Main Article Content
Assisted reproductive technologies (ART), such as in-vitro fertilisation (IVF), have revolutionised the treatment of infertility, with an estimated 8 million babies born worldwide. However, the long-term health outcomes for women and their offspring remain an area of concern. Linking IVF treatment data to long-term health data is the most efficient method for assessing such outcomes.
To describe the creation and performance of a bespoke population-based data linkage of an ART clinical quality registry to state-based and national administrative datasets.
The linked dataset was created by deterministically and probabilistically linking the Australia and New Zealand Assisted Reproduction Database (ANZARD) to New South Wales (NSW) and Australian Capital Territory (ACT) administrative datasets (performed by NSW Centre for Health Record Linkage (CHeReL)) and to national claims datasets (performed by Australian Institute of Health and Welfare (AIHW)). The CHeReL's Master Linkage Key (MLK) was used as a bridge between ANZARD's partially identifiable patient data (statistical linkage key) and NSW and ACT administrative datasets. CHeReL then provided personal identifiers to the AIHW to obtain national content data. The results of the linkage were reported, and concordance between births recorded in ANZARD and perinatal data collections (PDCs) was evaluated.
Of the 62,833 women who had ART treatment in NSW or ACT, 60,419 could be linked to the CHeReL MLK (linkage rate: 96.2%). A reconciliation of ANZARD-recorded births among NSW residents found that 94.2% (95% CI: 93.9--94.4%) of births were also recorded in state/territory-based PDCs. A high concordance was found in plurality status and birth outcome (≥99% agreement rate, Cohen's kappa ranged: 0.78--0.98) between ANZARD and PDCs.
The data linkage resource demonstrates that high linkage rates can be achieved with partially identifiable data and that a population spine, such as the CHeReL's MLK, can be successfully used as a bridge between clinical registries and administrative datasets.
Infertility affects one in six couples , resulting in significant personal suffering, and representing an important and increasingly prevalent public health problem [2, 3]. Fortunately, there are a number of Medically Assisted Reproduction (MAR) treatments that allow many infertile individuals to achieve parenthood. The most advanced of these are assisted reproductive technologies (ART), such as in vitro fertilisation (IVF), which involve the fertilisation of human eggs outside of the body before transferring the resulting embryos into the uterus in the hope of achieving a pregnancy. ART represents one of the most significant medical and social achievements of the past century, leading to the birth of an estimated 8 million babies over the last four decades . Non-ART treatment using ovulation induction (OI) with or without intrauterine insemination (IUI) is a more traditional form of MAR treatment in which fertilisation occurs within the woman’s reproductive tract, but is still widely used as part of evidence-based management . The increasing demand for MAR treatment, both ART and non-ART, reflects the social trend to delayed childbearing, changes in family structures, rising levels of sexually transmitted disease, obesity, and declining sperm quality [6–10].
Australia has one of the highest rates of ART utilisation per capita in the world . Over the last two decades, Australia has experienced a 192% increase in ART utilisation, and in 2018, around 4.9% of Australian children were conceived using ARTs [2, 11–13]. However, little is known about the number of children conceived through OI/IUI (non-ART) treatment or by spontaneous conception for women with a history of subfertility.
While evidence regarding the health outcomes of ART-conceived children is generally reassuring, several studies have suggested a higher risk of poorer perinatal outcomes, and longer term metabolic risks [14–18]. This is primarily because they are at a greater risk of being born as part of multiple gestation pregnancies (e.g. twins and triplets), but even singletons are at a marginally higher risk of low birth weight, small for gestational age, congenital anomalies, perinatal death (stillbirth and neonatal death), as well as maternal morbidity compared to spontaneously conceived children [14–16, 18]. The reasons for these increased risks to mothers and babies are not well understood. Interestingly, it appears that couples who experience subfertility but achieve a spontaneous conception also have similar adverse risk profiles [14, 19, 20]. Furthermore, there is a lack of evidence on the health outcomes of children conceived using OI/IUI (non-ART) treatment .
The lack of clarity on the potential risks of MAR treatments (ART and non-ART) and the possible confounding role of subfertility is an enduring evidence gap when advising patients, clinicians and policymakers on the use of MAR treatments.
To address this gap, the National Perinatal Epidemiology and Statistics Unit (NPESU) of the University of New South Wales created a MAR data linkage by linking a regional ART treatment registry (Australian and New Zealand Assisted Reproduction Database, ANZARD) to a number of other jurisdiction-based and national administrative databases. The resulting dataset contains longitudinal health records for women who have either undergone MAR (ART and non-ART), or who have conceived spontaneously, and their resulting children. The overarching objective of establishing the MAR data linkage resource was to quantify the risk of adverse health outcomes in children conceived from ART and non-ART treatments after accounting for confounders, in particular underlying subfertility, and to assess if specific forms of ART contribute differently to these outcomes.
Central to the MAR data linkage resource is the ANZARD, which is the oldest national ART registry in the world incorporating all accredited fertility clinics operating in Australia and New Zealand (currently over 90 clinics) and providing demographic, treatment, laboratory and outcome data on all ART cycles and donor insemination (DI) cycles (currently over 80,000 cycles per year) . The submission of data to ANZARD is a requirement of a clinic’s accreditation to practice and thus complete ascertainment of ART cycles is assumed . ANZARD does not currently collect data from non-ART treatments such as OI and IUI.
This paper describes the data linkage methodology and results between the ANZARD and the state/territory and national data sources, and describes the concordance between the births recorded in ANZARD and those in the state perinatal data collections (PDCs).
The MAR data linkage
New South Wales (NSW) and Australian Capital Territory (ACT) are two of the eight Australian states and territories, with their combined population of approximately 8 million residents accounting for one-third of the total Australian population . ANZARD was linked to NSW and ACT perinatal, births, deaths, hospital admissions, and congenital anomaly routinely collected databases, as well as national medical and pharmaceutical claims databases.
The linkage was possible because since 2009 ANZARD has collected the first two letters of female patients’ first and last names. These personal identifiers were combined with the female patients’ date of birth (DOB), residential postcode, and their partners’ DOB to form a Statistical Linkage Key (SLK). Combinations of the components of the SLK were the foundation for linkage with the administrative datasets .
The NSW Ministry of Health’s Centre for Health Record Linkage (CHeReL) and the Australian Institute of Health and Welfare (AIHW) undertook the required linkages before transferring the data into a secure research environment for cleaning and analysis by the researchers. Figure 1 summarise the data linkage process of ANZARD to 5 NSW and ACT administrative and 2 Commonwealth datasets.
A key strategy to enable the linkage of ANZARD’s partially identifiable data (components of the SLK) to the administrative datasets was the ability of the CHeReL to use their Master Linkage Key (MLK) as a bridge between ANZARD and the NSW and ACT Perinatal Data Collections (PDC) to identify births to women who had conceived using ART and those who had conceived spontaneously (without ART). The MLK is constructed by the CHeReL using probabilistic record linkage methods and ChoiceMaker software using a best practice approach to privacy-preserving record linkage. The MLK comprises over 188 million records containing personal and demographic information, but no health information, on over 15 million people in NSW and ACT from a range of population-based health and health-related data collections . The CHeReL uses the following personal information to link records for the same person to create the MLK: full name, address, sex, DOB, country of birth, and uses relevant event information such as hospital code, medical record number, event dates (e.g., hospital dates of admission and discharge), hospital transferred to, hospital transferred from, and date of death. The entire linked NSW and ACT administrative data has less than 5/1000 missed links and 3/1000 false positive links . In addition to person links, the MLK contains a family structure, by virtue of data sources that contain details of a child and up to two parents.
The data linkage between ANZARD and NSW/ACT administrative data and the Commonwealth data involved three stages that enabled the construction of the MAR data linkage while abiding by the principles of data separation to protect patient privacy.
Stage 1: Data linkage between ANZARD and NSW and ACT administrative data
The ANZARD Data Manager (who is independent from the research team) transferred to CHeReL a cycle ID (an anonymous unique cycle identifier) and a patient ID (an anonymous unique patient identifier) together with the SLK components associated with the 638,036 ART and DI cycles (with 195,490 SLKs) performed in Australia between 1st January 2009 and 31st December 2016.
The CHeReL then deterministically linked ANZARD personal SLK identifiers (SLK person ID) to the MLK identifiers (MLK person ID) for females. Several strategies were adopted to improve the data linkage rate between the ANZARD identifiers and the MLK. Because linkage using the SLK may have low sensitivity depending on data quality, the deterministic linkage on person characteristics was combined with event-based linkage using ANZARD and hospital procedures dates for selected procedure codes relating to oocyte retrieval. An ART cycle involves mostly outpatient services; however, almost all egg retrieval procedures are undertaken under sedation as part of an inpatient admission which is recorded in the hospital admission data collections. Thus, procedure codes related to egg retrieval procedures were used for the event-based linkage to supplement the linkage based on personal identifiers. Matches were initially performed on all personal identifiers and event information; then, restrictions were progressively relaxed to allow a higher rate of matching to the SLK. Where multiple MLK person IDs matched to a single ANZARD SLK person ID clerical review was performed. Details of the results of each pass in the linkage process between the ANZARD SLK and MLK person ID and the corresponding linkage rate are shown in the results section below.
The MAR data linkage is a birth cohort for all births in NSW/ACT, comprising births that were conceived through ART treatment, non-ART treatment, or spontaneous conception. All MLK person IDs that linked to PDC records indicated that a woman had given birth between 1st January, 2009 and 31st December, 2017 in NSW and women who gave birth between 1st January, 2009 and 31st December, 2016 in the ACT. These PDC records (including those with a link to an ANZARD SLKs and those that did not) were then linked to individual-level content data from the various NSW and ACT administrative databases by the CHeReL and ACT Health (see more details of the administrative databases included in Table S1, Supplementary Appendix).
Once the linkage rate and accuracy were considered to be maximised based on available identifiers, the CHeReL created a Project Person Number (PPN) for each woman. The PPNs were later used by the research team to merge all datasets. The CHeReL loaded the PPNs and the content data from the NSW and ACT administrative databases into the Sax Institute’s Secure Unified Research Environment (SURE) . The SURE is a central, secure, online remote-access computing environment for analysing sensitive human research data.
Stage 2: Data linkage between the NSW and ACT MLKs for mothers who gave birth and the Commonwealth Medicare Benefits Schedule and Pharmaceutical Benefits Scheme data
The CHeReL sent identifying information only (mother’s name, her DOB and residential address) representing the 606,658 (include duplicates) women who gave birth between 1st January, 2009 and 31st December, 2017 in NSW and 1st January, 2009 and 31st December, 2016 in ACT and the corresponding PPNs to the AIHW Data Integration Unit.
The AIHW Data Integration Unit undertook a probabilistic linkage between the MLKs and the personal identifiers from the Medicare Enrolment File (MEF) of 32,378,696 individuals registered to Australia’s national health care scheme, Medicare. The MEF linkage procedure involved creating record pairs between MLKs and MEF’s personal identifiers based on a combination of seven personal identifiers: surname; given name; sex; day, month, and year of birth; day, month, and year of death when applicable; residential postcode; upper case of the first six characters of the address after removing the punctuations and words such as unit, flat, PO box, etc. A total of 18 passes were undertaken to create the final linked dataset.
Following the completion of the probabilistic linkage, a sample-based clerical review including 32 batches each containing between 78,437 and 1,250,122 records was performed to determine the linkage status for record pairs with similar linkage weights.
Once all linkages were maximised, the AIHW retrieved the requested individual-level content data from the Medicare Benefits Schedule (fertility-related services) and Pharmaceutical Benefits Scheme (fertility medicines and other medicines) and uploaded the PPNs and the content data from the Medicare Benefits Schedule and Pharmaceutical Benefits Scheme into SURE.
Stage 3: Retrieved ANZARD treatment and outcome data
The CHeReL sent the PPNs and both linked and unlinked ANZARD’s unique patient IDs back to the study-independent ANZARD Data Manager. The ANZARD Data Manager removed all personal identifiers from the ANZARD content data and attached the PPNs and ANZARD’s unique patient ID to the ANZARD content data. The ANZARD Data Manager loaded the ANZARD content data (with PPNs and ANZARD’s unique patient ID) for all ANZARD treatment performed during the study period to SURE ready for the researchers to merge all data collections using the common PPN.
Final MAR data linkage
Table S1 (Supplementary Appendix) describes all data sources included in the final MAR data linkage. Briefly, the MAR data linkage constitutes 581,241 mothers who had conceived 874,922 babies born between 2009 and 2017 plus 261,711 siblings of these babies born before 2009 in NSW, and 27,631 mothers who had conceived 36,964 babies born between 2009 and 2016 plus 8,088 siblings of these babies born before 2009 in the ACT. The resulting longitudinal health record provided up to 10.25 years of follow-up. Of these NSW/ACT mothers, 37,443 (6.2%) had at least one ANZARD ART treatment cycle record. ANZARD was only linked to state and Commonwealth administrative datasets where a woman was identified as giving birth in NSW or ACT. Unlinked ANZARD records for treatment information for women who had undergone ART and DI treatment and who had not given birth were also uploaded to SURE.
The MAR data linkage resource includes a wide range of data sources. For mothers, information is available on their use of fertility medicines and other medicines, their use of fertility-related services, hospital admissions, history of health conditions, socio-demographics, Aboriginal and/or Torres Strait Islander status, and pregnancy-related risk factors such as hypertension and gestational diabetes (Table 1, Supplementary Appendix). For their offspring, information is available on birth outcomes, birth defects, long-term and short-term adverse health conditions, hospital admissions, death, and cause of death (Table 1, Supplementary Appendix).
Agreement between births in ANZARD and PDC
A primary research question to be addressed by the MAR data linkage is whether health outcomes are different between ART conceived versus non-ART conceived children (from other fertility treatment or spontaneously conceived). Therefore, an assessment was undertaken of the concordance between ANZARD recorded births and those recorded in the PDCs to identify births to women who had ART treatment, who may have conceived using non-ART treatment or spontaneously. This agreement analysis was only conducted for births resulting from ART or DI treatment by NSW residents and birthing in NSW or ACT because the ACT PDC data only covers births delivered in ACT public hospitals. Therefore, births to women residing in the ACT and who gave birth in a private hospital are missing from the ACT PDC, estimated to be about 20–25% of ACT births . The PDCs encompass all live births and stillbirths of at least 20 weeks gestation or ≥400 grams birth weight. The record of birth and pregnancy information by ANZARD relies on the ART clinic staff following-up with women after their ART treatment, while the PDC relies on the attending midwife or medical practitioner completing a record of birth and pregnancy information.
We relied on baby’s DOB, gestational age, embryo transfer or DI date, and the age of embryo at transfer to match each ANZARD birth to a PDC birth of the corresponding mother. Where the babies’ DOB in ANZARD did not exactly match with a PDC-recorded month and year of birth, we progressively relaxed the matching requirements, allowing a grace period of 16 days (see criterion 1 in Table 1). For the remaining unmatched ANZARD births, we then used the embryo transfer or DI date (from ANZARD data) to extrapolate a birth window of an expected DOB to match to the PDC’s DOB (see criterion 2 in Table 1). We progressively relaxed the birth window to account for uncertainty in the embryo transfer or DI date and ANZARD recorded DOB by a grace period of 10 days (Table 1). This was necessary because only the babies’ month and year of birth were provided by the PDC.
|Criteria||embryo stage||Birth window|
|1||1. Babies’ month and year of birth recorded in ANZARD matched exactly with that of the PDC|
|2. For the remaining unmatched ANZARD births, we progressively relaxed by a grace period of 16 days|
|Lower limit: Baby’s DOB from ANZARD data - grace period1|
|Upper limit: Baby’s DOB from ANZARD data + grace period1|
|2||Cleavage (3 day old embryo)||Lower limit: Embryo transfer date - 3 days - 14 days +7xgestational weeks - grace period2|
|Upper limit: Embryo transfer date - 3 days - 14 days +7xgestational weeks +grace period2|
|Blastocysts (5 day old embryo)||Lower limit: Embryo transfer date - 5 days - 14 days +7xgestational weeks - grace period2|
|Upper limit: Embryo transfer date - 5 days - 14 days +7xgestational weeks +grace period2|
|Both cleavage and blastocysts||Lower limit: Embryo transfer date - 5 days - 14 days +7xgestational weeks - grace period2|
|Upper limit: Embryo transfer date - 3 days - 14 days +7xgestational weeks +grace period2|
|Donor Insemination (DI)||Lower limit: DI date - 14 days +7xgestational weeks - grace period2|
|Upper limit: DI date - 14 days +7xgestational weeks + grace period2|
The data concordance rate was calculated as a total number of births in an agreement between ANZARD and PDC data divided by a total number of ANZARD treatment records from NSW residents with an ANZARD birth recorded or without an ANZARD birth recorded due to loss of follow-up. Fact of birth plus key birth outcomes (live or stillbirth) and plurality (singleton or multiple births) were chosen to evaluate concordance because these are recorded in ANZARD and the PDCs.
To examine the impact of these grace periods (criterion 1: ANZARD recorded DOB and criterion 2: embryo transfer or DI date) on the concordance rate, we performed several sensitivity analyses by varying the grace period of criterion 1 from 16 days to 5 or 31 days; and by varying the grace period of criterion 2 from 10 days to 5 or 15 days.
Concordance of plurality status and birth outcomes between ANZARD and the PDCs
For births in the agreement between the PDC and ANZARD birth records for NSW residents, we also examined agreement in the plurality (i.e., singleton, twins, and triplets) and birth outcome status (live birth or perinatal death) fields. We relied on the birth registry’s plurality status, and death registry’s perinatal death information if information in the PDC differed from that of the birth or death registries. We used both PDC and the Registry of Births, Deaths and Marriages (RBDM) (birth and death registries) as a gold standard reference when estimating sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV). The RBDM is a statutory registry in NSW and ACT. Agreement rate, sensitivity, specificity, PPV, NPV, area under the receiver operating characteristic curve (AUC), and Cohen’s Kappa statistics [27, 28] were reported.
Results of the deterministic data linkage for women recorded by ANZARD SLKs and NSW/ACT MLKs
Between 1st January, 2009 and 31st December, 2016, a total of 195,490 ANZARD SLKs (each representing a woman receiving ART treatment in Australia and New Zealand) were deterministically linked to MLKs of females recorded by CHeReL. Table 2 presents the linkage results of the nine passes of this data linkage process. Of the 195,490 ANZARD SLKs, 64,206 were linked to a single MLK person ID. There were 160 ANZARD records that linked to multiple MLK records. For these, a clerical review was performed to identify duplicates and use other data items to identify likely links. The linkage completed by the CHeReL was also checked for false-positive links by selecting a random sample of 1000 Person IDs from the linked data for review. The results indicated a false positive rate of 5/1000 Person IDs. Table 2 presents the linkage rate between ANZARD SLKs and the MLK for female patients with a NSW or ACT residential postcode (aligning with the state administrative datasets including the PDCs), or an unknown postcode. Of the 195,490 ANZARD SLKs, there were 62,833 ANZARD SLKs with a NSW or ACT residential postcode. Of these, 60,419 (including NSW or ACT) were linked to an MLK person ID, resulting in a 96.2% linkage rate between women residing in NSW or ACT who underwent ART treatment and being able to be identified in the MLK (i.e., 60,419/62,833).
|Pass||Description||ANZARD SLK IDs||Multi-links to MLK|
|1||First name code, surname code, DOB, partner DOB, postcode, ANZARD procedure date matches APDC subset dates||22,983||35.8%||11|
|2||First name code, surname code, DOB, partner DOB, postcode, ANZARD procedure date not NULL, and MLK in procedure subset||156||0.2%||0|
|3||First name code, surname code, DOB, partner DOB, postcode||9,385||14.6%||6|
|4||First name code, surname code, DOB, postcode, ANZARD procedure date matches APDC subset dates||17,339||27.0%||19|
|5||First name code, surname code, DOB, postcode, where either ANZARD or MLK partner DOB is NULL||12,296||19.2%||118|
|6||First name code, surname code, DOB, edit distance between partner DOB and MLK partner DOB ≤ 2, postcode||105||0.2%||0|
|7||First name code, surname code, DOB, partner DOB, edit distance between postcode and MLK postcode ≤ 1||241||0.4%||2|
|8||First name code, surname code, DOB, postcode in NSW or ACT, MLK in procedure subset||688||1.1%||2|
|9||First name code, surname code, DOB, partner DOB||1,013||1.6%||2|
|Total Linked ANZARDSLKs to NSW and ACT MLKs||64,206|
|Calculating linkage rate between ANZARD SLKs and MLKs|
|Total number of ANZARD SLK (with a NSW/ACT postcode) (A)||62,833|
|Total number of ANZARD SLKs (with a NSW/ACT postcode) linked to MLKs (B)||60,419|
|Linkage rate between ANZARD SLKs and MLKs (B/A)||96.2%|
Results of the probabilistic data linkage between the NSW/ACT MLKs for Mothers who gave birth and the Commonwealth Medicare Benefits Schedule and Pharmaceutical Benefits Scheme data
Overall, there were 606,658 PPNs from mothers who gave birth in NSW between 2009 and 2017, and from mothers who gave births in the ACT between 2009 and 2016, respectively. Of these, 597,549 (98.49%) were linked to the MEF’s personal identifiers representing a linkage accuracy of 99% between the PDCs and national Medicare enrolments.
Overall concordance rate for births recorded in ANZARD and the PDC data
To assess the concordance rate between ANZARD and the PDC, the cohort was restricted to ANZARD’s cycles performed between 2009 and 2015 (rather than 2016) because ACT PDC data was only available up to 31st December, 2016, and one year of follow-up after an ART or DI cycle was needed to capture births in the PDCs. Between 1st January, 2009 and 31st December, 2015, there were 400,853 cycles of ART or DI treatment records recorded in ANZARD (Figure 2). Of these, 28,937 cycles were from women residing in NSW with a birth outcome or unknown outcome due to loss follow-up. Of these, 27,248 were matched to an NSW or ACT PDC birth record via baby’s DOB, gestational age, embryo transfer or DI date, and the age of embryo at transfer, resulting in a concordance rate of 94.2% (27,248/28,937) (95% CI: 93.9–94.4%). Of these 27,248 births from women with ART or DI cycles in NSW, 158 (0.58%) had a cross-state delivery in an ACT public hospital. The concordance rate remained stable by year (Figure S1, Supplementary Appendix).
Our sensitivity analysis, in which we used different combinations of grace periods, showed consistent results of the concordance rate between ART/DI treatment cycles and NSW and ACT PDC records at 94.2% over our study period (Table S2, Supplementary Appendix).
Concordance of plurality status and birth outcomes between ANZARD and PDC data for the final MAR data linkage
Of the 27,248 ANZARD births (from NSW and ACT residents) linked to a NSW/ACT PDC birth, there was 99.7% (27,170/27,248) (95% CI: 99.6–99.8%) agreement with Cohen’s kappa of 0.977 (95% CI: 0.971–0.982) in plurality recording between ANZARD and PDC data (Table S3, Supplementary Appendix). There was also a high degree of agreement (≥99%, Cohen’s kappa ranged: 0.78–0.90) for live birth and perinatal death status between ANZARD and PDC records for the 25,758 singleton births, with a range of 87.0–99.9% for PPV and a range of 87.0–99.9% for NPV; for 1,412 plural births, with a range of 87.0–99.3% for PPV and a range of 88.9–99.2% for NPV (Table 3).
|Birth outcomes||Agreement (95% CI)||Sensitivity (95% CI)||Specificity (95% CI)||PPV (95% CI)||NPV (95% CI)||AUC (95% CI)||Kappa statistics (95% CI) 4|
|Singleton births||(N = 25,758)|
|(99.72, 99.83)||(99.8, 99.9)||(84.5, 93.0)||(99.8, 99.9)||(82.0, 91.1)||(0.93, 0.97)||(0.84, 0.91)|
|(99.72, 99.83)||(88.3, 95.7)||(99.8, 99.9)||(82.0, 91.1)||(99.8, 100)||(0.95, 0.98)||(0.87, 0.93)|
|Plural births||(N = 1,412)|
|(98.5, 99.3)||(99.5, 99.9)||(57.3, 80.1)||(98.8, 99.5)||(77.4, 95.8)||(0.79, 0.90)||(0.69, 0.86)|
|(98.6, 99.3)||(58.7, 81.7)||(99.5, 99.9)||(75.1, 94.6)||(98.9, 99.6)||(0.80, 0.91)||(0.70, 0.86)|
This paper describes the creation of a bespoke linked dataset (MAR data linkage) of a clinical quality registry (ANZARD) with state/territory and national administrative datasets. Despite limited personal identifiers being present in the registry, of the 62,833 women who had ART treatment in NSW or ACT, 60,419 could be linked to the CHeReL MLK population spine, representing a linkage rate: 96.2%. This means that only 3.8% of women who had ART treatment in NSW/ACT could not be linked to the NSW/ACT MLK created by CHeReL. This linkage rate is similar to other studies that have used limited identifiers as part of an SLK used to provide partial identifiers are part of clinical registries [29–31].
A reconciliation of the ART/DI cycles performed to women who resided in NSW and who were recorded as having a birth in ANZARD, found that 94.2% of the births were recorded in NSW and ACT PDCs. Possible reasons for the 5.8% of missing ANZARD births in the PDC, would include women who resided in NSW for ART/DI treatment but birthed in a private ACT hospital, in another Australian states/territories, or overseas, missing links between the SLK and MLK (3.8% linkage error), or births being erroneously recorded in ANZARD. An evaluation of the small percentage of NSW or ACT women recorded in ANZARD who could not be linked to the MLK population spine was not conducted as part of this study and could be a small source of linkage bias. However, because of Australia’s universal health system it is unlikely that there would be a systematic bias in the linkage between ANZARD and the MLK based on demographics because the MLK contains 210 million records from 17 data collections with 15 average links per person .
A high concordance was found in plurality status (>99% agreement rate; Cohen’s kappa: 0.977 (95% CI: 0.971–0.982)) and birth outcome (≥99% agreement rate; Cohen’s kappa ranged: 0.78–0.90) between ANZARD and PDC birth records confirming the validity of the linkage. The high degree of concordance between births recorded in ANZARD and those recorded in jurisdictional perinatal data collections provides reassurance that fertility clinics in Australia are accurately recording the outcomes of ART treatment undertaken in their clinics, and that clinics are not artificially inflating their success rates in NSW/ACT. The accreditation of each fertility clinic in Australia is managed by the industry’s Fertility Society of Australia and New Zealand under its Reproductive Technology Accreditation Committee’s voluntary Code of Practice, under which clinics must submit their treatment and outcomes data to ANZARD . The results of this concordance study and the high linkage rate reflects positively on the Fertility Society of Australia and New Zealand model of industry regulation being connected with clinical registry management.
Our linkage rate (96.2% at woman-level, 94.2% at birth-level) was higher than that achieved by the States Monitoring Assisted Reproductive Technology collaboration’s linkage of the U.S. state-based and national ART data linkage to vital birth registrations (80–90.2%) [34–36] and the linkage to pregnancy data (89.7%) in the Massachusetts Outcomes Study of Assisted Reproductive Technology . The Committee of Nordic Assisted Reproductive Technology and Safety was able to achieve a very high linkage between IVF registries and birth registries because of the existence of national personal identifiers .
Most of the earlier U.S. linkages conducted by the States Monitoring Assisted Reproductive Technology collaboration were cycle-based and adopted a deterministic or probabilistic linkage strategy to link treatment cycles to vital records based on less specific maternal or infant variables (e.g., DOB of mothers and infants, mother’s postcode, or plurality, etc.) due to a lack of mothers’ or infants’ identifiers [34–36]. The results of the latest U.S. linkage by the Massachusetts Outcomes Study of Assisted Reproductive Technology that included maternal or parental identifiers (i.e. mother’s first and last name, and father’s last name) in the linkage strategy is the most comparable to that of the MAR data linkage (89.7% vs. 94.2%, at birth-level) . The MAR linkage also shows a comparable rate of agreement of plurality status (>99%) and birth outcome (≥99%) but a higher sensitivity (71.2–99.9% vs. 27.1–47.4%) and PPV (87.0–99.9% vs. 41.3–69.2%) for live birth/fetal death outcomes between the ANZARD and the PDC data [37, 39, 40].
The MAR linkage contains up to 10.25 years of follow-up allowing the assessment of the short-term and long-term health risks for women and ART conceived children. Additionally, the prognostic value of the type of ART treatment performed, (e.g., use of fresh or frozen embryo, sperm injection, extended embryo culture) will be able to be assessed. Furthermore, the health of children born from non-ART treatments will be evaluated using national medicines and medical services claims data to identify children conceived using ovulation induction and ovarian stimulation. Moreover, because of the longitudinal nature of the datasets, women with a history of subfertility, but who subsequently conceived naturally can be identified, allowing the role of subfertility in health outcomes to be assessed – a confounder that is often elusive in studies of ART conceived children. Finally, sibship studies will also be possible because children born to the same mothers (including siblings from plural births or from singleton births since 1994) can be identified.
The MAR data linkage demonstrates that very high linkage rates can be achieved with partially identifiable data, and that a population spine such as the CHeReL’s MLK can be successfully used as a bridge between clinical registries and administrative datasets. The high concordance between births recorded in ANZARD and perinatal data collections provides reassurance about the accuracy of ART treatment outcomes recorded in ANZARD. The MAR data linkage will provide invaluable information on the safety and effectiveness of ART and non-ART treatment, and the possible effect of subfertility when advising patients, clinicians, and policymakers on fertility treatments for Australia and beyond.
Statement on conflicts of interest
G.C. is an employee of The University of New South Wales(UNSW) and Director of the National Perinatal Epidemiologyand Statistics Unit (NPESU), UNSW. The NPESU managesthe Australian and New Zealand Assisted ReproductionDatabase with funding support from the Fertility Society ofAustralia and New Zealand.
The MAR data linkage was funded by the Australian National Health and Medical Research Council (NHMRC #1127437). This study was approved by the NSW Population and Health Services Research Ethics Committee (2017/HRE1202), the ACT Health Human Research Ethics Committee (ETH.2.218.032), the Calvary Public Hospital Human Research and Ethics Committee (3-2018), the Australian Institute of Health and Welfare Ethics Committee (AIHW) (EO2017/4/420), and the ANZARD Management Committee.
This study was funded through Australian National Health andMedical Research Council (NHMRC) grant: APP1127437.
The authors wish to acknowledge the provision of data to the Australian and New Zealand Assisted Reproduction Database (ANZARD) clinical registry by Australian and New Zealand fertility clinics, and the Fertility Society of Australia and New Zealand for its continued support of ANZARD. The authors gratefully acknowledge the assistance of the Centre for Health Record Linkage (CHeReL) and the Australian Institute of Health and Welfare (AIHW) Data Integration Unit to link the ANZARD data to various state/territory and Commonwealth administrative data. The authors are grateful to the NSW Ministry of Health, ACT Heath and the AIHW for the provision of NSW and ACT Perinatal Data Collection; NSW Admitted Patient Data Collection and ACT Admitted Patient Care; NSW and ACT Registry of Births, Death and Marriage; NSW and ACT Cause of Death Registry Unit Record File; NSW Register of Congenital Condition; Commonwealth Pharmaceutical Benefits Scheme; Commonwealth Medicare Benefits Schedule from which this study was sourced. The authors are grateful this study was funded through an Australian National Health and Medical Research Council (NHMRC) grant: APP1127437.
|ACT||Australian Capital Territory|
|ACT APC||Australian Capital Territory Admitted Patient Care|
|AIHW||Australian Institute of Health and Welfare|
|ANZARD||Australia and New Zealand Assisted Reproduction Database|
|ART||Assisted reproductive technologies|
|CHeReL||New South Wales Centre for Health Record Linkage|
|DOB||Date of birth|
|MAR||Medically Assisted Reproduction|
|MEF||Medicare Enrolment File|
|MLK||Master Linkage Key|
|NPESU||National Perinatal Epidemiology and Statistics Unit|
|NSW||New South Wales|
|NSW APDC||New South Wales Admitted Patient Data Collection|
|PDC||Perinatal data collections|
|PPN||Project Person Number|
|RBDM||Registry of Births, Deaths and Marriages|
|SLK||Statistical Linkage Key|
Inhorn MC, Patrizio P. Infertility around the globe: New thinking on gender, reproductive technologies and global movements in the 21st century. Hum Reprod Update. 2015;21:411–26. 10.1093/humupd/dmv016.https://doi.org/10.1093/humupd/dmv016
Newman JE, Paul RC, Chambers GM. Assisted reproductive technology in Australia and New Zealand 2018. 2020. https://npesu.unsw.edu.au/data-collection/australian-new-zealand-assisted-reproduction-database-anzard.
Sun H, Gong T-T, Jiang Y-T, Zhang S, Zhao Y-H, Wu Q-J. Global, regional, and national prevalence and disability-adjusted life-years for infertility in 195 countries and territories, 1990-2017: Results from a global burden of disease study, 2017. Aging (Albany NY). 2019;11:10952–91. 10.18632/aging.102497.https://doi.org/10.18632/aging.102497
De GC, Calhaz-Jorge C, Kupka MS, Wyns C, Mocanu E, Motrenko T, et al. ART in Europe, 2014: Results generated from European registries by ESHRE: The European IVF-monitoring Consortium (EIM) for the European Society of Human Reproduction and Embryology (ESHRE). Hum Reprod 2018. 10.1093/humrep/dey242.https://doi.org/10.1093/humrep/dey242
Kulkarni AD, Jamieson DJ, Jones HW, Kissin DM, Gallo MF, Macaluso M, Adashi EY. Fertility treatments and multiple births in the United States. N Engl J Med. 2013;369:2218–25. 10.1056/NEJMoa1301467.https://doi.org/10.1056/NEJMoa1301467
Beaujouan E. Latest-Late Fertility? Decline and Resurgence of Late Parenthood Across the Low-Fertility Countries. Popul Dev Rev. 2020;46:219–47. 10.1111/padr.12334.https://doi.org/10.1111/padr.12334
Levine H, Jørgensen N, Martino-Andrade A, Mendiola J, Weksler-Derri D, Mindlis I, et al. Temporal trends in sperm count: A systematic review and meta-regression analysis. Hum Reprod Update. 2017;23:646–59. 10.1093/humupd/dmx022.https://doi.org/10.1093/humupd/dmx022
Schmidt L, Sobotka T, Bentzen JG, Nyboe Andersen A. Demographic and medical consequences of the postponement of parenthood. Hum Reprod Update. 2012;18:29–43. 10.1093/humupd/dmx022.https://doi.org/10.1093/humupd/dmx022
Talmor A, Dunphy B. Female obesity and infertility. Best Pract Res Clin Obstet Gynaecol. 2015;29:498–506. 10.1016/j.bpobgyn.2014.10.014.https://doi.org/10.1016/j.bpobgyn.2014.10.014
TSEVAT DG, WIESENFELD HC, Parks C, PEIPERT JF. Sexually Transmitted Diseases and Infertility. Am J Obstet Gynecol. 2017;216:1–9. 10.1016/j.ajog.2016.08.008.https://doi.org/10.1016/j.ajog.2016.08.008
Adamson GD, de Mouzon J, Chambers GM, Zegers-Hochschild F, Mansour R, Ishihara O, et al. International Committee for Monitoring Assisted Reproductive Technology: World report on assisted reproductive technology, 2011. Fertil Steril. 2018;110:1067–80. 10.1016/j.fertnstert.2018.06.039.https://doi.org/10.1016/j.fertnstert.2018.06.039
Canberra:Australia: Australia’s mothers and babies 2018: in brief; 2020.
Palomba S, Homburg R, Santagni S, La Sala GB, Orvieto R. Risk of adverse pregnancy and perinatal outcomes after high technology infertility treatment: A comprehensive systematic review. Reprod Biol Endocrinol. 2016;14:76. 10.1186/s12958-016-0211-8.https://doi.org/10.1186/s12958-016-0211-8
Zhao J, Yan Y, Huang X, Li Y. Do the children born after assisted reproductive technology have an increased risk of birth defects? A systematic review and meta-analysis. J Matern Fetal Neonatal Med. 2018:1–12. 10.1080/14767058.2018.1488168.https://doi.org/10.1080/14767058.2018.1488168
Giorgione V, Parazzini F, Fesslova V, Cipriani S, Candiani M, Inversetti A, et al. Congenital heart defects in IVF/ICSI pregnancy: Systematic review and meta-analysis. Ultrasound Obstet Gynecol. 2018;1 (51):33–42. 10.1002/uog.18932.https://doi.org/10.1002/uog.18932
Dayan N, Joseph KS, Fell DB, Laskin CA, Basso O, Park AL, et al. Infertility treatment and risk of severe maternal morbidity: A propensity score–matched cohort study. CMAJ. 2019;191:E118-27. 10.1503/cmaj.181124.https://doi.org/10.1503/cmaj.181124
Lacamara C, Ortega C, Villa S, Pommer R, Schwarze JE. Are children born from singleton pregnancies conceived by ICSI at increased risk for congenital malformations when compared to children conceived naturally? A systematic review and meta-analysis. JBRA Assist Reprod. 2017;21:251–9. 10.5935/1518-0557.20170047.https://doi.org/10.5935/1518-0557.20170047
Messerlian C, Gaskins AJ. Epidemiologic Approaches for Studying Assisted Reproductive Technologies: Design, Methods, Analysis and Interpretation. Current epidemiology reports. 2017;4:124–32. 10.1007/s40471-017-0105-0.https://doi.org/10.1007/s40471-017-0105-0
Reigstad MM, Larsen IK, Myklebust TÅ, Robsahm TE, Oldereid NB, Brinton LA, Storeng R. Risk of Cancer in Children Conceived by Assisted Reproductive Technology. Pediatrics. 2016;137:e20152061. 10.1542/peds.2015-2061.https://doi.org/10.1542/peds.2015-2061
Palomba S, Santagni S, Daolio J, Gibbins K, Battaglia FA, La Sala GB, Silver RM. Obstetric and perinatal outcomes in subfertile patients who conceived following low technology interventions for fertility enhancement: A comprehensive review. Arch Gynecol Obstet. 2018;297:33–47. 10.1007/s00404-017-4572-9.https://doi.org/10.1007/s00404-017-4572-9
Ethical guidelines on the use of assisted reproductive technology in clinical practice and research. Canberra: National Health and Medical Research Council; 2017.
Australian Bureau of Statistics. National, state and territory population. June 2020. https://www.abs.gov.au/statistics/people/population/national-state-and-territory-population/jun-2020.
Irvine K, Hall R, Taylor L. Centre for Health Record Linkage. International journal of population data science 2019. 10.23889/ijpds.v4i2.1142.https://doi.org/10.23889/ijpds.v4i2.1142
Taylor LK, Irvine K, Iannotti R, Harchak T, Lim K. Optimal strategy for linkage of datasets containing a statistical linkage key and datasets with full personal identifiers. BMC Med Inform Decis Mak. 2014;14:85. 10.1186/1472-6947-14-85.https://doi.org/10.1186/1472-6947-14-85
Sydney: Australia: SaxInstitute;.
Cohen J. A Coefficient of Agreement for Nominal Scales. Educational and Psychological Measurement. 1960;20:37–46. 10.1177/001316446002000104.https://doi.org/10.1177/001316446002000104
Fleiss JL, Levin B, Paik MC, Shewart WA, Wilks SS. Statistical Methods for Rates and Proportions 2003. 10.1002/0471445428.https://doi.org/10.1002/0471445428
Coulson TG, Bailey M, Reid C, Shardey G, Williams-Spence J, Huckson S, et al. Linkage of Australian national registry data using a statistical linkage key. BMC Med Inform Decis Mak. 2021;21:37. 10.1186/s12911-021-01393-1.https://doi.org/10.1186/s12911-021-01393-1
Karmel R, Anderson P, Gibson D, Peut A, Duckett S, Wells Y. Empirical aspects of record linkage across multiple data sets using statistical linkage keys: The experience of the PIAC cohort study. BMC Health Serv Res. 2010;10:41. 10.1186/1472-6963-10-41.https://doi.org/10.1186/1472-6963-10-41
Taylor LK, Irvine K, Iannotti R, Harchak T, Lim K. Optimal strategy for linkage of datasets containing a statistical linkage key and datasets with full personal identifiers. BMC Med Inform Decis Mak. 2014;14:85. 10.1186/1472-6947-14-85.https://doi.org/10.1186/1472-6947-14-85
Centre for Health Record Linkage. Master Linkage Key (MLK). 2021. https://www.cherel.org.au/master-linkage-key.
Reproductive Technology Accreditation Committee. Code of Practise for Assisted Reproductive Technology Units in Australia and New Zealand. 2017. https://www.fertilitysociety.com.au/wp-content/uploads/2017-RTAC-ANZ-COP-FINAL-1.pdf.
Sunderam S, Schieve LA, Cohen B, Zhang Z, Jeng G, Reynolds M, et al. Linking birth and infant death records with assisted reproductive technology data: Massachusetts, 1997-1998. Matern Child Health J. 2006;10:115–25. 10.1007/s10995-005-0013-7.https://doi.org/10.1007/s10995-005-0013-7
Zhang Y, Cohen B, Macaluso M, Zhang Z, Durant T, Nannini A. Probabilistic linkage of assisted reproductive technology information with vital records, Massachusetts 1997-2000. Matern Child Health J. 2012;16:1703–8. 10.1007/s10995-011-0877-7.https://doi.org/10.1007/s10995-011-0877-7
Mneimneh AS, Boulet SL, Sunderam S, Zhang Y, Jamieson DJ, Crawford S, et al. States Monitoring Assisted Reproductive Technology (SMART) Collaborative: Data collection, linkage, dissemination, and use. J Womens Health (Larchmt). 2013;22:571–7. 10.1089/jwh.2013.4452.https://doi.org/10.1089/jwh.2013.4452
Kotelchuck M, Hoang L, Stern JE, Diop H, Belanoff C, Declercq E. The MOSART database: Linking the SART CORS clinical database to the population-based Massachusetts PELL reproductive public health data system. Matern Child Health J. 2014;18:2167–78. 10.1007/s10995-014-1465-4.https://doi.org/10.1007/s10995-014-1465-4
Cambridge: Cambridge University Press; 2019.
Stern JE, Gopal D, Liberman RF, Anderka M, Kotelchuck M, Luke B. Validation of birth outcomes from the Society for Assisted Reproductive Technology Clinic Outcome Reporting System (SART CORS): Population-based analysis from the Massachusetts Outcome Study of Assisted Reproductive Technology (MOSART). Fertil Steril. 2016;106:717–722.e2. 10.1016/j.fertnstert.2016.04.042.https://doi.org/10.1016/j.fertnstert.2016.04.042
Cohen B, Bernson D, Sappenfield W, Kirby RS, Kissin D, Zhang Y, et al. Accuracy of assisted reproductive technology information on birth certificates: Florida and Massachusetts, 2004-06. Paediatr Perinat Epidemiol. 2014;28:181–90. 10.1111/ppe.12110.https://doi.org/10.1111/ppe.12110
This work is licensed under a Creative Commons Attribution 4.0 International License.