, 2,615 residents of Greater

Manchester and 3,227 residents of Newcastle-upon-Tyne,United Kingdom, were all sufficiently healthy and motivated totravel independently to the University of Newcastle-upon-Tyneor the University of Manchester, where they were givencognitive tests in groups of 10 to 20. There were 1,711 menaged between 49 and 93 years (M 65.6, SD 7.7) and 4,131women aged between 49 and 92 years (M 64.4, SD 7.8).They were each paid 5 (UK) per session to cover expenses. Asearch by Her Majestys Registry Office UK obtained exactdates and proximate causes for all 2,342 deaths between 1983,when the study began, and the census date, July 2004. Of 3,204dropouts, 1,208 also died before the census date. Most dropoutsonly appeared as failures to answer invitations for testingand so could only be dated from the last session attended.Consequently, to compare rates of changes preceding dropoutand death, we had to use the same dating. We could notcompare changes preceding dropouts from different causes

Journal of Gerontology: PSYCHOLOGICAL SCIENCES Copyright 2008 by The Gerontological Society of America2008, Vol. 63B, No. 5, P271P278

P271

because many participants did not reveal them. However, aprevious survey found that, although most respondents before1993 cited poor health as their reason for dropping out, manycited positive reasons such as taking up new employment(Rabbitt et al., 1994). Thus, data pooled over all dropoutsunderestimates the effects of illness and increasing frailty. Theremaining 1,996 participants did not drop out before July 2003and also survived the July 2004 census of deaths. During thecourse of the study from 1983 to 2004, 7 participants wereidentified as suffering from dementias, either by death certifi-cates or information from relatives, and we excluded these.Since 2004, all survivors have been screened every 6 monthson the Mini-Mental State Examination and other assessmentprotocols for dementias, and so far nine cases have been iden-tified. We have also retrospectively excluded these. Although itis impossible to be certain that no individuals contributing datadid not also suffer from dementia, the latter figure suggests thatthe incidence was sufficiently low as to have hardly any effecton the comparisons described.

MaterialsSome of us have given details of the entire study elsewhere

(Rabbitt, Diggle, Holland, McInnes, Bent, et al., 2004). Thedata analyzed here are from tests of general fluid intelligence,that is, the Heim (1970) AH4-1 intelligence test; of vocabulary,that is, the Raven (1965) Mill Hill B Vocabulary Test; andcumulative verbal learning, that is, the Cumulative VerbalLearning task (CVL task) administered on four successiveoccasions at 4-year intervals between 19831985 and 2003.The AH4-1 test consists of 64 logic, verbal comparisons, andarithmetic problems. Scores are the percentages of correctanswers given within 10 minutes. The Mill Hill B VocabularyTest requires correct definitions for each of 34 words with notime limit. For the CVL task, 15 three-syllable words matchedfor frequency (1/100,000) and concreteness are projected, oneat a time, on a screen at a rate of 1/1.5 s. Participants recall asmany words as possible and the words are then shown threetimes in different random orders and recalled without sight ofprevious attempts. Scores are percentages of correct answers.

Levels of socioeconomic advantage (SEA) are indexed bythe Office of Population Censuses and Surveys (1980) in theirclassification of occupational categories. These categories areas follows: C1, made up of professionals such as doctors,lawyers, senior managers, and academics; C2, which consists ofother professionals such as schoolteachers, junior managers,and pharmacists; C3N, made up of skilled nonmanual workerssuch as secretaries and clerical workers; C3M, consisting ofskilled manual workers such as plumbers, craftsmen, joiners,fitters, and machinists; C4, made up of nonskilled, nonmanualworkers such as security guards; and C5, which consists ofnonskilled manual workers such as cleaners. We categorizedthose participants who did not reveal occupations as non-responders (labeled NR). We include SEAs, city of residence,gender, and recruitment cohort in all analyses.

We divided participants into 11 groups according to theirhistories of survival, dropout, and death, logged with respect tothe four quadrennial assessment time points: T1, T2, T3, andT4. The groups, labeled as D (for death), W (for withdrawal), orWD (for withdrawal followed by death), are as follows: D1completed T1 but died before T2; D2 completed T1 and T2 butdied before T3; D3 completed T1, T2, and T3 but died beforeT4; D4 completed T1, T2, T3, and T4 but died before the closeof the census of deaths in 2004; WD1 completed T1 butwithdrew before T2 and also died before the 2004 census; WD2completed T1 and T2 but withdrew and then died before the2004 census; WD3 completed T1, T2, and T3 but withdrewand then died before the 2004 census; W1 completed T1 andwithdrew before T2 but survived the 2004 census; W2 com-pleted T1 and T2, withdrew before T3, but survived the 2004census; W3 completed T1, T2, and T3, withdrew before T4, butsurvived the 2004 census. Group C was a control group whosemembers completed all four assessments and survived the 2004census.

The average raw scores for each of these death and dropoutgroups at each of the four testing occasions, that is, T1, T2, T3,and T4, are plotted for the AH4-1 test in Figure 1, for the CVLtask in Figure 2, and for the Mill Hill B Vocabulary Test in

Figure 1. Profiles of percentage scores on the Heim AH4-1 test(AH41) of fluid intelligence across testing sessions for survivor (C),death (D), and dropout (W) groups.

Figure 2. Profiles of percentage scores on the Cumulative VerbalLearning task across testing sessions for survivor (C), death (D), anddropout (W) groups.

RABBITT ET AL.P272

Figure 3. Note that we have not taken effects of differences inage, recruitment cohort, and demographics into consideration.Although these means illustrate the logic of grouping individ-uals in this way, they only provide very approximate indica-tions of specific comparisons between data points. Results ofcomparisons between these means, after age, cohort, anddemographics have been considered, are described in thefollowing paragraphs.

Figures 1, 2, and 3 show that, for all tests, and most clearlyfor the AH4-1, there is a clear vertical separation between mostgroups from Group C through Group WD3. We exploit this inthe pattern-mixture mixed-effects model by introducing anindicator for the group to which the individual belongs.

RESULTSTable 1 shows details of the demographic and age categories,

and death and withdrawal categories, and mean scores of thesesubgroups on the AH4-1 test, the Mill Hill B Vocabulary Test,and the CVL task.

Analyses

Methodology. Our analyses were based on the modelsdescribed by some of us elsewhere (Rabbitt, Diggle, Smith,Holland, & McInnes, 2001; Rabbitt, Diggle, Holland, McInnes,Bent, et al., 2004; Rabbitt et al., 2005). We consider age,gender, socioeconomic status, and whether participants aretaking the test for the first, second, third, or fourth time (thepractice effect). We also include cities of residence and years ofrecruitment of cohorts to adjust for unidentified confoundingfactors. The model can be considered in two parts: a model forthe average response over time for a subject with given valuesof all explanatory variables, and a model for the randomvariation about the main response.

The aim is to determine the effect of imminent death ona participants cognitive performance. However, some partic-ipants who withdraw survive beyond the census date whereasothers do not, giving three mechanisms by which they fail to

complete the tests: death, withdrawal, and withdrawal followedby death. The statistical method determines whether or notthese three mechanisms have similar effects on the test scores.

There are at least two possible approaches to handlingmissing data. One uses a selection model, and the other usesa pattern-mixture model. We are interested in a retrospectiveanalysis of the effects of death and dropout, and so a pattern-mixture model is an ideal tool. The probability density functionis factored as f(y, d) f(yjd)f(d), where y represents the responseand d represents the dropout data. We will be interested in thefirst factor on the right-hand side of the equation, whichrepresents the response conditional on the dropout pattern. Inthis data set, dropouts occur between test sessions or betweenthe last session and the end of the study, but they are also ofthree types: death, dropout followed by death, and dropout withsurvival after census. We can thus classify volunteers intogroups by dropout type or pattern. We model the conditionaldensity as we did in Rabbitt and colleagues (2001), that is,

Yij lij Ai Bixij1 Eij;where the mean value of Yij is given by

lij b0 b1xij1 b2xij 2 . . . bpxijp c2gi 2 . . . crgir:Index i is the identifier of the volunteer, index j gives the

occasion of the test (j 1, . . ., 4), and xijk denotes the values ofthe explanatory variables, where k runs from 1 to p and gim is an

Table 1. Means and Standard Deviations of Scores on

Intelligence, Verbal Learning, and Vocabulary Tests for

Age and Demographic Groups

Subset M(SD) AH4-1 M(SD) CVL M(SD) MHB

Manchester (2,615) 49.6 (17.7) 71.6 (13.9) 54.8 (17.9)

Newcastle (3,227) 47.5 (17.3) 67.5 (14.8) 45.2 (17.9)

Men (1,711) 50.9 (17.7) 66.2 (15.1) 52.1 (18.4)

Women (4,131) 47.4 (17.4) 70.4 (14.2) 48.5 (18.5)

Oclass C1 (261) 61.1 (14.0) 72.2 (13.4) 63.3 (15.8)

Oclass C2 (1,849) 56.2 (16.6) 73.1 (13.4) 58.3 (16.6)

Oclass C3(NM)(2,061) 48.4 (14.9) 70.1 (13.6) 48.1 (16.2)

Oclass C3M (771) 39.2 (15.8) 63.0 (14.4) 40.6 (16.7)

Oclass C4 (433) 34.4 (14.0) 61.7 (14.7) 35.3 (16.3)

Oclass C5 (47) 31.0 (16.6) 55.5 (17.4) 29.9 (16.7)

Oclass Na (420) 40.5 (18.3) 62.6 (18.0) 42.7 (20.1)

Age 4960 (1,347) 55.4 (17.1) 75.5 (12.9) 51.7 (18.0)

Age 6170 (2,849) 49.5 (16.6) 69.7 (13.8) 49.4 (18.2)

Age 7180 (1,477) 41.8 (16.5) 63.5 (14.7) 48.1 (19.4)

Age 80 (169) 33.7 (15.5) 57.5 (16.7) 47.8 (20.0)Pattern

C (1,504) 56.3 (16.2) 74.9 (12.1) 53.0 (17.4)

D1 (365) 43.0 (17.4) 63.4 (16.8) 48.3 (19.1)

D2 (408) 46.7 (16.6) 65.3 (14.5) 50.8 (19.2)

D3 (246) 49.9 (14.8) 70.1 (12.7) 52.8 (18.7)

D4 (115) 53.6 (16.3) 70.2 (12.9) 50.4 (18.8)

WD1 (745) 39.4 (17.5) 62.5 (15.9) 45.2 (19.5)

WD2 (354) 42.9 (15.4) 63.2 (14.8) 47.0 (19.4)

WD3 (109) 48.7 (16.0) 67.9 (14.1) 50.5 (19.5)

W1 (1,013) 45.2 (17.1) 69.6 (14.3) 45.6 (18.2)

W2 (595) 49.8 (16.8) 71.7 (13.2) 50.4 (17.8)

W3 (388) 51.4 (15.6) 71.4 (13.3) 52.9 (16.6)

Note: AH4-1 Heim AH4-1 test of fluid intelligence; CVL CumulativeVerbal Learning task; MHB Mill Hill B Vocabulary Test; Oclass occupa-tional class (see text for details on the C designations). For patterns, C control, D death, and W withdrawal (see text for particular designations).

Figure 3. Profiles of percentage scores on the Mill Hill BVocabulary Test (MHB) across testing sessions for survivor (C), death(D), and dropout (W) groups.

DEATH, DROPOUT, AND COGNITIVE CHANGE P273

indicator, a value 1 denoting to which of r type or patterngroups the ith volunteer belongs. Only one of these indicators isnonzero. The first two variables xij1 and xij2 are related to ageand age squared, with a value of zero corresponding to age 49,which is close to the mean age.

As in Rabbitt and colleagues (2001), the random effects(Level 2) are given by Ai Bixij1 , where (Ai, Bi) are bivariatenormal with a mean of zero and covariance matrix

r2

A rABrAB r 2B

and are independent of the Level 1 error, Eij, which fol-lows a univariate normal distribution with a mean of zero andvariance rE

2.This allows us to quantify the mean effect of death or

withdrawal, including also the test occasion by which this hasoccurred. It also allows us to make a direct comparison of theeffects of death and of withdrawal, as we shall see.

We have also considered interactions between the variousexplanatory variables and these are included in the final models.

Analysis of AH4-1 intelligence test scores. To illustrategeneral trends common to all tasks, we give analyses of AH4-1scores in detail. Scores for vocabulary and cumulative learningare jointly discussed in the subsequent text. Table 2 showsresults from a linear mixed-effects pattern-mixture model com-paring percentages of correct unadjusted AH4-1 scores for thedeath and dropout groups after the effects of age, gender,occupational category, city of residence, recruitment cohort,and practice are considered. Age was centered at 49, the lowestrecorded in the study. Variance between individuals hasbeen modeled with random effects because these are longitu-dinal data and measurements from the same individual arecorrelated.

Significant linear and quadratic terms show an acceleratingdecline in mean scores with increasing age. Because there is noAge3 Death or Age3 Dropout Group interaction, there is noevidence that the cognitive effects of death or dropout differwith the ages at which they occur. There is no interactionbetween group and any of the variables of age, gender, city,occupational class, or cohort (entry year), but there is asignificant constant effect of group at each session time, as seenin the plot of the raw data. (It is not possible to look for aninteraction with session time because these regression coeffi-cients would be unidentifiable.)

There are, however, some significant interactions of age withpractice gains. The interaction between age and the differencebetween scores at T1 and T2 is not significant, but theinteractions between age and the T3 versus T1 difference (p.0086) and between age and the T4 versus T1 difference (p .0089) are significant. In survivors, deceased individuals, anddropouts alike, the effect of age on test scores increases withthe interval over which it is measured. Overall, men (M 50.9,SD 17.7) scored higher than women did (M 47.4, SD 17.4). The significant Age 3 Sex interaction shows that evenwhen longevity has been taken into account, women declineless as they age. Mancunians or persons from GreaterManchester (M 49.6, SD 17.7) score higher thanNovocastrians or persons from Newcastle (M 47.5, SD 17.3). Scores range from 61.1, SD 14.0, for C1 (the mostadvantaged SEA group), to 31.0, SD 16.6, for SEA C5. Thisis reflected in the main model, shown in Table 2. There, we cansee that the mean score of C5 is over 14% lower than that ofC3 (both C3N and C3M), the baseline class, and that of C1 is9.5% higher than C3 (C3N and C3M). There are significantdifferences between recruitment waves with performance forentry years 1990 and 1992 that are greater than those for 1985(baseline). There is a highly significant practice gain between

Table 2. Model for Analysis of Scores on the AH4-1

Intelligence Test

Item CE SE df t p

(Intercept) 53.63 0.75 6732 70.68 ,.0001Age 0.77 0.04 6732 19.16 ,.0001Age2 0.014 0.00 6732 5.78 ,.0001Gender 2.25 0.44 5814 5.09 ,.0001City 1.42 0.66 5814 2.13 .03Oclass C1 9.58 0.96 5814 9.91 ,.0001Oclass C2 6.14 0.46 5814 13.26 ,.0001Oclass C3M 8.56 0.60 5814 14.09 ,.0001Oclass C4 11.96 0.74 5814 16.06 ,.0001Oclass C5 14.85 2.07 5814 7.14 ,.0001Oclass Na 5.42 0.79 5814 6.78 ,.0001T2 3.71 0.18 6732 20.43 ,.0001T3 5.54 0.29 6732 19.08 ,.0001T4 7.22 0.52 6732 13.78 ,.0001

Entry year

1982 2.49 1.73 5814 0.28 .7737

1983 1.86 0.89 5814 2.09 .03651984 1.16 0.85 5814 1.36 .17131986 0.27 0.68 5814 0.40 .68841987 1.59 1.10 5814 1.43 .15021988 0.55 0.74 5814 0.74 .45521990 9.28 2.79 5814 3.32 .0009

1991 1.86 0.86 5814 2.16 .0305

1992 3.08 0.82 5814 3.72 .0002

Pattern

D1 7.22 0.87 5814 8.21 ,.0001D2 5.03 0.79 5814 6.31 ,.0001D3 3.71 0.96 5814 3.86 .0001D4 0.41 1.34 5814 0.31 .7569

WD1 9.93 0.69 5814 14.28 ,.0001WD2 7.24 0.85 5814 8.47 ,.0001WD3 3.43 1.38 5814 2.48 .0129W1 8.15 0.60 5814 13.45 ,.0001W2 4.71 0.68 5814 6.85 ,.0001W3 2.46 0.80 5814 3.07 .0021

Interactions

Age 3 Gender 0.08 0.03 6732 2.66 .0076Age 3 T2 0.04 0.02 6732 1.82 .0686Age 3 T3 0.12 0.04 6732 2.62 .0086Age 3 T4 0.20 0.07 6732 2.61 .0089

Residual standard deviations: Levels 1 and 2

Level SD correlation with intercept

2 Intercept 13.086

2 Age 0.389 0.05

1 Residual 5.467

Note: CE coefficient; T1T4 Times 14; Oclass occupational class(see text for details on the C designations); Oclass Na occupational classnot specified. For patterns, D death and W withdrawal (see text for partic-ular designations).

RABBITT ET AL.P274

T1 and T2 followed by rather less substantial gains (less than2%) between T2 and T3 and between T3 and T4.

After we consider these effects and interactions, then we com-pare the death, withdrawal, or death followed by withdrawalgroups against those for the control group of survivors,Group C. Table 3 compares average scores throughout the studybetween Group C survivors and all others by using t valuescomputed from the main model. Group C survivors scorehigher than any of the members of the death and withdrawalgroups except Group D4, whose members also completed allfour assessments but then died within the 12 months before thedeath census in July 2004. Table 3 shows comparisons by two-tailed t tests between all other death and withdrawal groups.After applying a Bonferroni correction for multiple testing, wetake the threshold level for significance for these multiplecomparisons as p .001.

Comparing the effects of death and dropout. Specificcomparisons using t tests (calculated from the main model)compared differences in the amounts of practice gains betweensurvivors and death and withdrawal groups. After correction thethreshold for significance is p .001. Persons in Group D2completed assessments at T1 and T2 but then died before T3.From the raw data their nonsignificant gain in average scoresbetween T1 and T2 was 0.03 points. Again from the raw data,members of Group C, who completed all four assessments andsurvived the census date, shows a gain of 1.34 points betweentest sessions T1 and T2. From the main model, over all testsessions, we see that Group D2 and Group C have significantlydifferent mean scores (5%), p , .0001. Similarly, Group D1and Group C have significantly different scores (7%), p ,.0001, with a rather larger loss of score than Group D2, aswould be expected from their more imminent time of death.Group D3 again has a significant loss as compared with GroupC but with a smaller mean difference than those of D1 and D2.

Similar results can be seen for comparisons between GroupC and (dropout and survivor) Groups W1 through W3. Rathermore pronounced results of a similar nature are also seen withGroups WD1 through WD3 (participants who dropped out andsubsequently died).

Both from the raw data (Figure 1) and from the main model(Table 2), we see that death, dropout, and dropout with sub-sequent death persons do indeed model in much the same wayas do overall survivors who did not withdraw.

Analyses for cumulative verbal learning and vocabulary. Tables 4 and 5 and Tables 6 and 7 show the same analyses forcorrect percentage scores on the CVL task and Mill Hill BVocabulary Test, respectively.

For the CVL task, significant linear and quadratic age termsshow that mean decline accelerates as calendar age increasesover all survival, dropout, and death groups. In the Mill Hill BVocabulary Test, the mean effects of age and of age squared arenot significant. In both tests, because age does not significantlyinteract with death or withdrawal group membership, there is

Table 4. Model for Analysis of Scores on the Cumulative

Verbal Learning Task

Item CE SE df t p

(Intercept) 68.96 0.70 5603 97.93 ,.0001Age 0.82 0.04 5603 19.06 ,.0001Age2 0.01 0.01 5603 3.95 .0001Gender 4.18 0.46 4774 10.28 ,.0001City 0.53 0.76 4774 0.70 .4824Oclass C1 3.14 0.90 4774 3.48 .0005

Oclass C2 2.93 0.41 4774 7.12 ,.0001Oclass C3M 4.59 0.55 4774 8.23 ,.0001Oclass C4 6.09 0.68 4774 8.87 ,.0001Oclass C5 11.44 1.95 4774 5.85 ,.0001Oclass Na 3.17 0.93 4774 3.38 .0007T2 0.56 0.24 5603 2.32 .0202T3 4.79 0.35 5603 13.57 ,.0001T4 0.21 0.60 5603 0.35 .7237Entry year

1982 1.77 1.53 4774 1.15 .24901983 5.68 0.87 4774 6.52 ,.00011984 1.23 0.85 4774 1.45 .14681986 1.41 0.59 4774 2.37 .01761987 0.59 0.98 4774 0.60 .54791988 0.48 0.71 4774 0.68 .4957

1991 0.62 0.97 4774 0.63 .5237

1992 0.75 1.49 4774 0.50 .6138

Pattern

D1 5.25 0.93 4774 5.64 ,.0001D2 4.56 0.77 4774 5.91 ,.0001D3 0.78 0.92 4774 0.84 .4004D4 0.17 1.18 4774 0.14 .8811WD1 7.05 0.68 4774 10.34 ,.0001WD2 6.06 0.75 4774 8.06 ,.0001WD3 2.40 1.16 4774 2.06 .0391W1 4.78 0.58 4774 8.22 ,.0001W2 2.56 0.60 4774 4.24 ,.0001W3 1.58 0.65 4774 2.41 .0157

Interactions

Age 3 City 0.36 0.03 5603 10.41 ,.0001Age 3 T2 0.08 0.03 5603 2.46 .0137Age 3 T3 0.18 0.05 5603 3.25 .0012Age 3 T4 0.36 0.09 5603 3.92 .0001



2 Intercept 10.607

2 Age 0.444 0.533

1 Residual 7.433


Table 3. Specific Comparisons of Mean AH4-1 Intelligence Test

Scores Between Death and Dropout Groups

AH4-1 D1 D2 D3 D4 W1 W2 W3 WD1 WD2 WD3

C .0001 .0001 .0001 .0001 .0001 .0001 .0001

D1 .0001 .0001 .001

D2 .0002

D3

D4

W1 .0001 .0001 .0009

W2

W3

WD1 .0001

Note: Values of t statistics are derived from the main model. After correc-

tion, p , .001 is taken as the threshold for significance. C control group;D death and W withdrawal (see text for particular designations).


no evidence that approaching death causes more rapid declinein older than in younger participants. Women score higher thanmen on the CVL task, but a similar result on the Mill Hill BVocabulary Test has only a 5% significance level. On bothtasks there are no significant Sex 3 Age interactions and sothere is no evidence that men and women decline at differentrates. Mancunians score significantly higher than Novocastrianson the Mill Hill B Vocabulary Test, but persons from thedifferent cities do not differ on the CVL task. Greater SEA(occupational class) is associated with better overall perfor-mance but there is no Occupational Class 3 Test Sessionsinteraction and so no evidence that occupational class affectsrates of decline. On both tasks, members of Group C, whocontinued and survived the census date, have a mean score thatis higher than that of all other groups. Differences in meangains again reflect the time to death or withdrawal.

Tables 5 and 7 show the results when t values from the mainmodels for the CVL task and the Mill Hill B Vocabulary Testare used to compare overall scores between all death or with-drawal groups throughout the study. The corrected threshold forsignificance is p .001.

For the CVL task, Tables 4 and 5 show that overall averagescores for the D3 and D4 groups are significantly higher thanfor the D1 and D2 groups and that the D2 group scores higherthan the D1 group, although this last item is not significant atthe corrected level of p .001. In other words, mean declines inverbal learning scores accelerate as death approaches. There aresimilar graded effects of dropout, with scores for the W3 groupbeing higher than those for the W2 or W1 group and scores forthe W2 group being higher than those for W1. The D1 and theW1 groups do not differ, suggesting that over this brief intervalthe effects of impending withdrawal are as severe as those ofimpending death. For the Mill Hill B Vocabulary Test, Table 6shows that the mean differences follow similar patterns. Evenscores on a test of crystallized intelligence, production vocabu-lary, fall as death approaches. There are no differences betweenthose groups who drop out and then die shortly thereafter andthose who die without first dropping out.

GENERAL DISCUSSIONThese analyses replicate the main findings of Rabbitt and

colleagues (2005) for AH4-1 intelligence test scores on a much

larger sample, extend and compare these to the CVL task andproduction vocabulary (Mill Hill B Vocabulary Test), andexamine the time courses of changes in greater detail, con-cluding that mean scores on all analyzed cognitive tests declineaccording to nearness of death or dropout.

Time Courses of Effects of Death and DropoutThere are significant declines up to 8 years preceding death,

but participants who survived for more than 8 years after their

Table 6. Model for Analysis of Scores on the Mill Hill B

Vocabulary Test

Item CE SE df t p

(Intercept) 57.88 0.82 6797 70.51 ,.0001Age 0.01 0.04 6797 0.02 0.97

Age2 0.01 0.01 6797 0.87 .3789Gender 1.02 0.47 5837 2.12 .0339City 6.08 0.72 5837 8.48 ,.0001Oclass C1 12.67 1.03 5837 12.24 ,.0001Oclass C2 8.91 0.49 5837 18.06 ,.0001Oclass C3M 7.36 0.65 5837 11.21 ,.0001Oclass C4 11.69 0.80 5837 14.46 ,.0001Oclass C5 16.88 2.26 5837 7.46 ,.0001Oclass Na 3.57 0.86 5837 4.12 ,.0001T2 0.13 0.26 6797 0.51 .6039

T3 1.40 0.37 6797 3.70 .0002T4 3.24 0.66 6797 4.89 ,.0001

Entry year

1982 1.82 1.83 5837 0.99 .3210

1983 0.07 0.95 5837 0.08 .9331

1984 2.36 0.90 5837 2.59 .00941986 1.81 0.74 5837 2.45 .01411987 3.75 1.18 5837 3.18 .00151988 0.12 0.80 5837 0.15 .87511990 1.05 2.94 5837 0.35 .7209

1991 3.34 0.93 5837 3.56 .00041992 3.06 0.90 5837 3.38 .0007

Pattern

D1 5.41 0.98 5837 5.51 ,.0001D2 3.57 0.85 5837 4.20 ,.0001D3 2.31 1.01 5837 2.27 .0227D4 1.02 0.73 5837 1.39 .4652WD1 7.90 0.76 5837 10.29 ,.0001WD2 6.09 0.92 5837 6.61 ,.0001WD3 1.78 1.47 5837 1.21 .2239W1 7.26 0.65 5837 11.09 ,.0001W2 4.67 0.72 5837 6.48 ,.0001W3 1.73 0.84 5837 2.05 .0402

Interaction

Age 3 T2 0.01 0.03 6797 0.41 .6755Age 3 T3 0.04 0.06 6797 0.72 .4687Age 3 T4 0.31 0.09 6797 3.25 .0011



2 Intercept 13.513

2 Age 0.097 0.845

1 Residual 9.163


Table 5. Specific Comparisons of Mean CVL Task Scores

Between Death and Dropout Groups

CVL D1 D2 D3 D4 W1 W2 W3 WD1 WD2 WD3

C .0001 .0001 .0001 .0001 .0001 .0001

D1 .0002 .0004 .0066 .0003

D2 .0005 .0009 .0007

D3

D4

W1 .0007 .0001

W2

W3

WD1 .0002

WD2


tion, p , .001 is taken as the threshold for significance. CVL task Cumula-tive Verbal Learning task; C control group; D death and W withdrawal(see text for particular designations).

RABBITT ET AL.P276

first assessment (T1) performed almost as well as those whosurvived the final testing session. Because there is no Age 3Death or Age 3 Dropout interaction, there is no evidence, inthis study of atypically healthy participants, of change in thesizes or the time courses of the cognitive effects of approachingdeath between the ages of 49 and 93 years.

A new finding is that amounts of decline preceding dropoutclosely resemble those before death in all tests. Because onlysome participants gave reasons for dropout, we could notcompare the decline of individuals who dropped out for differ-ent reasons. However, some of us (Rabbitt et al., 1994) foundthat the group of dropouts who gave reasons for withdrawalincluded relatively healthy and able people who withdrew forreasons such as employment. Because these analyses includesuch robust dropouts, the actual similarity of declines pre-ceding dropout caused by illness or frailty to those precedingdeath must be even closer than these analyses suggest. Thefinding that amounts of declines preceding impending deathand dropout are so strikingly similar suggests that they are bothcaused by declining health. On this interpretation, death anddropout are empirically useful, but rough and indirect, markersfor worsening health. The effects of approach to death, on theirown, tell us little about the functional causes of cognitivedecline. To learn more we must study precisely how particularterminal illnesses affect our brains and central nervous systems.

Are Different Cognitive Abilities DifferentiallyAffected by Death and Dropout?

Declines with age for production vocabulary (Mill Hill BVocabulary Test) are much less (and nonsignificant) than thosefor intelligence and cumulative verbal learning. Vocabulary,a skill crystallized because it is acquired early in life andmaintained by continual practice into old age (Horn, 1987;Horn, Donaldson, & Engstrom, 1981), is not only relativelyresistant to normal aging but also to the pathologies thataccompany aging and that may terminate in death. The effect ofdeath or withdrawal group is also weaker in Mill Hill BVocabulary Test.

This result differs from previous findings that vocabulary testscores may be especially sensitive to approaching death (Berg,1987; Birren, 1965; Siegler, McCarty, & Logue, 1982). It isnoteworthy that these previous studies were cross-sectional andused relatively elderly samples and brief census periods of24 months or less. It is therefore possible that participants inthese studies were quite near to their deaths but that this couldnot be observed because the census periods were so short. It istherefore plausible that declines in vocabulary only becomemarked very shortly before death and so were less pronouncedin the present study, in which the times of measurement were4 years or longer. It seems likely that individuals who are, asyet, only experiencing declines in so-called fluid abilities arerelatively further from death than those whose terminal pathol-ogies have become severe enough to affect even their crys-tallized abilities. Accepting this context declines in vocabularyare particularly sensitive markers of approaching death.

Effects of Gender and Demographic VariablesWomen perform better than men on the CVL task and on the

Mill Hill B Vocabulary Test (the latter at a marginal signifi-cance level), and men perform better than women on the AH4-1

intelligence test. On the AH4-1 task, although women scorelower on average than men do, they also decline more slowly asthey grow older. This is informative because the effects ofdifferences in death and dropout have been taken into con-sideration, as the slower declines of women of advancing agecannot be attributed simply to their greater longevity or lowerdropout rate. This raises a fruitful topic for further investigation.It is possible that examining the differences in the nature ofterminal pathologies in men and women would be informative.

A different point is that significant gender advantages inbaseline scores are opposite on different cognitive tests. Thismeans that if we do not take into account the typical progres-sive, age-related increases in the proportion of women to menduring longitudinal studies, then our speculations as to whichcognitive abilities are most and least sensitive to age andpathology will be insecure.

Recruitment cohorts differ markedly in cognitive test scoresand, as exploratory analyses show, also in average levels ofSEA. Because mortality is strongly associated with level ofgeneral intellectual ability (Hart et al., 2003) and with SEA(e.g., Nagi & Stockwell, 1973), it is essential that researcherscheck for differences between recruitment cohorts whenanalyzing the effects of impending death on cognitiveperformance.

Methodological IssuesMost studies have found that particular pathologies that

become common in later life, such as diabetes (e.g., Bent,Rabbitt, & Metcalf, 2000), hypertension and other cardiovas-cular problems (Fahlander et al., 2000; Hertzog, Schaie, &Gribbin, 1978; Lopez et al., 2003), respiratory problems(Holland & Rabbitt, 1991), and undifferentiated healthproblems (McInnes & Rabbitt, 1997), have significant butsurprisingly small effects on cognitive performance. Moststudies have compared patients and healthy controls only ata single time point, and the few longitudinal studies haveignored death and dropout. The current analyses suggest that ifindividuals in so-called patient groups who die or withdraware not included in comparisons against healthy controls, thenthe true effects of pathologies must be severely underestimated.If dropouts and deaths are excluded from analyses then the

Table 7. Specific Comparisons of MHB Scores Between

Death and Dropout Groups

MHB D1 D2 D3 D4 W1 W2 W3 WD1 WD2 WD3

C .0001 .0001 .0001 .0001 .0001 .0001

D1 .001

D2

D3

D4

W1 .0002 .0001 .0009

W2 .001

W3

WD1 .0004

WD2

WD3


tion, p , .001 is taken as the threshold for significance. MHB Mill Hill BVocabulary Test; C control group; D death and W withdrawal (see textfor particular designations).


comparisons will involve only patients whose conditions are, asyet, relatively mild. The effects of terminal declines and of earlystages of pathologies can only be compared if all deaths anddropouts are logged and taken into account.

CORRESPONDENCE

Address correspondence to P. M. A. Rabbitt, Department of Experi-mental Psychology, University of Oxford, England. E-mail: [email protected]

REFERENCES

Bent, N., Rabbitt, P. M. A., & Metcalf, D. (2000). Diabetes mellitus and therate of cognitive change. British Journal of Clinical Psychology, 39,349362.

Berg, S. (1987). Intelligence and terminal decline. In G. Maddox and E. W.Busse (Eds.), Aging: The universal human experience (pp. 234259).New York: Springer.

Birren, J. E. (1965). Age changes in speed of behavior: Its critical natureand physiological correlates. In A. T. Welford and J. E. Birren (Eds.),Behavior, aging and the nervous system (pp. 171189). Springfield, IL:Charles C Thomas.

Fahlander, K., Wahlin, A., Grut, M., Forsell, Y., Hill, R. D., Winblad, B.,et al. (2000). The relationship between signs of cardiovasculardeficiency and cognitive performance in old age. Journal ofGerontology: Psychological Sciences, 55B, P259P265.

Hart, C. L., Taylor, M. D., Davey-Smith, G., Whalley, L. J., Starr, J. M.,Hole, D. J., et al. (2003). Childhood IQ, social class, deprivation andtheir relationships with mortality risk in later life: Prospectiveobservational study linking the Scottish Mental Survey 1932 and theMidspan Studies. Psychosomatic Medicine, 65, 877883.

Hertzog, C., Schaie, K. W., & Gribbin, K. (1978). Cardiovascular diseaseand changes in intellectual functioning from middle to old age.Gerontology, 33, 872883.

Holland, C. M., & Rabbitt, P. M. A. (1991). The course and causes ofcognitive change with advancing age. Reviews in Clinical Gerontology,1, 8196.

Horn, J. L. (1987). A context for understanding information processingstudies of human abilities. In P. A. Vernon (Ed.), Speed of informationprocessing and intelligence (pp. 201238). Norwood, NJ: Ablex.

Horn, J. L., Donaldson, G., & Engstrom, R. (1981). Application, memoryand fluid intelligence decline in adulthood. Research on Aging, 3, 3384.

Houx, P. J. (1991). Rigorous health screening reduces age effects ona memory scanning task. Brain and Cognition, 15, 246260.

Johansson, B., & Berg, S. (1989). The robustness of the terminal declinephenomenon: Longitudinal data from the digit-span memory test.Journal of Gerontology: Psychological Sciences, 44B, P184P186.

Kleemeier, R. W. (1962). Intellectual changes in the senium. Proceedingsof the Social Statistics Section of the American Statistical Association,23, 290295.

Lachman, R., Lachman, J. L., & Taylor, D. W. (1982). Reallocation ofmental resources over the productive lifespan: Assumptions and taskanalyses. In F. I. M. Craik & S. Trehub, (Eds.), Aging and the cognitiveprocess (pp 304350). New York: Plenum Press.

Lieberman, M. A. (1965). Psychological correlates of impending death:Some preliminary observations. Gerontology, 20, 181190.

Lopez, O. L., Jagust, W. J., Dulberg, C., Becker, J. T., De Kosky, S. T.,Fitzpatrick, A., et al. (2003). Risk factors for mild cognitive impairmentin the Cardiovascular Health Study cognition study. Archives ofNeurology, 60, 13941399.

McInnes, L., & Rabbitt, P. M. (1997). The relationship between functionalability and cognitive ability among elderly people. In Facts andresearch in gerontology (pp. 3445). Paris: Sardi.

Nagi, M. H., & Stockwell, E. G. (1973). Socioeconomic differentials inmortality by cause of death. Health Services Reports, 88, 449456.

Office of Population Censuses and Surveys. (1980). Classification ofoccupations 1980. London, England: Her Majestys Stationery Office.

Palmore, E., & Cleveland, W. (1976). Aging, terminal decline and terminaldrop. Gerontology, 31, 7681.

Pincus, T., Callaghan, L. F., & Birkhauser, R. V. (1987). Most chronicdiseases are reported most frequently by individuals with fewer than 12years formal education in the age 1864 US population. Journal ofChronic Disorders, 40, 865874.

Rabbitt, P. (2002). Cognitive gerontology. In H. Pashler & J. Wixted (Eds.),Stevens handbook of experimental psychology (Vol. 4., pp. 203254).New York: Wiley.

Rabbitt, P., Diggle, P., Holland, F., & McInnes, L. (2004). Practice anddrop-out effects during a 17-year longitudinal study of cognitive aging.Journal of Gerontology: Psychological Sciences, 59B, P84P97.

Rabbitt, P., Diggle, P., Smith, D., Holland, F., & McInnes, L. (2001).Identifying and separating the effects of practice and of cognitiveageing during a large longitudinal study of elderly community residents.Neuropsychologia, 39, 532543.

Rabbitt, P., Lunn, M., & Wong, D. (2005). Neglect of dropoutunderestimates effects of death in longitudinal studies. Journal ofGerontology: Psychological Sciences, 60B, P106P109.

Rabbitt, P., Watson, P., Donlan, C., Bent, N., & McInnes, L. (1994).Subject attrition in a longitudinal study of cognitive performance incommunity resident elderly people. In B. J. Vellas, J. L. Albarende, &P. J. Garry (Eds.), Facts and research in gerontology: Epidemiologyand aging (pp. 2934). Paris: Serdi.

Rabbitt, P., Watson, P., Donlan, C., McInnes, L., Horan, M., Pendleton, N.,et al. (2002). Effects of death within 11 years on cognitive performancein old age. Psychology and Aging, 17, 114.

Rabbitt, P. M. A., Diggle, P., Holland, F., McInnes, L., Bent, N., Abson,V., et al. (2004). The University of Manchester longitudinal study ofcognition in normal healthy old age, 1983 through 2003. Aging,Neuropsychology and Cognition, 11, 245279.

Raven, J. C. (1965). The Mill Hill Vocabulary Scale. London: Lewis.Reimanis, G., & Green, R. (1971). Immanence of death and intellectual

decrement in the aging. Developmental Psychology, 5, 270272.Riegel, K. F., & Riegel, R. M. (1972). Development, drop and death.

Developmental Psychology, 6, 306319.Riegel, K. F., Riegel, R. M., & Meyer, G. (1967). A study of the drop-out

rates in longitudinal research on aging and the prediction of death.Journal of Personality and Social Psychology, 5, 342348.

Roth, M. (1986). The association of clinical and neuropsychologicalfindings and its bearing on the classification and etiology ofAlzheimers disease. British Medical Bulletin, 42, 4250.

Schaie, K. W., Labouvie, G. V. & Barrett, T. J. (1973). Selective attritioneffects in a fourteen-year study of adult intelligence. Gerontology, 28,328334.

Siegler, I. C., & Botwinick, J. (1979). A long-term longitudinal study ofintellectual ability of older adults: The matter of selective subjectattrition. Gerontology, 34, 242245.

Siegler, I. C., McCarty, S. M., & Logue, P. E. (1982). Wechsler MemoryScale scores: Selective attrition and distance from death. Gerontology,37, 176181.

Small, B. J., Fratiglioni, L., von Strauss, E., & Backman, L. (2003).Terminal decline and cognitive performance in very old age.Psychology and Aging, 18, 193202.

Snowden, D. A., Ostwald, S. K., Kane, R. E. L., & Keenan, N. L. (1989).Years of life with good and poor mental and physical functions in theelderly. Journal of Clinical Epidemiology, 42, 10551066.

Van Boxtel, M. P. J., Buntink, F., Houx, P. J., Metsemakers, J. F. M.,Knottmerus, A., & Jolles, J. (1998). The relation between morbidityand cognitive performance in a normal aging population. Journal ofGerontology: Medical Sciences, 55A, M147M154.

Received June 28, 2005Accepted January 30, 2008Decision Editor: Thomas M. Hess, PhD

RABBITT ET AL.P278

rabbitt (2008) j gerontol psychol sci death dropout longitudin measures cognitive change

Documents