Mortality among prisoners: how accurate is the Australian National Death Index?

OBJECTIVE: The effectiveness of the National Death Index (NDI) in identifying deaths among different populations has been studied before. Prisoners represent a segment of the population that might be expected to be difficult to link. The objective of this research was to evaluate the accuracy of the Australian NDI to ascertain the mortality status of prisoners in New South Wales, Australia. METHODS: Data on 311 inmates who had died in custody between January 1988 and March 2003, and on 7,558 inmates known to be alive at the end of May 2003, were matched to the NDI database. Matching was undertaken using a probabilistic linkage method followed by a clerical review. The underlying causes of death reported by the NDI were compared against those reported by the NSW Coroner and held at the Legal Branch of the Department of Corrective Services. RESULTS: Linkage correctly identified 275 of the 311 known deaths (sensitivity: 88.4%). Of the 7,558 prisoners known to be alive in custody at the end of May 2003, linkage to the NDI incorrectly identified 23 as dead (specificity: 99.7%). Cause of death was accurately reported by the NDI for 95.5% of deaths identified by the linkage. CONCLUSION: The results of this study show that the NDI is an accurate and feasible means for evaluating mortality among prisoners. The nature of the prisoner population makes the linkage a difficult task for the NDI. The success of the data linkage depends on the extent and accuracy of information contained in the index file and type and quality of information available.     

Accuracy of the Australian National Death Index: comparison with adjudicated fatal outcomes among Australian participants in the Long-term Intervention with Pravastatin in Ischaemic Disease (LIPID) study

OBJECTIVE: To assess the accuracy of the Australian National Death Index (NDI) in identifying deaths and recording cardiovascular and cancer causes of death. METHODS: Adjudicated mortality data from Australian participants in the Long-term Intervention with Pravastatin in Ischaemic Disease (LIPID) study up until September 1999 were used as reference. Nine hundred and eighty deceased subjects and 4,841 surviving subjects were matched to the NDI by name, date of birth, date of death and postcode. Matching rigour was confirmed by manual review. Deaths ascribed to cardiovascular and cancer causes within the NDI were also compared against LIPID-adjudicated causes. RESULTS: The NDI displayed 93.7% sensitivity and 100% specificity for the identification of deaths. Mis-recording of identifiers was responsible for 69% of known deaths not matching to the NDI and, if eliminated, would have increased the sensitivity to 98.0%. Among deceased subjects who matched to the NDI, cause of death was recorded in 96.2%. The sensitivity and specificity for cardiovascular deaths were 92.5% and 89.6%, respectively, and for cancer deaths 95.2% and 99.2%, respectively. CONCLUSION: Much of the inaccuracy of the NDI could potentially be overcome by the use of unique identifiers. Among deaths identified by the NDI, those due to cardiovascular disease are more likely to be inaccurately recorded than cancer-related deaths, probably because less uncertainty surrounds the presence or absence of terminal malignant disease.     

A feasible method for linkage studies avoiding clerical review: linkage of the national HIV/AIDS surveillance databases with the National Death Index in Australia

OBJECTIVE: To assess the sensitivity and specificity of linkage of HIV/AIDS diagnoses in Australia to the National Death Index (NDI). METHODS: An aggregated file containing 19,772 matched HIV/AIDS diagnoses reported to the national HIV/AIDS databases from 1980 to 30 June 2004 was linked to the NDI using probabilistic linkage methods based on the namecode, date of birth, and sex as identifiers. Based on the 6,900 HIV/AIDS known deaths reported by 1 January 2003 and 1,455 known non-deaths with an active follow-up beyond 1 January 2003, the different combinations of weights assigned to matched pairs were examined to obtain maximum sensitivity and specificity. RESULTS: The trade-off between sensitivity and specificity was used to obtain an optimal linkage. The optimal linkage was found to link 5,658 of the 6,900 HIV/AIDS known deaths (a sensitivity of 82%), and 116 false positives of the 1,455 known not deaths (specificity of 92%). Causes of deaths were recorded for 86.5% of deaths that were linked to the NDI. CONCLUSION: This is a feasible method for conducting linkage studies if both the identifying deaths and non-deaths are available. The relatively poor sensitivity could be due to limited identifiers available for linkage on the HIV/AIDS databases.     

Community-based recruitment strategies for a longitudinal interventional study: the VECAT experience

This article examines different recruitment strategies for the VECAT Study, a 4-year, double-masked, placebo-controlled, randomized clinical trial of vitamin E in the prevention of cataract and age-related maculopathy. Five recruitment methods were employed: newspaper advertising, radio advertising, approaches to community groups, approaches via general practices, and an electoral roll mail-out. Participants (1204) from the community in Melbourne, Australia were recruited and enrolled within 15 months (age range: 55-80 years, mean 66 years; gender ratio: 57% female, 43% male). The electoral roll mail-out and newspaper advertising were the most efficient methods of recruitment in terms of absolute numbers of participants recruited and cost per participant. Recruitment for the VECAT study was successfully completed within the planned period. Although the electoral roll mail-out and newspaper advertising were the most efficient for this study, other methods may be of value for studies with different subject selection criteria.     

Tracing and recruiting a cohort with community acquired hepatitis 25 years later

OBJECTIVE: We describe the methods used to trace and recruit a cohort (including injecting drug users) 25 years after admission to hospital in Melbourne with hepatitis. METHODS: Information recorded in the original medical record was used for tracing purposes. Subjects were located using the electoral roll, CD-ROM telephone directory, the Health Insurance Commission database, Hepatitis Foundation newsletters, advertising and features in the press and local radio. Other approaches included contacting local medical officer or next of kin from the original admission, and the National Death Registry. RESULTS: Tracing was undertaken on all members of the cohort; two-thirds were located. Methods were applied sequentially and 36% were found by CD-ROM telephone directory and the electoral roll, 33% from Health Insurance records, 15% by letter to last known address, 10% from contact of original next of kin, 3% from radio publicity and 1% from media advertisements. Of those who were located, 85% participated in the study. CONCLUSIONS: Accurate ascertainment of birth date and full name from original case records was essential for tracing purposes. Use of Health Insurance records and CD-ROM telephone directory and the electoral roll were the most successful means of locating subjects while newspaper advertising was non-productive and expensive. History of IDU was the major association with failure to trace and unwillingness to enrol, but despite this tracing and recruitment was high in this group. The high participation rate in those located appears to be due to the experience of the staff member making the initial contact.     

Effectiveness of the National Death Index for establishing the vital status of older women in the Australian Longitudinal Study on Women's Health

OBJECTIVE: To assess the effectiveness of the National Death Index (NDI) in identifying participants in the oldest cohort of the Australian Longitudinal Study on Women's Health (ALSWH) who had died between 1996 and 1998. METHODS: Identifying information for each woman was matched with the NDI using a probabilistic algorithm and clerical review. Differences in full name, date of birth, State of residence and date of last contact were used to assess the probability of a true match. RESULTS: NDI identified 410 matches of death records for 409 women; 386 were categorised as true matches and 23 were doubtful matches. Responses to the follow-up survey confirmed that for six of the doubtful matches the women had died, 16 were alive and the vital status of one woman remained unconfirmed at 30 June 1998. Twelve deaths, known to have occurred before July 1998, were not identified through NDI. The sensitivity of the NDI for identifying known deaths was 95%. Detailed identifying information, particularly the middle name, was important for accurate identification of the vital status. CONCLUSIONS: Using surname, all given names, gender, date of birth, State of residence and age at last contact as matching variables, the NDI was an effective tool for identifying women who had died. IMPLICATIONS: Routinely collected mortality data in the NDI are useful for the practice of epidemiology.     

Utility of the National Death Index in ascertaining mortality in acquired immunodeficiency syndrome surveillance

To assess the utility of the National Death Index (NDI) in improving the ascertainment of deaths among people diagnosed with acquired immunodeficiency syndrome (AIDS), the authors determined the number and characteristics of additional deaths identified through NDI linkage not ascertained by using standard electronic linkage with Florida Vital Records and the Social Security Administration's Death Master File. Records of people diagnosed with acquired immunodeficiency syndrome between 1993 and 2007 in Florida were linked to the NDI. The demographic characteristics and reported human immunodeficiency virus (HIV) transmission modes of people whose deaths were identified by using the NDI were compared with those whose deaths were ascertained by standard linkage methods. Of the 15,094 submitted records, 719 had confirmed matches, comprising 2.1% of known deaths (n = 34,504) within the cohort. Hispanics, males, people 40 years of age or older, and injection drug users were overrepresented among deaths ascertained only by the NDI. In-state deaths comprised 59.0% of newly identified deaths, and human immunodeficiency virus was less likely to be a cause of death among newly identified compared with previously identified deaths. The newly identified deaths were not previously ascertained principally because of slight differences in personal identifying information and could have been identified through improved linkages with Florida Vital Records.     

Sexual Abuse in Childhood and Physical and Mental Health in Adulthood: An Australian Population Study

Although childhood sexual abuse (CSA) is associated with a wide range of health problems later in life, there is also evidence of substantial individual differences. This study describes the mental and physical health of a population sample of Australians, randomly selected from the Commonwealth electoral roll, who have reported their CSA histories. Some 58% of those located from the electoral roll agreed to a telephone interview (n=1,784). Health status was measured using the Short Form 36 questionnaire. Men who had experienced non-penetrative and penetrative sexual abuse in childhood had 2.25 (95% CI=1.32-3.82) and 5.93 (95% CI=2.72-12.95) times respectively the rate of impaired mental health, but no higher rates of impaired physical health. Women who had experienced non-penetrative and penetrative sexual abuse in childhood had 1.87 (95% CI=1.19-2.95) and 3.15 (95% CI=1.87-5.33) times respectively the rate of impaired mental health and 1.87 (95% CI=1.19-2.92) and 2.31 (95% CI=1.34-3.97) times respectively the rate of impaired physical health. However, participants who had experienced CSA were no less likely than those who had not experienced CSA to be in optimum physical and mental health. None of the possible confounding or moderating variables tested appeared to mitigate the impact of CSA on health outcomes. Those with the highest levels of mental and physical health appear to be unaffected by the experience of CSA.     

Mortality ascertainment of participants in the National Wilms Tumor Study using the National Death Index: comparison of active and passive follow-up results

Long term studies of childhood cancer survivors are hampered by difficulties in tracking young adult participants. After performing a National Death Index (NDI) search we sought to identify which factors best predicted a match among known decedents from the National Wilms Tumor Study (NWTS) and to determine if record linkage could substitute for missing follow-up in a cohort of NWTS survivors. To our knowledge, this is the first study to compare passive mortality follow-up using the NDI to active follow-up of a childhood and young adult population. Records for 984 known decedents and 3,406 subjects whose January 1, 2002 vital status was unknown were sent to the NDI in June 2003. In April 2005 NWTS follow-up records were used to reassess January 1, 2002 vital status. Matches were established for 709 of 789 known decedents (sensitivity 89.9%) with a date of death between 1979 and 2001, the calendar period covered by the NDI at the time of the search. No matches were identified among 1,052 subjects known to be alive in 2002 (specificity 100%). Factors associated with decreased sensitivity were an unknown social security number (sensitivity 87.8%), Hispanic ethnicity (76.4%) and foreign birth (56.5%). For 2,351 subjects with 2002 vital status unknown who had 13,166 pre 2002 person-years of missing observation, only 18 deaths were ascertained by the NDI whereas 79.3 were expected based on NWTS mortality data. Mortality analyses based strictly on NDI search results and those based on NWTS follow-up augmented with NDI search results yielded inflated estimates of the 15 year survival rate when compared with estimates based on NWTS active follow-up. Match rates were comparable to those observed in adult populations. Since the same selection factors were likely associated with NDI failure to match and NWTS loss to follow-up, use of the NDI to fill in missing follow-up data appears unwarranted.     

Development of a computer linkage system for a blood recipient notification program in Nova Scotia

OBJECTIVES: To assess the potential uses of computer-assisted record linkage in the surveillance of infectious diseases, using the Nova Scotia blood recipient notification program as the example. METHODS: We developed a computer-assisted, multiple-pass, probabilistic record linkage to link records for blood recipients identified by the Nova Scotia notification program (Nova Scotia Phase I Blood Bank File information) with corresponding Nova Scotia Health Card Registration File records to obtain current mailing addresses to contact potentially living recipients. We used variables available from both files (e.g., name, date of birth, gender, and health care registration number) to link records, after eliminating duplicates/deceased cases. RESULTS: Among 23,925 eligible records in the Nova Scotia Phase I Blood Bank File (1984-1990), there were 1,818 (7.8%) duplications and 8,675 deceased cases, leaving 13,432 cases for linkage. 8,713 (65%) cases were successfully linked to the 1998 Health Card Registration Data File for current mailing addresses. INTERPRETATION: Multiple-pass linkage seems acceptable for maximizing detection of correctly matched records for look-back projects. To overcome quality/lack of information obstacles, future look-back linkages should explore the use of supplementary data files (tax files, voter lists, license files, other provincial databases) to obtain most current addresses.     

Socioeconomic differences in food preference and their influence on healthy food purchasing choices

Background: Studies have shown that socioeconomic groups differ in their dietary behaviours, and it has been suggested that these differences partly account for health inequalities between social groups. To-date, however, we have a limited understanding of why socioeconomic groups differ in their dietary behaviours. This paper addresses this issue by examining the relationship between socioeconomic status, food preference (likes and dislikes) and the purchase of 'healthy' food (i.e. food consistent with dietary guideline recommendations). Methods: This study was based on a dual-sample, dual-method research design. One sample was systematically selected from the Australian Commonwealth electoral roll and the data collected using a mail-survey methodology (81% response rate, n =403). The second consisted of a convenience sample of economically disadvantaged people recruited via welfare agencies (response rate unknown, n =70). A mail survey methodology was deemed inappropriate for this sample, so the data were collected by personally delivering the questionnaire to each respondent. Results: Socioeconomic groups differed significantly in their food purchasing choices and preferences. The food choices of respondents in the welfare sample were the least consistent with dietary guideline recommendations, and they reported liking fewer healthy foods (all results were independent of age and sex). Notably, socioeconomic differences in preference explained approximately 10% of the socioeconomic variability in healthy food purchasing behaviour. Conclusion: Whilst it is not clear why socioeconomic groups differ in their food preferences, possible reasons include: reporting bias, differential exposure to healthy food as a consequence of the variable impact of health promotion campaigns, structural and economic barriers to the procurement of these foods, and subculturally specific beliefs, values, meanings, etc.     

Participation of older males in a study on photography as an aid to early detection of melanoma

OBJECTIVE: To examine the acceptability of photography as an aid to skin examinations in men over 50 years of age. METHODS: A randomised trial of men selected from the electoral roll. All participants were photographed, but only half received their photographs. Skin examinations by GPs at years one and two. RESULTS: 55% of men consented to have photographs taken and 51% did so. 86% of respondents had risk factors for melanoma (compared to 68% of non-responders) and 47% had two or more risk factors (compared to 23% of non-responders). At year one, 91% of participants remaining in study regions had been examined. Photographs were lost by only six participants. CONCLUSIONS: Men over 50 years of age respond to personalised health messages about melanoma and respondents include a high proportion of males with risk factors for melanoma. IMPLICATIONS: These initial results suggest that photography may be a logistically acceptable approach for assisting in the early detection of melanoma.     

Comparability of national death index plus and standard procedures for determining causes of death in epidemiologic studies

PURPOSE: To determine whether causes of death obtained through National Death Index (NDI) Plus are comparable to those obtained by requesting death certificates from state vital statistics offices and having deaths coded by contractor nosologists. METHODS: The authors compared underlying cause of death codes obtained from NDI Plus with those assigned by contractor nosologists for a sample of 250 known decedents. RESULTS: The underlying cause of death codes differed for 18 (7%) of 249 successful matches. Independent coding by an expert National Center for Health Statistics (NCHS) nosologist trainer revealed that seven of these had an NDI Plus code that matched the code provided by the NCHS nosologist and a contractor nosologist code that did not match the NCHS nosologist code, seven had a contractor nosologist code that matched the NCHS nosologist code and an NDI Plus code that did not match the NCHS nosologist code, and four had both an NDI Plus and a contractor nosologist code that did not match the NCHS nosologist code. The level of disagreement with the NCHS nosologist and the organ systems involved were similar for NDI Plus and the contractor nosologist. CONCLUSIONS: The authors report that NDI Plus provides comparable information within a substantially shorter time period for most states and, for known decedents, at about half the cost of standard procedures.     

Selection bias from sampling frames: telephone directory and electoral roll compared with door—to—door population census: results from the Blue Mountains Eye Study

Abstract: Many Australian public health research studies use the telephone directory or the electoral roll as a sampling frame from which to draw study subjects. The sociode–mographic, disease–state and risk–factor characteristics of subjects who could be recruited using only the telephone directory or only the electoral roll sampling frames were compared with the characteristics of subjects who would have been missed using only these sampling frames, respectively. In the first phase of the Blue Mountains Eye Study we interviewed and examined 2557 people aged 49 and over living in a defined postcode area, recruited from a door–to–door census. This represented a participation rate of 80.9 per cent and a response rate of 87.9 per cent. The telephone directory was searched for each subject's telephone number and the electoral roll was searched for each subject. Subject characteristics for those who were present in each of these sampling frames were compared with the characteristics of those subjects not included in the sampling frames. The telephone directory listed 2102 (82.2 per cent) of the subjects, and 115 (4.5 per cent) had no telephone connected. The electoral roll contained 2156 (84.3 per cent) of the subjects, and 141 subjects (5.5 per cent) could not be found in either the electoral roll or the telephone directory. Younger subjects, subjects who did not own their own homes and subjects born outside of Australia were significantly less likely to be included in either of these sampling frames. The telephone directory was also more likely to exclude subjects with higher occupational prestige, while the electoral roll was more likely to exclude unmarried persons and males. Researchers using the telephone directory and electoral roll to select subjects for study should be aware of the potential selection bias these sampling frames incur and need to take care when generalising their findings to the wider community.     

Evaluation of a mobile screening service for abdominal aortic aneurysm in Broken Hill,a remote regional centre in far western NSW

Objectives: To evaluate the feasibility of a mobile screening service model for abdominal aortic aneurysm (AAA) in a remote population centre in Australia. Design: Screening test evaluation. Setting: A remote regional centre (population: 20 000) in far western NSW. Participants: Men aged 65–74 years, identified from the Australian Electoral roll. Interventions: A mobile screening service using directed ultrasonography, a basic health check and post‐screening consultation. Main outcome measures: Attendance at the screening program, occurrence of AAA in the target population and effectiveness of screening processes. Results: A total of 516 men without a previous diagnosis of AAA were screened, an estimated response rate of 60%. Of these, 463 (89.7%) had a normal aortic diameter, 28 (5.4%) ectatic and 25 (4.9%) a small, moderate or significant aneurysm. Two men with AAA were recommended for surgery. Feedback from participants indicated that the use of a personalised letter of invitation helped with recruitment, that the screening process was acceptable and the service valued. Conclusions: It is feasible to organise and operate a mobile AAA screening service from moderate sized rural and remote population centres. This model could be scaled up to provide national coverage for rural and remote residents.     

An evaluation of the Social Security Administration master beneficiary record file and the National Death Index in the ascertainment of vital status

To evaluate the relative utility of the Social Security Administration and National Death Index as sources of mortality data, the vital status of 12,866 participants in the Multiple Risk Factor Intervention Trial was identified from these sources and compared to the known mortality experience. The SSA correctly identified 87.8 per cent of the 409 deaths that occurred between 1974 and 1980. Underreporting of deaths by the SSA occurred for participants with certain demographic characteristics, especially marital status. For the years 1979 and 1980, the period for which the SSA and NDI have comparable data, the SSA correctly identified 93.2 per cent and the NDI correctly identified 98.4 per cent of the 191 known deaths. The NDI matching process resulted in a large number of false positive matches.     

Socio-economic status and dietray habits in the elederly: results form a large random survey

Socio-economic status differences in the usual food consumption patterns of a representative sample of 2195 people aged 65 years and over residing in Adelaide, South Australia, were examined. The dietary instrument used was a self-completed semi-quantitative food frequency questionnaire which included additional questions concerning food preparation and cooking practices. Subjects were randomly selected from the State Electoral Roll and there was a 77% response rate. Members of higher social class groups reported higher consumption of high fibre and vitamin C rich foods (fruits, fruit juice, vegetables, wholegrain bread) and less frequent intakes of several fried and high fat foods and salt. They were also more likely to report nutritionally favourable vegetable cooking practices.     

Ascertainment of vital status through the National Death Index and the Social Security Administration

Ascertainment of the vital status of individuals is of central importance to epidemiologic studies which monitor mortality as an end point. Utilizing identifying information collected in 1973-1974, the Hypertension Detection and Follow-up Program, a prospective, multicenter study, followed 25,362 individuals to determine eight-year mortality. In the most recent follow-up, there were 617 individuals whose vital status was not known. Available identifying information on these and on all 1,322 participants known to have died in 1979-1981 was submitted to the National Death Index (NDI) for possible confirmation of vital status. A subset of individuals who had Social Security numbers (490 lost to follow-up and 1,154 known deaths) was also submitted to the Social Security Administration (SSA). The NDI correctly identified 87.0% of the known deaths. Of the 1,154 known deaths (those with known Social Security numbers) submitted to both agencies, the NDI identified 93.1% and the SSA 83.6%. Significant variations by race and sex were noted in the identification rates, in part because of Social Security number discrepancies. False matches through the NDI matching process occurred for 10.4% of the known deaths. In the more restrictive SSA search, only 0.5% false matches resulted. For those lost to follow-up, vital status was ascertained in 57.1%. This paper describes the relative efficacy and attributes of the use of these systems to ascertain vital status.     

Medicare claims data reliably identify treatments for basal cell carcinoma and squamous cell carcinoma: a prospective cohort study

Objective: To investigate the accuracy of Medical Benefit Schedule (MBS) item numbers to identify treatments for basal cell carcinomas (BCC) and squamous cell carcinomas (SCC). Methods: We linked records from QSkin Study participants (n=37,103) to Medicare. We measured the proportion of Medicare claims for primary excision of BCC/SCC that had corresponding claims for histopathology services. In subsets of participants, we estimated the sensitivity and external concordance of MBS item numbers for identifying BCC/SCC diagnoses by comparing against ‘gold‐standard’ histopathology reports. Results: A total of 2,821 (7.6%) participants had 4,830 separate Medicare claims for BCC/SCC excision; almost all (97%) had contemporaneous Medicare claims for histopathology services. Among participants with BCC/SCC confirmed by histology reports, 76% had a corresponding Medicare claim for primary surgical excision of BCC/SCC. External concordance for Medicare claims for primary BCC/SCC excision was 68%, increasing to 97% when diagnoses for intra‐epidermal carcinomas and keratoacanthomas were included. Conclusions: MBS item numbers for primary excision of BCC/SCC are reasonably reliable for determining incident cases of keratinocyte skin cancers, but may underestimate incidence by up to 24%. Implications: Medicare claims data may have utility in monitoring trends in conditions for which there is no mandatory reporting.     

Health services research using linked records: who consents and what is the gain?

OBJECTIVE: To assess consent to record linkage, describe the characteristics of consenters and compare self-report versus Medicare records of general practitioner use. METHOD: Almost 40,000 women in the Australian Longitudinal Study on Women's Health were sent a request by mail for permission to link their Medicare records and survey data. RESULTS: 19,700 women consented: 37% of young (18-23 years), 59% of mid-age (45-50 years) and 53% of older women (70-75 years). Consenters tended to have higher levels of education and, among the older cohort, were in better health than non-consenters. Women tended to under-report the number of visits to general practitioners. CONCLUSIONS: Record linkage of survey and Medicare data on a large scale is feasible. The linked data provide information on health and socio-economic status which are valuable for understanding health service utilisation. IMPLICATIONS: Linked records provide a powerful tool for health care research, particularly in longitudinal studies.