Age Validation in the New England Centenarian Study
by T.T. Perls, K. Bochen, M. Freemen, L. Alpert, and M.H. Silver
[ References ]
The New England Centenarian Study (NECS) is an attempt to find all centenarians in a population-defined area of suburban Boston. By doing so, we hope to minimize selection bias in the study. To locate our potential subjects we use the annual town censuses of eight towns (Belmont, Cambridge, Somerville, Dedham, Quincy, Framingham, Waltham, and Lexington, see Table 1). We chose these particular towns because of their local proximity to the Harvard Division on Aging and our preliminary work which indicated the high sensitivity of their censuses. The censuses are collected by each town's elections office. Last year's accounting of who lives in nursing homes and assisted living apartments is given to the administrators of these settings and the information is updated annually. The censuses are publicly available and they list name, address, birth date, occupation and political party affiliation.
We measure the sensitivity of these lists by monitoring other sources for the detection of centenarians including local obituary listings, councils on aging and features in local newspapers. We regularly check in with the nursing homes' social workers to ascertain new centenarian residents and correspond with physicians and visiting nurses who have large geriatrics care panels.
Subjects and their families are contacted by mail either directly or via a nursing home's social worker. A letter includes a description of the study and a postage-paid response card which enables the subject or their proxy to indicate that they are indeed a centenarian, whether or not they have a living sibling and if they would allow us to contact them for enrollment. If no response is received, a follow-up letter is sent. Attempts to contact the subject or their family by phone are made if previous attempts by mail have not been successful. In that case we also contact neighbors to determine if the subject is even alive and finally, if need be, we visit the address. Another valuable resource is the Massachusetts Department of Vital Records which keeps a computerized list of recent deaths.
Upon initial contact with the subject or family, we ask if they possess a birth certificate. If they do not, then we obtain the necessary information to obtain a birth certificate ourselves. Even if the subject and/or their family are inclined to not participate in the study, we still ask them the favor of giving us this information so that we can, at the least, verify the person's age. If after all efforts are exhausted, we are still unable to obtain a birth certificate, than we request multiple corroborative pieces of evidence indicating the person's age. These might include: military certificates, an old passport, school report card, family bible, and baptismal or other church certificate. In all cases, if the subject had children or siblings, careful note is made of their birth dates and birth order to see if their ages make sense in relation to the centenarian's age.
After being satisfied that we had made a reasonable attempt at finding all the centenarians in our sample area and obtaining their birth certificates, we chose a date in time to count the centenarians. This date was December 31, 1996. During the following two days, we telephoned all of the subjects and their families to verify that they were indeed alive on December 31, 1996.
Table 1 provides population size segregated by age group according to 1995 data provided by the Massachusetts Institute for Social and Economic Research. The estimated frequencies of those age 65 and older and those age 85 and older are similar to other western countries such as Germany, Denmark and Sweden (Kannisto 1995). From the censuses of the eight towns comprising our study sample, we obtained 285 names of people who had birthdays prior to January 1, 1897. By our surveillance methods, we obtained another four names; one from a nursing home social worker indicating that they had admitted a centenarian recently (who came from a town outside of our sample area), another from a local newspaper which had a feature article on the birthday of a centenarian previously unknown to us and two names from local newspaper obituaries. Therefore we had a list of 289 potential centenarians in the 8 towns.
Figure 1 summarizes our investigation of these 289 persons of whom 186 (64%) were found to have died. Many of these people had died prior to 1996 but had not been deleted from the censuses. Thirteen (4%) had moved to outside the sample area. At the time of manuscript submission, we had not been able to determine the status of 10 (4%) people listed. In all 80 people (28%) were found to be alive with a census record indicating a birth date prior to January 1, 1897.
Of the 80 people found to be alive, 13 (16%) were found to have incorrect birth dates listed on the censuses. In these instances, we contacted the potential subject or their proxy and were told that the person was not a centenarian. Several were younger; the birth date was simply incorrectly noted in the census. In about half of the cases, the census had stated an incorrect century of birth, though the month and day were correct.
Population totals for the eight towns of the sample also stratified by age groups age 65 years and older
Status of 289 people listed in the censuses as having a birth date prior to January 1, 1997.
Forty-two (53%) were alive and enrolled. Another 14 (18%) people were located who (or their proxy) stated that they were 100 years old and were willing to participate in the study. Eleven (14%) of the people (or their proxies) refused to participate in the study but also stated that the person listed was indeed one hundred years old or older. Therefore, among those with census-recorded birth dates prior to January 1, 1897, the specificity of the censuses for stating the number of people alive and living in the eight towns was 28% (80/289) to 31% (90/289 which includes the 10 people we have not been able to locate). Using additional sources to locate centenarians, as noted in the methods, we have found only four centenarians not listed in the censuses and therefore the sensitivity of the censuses approaches 100%.
We attempted to obtain proof of birth date for the 67 subjects (the sum of: 42 enrolled subjects plus 14 'willing to enroll subjects' plus the 11 people who refused to participate in the study) who verbally substantiated the census data indicating a birth date prior to January 1, 1897. This effort is summarized in Figure 2. Our success rate for obtaining proof was highest among those already enrolled (of 42 subjects, 34 birth certificates plus one old passport were obtained). In other words, we had an 83% success rate in obtaining proof if given the opportunity to obtain the necessary information (names of parents including the mother's maiden name, date and place of birth). Because of the Armenian massacre in 1919, in which all Armenian birth records were destroyed, it was impossible to obtain birth certificates for three potential centenarians who were Armenian.
Seven of the 14 subjects who we were in the process of enrolling lived in nursing homes. It is a significant challenge to enroll subjects living in nursing homes because we are at the mercy of the homes' social workers to forward our correspondence to the proxies and/or to the centenarians. Until permission is obtained from the proxy or centenarian, we are unable to obtain even enough information to search for a birth certificate. We had a 43% success rate (n=6) in obtaining birth certificates for this group. Five people (and their families) that refused to participate were still willing to give us the information we needed to track down a birth certificate, however, six (55%) were not willing. Therefore a total of 45 birth certificates and one old passport (issued in 1950) were obtained as proof of age for 46 centenarians living in our sample area.
Subject status of centenarians (assuming those even without birth certificates are age 100 or older) within the 8 town sample and living situation
Results of initiative to obtain birth certificates on all people claiming to be centenarians or said to be centenarians by their proxies.
Of the 67 people stated to be centenarians by either a proxy or the centenarian themselves, we obtained 45 birth certificates and one old passport. We did not encounter documentation disproving the ages of the remaining 21 people. Nor did we encounter a situation in which the children were either too young or too old to have centenarian parents. However, to be conservative in our estimate, we can safely state that there were 46 centenarians living in the eight towns. As noted in Table 1, there are approximately 460,829 people in the 8 towns and therefore approximately 1 centenarian per 10,000 people. Given the 1990 U.S. census estimate of 6,016,425 people in Massachusetts, our ratio would translate into 601 people for the State.
One must be cautious in applying these ratios to the State as a whole, given that urban communities such as the ones that make up our sample, may be different in their tendency to have centenarians compared to, for instance, rural communities. For instance, we suspected there would be a significant difference between rural and urban towns in terms of the density of nursing home beds. This would be important since, as Table 2 indicates, a significant proportion of centenarians in each town, ranging from 25% to 100%, live in nursing homes (mean of 61% for the 8 towns combined). According to the Massachusetts Department of Public Health, as of November, 1996, there were a total of 58,809 nursing home beds in the State. Again, given the 1990 U.S. census estimate of 6,016,425 people in Massachusetts, the per capita nursing home bed rate for the State would be approximately one bed per 102 people. The State's rate was actually higher than the rate for the 8 towns combined, which was one bed per 110 people and thus we do not suspect that our estimation of centenarian prevalence is inflated on the basis of nursing home bed density in our population-based sample.
Using surveillance methods other than the censuses, we found the sensitivity of the censuses to be close to 100%. Several factors contribute to this exceptional rate including the requirement to be counted by the census in order to vote (a right which older people take very seriously), nursing home administrators' participation in the census collection, annual performance of the census, and the procedure of follow-up by census officials when there is missing data relative to the previous year's results.
In contrast to the sensitivity, the specificity of the census was very poor at a rate of 28% to 31%. A large proportion of those listed in the census had already died. Though an important reason for this is the high mortality rate among centenarians (about 50% annually in our experience), we also found that many had died prior to 1996. One town official admitted to us they were less diligent about following up on potential deaths since it is to the town's advantage to report as high a prevalence of elderly as possible in order to qualify for State programs for older people. Therefore, we can rely upon the censuses in these towns to find nearly all of the centenarians. On the other hand, the specificity is so poor that significant effort must be made to determine who on the list is actually an alive centenarian.
For this New England sample of centenarians, if we had the opportunity to obtain enough information to obtain a birth certificate, our success rate was 81%. This was the rate achieved for the 42 alive subjects enrolled in the study as of December 31, 1996. For one subject we accepted an American passport issued in 1950 as proof of age. She was born in Quincy, Massachusetts but for unclear reasons, the town could not locate proof of her birth. This subject also has two alive sisters in their mid and late nineties who corroborate her claim of being 103 years old.
The forms of proof-of-age that are acceptable in a centenarian survey vary according to circumstances such as what is reasonably possible for the country in which the survey is performed and the age claimed by the subject and/or their family. There is currently a need for groups performing such studies to reach a consensus regarding acceptable criteria. In our case, we accepted birth certificates and old passports (the older the better, but in general, issued before 1963 when Medicare was created) as proof. The age of children and siblings was always helpful as additional supportive evidence. Military certificates, baptism records, and entries into a family bible at the time the person was born should be acceptable. For these secondary sources of proof, multiple corroborating pieces of evidence would be preferable to only one supportive document. Again, a consensus on the validity of these latter forms of proof must still be reached. Potential forms of proof which we did not accept included naturalization certificates and old photographs with details written on the back. In the case of naturalization certificates, it may have been common for young people to exaggerate their age while immigrating into America in order to qualify for work. In the case of the photographs, it would be unclear when the information written on the back of the photo was actually written.
New England may be special among areas in the United States in terms of our ability to obtain birth certificates. Because New England is the oldest settled area of the country, it has a longer history of keeping birth records. Of those centenarians born in Boston, we had a 100% birth certificate retrieval rate. However, according to the 1990 U.S. census, one third of the current Massachusetts population is foreign born: of those foreign born approximately 21% come from Ireland, 14.5% from England, and 10% from French Canada. In our study, 52% of our subjects are foreign born. Especially in the cases of Ireland and Quebec/Nova Scotia, we still had very high rates of birth certificate retrieval. The experience of obtaining proof-of-age for the oldest old in other parts of the United States has yet to be determined. It must also be noted that the vast majority of our experience has been with Caucasians. In the two cases we have enrolled African Americans (one born in Cambridge, the other born in the West Indies) we were able to obtain birth certificates but this is certainly not an adequate sample to predict success of birth certificate retrieval according to race.
There were an additional 21 people who were potentially centenarians but we either could not or did not have the opportunity to obtain adequate proof of their ages. The majority of these cases were those living in nursing homes and contact had not yet been made with their families. Others included those who refused to participate and would not give us the information we needed to obtain a birth certificate and those who were Armenian. To err on the side of being conservative, when calculating the prevalence of centenarians, we only included those with acceptable proof-of-birth.
The age range of the people for whom we obtained proof-of-age was 100 years to 107 years. During our two year experience of enrolling subjects however, we have had the opportunity to enroll a subject age 109 years and another, age 111 years. We were able to obtain birth certificates for these individuals and the ages of their children were consistent with the reported age as well. Beyond these two subjects, we do not have an appreciable experience with 'super centenarians' (those over the age of 110 years). Given the controversy over the now infamous 130-160 year olds of the Russian Caucasus and the case of Shigechiyo Izumi whose age at death (120 years and 237 days) is still mistakenly listed as the oldest ever in The Guinness Book of Records (Wilmoth and Lundstrom 1996; Matuzaki 1989) it is probably prudent to be skeptical of claims which involve ages above 110 years. When such cases are claimed, it is important to undertake a much more thorough determination of age validity as has been done in the cases of Jeanne Calment who recently died at age 122 and Chris Mortensen, a Danish man, who immigrated to the US and who recently died at age 115 (Robine and Allard 1995; Wilmoth et al. 1996).
We were concerned that perhaps our eight town sample was not generalizable to the State as a whole because of a relatively high prevalence of nursing home beds. Since we had found that about 60% of the centenarians live in nursing homes, it seemed reasonable that where there were more nursing home beds, there would also be more centenarians. In fact we found the State had a greater per capita nursing home bed rate than our sample area. Also of note is that while we did find that the majority of centenarians (60%) live in a nursing home setting, a remarkable 12% live at home alone. The remaining 27% live with family. One of our subjects living with family would qualify as living alone (thus increasing the prevalence to 13%) since she was functionally independent and was actually looking after her 75 year old son.
We noted a centenarian prevalence rate within our eight town sample of 1 per 10,000. This ratio was calculated using our 1996 search for centenarians and 1995 estimates for the population sizes of the eight towns. The populations for the eight towns were probably larger at the end of 1996 and therefore, the prevalence rate was probably slightly lower. This prevalence rate is about twice of what has been observed for other industrialized countries where the estimate is 50 per million (Wilmoth 1994; Kannisto 1994) . On the other hand, our observed prevalence rate is comparable to rates found in the national Danish, French and Italian centenarian studies (correspondence with James Vaupel, Michel Allard and Claudio Franceschi respectively) though a little higher. A higher rate could be explained by the higher life expectancy after age 80 in the United States compared to other industrialized nations (Manton and Vaupel 1995).
Centenarians represent an extremely valuable resource for the study of successful aging and factors associated with achieving extreme longevity. Previous studies of centenarians have been faulted for not verifying the ages of their subjects, but in our experience with centenarians ages less than 110 years, we have not encountered subjects who misrepresent their age. Additional studies which include minorities are necessary to determine the validity of age reports in groups where cultural differences may yield different results. The low specificity of the censuses used in this study and the potential for selection bias underscores the importance of performing population-based studies when determining prevalence rates and socio-demographic and medical characteristics of distinct groups such as centenarians.
We greatly appreciate the assistance of Robert Condon Jr., of the Massachusetts Department of Public Health, Thomas Chung, Ph.D. from the Massachusetts Executive Office of Elder Affairs and Alice Rarig of the Massachusetts Institute for Social and Economic Research, for providing us with nursing home and State population data. Support: Alzheimer's Association Darrell and Jane Phillippi Faculty Scholar Award, AARP/Andrus Foundation Research Award, and a National Institute on Aging sponsored Physician Scientist Award, AG-00294.
[ Home | Contents | Return to previous page ]