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

A comparison of accuracy and computational feasibility of two record linkage algorithms in retrieving vital status information from HIV/AIDS patients registered in Brazilian public databases

BACKGROUND AND OBJECTIVE

While cross-referencing information from people living with HIV/AIDS (PLWHA) to the official mortality database is a critical step in monitoring the HIV/AIDS epidemic in Brazil, the accuracy of the linkage routine may compromise the validity of the final database, yielding to biased epidemiological estimates. We compared the accuracy and the total runtime of two linkage algorithms applied to retrieve vital status information from PLWHA in Brazilian public databases.

METHODS

Nominally identified records from PLWHA were obtained from three distinct government databases. Linkage routines included an algorithm in Python language (PLA) and Reclink software (RlS), a probabilistic software largely utilized in Brazil. Records from PLWHA¹ known to be alive were added to those from patients reported as deceased. Data were then searched into the mortality system. Scenarios where 5% and 50% of patients actually dead were simulated, considering both complete cases and 20% missing maternal names.

RESULTS

When complete information was available both algorithms had comparable accuracies. In the scenario of 20% missing maternal names, PLA² and RlS³ had sensitivities of 94.5% and 94.6% (p > 0.5), respectively; after manual reviewing, PLA sensitivity increased to 98.4% (96.6-100.0) exceeding that for RlS (p < 0.01). PLA had higher positive predictive value in 5% death proportion. Manual reviewing was intrinsically required by RlS in up to 14% register for people actually dead, whereas the corresponding proportion ranged from 1.5% to 2% for PLA. The lack of manual inspection did not alter PLA sensitivity when complete information was available. When incomplete data was available PLA sensitivity increased from 94.5% to 98.4%, thus exceeding that presented by RlS (94.6%, p < 0.05). RlS spanned considerably less processing time compared to PLA.

CONCLUSION

Both linkage algorithms presented interchangeable accuracies in retrieving vital status data from PLWHA. RlS had a considerably lesser runtime but intrinsically required manually reviewing a fastidious proportion of the matched registries. On the other hand, PLA spent quite more runtime but spared manual reviewing at no expense of accuracy.

Assessing Hepatitis C Burden and Treatment Effectiveness through the British Columbia Hepatitis Testers Cohort (BC-HTC): Design and Characteristics of Linked and Unlinked Participants

Background

The British Columbia (BC) Hepatitis Testers Cohort (BC-HTC) was established to assess and monitor hepatitis C (HCV) epidemiology, cost of illness and treatment effectiveness in BC, Canada. In this paper, we describe the cohort construction, data linkage process, linkage yields, and comparison of the characteristics of linked and unlinked individuals.

Methods

The BC-HTC includes all individuals tested for HCV and/or HIV or reported as a case of HCV, hepatitis B (HBV), HIV or active tuberculosis (TB) in BC linked with the provincial health insurance client roster, medical visits, hospitalizations, drug prescriptions, the cancer registry and mortality data using unique personal health numbers. The cohort includes data since inception (1990/1992) of each database until 2012/2013 with plans for annual updates. We computed linkage rates by year and compared the characteristics of linked and unlinked individuals.

Results

Of 2,656,323 unique individuals available in the laboratory and surveillance data, 1,427,917(54%) were included in the final linked cohort, including about 1.15 million tested for HCV and about 1.02 million tested for HIV. The linkage rate was 86% for HCV tests, 89% for HCV cases, 95% for active TB cases, 48% for HIV tests and 36% for HIV cases. Linkage rates increased from 40% for HCV negatives and 70% for HCV positives in 1992 to ~90% after 2005. Linkage rates were lower for males, younger age at testing, and those with unknown residence location. Linkage rates for HCV testers co-infected with HIV, HBV or TB were very high (90–100%).

Conclusion

Linkage rates increased over time related to improvements in completeness of identifiers in laboratory, surveillance, and registry databases. Linkage rates were higher for HCV than HIV testers, those testing positive, older individuals, and females. Data from the cohort provide essential information to support the development of prevention, care and treatment initiatives for those infected with HCV.

Determinação de um ponto de corte para a identificação de pares verdadeiros pelo método probabilístico de linkage de base de dados

O objetivo deste estudo foi propor pontos de corte nos escores calculados no processo de linkage probabilístico, para as diversas topografias de câncer. Neste estudo foi utilizada a base de dados do RCBP-SP, composta por 343.306 casos incidentes de câncer do município de São Paulo, registrados no período de 1997 a 2005, com idades que variaram de menos um a 106 anos, de ambos os sexos. Para o linkage probabilístico, realizado no programa Reclink III, foram utilizadas a base de dados do PRO-AIM e APAC-SIA/SUS. Foram calculados os valores da área sob a curva, sensibilidade e especificidade para determinar o ponto de corte do escore de maior precisão na identificação dos pares verdadeiros. Na análise das topografias, verificou-se que o ponto de corte no escore 18 apresentou boa acurácia, com valores de sensibilidade que variaram de 73,7 a 96,7% e de especificidade de 98,5 a 99,4%. Conclui-se que, acima do escore 18 encontravam-se quase a totalidade dos pares verdadeiros, enquanto que abaixo deste, menos de 1% dos registros vinculados eram verdadeiros.

Examining the quality of name code record linkage: what is the impact on death and cancer risk estimates? A validation study

OBJECTIVE

To examine the validity and impact of record linkage using name code compared to full name records.

METHODS

A registry of 45,419 opioid substitution clients (1985-2007) was linked with national population-based death and cancer registries using registrant's name, date of birth, sex, state, postcode and date of death. Records were linked using full name and then using the first two letters of the given and surname (2×2 name code). Sensitivity and specificity were computed and regression analysis used to identify factors related to linkage accuracy. Standardised mortality ratios (SMR) and standardised cancer incidence ratios (SIR) were estimated.

RESULTS

The sensitivity and specificity of name code compared to full name linkage were 65.31% and 99.91% for death records and 76.81% and 99.89% for cancer records. Registrants' age and sex and accuracy of the registries were associated with risk of false linkages. Death and cancer risks (SMR 6.98, 95%CI 6.77-7.19; SIR 1.16, 95%CI 1.08-1.24) were significantly under-estimated using name code linkage (SMR 4.39, 95%CI 4.23-4.56; SIR 0.92, 95%CI 0.85-0.99).

CONCLUSION

Record linkage using 2×2 name code has low sensitivity but high specificity, resulting in conservative estimates of death and cancer risk. This may translate to meaningful differences in outcomes.

Evaluating health outcomes of criminal justice populations using record linkage: the importance of aliases

Individuals in contact with the criminal justice system are a key population of concern to public health. Record linkage studies can be useful for studying health outcomes for this group, but the use of aliases complicates the process of linking records across databases. This study was undertaken to determine the impact of aliases on sensitivity and specificity of record linkage and how this affects ascertainment of mortality. Records for a cohort of prisoners were linked to methadone maintenance treatment (MMT) and mortality records. The record linkage was conducted in two stages. First, the linkage was undertaken using the participant's name and date of birth as recorded in a prior study. Then, a second linkage was undertaken using these identifiers, plus all known aliases. Sensitivity was 64%, and specificity 100%, for the first linkage. When aliases were added to the linkage, sensitivity increased to 86% and specificity remained 100%. The standardized mortality ratio was 4.3 for the first linkage, increasing to 6.1 when aliases were used. These results suggest that the potential effects of participant aliases on linkage outcomes, and methods for mitigating these effects should be carefully considered when planning and undertaking record linkage studies with criminal justice populations.

Accuracy of a probabilistic record linkage strategy applied to identify deaths among cases reported to the Brazilian AIDS surveillance database

Since record linkage errors can bias measures of disease occurrence and association, it is important to assess their accuracy. The aim of this study is to assess the accuracy of a multiple pass probabilistic record linkage strategy to identify deaths among persons reported to the Brazilian AIDS surveillance database. An HIV/AIDS national surveillance database (N = 559,442) was linked to a total of 6,444,822 deaths registered (all causes) in the Brazilian mortality database. To estimate standard measures of accuracy, we selected all AIDS cases with a date of death registered in the surveillance database from 2002 to 2005 (N = 19,750) and 38,675 cases known to be alive in 2006. The linkage strategy presented a sensitivity of 87.6% (95%CI: 87.1-88.2), a specificity of 99.6% (95%CI: 99.6-99.7), and a positive predictive value of 99.2% (95%CI: 99.1-99.3). We observed a small variation in the validity measures according to some putative predictors of mortality. Our findings suggest that even large and heterogeneous databases can be linked with a satisfactory accuracy.

Hospitalizations in a cohort of HIV patients in Australia, 1999-2007

OBJECTIVES

To describe hospitalization rates, risk factors and associated diagnoses in people with HIV in Australia between 1999 and 2007.

DESIGN

Retrospective cohort study of people with HIV (n = 842) using data linkage between the Australian HIV Observational Database and administrative hospital morbidity data collections.

METHODS

Incidence rate ratios with 95% confidence intervals were estimated using Poisson regression models to assess risk factors for hospitalization. Predictors of length of stay were assessed using generalized mixed models. The association between hospitalization and mortality was assessed using Cox regression.

RESULTS

In 4519 person-years of observation, there were 2667 hospital admissions; incidence rate of 59 per 100 person-years. Hospitalization rates were 50-300% higher in this cohort than comparable age and sex strata in the general population. Older age (incidence rate ratio 1.46, 95% confidence interval 1.28-1.65 per 10-year increase) and prior AIDS (incidence rate ratio 1.71, 95% confidence interval 1.24-2.35) were significantly associated with hospitalization. Other predictors of hospitalization included lower CD4 cell counts, higher HIV RNA, longer duration of HIV infection and experience with more drug classes. Lower CD4 cell counts, older age and hepatitis C virus antibody positivity were independently associated with longer hospital stay. Non-AIDS diseases were the principle reason for admission in the majority of cases. Mortality was associated with more frequent hospitalization during the study period.

CONCLUSION

Hospitalization rates are higher in people with HIV than the general population in Australia and are associated with markers of advanced HIV disease despite the widespread use of combination antiretroviral therapy.

Changes in mortality following HIV and AIDS and estimation of the number of people living with diagnosed HIV/AIDS in Australia, 1981-2003

OBJECTIVE

To investigate changes in mortality following HIV and AIDS in Australia.

METHODS

The results of a linkage between HIV/AIDS diagnoses and the National Death Index (NDI) to the end of 2003 were used to estimate mortality rates following HIV/AIDS. Standardised Mortality Ratios (SMRs) were calculated for deaths following HIV, with and without AIDS, in three periods of treatment; before antiretroviral therapy (< or =1989), pre- and early-HAART (1990-1996) and HAART (1997-2003). Crude mortality rates were calculated as the number of deaths per 1000 person-years. The total number of people living with HIV/AIDS was estimated.

RESULTS

There were 1789 deaths following HIV without AIDS and 6730 deaths after AIDS. For deaths following HIV without AIDS, the SMRs were 2.99, 1.22 and 1.6 during the periods before 1990, 1990-1996 and 1997-2003. For deaths after AIDS the SMRs were 137.84, 28.64 and 4.55 in the periods one to three, respectively. The crude death rate following HIV without AIDS increased from 16.8 before 1986 to 19.6 in 2003. Death rates after AIDS decreased from 958.7 up to 1986 to 60.4 in 2003. The number of new HIV diagnoses increased to 1276 in 1990 then decreased to 780 in 2003, while AIDS diagnoses increased to 950 in 1994 then decreased to 252 in 2003. The total number of people living with HIV was estimated to be 7873 in 1989, and 12828 in 2003.

CONCLUSION

Mortality following AIDS decreased while deaths before AIDS remained low. The number of people living with HIV/AIDS has increased.