Health Screening and Random Recruitment for Cognitive Aging Research

Abstract

A survey of 197 cognitive aging studies revealed infrequent use of structured health assessments and random recruitment. In this study, a health screening questionnaire developed to identify subjects with medical problems that might impair cognition was administered to 315 adults aged 60 and older who were recruited by random digit dialing. On the basis of self-reported medical problems, 35% of the subjects were excluded. Those excluded were older (p < .001) and tended to be male but did not differ in education from those who passed the screening. Subjects who passed the screening and decided to participate in a neuropsychological research project were younger (p < .001), better educated (p < .001), and more likely to be male (p < .001) than nonparticipants. These findings suggest that careful assessment, selection, and description of subjects is needed to aid interpretation of cognitive aging research. Further attention to health status is needed to aid interpretation of cognitive aging research. Although random recruitment of the elderly is feasible, obtaining representative samples may require stratification on demographic variables.

The health status of elderly samples and the method by which they are recruited may have significant effects on the results of research on cognition (Browning & Spilich, 1981; Camp, West, & Poon, 1989; Nesselroade, 1988; Poon, Krauss, & Bowles, 1984). Failure to consider health status or other demographic variables related to cognition during sample recruitment may result in confounding of these variables with age (Poon et al., 1984).

To determine the current practices in assessment of health status and subject recruitment, we surveyed all cognitive studies that included a “normal” sample of elderly persons (age60 or older) appearing between 1984 and mid-1990 in the Journal of Clinical and Experimental Neuropsychology, the Journal of Gerontology, Neuropsychologia, and Psychology and Aging. Of the 197 studies reviewed, only 10% described specific exclusion criteria based on health conditions related to cognitive functioning. Most articles (64%) described subjects’ health status in general terms (e.g., “healthy,” “neurologically normal”). Most frequently, these descriptions were based on global self-ratings obtained after entry into the study. No information on health status was provided in 25% of the articles. The number of subjects excluded on the basis of health status appeared in two articles (1%).

The infrequent screening of subjects for specific medical conditions affecting cognition may reflect the lack of practical, explicit methods to accomplish this. Self-report would appear to be a practical and valid method for assessing the presence of medical conditions (Ferraro, 1980; Ford et al., 1988; LaRue, Bank, Jarvik, & Hetland, 1979; Maddox & Douglass, 1973); however, there are no standard instruments for screening subjects specifically for conditions related to cognition.

Of the studies we reviewed, 36%provided no information on subject recruitment. Four studies (2%) described using a random or quasi-random method, such as randomly selecting homes from electoral rolls. The remaining 62% selected subjects through nonrandom recruitment, through senior centers and community agencies, for example. Some neglect of random recruitment methods in cognitive aging studies may stem from limited information on their efficacy. Only a few studies have described the use of random telephoning to establish an elderly control sample (Olsen & Mandel, 1988; Robertson, Grufferman, & Cohen, 1988; Wallace & Relief, 1987).

In this article, we present a health screening questionnaire designed to identify subjects with medical problems that might affect cognition, and describe its use in the context of random recruitment for a neuropsychological research project. To determine the efficacy of this recruitment method in yielding a representative sample, we examine the relationship between demographic variables, eligibility, and participation rates. Reports on neuropsychological data obtained from subjects participating in this study appear elsewhere (Christensen, Multhaup, Nordstrom, & Voss, 1990, 1991a, 1991b).

Method

Subjects

Subjects were recruited from a list of randomly generated telephone numbers from working exchanges in a two-county area that included Minneapolis and St. Paul, Minnesota.

Instrument

A health screening questionnaire was developed to identify subjects with medical problems that would be likely to affect cognition. The guiding principles in selecting questions were to identify conditions that were most likely to affect cognition without excluding an inordinate number of subjects, to keep the number of questions relatively low, and to describe as many medical conditions as possible in lay terms. Conditions, such as thyroid disease, that could be identified by having the subjects provide a list of current medications were not included in the questionnaire. The list of questions was developed in consultation with two psychiatrists, a neurologist, and two internists. A copy of the questionnaire appears in the Appendix.

Procedure

Trained telephone interviewers from the University of Minnesota Center for Survey Research called the random sample of telephone numbers using a scripted interview. They asked to speak to anyone in the household who was 60 years of age or older. Age-eligible adults were told,

The University is preparing to do an important health research project focusing on older adults. In order to get accurate results, they need to get a wide variety of people to participate. I’m going to ask you some questions about many specific medical conditions people may have. If you don’t know what any of these are, please say so and we’ll skip over them. This will only take a few minutes, and it is just to get an idea of your overall level of health.

As soon as the subject affirmed an exclusion criterion, the interviewer obtained information on age and education and thanked the individual for participation. Gender was determined by the interviewer on the basis of voice characteristics. Individuals not excluded on the basis of a health problem were asked to participate in a study involving paper and pencil tests, for which they would be paid $25.00. Within two days of telephone screening, letters describing the study were mailed to willing participants. Three days after these letters were mailed, subjects were contacted by telephone to schedule testing at a community location close to the subject’s home, such as a church, community center, or library.

Results

A total of 2,777 phone numbers were called. Of these, 1,544 (56%) did not have an age-eligible resident, 715 (26%) were disconnected or were not residential, 77 (3%) were eliminated after 10 unanswered calls, and 12 (.4%) were still being contacted at the end of the study. We encountered 112 refusals (6%of completed calls), 48 of these before age was given, 55 (15%of the known age-eligible subjects) at the start of the health screening, and 9 at an unknown stage of the interview. Among the 315 subjects who completed the health screening questionnaire, complete data on age, education, and gender were available for 301 (see Table 1). Data on ethnic group membership were not obtained.

Table 1.

Age, Education, and Gender Distributions in Total, Eligible, and Participant Samples

Characteristic	Total sample (N = 301)	Eligible sample (N = 203)	Participant sample (N = 101)
Age
60–69	52	58	67
70–79	34	31	27
80+	14	11	6
Education
Eighth grade or less	8	5	3
Some high school	11	11	8
High school graduate	36	37	30
Technical school	8	8	9
Some college	18	18	21
College graduate	12	13	19
Postgraduate	7	7	11
Gender
Male	32	29	38
Female	68	71	62

Note. All figures are percentages. Total sample refers to all subjects screened. Eligible sample refers to subjects who passed the health screening. Participant sample refers to subjects who participated in neuropsychological testing.

Health Screening

Of the 315 subjects completing the health screening, 206 (65%) passed and 109 (35%)were excluded on the basis of a self-reported medical condition. Complete data on age, education, and gender were available for 301 subjects, 203 of whom passed the health screening. Pearson product-moment correlations revealed a significant relationship between eligibility and age (r = −.19, p <.01), but not between eligibility and education (r = .07, p > .05) or gender (r = .10, p > .05). Eligibility was coded 1 for eligible and 0 for ineligible. Gender was coded 1 for female and 0 for male. Stepwise multiple regression analysis yielded an R² of .05, F(2,298) = 8.04, p < .0005, and agreed with the simple correlational results for age (β = −.20, p < .0005) and education (β = .06, p > .05). Being female was a significant predictor of eligibility in the multiple regression analysis (β = .13, p < .05), whereas gender failed to reach significance in the simple correlation. The selection of predictor variables was unchanged when multiple regression analyses were repeated with the exclusion of subjects with standardized residual values greater than 2.0.

Random Recruitment

Of the 206 subjects passing the health screening, 101 (49%) participated in the neuropsychological research project (see Table 1). Among the refusals, 70 declined participation at the time of initial telephone screening, and 35 declined at the time of the scheduling call. Pearson product-moment correlations within the eligible sample revealed significant relationships between participation and age (r = −.24, p < .001), education (r = .29, p < .001), and gender (r = −.20, p < .001). Participation was coded 1 for participants and 0 for nonparticipants. Gender was coded 1 for female and 0 for male. Multiple regression analysis yielded an R² of. 14, F(3,199) = 11.16, p < .00005, and agreed with the simple correlational results for age (β = −.19, p < .01), education (β = .25, p < .0005), and gender (β = .14, p < .05). The selection of predictor variables was unchanged when multiple regression analyses were repeated with the exclusion of subjects with standardized residual values greater than 2.0.

Discussion

Poon et al. (1984) have demonstrated the risks of under- and overestimation of the true effects of age when sample characteristics are not given adequate consideration. Nevertheless, our survey of recent articles on aging and cognition revealed inconsistent reporting of subjects’ health and recruitment methods, two factors likely to influence outcome. In particular, exclusion of subjects with health conditions known to affect cognition and use of random recruitment were rare.

In the current study, use of a health screening questionnaire designed to identify subjects with medical problems that can impair cognition resulted in the exclusion of 35%of a sample of randomly recruited subjects.¹ Subjects who were excluded were older and demonstrated a trend toward a higher proportion of males than those who passed the screening but did not differ from those passing the screening with respect to education. The findings that a large proportion of potential subjects were excluded and that exclusion was related to age even in this age-restricted sample call attention to the need for health status to be addressed in cognitive aging research. Health screening is important when research goals include minimizing the possibility of confounding age and disease. Future studies examining the relationship between this questionnaire, physician-reported health status, global health ratings, and cognitive performance should help to clarify the utility of this particular questionnaire.

Recruitment by random digit dialing yielded an acceptable rate of refusal (Fowler, 1988) for the initial health screening. In contrast, only half of the subjects who were eligible on the basis of health screening agreed to participate in testing. In addition, those subjects who agreed to participate were younger, more highly educated, and more likely to be male than were those who refused participation. These findings suggest that random recruitment of older adults is feasible but that stratification on these demographic variables would be required to achieve a representative sample.

Appendix. Health Screening Questionnaire

The questionnaire is presented here in the format that was used in the current study. Based on the data from this study and suggestions from other investigators, the questionnaire has been revised to include additional criteria and to present items in an order that reflects their likelihood of endorsement. The change in order is designed to allow earlier identification of subjects who meet exclusion criteria. The item number for the revised questionnaire appears in parentheses after the item. Information on these changes is provided for investigators who may be considering the use of the questionnaire.

Question (new item number)	Response exclusion criteria
1. Have you ever had a stroke or a T.I.A.? (Probe: T.I.A. stands for transient ischemic attack) (1)	Yes
2. Have you ever had seizures? (22)	Yes
3. Do you have Parkinson’s disease? (23)	Yes
4. Do you have multiple sclerosis, cerebral palsy, or Huntington’s disease? (33)	Yes
5. Have you had encephalitis or meningitis? (18)	Yes
6. Have you ever had brain surgery? (24)	Yes
7. Have you ever undergone surgery to clear arteries to the brain? (25)	Yes
8. Do you have diabetes that requires insulin to control? (13)	Yes
9. Do you have hypertension that is not well controlled? (14)	Yes
10. Have you had a cancer other than skin cancer diagnosed within the last 3 years? (3)	Yes
11. Do you have shortness of breath while sitting still? (4)	Yes
12. Do you use home oxygen? (5)	Yes
13. a. Have you ever had a heart attack?
b. (If yes) Did you have any change in your memory, ability to talk or solve problems 24 hours after your heart attack? (19)	Yes to a & b
14. Are you receiving kidney dialysis? (34)	Yes
15. Do you have a liver disease? (35)	Yes
16. Have you been hospitalized for mental or emotional problems in the past 5 years? (21)	Yes
17. Are you currently taking medications for mental or emotional problems? (20)	Yes
18. a. Do you drink beer, wine, or other alcoholic beverages every day or less often?
b. (If daily) How many drinks do you have each day? (10)	Every day to a and more than 3 to ba
19. a. Have you ever had a problem due to abuse of drugs or medications?
b. (If yes) Was this within the past 5 years? (11)	Yes to a & b
20. a. Have you ever been treated for alcohol or drug abuse?
b. (If yes) Was this within the past 5 years? (12)	Yes to a & bb
21. Have you ever been unconscious for more than 1 hour other than during surgery? (16)	Yes
22. Have you ever required overnight hospitalization because of a head injury? (17)	Yes
23. Have you ever had any illness that caused a permanent decrease in memory or other mental functions? (26)	Yes
24. Do you have trouble with your vision that prevents you from reading ordinary print even when you have glasses on? (7)	Yes
25. Do you have difficulty understanding conversations because of your hearing even if you wear a hearing aid? (6)	Yes
26. Are you able to write your name? (37)	No

The following 11 items were added to the questionnaire subsequent to the study reported in this article:

Question (new item number)	Response exclusion criteria
1. a. Have you been seen by a neurologist or neurosurgeon?
b. (If yes) Was this for a back or neck problem?
c. (If yes) Was this for a tension headache? (2)	Yes to a and no to b & c
2. Have you had heart surgery? (8)	Yes
3. Have you ever been resuscitated? (9)	Yes
4. Have you had a head injury with loss of consciousness greater than 5 minutes? (15)	Yes
5. Have you ever received electroshock therapy? (27)	Yes
6. Have you ever been diagnosed as learning disabled? (28)	Yes
7. Were you placed in special classes in school because of learning problems? (29)	Yes
8. Have you ever been diagnosed as having a brain tumor? (30)	Yes
9. Do you have difficulty using your hands?c (31)	Yes
10. a. Have you every had major surgery with general anesthesia?
b. (If yes) Did you have any change in your memory, ability to talk or solve problems 1 week after surgery? (32)	Yes to a & b
11. Do you have lupus? (36)	Yes

^aThis criterion was changed to “more than 2” in the revised questionnaire.

^bPart b of this question was omitted from the revised questionnaire, so that individuals with any history of treatment for alcohol or drug abuse are excluded.

^cThis criterion was included because of the use of timed tests requiring manual dexterity in this research.