This Article Abstract Full Text (PDF) Services Download to citation manager PubMed PubMed Citation

Age, Medical Conditions, and Gender as Interactive Predictors of Cognitive Performance

The Effects of Selective Survival

^a Leonard Davis School of Gerontology, University of Southern California, Los Angeles
^b Department of Preventive Medicine and Environmental Health, University of Iowa, Iowa City

Decision Editor: Toni C. Antonucci, PhD

	Abstract

TOP Abstract Study 1 Study 2 General Discussion References

 
The interaction of age, medical conditions, and gender on free-recall and mental status test performance was analyzed in two large survey samples of older adults. Age, gender, and the presence of medical conditions interacted with recall and mental status in Study 1 (n = 2,695) and mental status in Study 2 (n = 6,299). For men, those with one or more medical conditions declined more steeply with age than those with no conditions. For women, this relationship was reversed. The findings suggest survival effects, whereby those who lived to old age with medical conditions and were selected into the sample had high levels of cognitive functioning. The age at which these effects are seen vary with gender because women survive longer than men.

AS the proportion of the population over the age of 85 increases, the relevance of health and gender-related factors to old age differences in cognitive performance must be determined. Zelinski, Crimmins, Reynolds, and Seeman 1998 found that for mental status and Series 7 (subtraction by 7s from 100 for five trials, which taps numerical ability and working memory), high blood pressure and diabetes each interacted with age, with older participants showing less condition-related impairment than younger participants. These interactions suggest that those with medical conditions and poor cognitive performance are more likely to die at younger ages and that the oldest-old with the same conditions represent a progressively more select sample. It is likely that gender will also interact with the presence of medical conditions in affecting cognition, because men have shorter survival and health is an important determinant of survival.

Because this differential survival difference is expected to be more pronounced in the oldest age groups, the three-way interaction between age, medical conditions, and gender will be tested in two large survey samples in this article. Women are expected to show increasingly more condition-related impairment with age, whereas condition related-impairment will not increase as much with age in men. The dependent measures are two widely used in studies of cognition in epidemiological studies: free recall and mental status.

	Study 1

TOP Abstract Study 1 Study 2 General Discussion References

 
In Study 1, we analyzed data for the first wave of the Iowa 65+ Rural Health Study (RHS), one of four sites of the National Institute on Aging's Established Populations for Epidemiological Studies of the Elderly (EPESE). The rationale, measures, sampling, and other details of the study are presented elsewhere (Cornoni-Huntley, Brock, Ostfeld, Taylor, and Wallace 1986).

Measures
We assessed immediate oral free recall after an examiner read 20 common, one- to two-syllable English nouns to participants at 2-s intervals. We measured mental status by using a modified version of Pfeiffer 1975 Short Portable Mental Status Questionnaire (SPMSQ).

Participants self-reported high blood pressure, heart disease, diabetes, and stroke. If any of these medical conditions were reported, participants were given a score of 1 on a dichotomous medical conditions variable. Each of these diseases has been shown to affect cognitive performance independently (see Zelinski et al. 1998). Although self-reports of medical conditions could be considered to be less reliable than physician ratings, they are highly correlated (LaRue, Bank, Jarvik, and Hetland 1979).

We included depression and self-rated health as covariates in analyses, because both are associated with objective measures of physical health and with cognitive performance (LaRue et al. 1979). Depression was measured using the 20-item Center for Epidemiologic Studies–Depression Scale (CES-D; Radloff 1977). Self-rated health was measured on a 5-point scale ranging from poor to excellent, with respondents asked to compare themselves to people their own age. We conducted analyses on data of the 2,695 nonproxy respondents aged 65–102 with complete data on the measures in this study.

Analyses
We conducted hierarchical regression analysis for recall and mental status with forced separate entry of age, followed by gender, education, depression, self-rated health, and presence of medical conditions; the two-way interactions between age and gender, age and medical conditions, and gender and medical conditions; and the three-way interaction between age, gender, and medical conditions.

Results and Discussion
For recall, being younger, female, and having more education, lower depression score, and better health rating each predicted better recall, together accounting for 19% of the variance in performance. Medical condition was not a significant main effect. However, there was a significant three-way interaction between age, gender, and number of medical conditions, accounting for .1% of the variance in performance. The interaction is plotted in the top panel of Fig. 1. Among those with one or more conditions, men showed greater decline with age than women. Conversely, among those reporting no medical conditions, women declined more steeply with age than men.

View larger version (24K): [in this window] [in a new window]   Figure 1. Study 1, Rural Health Study. Top panel: predicted recall score as a function of age, gender, and medical conditions. Recall is of 20 words. Bottom panel: predicted mental status score as a function of age, gender, and medical conditions. Maximum mental status test score is 9.

A possible explanation for the steeper decline among men reporting medical conditions is that men reported more severe medical conditions than women. To test this, we ran analyses in which only those who reported a serious level of a medical condition were counted as having the condition. However, the results were identical to those reported, ruling out a severity explanation.

The steeper drop among men with one or more conditions appears to be largely driven by the relatively high performance of men with one or more conditions in the youngest age group. It could be that a survival effect already exists among these men—that is, men who have survived to their mid-to-late 60s with one or more conditions may already be a select group relative to those without medical conditions. In contrast to men, women with medical conditions showed a survival effect in the older age groups, consistent with the suggestion by Perls, Morris, Ooi, and Lipsitz 1993 that survival effects may appear in women at older ages than they appear in men. However, the severity analysis indicated that the survival effect among women was not due to disease severity.

	Study 2

TOP Abstract Study 1 Study 2 General Discussion References

 
In Study 2, we examined the first wave of the Asset and Health Dynamics of the Oldest-Old (AHEAD) study, a national survey of noninstitutionalized Americans age 70–103 from 1993 to 1994. The rationale, measures, sampling, and other details of the study have been presented elsewhere (Soldo, Hurd, Rodgers, and Wallace 1997).

Measures
We assessed immediate free oral recall by an examiner reading 10 common, English nouns to participants at 2-s intervals. The mental status test consisted of nine questions from the Telephone Interview for Cognitive Status (TICS; Brandt, Spencer, and Folstein 1988), with a maximum score of 10 points.

We constructed a dichotomous medical conditions variable as in Study 1. Depression was measured using a modified 8-item/8-point version of the CES-D (Radloff 1977) adapted especially for the AHEAD survey. Self-rated health was measured on a 5-point scale ranging from poor to excellent. A total of 6,299 respondents was selected into the analyses because they had complete data on the measures used.

As in Study 1, we used hierarchical regression analysis with the same ordering of variables, except that in this study, we added race (being White vs. being a minority) as a control variable after education. Virtually all members of the Study 1 sample were White, whereas 14% of Study 2 respondents were members of minority groups.

Results and Discussion
Being younger, female, having more education, and being White and having a lower depression score and a better health rating each reliably predicted better recall, accounting for 24% of the variance in recall. Medical conditions did not reliably predict recall. The only significant interaction was between age and gender, explaining .1% of the variance in performance. Women had higher scores than men in every age group, but the gap between men and women was widest in the youngest age group and narrowest in the oldest age group. The steeper slope of decline with age for women in both interactions supports the hypothesis of greater age-related decline in cognitive performance among women, perhaps due to a more pronounced survival effect among men. An alternative explanation for the Age x Gender interaction is that men had less serious forms of the same conditions. A severity analysis in which only those having serious forms of some of the conditions were coded as having the conditions produced identical findings, however.

For mental status, being younger, female, having more education, and being White and having a lower depression score and a better health rating each reliably predicted better mental status score and accounted overall for 27% of the variance. There were reliable Age x Gender and Age x Medical Conditions interactions, but they were limited by a significant three-way interaction between age, gender, and medical conditions, accounting for .1% of the variance in mental status. This interaction was similar in pattern to the one in the lower panel of Fig. 1 and is consistent with the suggestion that survival effects may appear at later ages in women than in men (Perls et al. 1993). We performed an analysis of severity as a plausible explanation for the superior performance of women with conditions in the oldest age groups, but results were identical to the main analysis.

	General Discussion

TOP Abstract Study 1 Study 2 General Discussion References

 
The hypothesis for the three-way interaction between age, gender, and medical conditions was that condition-related impairment would increase more with age in women than in men. Though the interaction reliably predicted recall in Study 1 and mental status in Studies 1 and 2, it was women who showed a more pronounced survival effect related to medical conditions. Because men showed evidence of a survival effect related to conditions in the youngest age group, it may be that both genders exhibit survival effects, but the effects appear at older ages in women (Perls et al. 1993).

Although the Age x Gender x Medical Conditions interaction was not reliable for recall in Study 2, the significant Age x Gender interaction revealed that declines in cognitive performance were less steep for men with age than for women. This interaction is consistent with the argument that survival effects vary by gender. The three-way interactions for recall in Study 1 and for mental status in Study 2 revealed this pattern in the youngest age group. Among the young-old, women with conditions performed worse than women with no conditions, whereas among men, those with conditions performed the same or better than those with no conditions.

This study provides some support for the hypothesis that survival effects may be present in cognition in older adults. When survival is considered in relation to medical conditions, men show evidence of a survival effect at younger ages, whereas the condition-related survival effect is stronger among women at older ages. The proportion of variance explained by the interactions supported in this study was very small and may not seem substantial as an explanation for age differences in cognition. However, the interactions were significant despite the fact that they had differential residual variances after partialling out the main effects, which makes significant interactions difficult to obtain in regression analysis (McClelland and Judd 1993). Also, in other fields, such as medicine and epidemiology, effects this size are seen as substantively useful with practical implications (Rosenthal 1990).

There are, however, some limitations to the present study. One is that the use of a dichotomous variable for the presence of medical conditions is less informative than separate evaluations of effects by specific conditions. Another involves the outcome measures. Free recall is generally used in studies of normal cognitive aging and mental status in studies of cognitive impairment, and these involve not only very different performance distributions but interpretations of performance as well. However, they are currently the most frequently used measures in epidemiological studies of cognition in the elderly. It is not clear whether other measures of attention, memory, or intelligence would show results similar to those reported.

Nevertheless, the results suggest that there is something special about many of those who attain very old age, and these results should encourage researchers to identify some of the unique characteristics of these men and women that confer a cognitive advantage.

	Acknowledgments

 
This research was supported in part by NIA Grant AG14203. We thank Eileen Crimmins for reviewing the manuscript.

Address correspondence to Elizabeth Zelinski, Leonard Davis School of Gerontology, University of Southern California, Los Angeles, CA 90089-0191. E-mail:

Received for publication April 12, 1999. Accepted for publication April 24, 2000.

	References

TOP Abstract Study 1 Study 2 General Discussion References

 

• Brandt J., Spencer M., Folstein M., 1988. The telephone interview for cognitive status. Neuropsychiatry Neuropsychology and Behavioral Neurology 1:111-117.
• Cornoni-Huntley J., Brock D. B., Ostfeld A. M., Taylor J. O., Wallace R. B., , ed.Established populations for epidemiologic studies of the elderly: A resource book 1986National Institutes of Health, Bethesda, MD.
• LaRue A., Bank L., Jarvik L., Hetland M., 1979. Health in old age: How do physicians' ratings and self-ratings compare?. Journal of Gerontology 34:687-691. [Abstract/Free Full Text]
• McClelland G. H., Judd C. M., 1993. Statistical difficulties of detecting interactions and moderator effects. Psychological Bulletin 114:376-390. [Medline]
• Perls T. T., Morris J. N., Ooi W. L., Lipsitz L. A., 1993. The relationship between age, gender and cognitive performance in the very old: The effect of selective survival. Journal of the American Geriatrics Society 41:193-201.
• Pfeiffer E., 1975. A short portable mental status questionnaire for the assessment of organic brain deficit in elderly patients. Journal of the American Geriatrics Society 23:433-441. [Medline]
• Radloff L. S., 1977. The CES-D Scale: A self-report depression scale for research in the general population. Applied Psychological Measurement 1:385-401.
• Rosenthal R., 1990. How are we doing in soft psychology?. American Psychologist 45:775-777.
• Soldo B. J., Hurd M. D., Rodgers W. L., Wallace R. B., 1997. Asset and Health Dynamics Among the Oldest-old: An overview of the AHEAD study. Journal of Gerontology 52B: (Special Issue) 1-20.
• Zelinski E. M., Crimmins E., Reynolds S., Seeman T., 1998. Do medical conditions affect cognition in older adults?. Health Psychology 17:504-512. [Medline]

This Article Abstract Full Text (PDF) Services Download to citation manager PubMed PubMed Citation