Although more expensive and time

Transcription

1 Article Methods Tracing 8,600 participants 36 years after recruitment at age seven for the Tasmanian Asthma Study Cathryn Wharton, Shyamali Dharmage, Mark Jenkins, Gillian Dite, John Hopper Centre for MEGA Epidemiology, Department of Public Health, University of Melbourne, Victoria Graham Giles Cancer Epidemiology Centre, The Cancer Council Victoria Michael Abramson Department of Epidemiology and Preventive Medicine, Monash University, Victoria E. Haydn Walters Department of Medicine, University of Tasmania Although more expensive and time consuming, prospective cohort studies are often preferred to other observational epidemiological designs as they permit temporal observations of exposure and outcomes. A high rate of participant follow-up is crucial in cohort studies to maximise participant response and reduce a response bias in the results. Ideally, participant contact details need to be continuously updated between follow-up studies to minimise attrition. Participant tracing can be a large obstacle to conducting cohort follow-up study and ultimately obtaining study data. Manual tracing methods of participants in follow-up studies have been published. 1-6 Large follow-up studies, such as for the Women s Health Australia Project, 5 the Nurses Health Study 7 and the Women s Health Initiative Study, 8 all surveyed participants on a regular basis (every six months to every three years) and relied on manual tracing methods to maintain contact details. While manual tracing methods might be reliable, they are quite often slow and expensive; large amounts of staff time are required to locate one individual at a time. With the increasing availability of many electronic databases, the option of using electronic tracing methods is now receiving much attention. However, little information is available on the utility of electronic tracing methods. This is particularly the case after long periods since the last contact, such as with the follow-up study of the Tasmanian Asthma Study (TAS). We are conducting a 36-year follow-up of the TAS where participants were first surveyed in 1968 (n=8,583) and only their name, date of birth and gender were recorded. There have been three follow-up surveys of subgroups of the cohort in 1974, 1981 and This longitudinal study now presents an excellent opportunity to assess the respiratory health in middle-aged adults with prospective data on exposures in childhood, adolescence and young adulthood. However, the success of this follow-up study first relied on tracing of the participants. We aimed to identify residential addresses of the original 8,583 participants in order to resurvey this cohort by mailed questionnaire. Given the large number of participants and Abstract Objective: To trace all participants 36 years after the original Tasmanian Asthma Study (TAS). Methods: In 1968, the TAS investigated asthma in 8,583 children who were born in We attempted to trace these participants in using names, dates of birth and gender. Current addresses were sought by computer linkage to the Commonwealth Electoral Roll, the Medicare database and the Tasmanian marriage records. Computer linkage was conducted with the National Death Index (NDI). Siblings of participants were also linked to the Commonwealth Electoral Roll and those identifi ed were sent a letter requesting the participant s address. The Australian Twin Registry (ATR) and the TAS substudy were used to locate participant addresses. Results: After three rounds of electoral roll linkage, 56% of all cohort members were traced. Name changes were identifi ed for 49% of the 3,477 females not initially matched to the electoral roll using linkage to marriage records. NDI linkage yielded a 0.7% match. Medicare linkage identifi ed addresses for 27% of the 1,982 remaining participants. Writing to siblings located 60% of 1,661 participants. One hundred and eighty-three participants were matched to the TAS and 23 twins matched to the ATR. Overall, 81.5% of the cohort members were identifi ed. Conclusions: With these methods, we have been able to trace a possible address for a large portion of the original participants, with the electoral roll linkage being the most useful. Implications: It is possible to trace Australians for follow-up studies using electronic linkage, although without unique identifi ers it is labour and resource intensive and requires matching to several databases. (Aust N Z J Public Health 2006; 30: ) Correspondence to: Dr Shyamali Dharmage, Centre for MEGA Epidemiology, Department of Public Health, University of Melbourne, Carlton, Victoria Fax: (03) ; s.dharmage@unimelb.edu.au Submitted: January 2005 Revision requested: July 2005 Accepted: December VOL. 30 NO. 2 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 105

2 Wharton et al. Article paucity of identifying information, our first tracing efforts relied on computerised record linkage because manual tracing methods would have been slow and costly. Computerised record linkage is the process of using computerised systems to match two or more records containing identifying information and determine if these records refer to the same individual. 9 As there was no single unique identifier for TAS participants, record linkage was conducted using a combination of name, gender and date of birth that together formed an identifier likely to be unique. 10 Several manual methods of tracing were also incorporated into this follow-up study. This paper presents the different methods used to trace participants, the processes, requirements and results. Figure 1: Sequence of tracing processes for TAS participants. Main participants (n= 8,583) National Death Index linkage No match Electoral Roll linkage Round 1 (original names) Males not matched Females not matched Tasmanian marriage records Deceased Methods The Tasmanian Asthma Study In 1968, 99% of all Tasmanian school children born in 1961 (n=8,583) (51% male: 49% female), their parents (n=16,266) and siblings (n= 21,043) were surveyed by questionnaire as part of the TAS. Name, date of birth and gender were recorded for each of the participants born in 1961 but no address information was collected. Subsequent follow-up surveys were conducted on subsets in 1974 (n=7,500), 1981 (n=850) and (n=1,500) using manual tracing methods. Address information was collected only for the 1981 and studies. In 2002, we began tracing the 8,563 original participants for a 36-year follow-up study. Each tracing method used, the costs, and time required are discussed below. The sequence followed is illustrated in Figure 1. Sibling addresses acquired by data matching to Commonwealth Electoral Roll Not matched Not traced Electoral Roll linkage Round 2 (new surname) Not matched Traced Electoral Roll linkage Round 3 (Six month update) Not matched Not matched Not traced Not traced Sibling mail out Medicare linkage Australian Twin Registry TAS follow-up Traced Traced Deceased Withdrawn 1. Linkage to the National Death Index The National Death Index (NDI) contains data on approximately 2.7 million deaths occurring in Australia since As seen in Table 1, information stored by the NDI includes identifying information as well as death data. NDI linkage was carried out by staff at the Australian Institute of Health and Welfare (AIHW) using first name, middle name or initial, surname, date of birth and gender. Probabilistic linkage rules were used where multiple passes grouped the data based on different characteristics and applied a weighting for the type of match. The linkage factored in minor alternatives of spelling and phonetic variations of names and all matched pairs were clerically reviewed to maximise the number of Further tracing required Follow-up postal survey sent to traced participants correct matches. Siblings and parents (who were surveyed in 1968) were also linked to the NDI to update their vital status, although parents dates of birth were not used in the data matching as this was not recorded at baseline. 2. Linkage to the Commonwealth Electoral Roll The Australian Electoral Commission (AEC) maintains the Commonwealth Electoral Roll and administers compulsory Table 1: Linkage to the National Death Index (NDI). Custodial Cost for TAS linkage Time required for TAS linkage Australian Institute of Health and Welfare (AIHW). Linkage is confi ned to health research. Requires approval from Human Research Ethics Committee (HREC). Submission must be made to AIHW Ethics Committee, which will assess it at a quarterly meeting. Study must be judged to be scientifi cally valid. Researchers must comply with legislation (AIHW Act and Privacy Act) and sign an Undertaking form. Instructions and application available on AIHW website. Information from matches includes: name, alternative names, date of birth or estimated year of birth, age at death, date of death, sex, marital status, Aboriginality, State/Territory of registration, registration number, ICD cause of death codes and other causes of death codes. $A250 application fee to AIHW Ethics Committee plus $A8,000 to match 45,900 TAS records. 2-3 weeks (after assessment at quarterly AIHW Ethics Committee meeting). 106 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 2006 VOL. 30 NO. 2

3 Methods Tracing study participants enrolment and voting for Australians aged 18 and over. Three rounds of probabilistic data matching to the Commonwealth Electoral Roll were conducted by AEC staff using participant name, date of birth and gender. An initial round of linkage was conducted with the recorded names; a second with new surnames for females identified from the Tasmanian marriage records; and a third round to comply with the AEC expiration which specify that extracted data must not be used after six months from the date of matching. The probabilistic linkage used a phonetic based algorithm to score each match and those with a match score of 95% or higher were used. Table 2 shows the and costs of data linkage to the electoral roll. 3. Linkage to the Tasmanian marriage records The Tasmanian Registry of Births, Deaths and Marriages is the repository and disseminator of records relating to births, deaths and marriages occurring in the State of Tasmania. The details for female TAS participants not traced in the first round of electoral roll matching were linked to the Tasmanian marriage records to identify a change in surname. The marriage record matching was a two-step process where registry staff first matched names electronically using deterministic linkage and then checked the date of birth manually to verify positive matches (see Table 3). The deterministic matching meant that slight variations in birth date or names, including hyphenated names, were not identified. For multiple marriages the most recent name was recorded. Linkage to the electoral roll was repeated with the new surname. 4. Linkage to the Medicare database The Health Insurance Commission (HIC) manages Medicare, the Pharmaceutical Benefits Scheme and the 30% Private Health Insurance Rebate. In June 2005, more than 20.5 million Australians were registered for Medicare benefits. 11 Unlike the AEC, the HIC does not release address information for medical research, but will conduct mail outs to potential participants on behalf of researchers (see Table 4). Deterministic computer linkage was carried out by staff of the HIC. Those participants who were traced using this method were sent the follow-up survey with a Table 3: Linkage to the Tasmanian marriage records. Custodial Cost for TAS linkage (including GST) Registrar of Births, Deaths and Marriages (Tasmanian Department of Justice). Access is permitted for medical organisations. Evidence of HREC approval is required. Conditions with the use of data apply. Groom s given and surname, date of marriage, registration number and year of registration. $A410 for electronic matching plus an additional $A660 for manual verifi cation of 3,500 TAS records a Time required for Three weeks once approved by the registry. TAS linkage Note: (a) This cost underestimated the time and resources required and would be more for future exercises. Table 2: Linkage to the Commonwealth Electoral Roll. Custodial Australian Electoral Commission (AEC). Access to electronic data is permitted for medical research. Evidence of HREC approval is required. Researchers must sign a safeguard agreement (which specifi es of access to data, including a statement that extracted data cannot be used after six months from the date of linkage). Current street address. Cost for TAS records: $A165 linkage (including 501-1,000 records: $A275 GST) 1,001-5,000 records: $A385 5,001-10,000 records: $A550 10,001-20,000 records: $A715 20,001-30,000 records: $A880 30,001-40,000 records: $A1,045 Time required for Three weeks once approved by AEC. TAS linkage covering letter written by the HIC explaining how the participants had been traced and assuring them that their privacy had been maintained. The study investigators only obtained participant addresses if participants returned their completed survey with their contact details. 5. Mail-out to located siblings Names and dates of birth of siblings of untraced TAS participants were linked to the Commonwealth Electoral Roll. If an address was identified the sibling was mailed a survey asking for contact details of their brother or sister (the main participant). Two repeat letters were sent to non-responding siblings (see Table 5). 6. Australian Twin Registry The Australian Twin Registry (ATR) is a volunteer registry of more than 30,000 pairs of twins and its primary function is to facilitate medical and scientific research that requires access to Table 4: Linkage to the Medicare database. Custodial Cost for TAS linkage Time required for TAS linkage Health Insurance Commission (HIC). Evidence of HREC approval is required. A Confi dentiality Deed Poll agreement must be signed by the researcher. Participant documentation is assessed by HIC Privacy Committee. HIC will only do a mail-out on researcher s behalf (address information is not released). HIC can advise which participants have been matched. Once the participant receives their letter it is their choice to respond and provide address details. $A1,350 (to match 2,000 TAS records and for a mail-out to 453 of the 542 matched records). Three months to arrange and carry out linkage. Additional time was required to do the mail-out VOL. 30 NO. 2 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 107

4 Wharton et al. Article Table 5: Tracing with a mail-out to siblings. Cost for TAS Time required for TAS tracing HREC approval was required to approach the siblings and for mail-out documentation. The mail-outs were prepared internally to reduce costs. The form asked for changes of name and the current addresses and phone numbers of the main participants, their parents and other siblings. $A7,000 for stationery and postage costs (letters sent to 4,521 siblings over three mail-outs totalling of 7,522 letters). Six months (for three mail-outs and data entry of addresses on returned forms). twins. 12 Names and dates of birth of twin participants identified at baseline were provided to the ATR to identify registered twins. The ATR staff contacted participants to obtain their consent to provide current contact details to TAS researchers (see Table 6) TAS follow-up In , a follow-up postal survey was conducted of 1,500 TAS participants. 13 For participants not traced by the methods already discussed the addresses from the study were checked against the national Electronic White Pages to determine if they were current. There was no cost for this process, however it required approximately one day of a staff member s time. An MS ACCESS database was used to record all information for each participant, including changes of names, addresses, phone numbers, as well as the tracing processes used, communication with participants and the vital status of each participant. In the instance where a participant was traced through multiple methods, then the source that first identified the address was noted as the source of tracing. If a participant was traced to multiple addresses then these were manually assessed and the address from the most reliable source was selected. This study was approved by the Human Research Ethics Committee of the University of Melbourne. Results The tracing outcomes for the various methods are illustrated in Figure 2. Linkage to the NDI yielded matches for 0.7% (n=62) for the main participants (aged years), 1% (n= 206) for their siblings, and 0.7% (n=108) for their parents (who were matched by name and gender only as dates of birth were not recorded in 1968). Using the original names of participants collected during the 1968 study, 44% (3,180) were matched to the electoral roll, with the majority being male (81%). Of the untraced females, 49% (1,713) had a surname update from the Tasmanian marriage records. Meanwhile, an address for 702 of these female participants was identified through the mail-out to siblings that was done concurrently with the marriage registry matching. Therefore, the details of 1,011 of these women were sent for a second round of matching to the electoral roll, which yielded a 71% match rate. For the six-month update, the third linkage to the electoral roll for 8,263 participants (excluding those deceased, withdrawn or with no date of birth) identified 62% of participants (n=5,099). Many of the participants remaining untraced had living siblings noted in The details of these siblings were sent for linkage to the electoral roll and 4,521 were matched (41%). These siblings were mailed a request for the contact details for 2,759 untraced main participants (with many of the untraced participants having more than one sibling who was posted a letter). The siblings returned 2,230 forms over a six-month period providing an address for 60% (1,661) of the main participants. The siblings also Figure 2: Outcomes of tracing processes for TAS participants. 1. National Death Index linkage 2. Commonwealth Electoral Roll data matching Number of participants sent for linkage or attempting to trace 8,583 main participants 21,043 siblings 16,266 parents Round 1 (names from 1968): 8,521 Round 2 (married names): 1,011 Round 3 (6 month update): 8,263* Number matched or located 62 main participants (0.7%) 206 siblings (1%) 108 parents (0.7%) 3,810 (81% M: 19% F) (45%) 719 (71%) 5,099 (59% M: 41% F) (62%) 3. Sibling mail out (concurrent with marriage records linkage and Medicare linkage) Letters to 4,521 siblings seeking 2,759 main participants 1,661 main participant s details provided (60%) with 2,230 forms returned. Table 6: Tracing using the Australian Twin Registry (ATR). Custodial The University of Melbourne. An application kit to access the ATR is available on the registry website. Evidence of HREC approval is required. The application is assessed by the ATR Executive Committee using a review process. ATR staff recruit twins on behalf of researchers. Changes of name, addresses, phone numbers. Cost for TAS $A220 Time required Eight hours of ATR staff time for TAS tracing 4. Tasmanian marriage records linkage 5. Medicare database linkage 6. Australian Twin Registry TAS sub-study 3,477 females 1, twins 183 listed on database 1,713 identified with a change of name (49%). (Addresses for 702 were also identified by a sibling.) 542 matched (27%)** (Addresses for 89 were also identified by a sibling.) 23 (48%) identified and 19 (40%) consented 59 on White Pages (32%) 124 not on White Pages * Round 3 of the electoral roll matching excluded 320 participants who were living overseas, withdrawn or deceased as notified by NDI matching or by a family member, or with no date of birth information. ** Participants addresses not provided from Medicare database. 108 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 2006 VOL. 30 NO. 2

5 Methods Tracing study participants informed us of additional deaths not identified by the NDI linkage including 102 main participants, 155 siblings and 1,059 parents, making the total known deaths 164, 361 and 1,167 respectively. Of the remaining 1,986 participants linked to the Medicare database, 27% (542) had an exact match. Siblings had already identified 89 of these participants so 453 were sent the followup survey through the HIC. To date, approximately 12 months after these mail-outs, 129 of the 453 participants had responded and provided their contact details on their returned survey. For the 48 twin participants still untraced, 23 were identified on the Australian Twin Registry database. Of these, 19 provided consent to participate in the current TAS follow-up study, one requested not to be contacted and three were living overseas. Addresses for 183 participants were obtained from the TAS database and 32% (n=59) of these were listed on the Electronic White Pages. Table 7 shows the source that first traced the main participants and the proportion traced by each method. The majority of participants were located using electoral roll matching (56%), which was the first method used and it picked up the less elusive individuals. Table 8 shows the vital status of all participants prior to the TAS 36-year follow-up mail-out. A likely address was identified for 6,998 (81.5%) of the main participants. When including withdrawn or deceased participants, the total traced increased to 84.4%. An additional 15.9% were not yet located and require additional investigation. Based on the addresses obtained from these tracing methods, 72% of the participants still resided in Tasmania while 28% had moved interstate (8% QLD, 8% VIC, 5% NSW, 3% WA, 2% SA, 1% ACT and 1% NT). These tracing processes took approximately 12 months with one full-time research assistant co-ordinating the sibling mail-outs, the linkages through each and the data management. The total cost for the electronic linkage exercises and manual tracing was approximately $20,000, not including the salary of the research assistant. The electoral roll data matching was the most cost-effective tracing method, where the cost for the computer linkage exercise ranged from an average of $0.10-$0.54 per traced participant. Discussion Even without a unique identifier, we have been able to trace 84% of the main participants in this 36-year follow-up study, although the processes were resource intensive. A likely current address has been traced for 82% of the main participants, with more than half of all participants identified through the Commonwealth Electoral Roll. The sibling mail-outs were also valuable in obtaining current information, particularly when addresses provided by the earlier Table 7: Source from which participant addresses were first identified. Number % of main identified participants n=8,583 Commonwealth Electoral Roll 4, % Sibling mail-outs 1, % Medicare database % TAS follow-up % Australian Twin Registry % Total 6, % AEC linkage were out of date. The drawback of this manual method is that it is resource intensive, with staff required to conduct the mail-outs, respond to incoming calls and manually enter addresses on a database. NDI linkage produced a low number of matches and many additional participants were reported as deceased by siblings. While the low match rate might be expected for the main participants (aged years) and their siblings, it was lower than anticipated for their parents, most of whom would be aged over 60 years. This could be attributed to lack of date of birth for parents and the absence of date of birth data on the NDI prior to the early-mid 1990s. Considering the large number of Australians enrolled for Medicare benefits, matching to the HIC database was low. This was probably a consequence of the limited data available with which to perform linkage. Results would have been improved using Medicare registration numbers, but the original TAS was conducted before Medibank or Medicare were implemented. The main limiting factor of this study was the lack of a unique personal identifier for the electronic linkage. We were restricted to linkage that relied on combinations of first names, middle names or initials, surname, gender and date of birth, which are prone to duplication and error, i.e. false positives and false negatives. 14,15 Since the AEC uses a system where matches with a score of 95 or greater out of 99 are accepted, it is likely that false matches will occur at some point. Unfortunately, verifying all information provided from electronic data matching with manual review would be extremely slow and costly. Therefore, we cannot comment on the number of incorrect matches until the current follow-up study has been completed. The use of electronic linkage to collect data on illness and health services utilisation, such as hospital admissions, mortality surveillance and tumour registries, has been documented. 9,15,16 Only a few studies have documented using this technique to trace research participants. The Chicago Detection Heart project 17 aimed Table 8: Vital status for TAS main participants before follow-up postal survey mail-out. Found Withdrawn a Deceased Total traced Yet to trace Participants 6, ,247 1,336 n=8,583 (81.5%) (1%) (1.9%) (84.4%) (15.6%) Note: (a) Withdrawn because of living overseas, injury, disability, poor health or disinterested VOL. 30 NO. 2 AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 109

6 Wharton et al. Article to follow-up 12,409 participants 26 years after baseline and traced 92% by matching name, sex, date of birth and social security number to data held by the Health Care Financing Administration (now referred to as Medicare and Medicaid). In a study of 4,305 New Zealand firefighters, 18 94% of the participants were traced using several national databases to match names, sex, date of birth and national index numbers, which are used for hospital admissions. Both of these studies had a unique identifier to use for the data linkage. The Cornella Health Interview Study 19 used deterministic followed by probabilistic matching methods with name, date of birth, address and phone number against the local Census data to trace 2,500 participants seven years after baseline and 98.7% of participants were located. In comparison to these studies, our lower tracing rate could be attributed to having less identifying information, no unique identifier, and a longer period of time since last contact with all participants. Unlike the New Zealand firefighters cancer study, 18 where 98% of the cohort were male, 49% of the TAS participants were female and more difficult to trace because of name changes. Also, the TAS cohort is now spread across the country and there are few national databases with contact information that are accessible for medical research. In recent years, Australian legislation has become more restrictive with the implementation of the Health Records Act. This has made it even more difficult to trace research participants using existing databases. Given that Australia is one of the most technically advanced countries in the world and has extensive amounts of information in national databases, it would be well suited to a centralised system that would be useful for health research. However, this is not possible under the current legislation. Arguments have been presented for State and national databases that could link health data and individuals using a unique identifier. 16,20 The 1986 Australia Smart Card Bill, which would have in effect facilitated linkage between relevant databases, was not passed because of concerns about privacy, mass surveillance and the huge costs of implementing such a system. Recently, there has again been a push for the planning of a national database and the success of the WA Health Services Research Linked Database is laying the foundations for such a change. 21 Our findings support the development of such a health records database as this would be a powerful tool for health research. Conclusion Using the methods discussed, we have been able to trace a possible address for 81.5% of the TAS participants for a 36-year follow-up study. Tracing Australian research participants with only name, date of birth and gender information is possible through electronic linkage and manual tracing, although this is a costly and time-consuming process. Implications The implementation of a national health database with a unique individual identifier would greatly improve the success in tracking participants in cohort studies, decrease attrition rates and potentially maximise the response rate to follow-up studies. Ultimately, this would improve the quality of epidemiological data. Acknowledgements Tracing for this current follow-up study has been funded by the University of Melbourne, Asthma Victoria and the Royal Hobart Hospital Research Foundation. We wish to thank: John Tonkin from the Health Insurance Commission; John Harding from the Australian Institute of Health and Welfare; Margaret Stretton, John Hardman and Jacqui Knobel of the Australian Electoral Commission; Tony Wright from the Tasmanian Department of Justice, Registry of Births, Deaths and Marriages; and Peggy Johnson and Palma Ragno from The Australian Twin Registry, for their assistance. References 1. Hunt JR, White E. Retaining and tracking cohort study members. Epidemiol Rev. 1998;20(1): Nordberg PM. Leave no stone unturned. Am J Epidemiol. 1992;136(9): Farrington DP. Minimising attrition in longitudinal research: Methods of tracing and securing cooperation in a 24-year follow-up study. In: Magnusson D, Bergmann LR, editors. Data Quality in Longitudinal Research. Cambridge (UK): Cambridge University Press; p Given BA, Keilman LJ, Collins C, Given CW. Strategies to minimise attrition in longitudinal studies. Nurs Res. 1990;39(3): Lee C, Dobson A, Brown W, Adamson L, Goldsworthy J. Tracking participants: lessons from the Women s Health Australia Project. Aust N Z J Public Health. 2000;24(3): Pirie PL, Thomson SJ, Mann SL, Peterson AV Jr, Murray DM, Flay BR, et al. Tracking and attrition in longitudinal school-based smoking prevention research. Prev Med. 1989;18(2): Colditz GA, Manson JE, Hankinson SE. The Nurses Health Study: 20-year contribution to the understanding of health among women. J Womens Health. 1997;6(1): Design of the Women s Health Initiative clinical trial and observational study. The Women s Health Initiative Study Group. Control Clin Trials. 1998;19(1): Howe G. Use of computerised Record Linkage in Cohort Studies. Epidemiol Rev. 1998;20(1): Gill LE, Baldwin JA. Methods and technology of record linkage: some practical considerations. In: Baldwin JA, Acheson ED, Graham WJ, editors. Textbook of Medical Record Linkage. Oxford (UK): Oxford Medical Publications; p Health Insurance Commission [homepage on the Internet]. Canberra (AUST): Australian Government; 2004 [2004 July]. HIC Organisation. Available from: Hopper JL. The Australian Twin Registry. Twin Res. 2002;5(5): Jenkins MA, Hopper JL, Bowes G, Carlin JB, Flander LB, Giles GG. Factors in childhood as predictors of asthma in adult life. Br Med J. 1994;309(6947): Smith ME, Newcombe HB. Accuracies of computer versus manual linkages of routine health records. Methods Inf Med. 1979;18(2): Holman CD, Bass AJ, Rouse IL, Hobbs MS. Population-based linkage of health records in Western Australia: development of a health services research linked database. Aust N Z J Public Health. 1999;23(5): Sibthorpe B, Kliewer E, Smith L. Record linkage in Australian epidemiological research: health benefits, privacy safeguards and future potential. Aust N Z J Public Health. 1995;19(3): Pirzada A, Yan LL, Garside DB, Schiffer L, Dyer AR, Daviglus ML. Response rates to a questionnaire 26 years after baseline examination with minimal interim participant contact and baseline differences between respondents and nonrespondents. Am J Epidemiol. 2004;159(1): Fawcett J, Garrett N, Bates MN. Follow-up methods for retrospective cohort studies in New Zealand. Aust N Z J Public Health. 2002;26(3): Garcia M, Schiaffino A, Fernandez E, Marti M, Salto E, Perez G, et al. The Cornella Health Interview Survey Follow-Up (CHIS.FU) Study: design, methods, and response rate. BMC Public Health. 2003;3(1): Kelman C, Smith L. It s time: record linkage the vision and the reality [comment]. Aust N Z J Public Health. 2000;24(1): Holman CD, Bass AJ, Rouse IL, Hobbs MS. Population-based linkage of health records in Western Australia: development of a health services research linked database. Aust N Z J Public Health 1999;23(5): AUSTRALIAN AND NEW ZEALAND JOURNAL OF PUBLIC HEALTH 2006 VOL. 30 NO. 2