Please be patient. Images under construction. Thank you.
Department of Psychology, University of Minnesota, Minneapolis, USA
|
|
|
|
|
Establishing the Registry Recruitment |
Demographic Characteristics |
Conclusion References |
Please be patient. Images under construction. Thank you. |
|
Abstract. A birth-record based Registry is nearing completion of some 8,000 pairs of twins born in Minnesota from 1936 to 1955, plus some 1,200 pairs of male twins born 1971-81. The middle-aged twins were recruited with graded incentives so that ease of recruitment could be measured; it was found that pairs concordant for ease of recruitment were no more similar than discordant pairs in education, socioeconomic status (SES), or a variety of personality and interest factors, i.e., that selection bias may not be a problem in research with adult twins when contacts are only by mail. A 50% increase in neonatal mortality from 1936-55 to 1971-81 was associated with an increase from 3.5 to 4.0 per thousand in the frequency of viable MZ twin births. The broad heritability of SES, educational attainment, fecundity, and risk for divorce ranges from 0.30 to 0.50, although all 4 variables are plainly multifactorial and the latter 2 both involve variance contributed by a second person. Investigators interested in making use of this research resource are invited to submit proposals.
Key words: Twins, Registries, Recruitment bias, Birth rate, Birthweight, Perinatal mortality, Divorce risk, Fecundity
Beginning in 1983, we have been constructing what we believe will be the largest birth-record-based twin registry in the United States. When completed in 1990, the Resgistry will comprise about 80% of the approximately 10,400 surviving intact pairs born in Minnesota from 1936 through 1955, about 8,400 pairs in all (see Fig. 1). In a separate project, we have located and added to the Registry about 91% (1,200 pairs) of the male pairs born from 1971 through 1981. We also have obtained the computerized birth records of all multiple births in Minnesota from 1959 through 1987 and hope to locate and add many of these families to the Registry in the future.
Fig. 1. All Minnesota births (x 10,000) plotted together with birth rates (x 1,000) of opposite-sex (O-S) and monozygotic (MZ) twin pairs for the older and younger cohorts of the Minnesota Twin Resgistry. MZ rates estimated by Weinberg's formula. Assuming O-S rates are typical of dizygotic (DZ) twins generally, the DZ birth rate has decreased considerably while the MZ rate has increased slightly.
The value of using twins in the study of both genetic and environmental influences, as well as their interaction, first pointed out by Galton [17] and elaborated by many others seems recently to be gaining wider appreciation in the scientific community [3, 4, 13, 14, 18, 22, 26]. It is our impression that psychological and biomedical researchers, in increasing numbers, are becoming interested in accessing twin samples and we anticipate that other investigators may wish to establish twin registries in their own areas. The purpose of this article is to describe the Minnesota Registry, to outline how it was established, and to explain how we propose to make it available to other researchers. We shall present some demographic data and examine evidence concerning the extent to which the monozygotic (MZ) and dizygotic (DZ) participants are representative of the population from which they were recruited.
Our first step was to arrange with the State Health Department, whose main offices are situated conveniently on the University campus, to allow us to xerox all birth certificates reporting multiple births (Minnesota birth records prior to 1959 have not been computerized.). Although the information contained in these birth records varies some from year to year, it typically includes the following: the child's name and sex, the name, age, race, and occupation of the father, the full maiden name, age, occupation, race, and address of the mother, the date, time, and place of birth, the number of living children born previously to this mother, whether this was a multiple birth and the order of birth of this child, the birth weight and length, and the name of the attending physician, midwife, or other informant. Pairs (sets) of records reporting one or more still-births or infant deaths, as well as records of illegitimate births, we have discarded from further consideration.
There exists a variety of methods for locating the present whereabouts of twins ascertained from birth records. Since some methods are more cost-effective than others, it is important to apply the cheaper methods first, saving the more costly ones for the obdurate cases. Our first step is to search the state's death records for 6 months following the date of birth to ascertain infant deaths. The next step depends on the twins' present age; if the parents are likely to be still living, it may be easiest to locate them, especially in the case of female twins both of whom may have changed their surnames through marriage. We find that many of the parents who still reside in Minnesota can be readily located with the help of telephone directories (or Directory Assistance), beginning in the area where the twins were born. Failing this, it may be helpful to call other persons in the area with the same surname who might be relatives; this works best when the birth was in a small community or when the father's surname was an unusual one. If the father has a common name (eg, in Minnesota, a name like Robert Hanson), but the mother's maiden name was uncommon, it may be possible to trace the family through the mother's relatives.
Next, one can employ a reverse directory (organized by location and address rather than by name of resident) to identify persons now living at the parent's former address. These individuals may know where the former tenant is now or may at least be able to provide the name of a long-resident neighbor who would have known the former resident and might now be able to provide a lead. Other possibilities include contacting the father's former employer or the birth hospital, if this information is on the birth certificate. On rare occasions, when dealing with smaller communities, we have identified the high school that the twins once attended and, with the school's help, located a classmate still living in the community and still in touch with the twins; the classmate who took charge of the most recent class reunion is an especially promising resource.
If one is attempting to locate adult twins directly, rather than through their families, after eliminating cases of infant death we proceed to search the marriage records to identify name changes of the female twins since some 86% of those over 30 will have married. The next step, because it is so cost-effective, may be to search the state's driving license records. In Minnesota one can conduct a computer search of these records, given the full name and birth date, for just 50 cents per name.
Where the twins now reside out of state, some of these techniques are not available. However, since all one needs to locate is one family member or other informant, we have managed to contact more than 80% of twins aged 30 to 50 who were identified from birth records and more than 90% of the families of twins born from 8 to 17 years previously. (In some unknown proportion of the missing cases, at least one twin will have died so that the pair, at least for our purposes, is of no interest.) We have used undergraduate students as "twin finders," making each student responsible for identifying and locating all twins in a particular birth year. The time investment per pair located has ranged from 15 to 45 minutes from one twin finder to another and is shorter for younger twins than for middle-aged ones; the median search time is less than the mean because the distribution is positively skewed. We could undoubtedly have increased somewhat the percentage of pairs located but found that the time and cost investment increased sharply above 80% for the older cohort and above 90% for the younger twins still living at home.
Our approach to recruitment has been to send the located subjects a 4-page Biographical Questionnaire (BQ) in the first mailing together with an introductory Newsletter, describing the project, and a copy of a letter signed by Minnesota's Governor urging participation. We find that 60% to 70% of individuals respond withing 6 to 8 weeks. Non-responders are then prompted by telephone. Many of these individuals have mislaid (or never received) our first mailing and readily agree to complete and return a second BQ that we send them. Others still have the first BQ in hand and our phone call suffices to induce them to complete it. We find that 2 telephone prompts, 6 to 8 weeks apart, yield return of the BQ from a total of about 80% of the twins originally contacted.
Mass mailing, in batches of at least 200 identical packets sorted in zip-code order, costs (in the US) about 1/5 of first class rates. Sorting is automatically achieved by computer generation of mailing labels in the desired order. Duplicate mailing labels are affixed to the booklets to be returned to insure correct identification. Mailing envelopes are labeled "Forwarding and Address Correction Requested," a service which incurs an additional charge but that more than pays for itself in tracking down participants who have moved since our last contact with them.
At the time of this first mailing, our computer database consists of the birth record information plus the address information obtained by our twin finders. Upon receipt of the BQ, its information is hand-entered into the database and provides, among other things, information about the current stricture of the twin's family. We then send a packet of test materials to the twins and, if they are married, to their spouses. For the first 6 birth-year cohorts, these were bound together in a rather forbidding booklet containing 6 separate inventories totaling nearly 700 items. A 4-page, computer-coded IBM-type answer-booklet was included and respondents were required to enter their answers by blackening the appropriate circle in the answer-booklet.
We soon discovered that, surprisingly, the technology of optical scanners is not far advanced and that the scanned answer sheets had to be edited and often corrected by hand. Moreover, many average citizens have had little experience with such answer sheets and find it easier and quicker to simply circle the chosen answer right in the test booklet. Our present practice, therefore, is to bind each inventory separately, labeled with the respondent's name and identification number (ID), to print them so that answers can be marked directly beside the test question, and then to hand-enter the results when the booklets are returned. One can devise computer entry programs that create on the monitor a display of each test page as the answers are entered, and that test on-line for answers that are out of range. The data are entered again later by a different clerk using a program that flags discrepancies on-line so that the second clerk can check that entry for accuracy.
The Biographical Questionnaire included the 5 questions listed in Table1, very similar to those employed by prior authors [9, 10, 35] in achieving an accuracy of some 95% in zygosity diagnosis. Since 74 pairs of Registry twins had participated in an earlier laboratory study, for which their zygosity was established by elaborate serological analysis [25], it was possible to validate our questionnaire method for these 148 individuals. When pairs with mean zygosity scores of 4 or higher are classified as monozygotic, and all others as dizygotic, 1 of 49 pairs of MZ twins and 2 of 25 pairs of DZ twins were misclassified, yielding an overall validity of 96%. Zygosity was determined in this manner for the 1,309 pairs, from the first 6 birth-year cohorts, of which at least one member returned a BQ. Of this total, 47% were male pairs and 48% were MZ pairs. Of the 791 pairs of which both twins returned the test data, 35% were males and 54% were MZ pairs. Test data was returned by 73% of individual female twins but by only 47% of the males.
Table 1 - Questions used for zygosity diagnosis*
- A. Your eye color:
- Same color (1) Similar, different shade (0) Different color (-5)
- B. Your (natural) hair color:
- Same color (1) Similar, different shade (0) Different color (-5)
- C. (Even identical twins sometimes differ quite a bit in height or weight as a result of accident or illness. In answering the questions below, try to ignore such differences.)
- 1. During your childhood, were you and your twin as alike as "two peas" or were you no more alike in appearance than ordinary brothers or sisters?
- Like "two peas" (3) Ordinary likeness (0) Quite unlike (-3)
- 2.When you were school age, were you similar enough in appearance so that people had difficulty telling you apart?
- Never (-2) Sometimes (2) Even family had difficulty (3)
- 3. Could you ever have fooled friends or family by pretending to be your twin?
- Yes (2) Maybe (0) No (-1)
*Weights assigned to each alternative answer are shown in parentheses. These weights were assigned rationally (eg, different eye or hair color ensures dizygosity; general physical similarity gets more weight than similarity on one trait) and the cutting score chosen inspection of the bimodal distribution for the total same-sex sample. Twins are classified as monozygotic if the sum of the weighted responses is +4 or higher.
On the second (and final) telephone prompting of persons who had not returned the BQ, they were told that participation of both twins would cause their names to be entered in a lottery with a prize of $1,000 and that there would be a separate lottery for each birth year so that participating pairs would have nearly one chance in 200 of winning. To pay each twin in 1600 pairs even $10, probably a threshold inducement, would have to cost $32,000 as compared to $6,000 for the lottery. By entering all participants' names in the lotteries but revealing this inducement only to the most reluctant, as a last-chance recruitment effort for those few, we managed to increase the number of tested individuals by about 17% and the number of concordant pairs by 30% (see Table 2) while at the same time giving the more cooperative pairs an equal opportunity to win the prizes.
It is important to note that the lottery was used as an incentive only with the minority of reluctant twins who had already indicated that they would not participate without some added compensation. Therefore, our participants are not selected for cupidity nor as persons who are especially attracted by games of chance. That is, our sample includes all those subjects who would normally participate in such studies, either out of interest or a desire to contribute to research, plus an additional group who would not otherwise have been sampled, and it is thus rather more representative of the general population than it would have been without the lottery.
|
Individuals |
Concordant pairs |
||||||
|---|---|---|---|---|---|---|---|
|
N |
% |
N |
% |
||||
|
Biographical Questionnaires sent |
3,258 |
1,629 |
|||||
|
BQ returned: |
No prompt |
2,385 |
73 |
1,027 |
63 |
||
|
One prompt |
2,573 |
79 |
1,136 |
70 |
|||
|
Lottery prompt |
2,601 |
80 |
1,163 |
71 |
|||
|
Test booklets sent |
2,585 |
1,155 |
|||||
|
Tests Returned: |
No prompt |
1,160 |
45 |
354 |
31 |
||
|
One prompt |
1,619 |
63 |
598 |
52 |
|||
|
Lottery prompt |
1,909 |
74 |
791 |
68 |
|||
*Telephone prompting increases return rates substantially. Offering a chance to win $1,000 in a lottery on the second prompt does not influence many of those who did not even return the BQ but it increases substantially the willingness of (especially male) twins to spend several hours filling out questionnaires.
Although this lottery inducement worked tolerably well, we have subsequently reverted to direct payment to reluctant subjects. This has the disadvantage that the more cooperative individuals, who comprise the vast majority, receive no monetary compensation. However, direct payment of $10 seems to attract as a large a proportion of the reluctant individuals as does the lottery incentive and lotteries, it turns out, are illegal in many jurisdictions. We rationalize the fairness issue by pointing out that we are dealing in a seller's market; the majority, who particippate for nonmonetary incentives, are apparently adequately compensated while the reluctant minority would not feel adequately compensated without the cash payment. The objective, after all, is to approximate as nearly as possible a final sample that is truly representative of the population of all twins originially identified.
Our purpose in establishing the Registry was two-fold. First, we wished to have ready access to large, relatively unselected samples of twins and their families as a research resource for ourselves and for other investigators. The Minnesota Study of Twins Reared Apart [3, 4] has generated many hypotheses that require testing on large, representative samples of middle-aged twins. Secondly, we wished to make first use of this resource by collecting such data from the located twins as was possible and useful by mail. This second purpose fed into the first in several ways. The biographical data make it possible (a) to establish twin zygosity with high confidence, (b) to determine their educational level, occupation and, from these, their socioeconomic status, and (c) to determine the make-up of their immediate families.
Another way in which our preliminary data collection contributes to the usefulness of the Registry can be derived from Festinger's theory of "cognitive dissonance" which predicts that persons who have invested of their time and effort in some enterprise will be disposed to regard that enterprise as worthwhile, a good investment, rather than to have to believe that their decision to participate had been foolish or misguided. This principle predicts that, having responded to our first modest request for biographical information, twins will be more likely to also respond to a second mailing asking them to complete the personality inventory and questionnaires about their occupational and leisure-time interests.
This principle will operate, however, only if the participants are treated considerately, provided with interesting feedback and with other evidence that the research to which they have contributed has scientific value. We send each participant a computer-generated, personalized report of his or her scores on some 60 factor-analytically derived dimensions of personality and interest, plus 10 superfactors based on the intercorrelations of the sub-scales. We also send out newsletters, irregularly, to all participants with reports on our various studies and other information of interest to twins. The increased media attention to twin research in recent years, especially to our Minnesota Study of Twins Reared Apart [3,4] has also helped to assure our participants that they are contributing to an interesting and scientifically productive effort.
Fig. 2Recruitment success for the first 6 cohorts, plotted by individual and by concordant pairs. 92% of the individual females returned a Biographical Questionnaire (BQ) while 63% of the female pairs were concordant for returning all tests. For males, 77% of individuals returned a BQ but only 34% of the pairs were concordant for testing.
The first 6 birth-year cohorts located were the same-sex twins born in 1936, 1937, 1938, 1949, 1954, and 1955, spanning the 20 years of the Registry. A total of 2,912 same-sex pairs were live-born in these six years, of which 779 pairs (27%) are known to have been broken by death of one or both members by the time of this study. We managed to locate both members of 1,673 pairs (78%) of the presumed survivors and, as shown in Table 2, BQs were sent to 3,258 individuals (in 44 pairs one twin was incapacitated or asked not to be contacted). Completed BQs were received from 2,385 (73%) of the individuals (and from both members of 63% or the pairs) within 8 weeks. One telephone prompt increased the return rate to 79% and 70%, for individuals and concordant pairs, respectively. A second telephone prompt, which offered a chance for the pair to compete in the $1,000 lottery, increased the BQ returns only to 80% and 71%. Recruitment success for individuals and for concordant pairs is graphed, separately for males and females, in Fig. 2.
The 700-item personality and interest test battery was then sent to the 2,585 individuals (including 1,155 concordant pairs) who had returned the BQ and also indicated their willingness to participate further. This set of 7 questionnaires required several hours of the subjects' time and therefore, not surprisingly, there was considerable attrition in responding, especially among male twins. Once again we employed a maximum of 2 telephone prompts of subjects who had not yet returned the completed tests. Once again, the reluctant subjects were informed, on the second prompting, that the names of all pairs in which both twins were participants would be entered in the lottery. As shown in Table 2, 45% of the individuals, and both members of 31% of the pairs, returned completed questionnaires without prompting. The first telephone prompt increased this yield to 63% and 52%, respectively, for individual twins and concordant pairs. The second or lottery prompt produced a further increase, to 74% and 68%. As is usually found in twin studies [27, 29], women were much more likely to participate than men (80% vs 56%.) A similar sex ratio was found for spouses of participating twins; 78% of the wives and only 58% of the husbands agreed to participate.
The effect of prompting on concordant BQ and test returns is shown graphically in Fig. 3, which illustrates that being DZ, rather than MZ, has an effect on participation similar to being male, rather than female. However, the demands of the long test battery caused greater attrition among males than females, possibly because more men than women find so much reading to be onerous; 2/3 of the female pairs concordant for th BQ also were concordant for returning the tests, as compared with about half of the males. Opposite-sex (OS) pairs were most comparable to same-sex (SS) DZ males.
In Fig. 4 are plotted, for the 1,436 pairs of SS twins from the 6 completed cohorts, of which at least one twin returned a BQ, the proportions of each of the possible patterns of recruitment response. Once again we see the similar effect of sex and zygosity upon recruitment. The only exception consists of the pairs concordant for returning the BQ and for not completing the testing. This group contains a higher proportion of males than of females, reflecting the men's greater reluctance to be tested, but a larger proportion of MZ than DZ twins, reflecting the MZ's stronger tendency to behave concordantly.
Effect of prompting, first 6 cohorts. For females, 34% returned a Biographical Questionnaire (BQ) and tests without prompting, 38% returned tests after prompting, 14% returned BQs without prompting but never returned tests, etc..
The lottery incentive was also employed in recruiting parents, SS siblings, and offspring of the Registry twins (a total of 1,085 spouses of the married twins returned test data when the twins themselves did). From the twin BQs we determined which pairs had both parents living and well, which had at least 2 SS singleton siblings, and which pairs each had one child aged 17 or older and of the same sex as the cotwin's offspring. Special BQs and test materials were sent to these family members, followed up by the usual two telephone prompts. Complete test data were obtained from 233 parents, 169 siblings, and 285 offspring, including 42 pairs who were offspring of DZ twins and are therefore first cousins, and 44 pairs who were offspring of MZ twins and are genetic half-sibs reared separately in intact biological families. Some 69% of the spouses of the married twins, and all the spouses of twins whose offspring participated, contributed complete test data.
Our subsequent experience with the next 4 birth-year cohorts indicates that siblings and offspring can be recruited with greater success by sending the materials directly to the relative (rather than to the twins to be passed on), and by offering a cash payment of $10 at the second telephone prompt. Already, 212 parents, 146 sibs, and 243 offspring have returned test materials from these four cohorts, about 69% of those solicited and about 31% better than the return rate achieved in the 6 cohorts for which complete returns are in. A total of some 445 parent, 315 SS sibs, and 528 offspring have provided test data thus far and this total should double by the time the registry is complete.
Fig. 4 All possible within-par patterns of recruitment response for the 6 completed birth-year cohorts, excluding pairs where neither twin returned even the Biographical Questionnaire (BQ), plotted for male vs. female and for MZ vs. DZ pairs. Being male has a similar effect on recruitment to being DZ, except for those pairs concordant for returning only the BQ.