Accuracy of reporting of Aboriginality on administrative health data collections using linked data in NSW, Australia.

BACKGROUND: Aboriginal people are under-reported on administrative health data in Australia. Various approaches have been used or proposed to improve reporting of Aboriginal people using linked records. This cross-sectional study used self-reported Aboriginality from the NSW Patient Survey Program (PSP) as a reference standard to assess the accuracy of reporting of Aboriginal people on NSW Admitted Patient (APDC) and Emergency Department Data Collections (EDDC), and compare the accuracy of selected approaches to enhance reporting Aboriginality using linked data. METHODS: Ten PSP surveys were linked to five administrative health data collections, including APDC, EDDC, perinatal, and birth and death registration records. Accuracy of reporting of Aboriginality was assessed using sensitivity, specificity, and positive and negative predictive values (PPVs and NPVs) and F score for the EDDC and APDC as baseline and four enhancement approaches using linked records: "Most recent linked record", "Ever reported as Aboriginal", and two approaches using a weight of evidence, "Enhanced Reporting of Aboriginality (ERA) algorithm" and "Multi-stage median (MSM)". RESULTS: There was substantial under-reporting of Aboriginality on APDC and EDDC records (sensitivities 84 and 77% respectively) with PPVs of 95% on both data collections. Overall, specificities and NPVs were above 98%. Of people who were reported as Aboriginal on the PSP, 16% were not reported as Aboriginal on any of their linked records. Record linkage approaches generally increased sensitivity, accompanied by decrease in PPV with little change in overall F score for the APDC and an increase in F score for the EDDC. The "ERA algorithm" and "MSM" approaches provided the best overall accuracy. CONCLUSIONS: Weight of evidence approaches are preferred when record linkage is used to improve reporting of Aboriginality on administrative health data collections. However, as a substantial number of Aboriginal people are not reported as Aboriginal on any of their linked records, improvements in reporting are incomplete and should be taken into account when interpreting results of any analyses. Enhancement of reporting of Aboriginality using record linkage should not replace efforts to improve recording of Aboriginal people at the point of data collection and addressing barriers to self-identification for Aboriginal people.

Cardiac procedures in ST-segment-elevation myocardial infarction - the influence of age, geography and Aboriginality.

BACKGROUND: Timely restoration of bloodflow acute ST-segment elevation myocardial infarction (STEMI) reduces myocardial damage and improves prognosis. The objective of this study was describe the association of demographic factors with hospitalisation rates for STEMI and time to angiography, Percutaneous Coronary Intervention (PCI) and Coronary Artery Bypass Graft (CABG) in New South Wales (NSW) and the Australian Capital Territory (ACT), Australia. METHODS: This was an observational cohort study using linked population health data. We used linked records of NSW and the ACT hospitalisations and the Australian Government Medicare Benefits Schedule (MBS) for persons aged 35 and over hospitalised with STEMI in the period 1 July 2010 to 30 June 2014. Survival analysis was used to determine the time between STEMI admission and angiography, PCI and CABG, with a competing risk of death without cardiac procedure. RESULTS: Of 13,117 STEMI hospitalisations, 71% were among males; 55% were 65-plus years; 64% lived in major cities, and 2.6% were Aboriginal people. STEMI hospitalisation occurred at a younger age in males than females. Angiography and PCI rates decreased with age: angiography 69% vs 42% and PCI 60% vs 34% on day 0 for ages 35-44 and 75-plus respectively. Lower angiography and PCI rates and higher CABG rates were observed outside major cities. Aboriginal people with STEMI were younger and more likely to live outside a major city. Angiography, PCI and CABG rates were similar for Aboriginal and non-Aboriginal people of the same age and remoteness area. CONCLUSIONS: There is a need to improve access to definitive revascularisation for STEMI among appropriately selected older patients and in regional areas. Aboriginal people with STEMI, as a population, are disproportionately affected by access to definitive revascularisation outside major cities. Improving access to timely definitive revascularisation in regional areas may assist in closing the gap in cardiovascular outcomes between Aboriginal and non-Aboriginal people.

Optimal strategy for linkage of datasets containing a statistical linkage key and datasets with full personal identifiers.

BACKGROUND: Linkage of aged care and hospitalisation data provides valuable information on patterns of health service utilisation among aged care service recipients. Many aged care datasets in Australia contain a Statistical Linkage Key (SLK-581) instead of full personal identifiers. We linked hospital and death records using a full probabilistic strategy, the SLK-581, and three combined strategies; and compared results for each strategy. METHODS: Linkage of Admitted Patient Data for 2000-01 to 2008-09 and Registry of Births, Deaths and Marriages death registration data for 2008-09 for New South Wales, Australia, was carried out using probabilistic methods and compared to links created using four strategies incorporating a SLK-581. The Basic SLK-581 strategy used the SLK-581 alone. The Most Recent SLK-581, Most Frequent SLK-581, and Any Match SLK-581 strategies leveraged probabilistic links between hospital records drawn from the Centre for Health Record Linkage Master Linkage Key. Rates of hospitalisations among people who died were calculated for each strategy and a range of health conditions. RESULTS: Compared to full probabilistic linkage, the basic SLK-581 strategy produced substantial rates of missed links that increased over the study period and produced underestimates of hospitalisation rates that varied by health condition. The Most Recent SLK-581, Most Frequent SLK-581, and Any Match SLK-581 strategies resulted in substantially lower rates of underestimation than the Basic SLK-581. The Any Match SLK-581 strategy gave results closest to full probabilistic linkage. CONCLUSIONS: Hospitalisation rates prior to death are substantially underestimated by linkage using a SLK-581 alone. Linkage rates can be increased by combining deterministic methods with probabilistically created links across hospital records.

Potential prevention of small for gestational age in Australia: a population-based linkage study.

BACKGROUND: Small for gestational age (SGA) infants are at increased risk of morbidity and mortality. We sought to identify risk factors associated with SGA and examined the potential for reducing the proportion of infants with SGA at a population level. METHODS: Birth and hospital records were linked for births occurring in 2007-2010 in New South Wales, Australia. The analysis was stratified into three groups: preterm births, term births to non-diabetic mothers and term births to diabetic mothers. Logistic regression was used to examine the association between SGA and a range of socio-demographic and behavioural factors and health conditions, with generalised estimating equations to account for correlation among births to the same mother. Model-based population attributable fractions (PAFs) were calculated for risk factors that were considered causative and potentially modifiable. RESULTS: Of 28,126 SGA infants, the largest group was term infants of non-diabetic mothers (88.5%), followed by term infants of diabetic mothers (6.3%) and preterm infants (5.3%). The highest PAFs were for smoking: 12.4% for preterm SGA and 10.3% for term SGA infants of non-diabetic mothers. Other risk factors for SGA that were considered modifiable included: illicit drug dependency or abuse in pregnancy in all three groups, and pregnancy hypertension and late commencement of antenatal care in term infants of non-diabetic mothers, but PAFs were less than 3%. CONCLUSIONS: There are opportunities for modest reduction of the prevalence of SGA through reduction in smoking in pregnancy, and possibly earlier commencement of antenatal care and improved management of high-risk pregnancies.

The impact of cross-jurisdictional patient flows on ascertainment of hospitalisations and cardiac procedures for ST-segment-elevation myocardial infarction in an Australian population.

Introduction: The patient journey for residents of New South Wales (NSW) Australia with ST-elevation myocardial infarction (STEMI) often involves transfer between hospitals and these can include stays in hospitals in other jurisdictions. Objective: To estimate the change in enumeration of STEMI hospitalisations and time to subsequent cardiac procedures for NSW residents using cross-jurisdictional linkage of administrative health data. Methods: Records for NSW residents aged 20 years and over admitted to hospitals in NSW and four adjacent jurisdictions (Australian Capital Territory, Queensland, South Australia, and Victoria) between 1 July 2013 and 30 June 2018 with a principal diagnosis of STEMI were linked with records of the Australian Government Medicare Benefits Schedule (MBS). The number of STEMI hospitalisations, and rates of angiography, percutaneous coronary intervention and coronary artery bypass graft were compared for residents of different local health districts within NSW with and without inclusion of cross-jurisdictional data. Results: Inclusion of cross-jurisdictional hospital and MBS data increased the enumeration of STEMI hospitalisations for NSW residents by 8% (from 15,420 to 16,659) and procedure rates from 85.6% to 88.2%. For NSW residents who lived adjacent to a jurisdictional border, hospitalisation counts increased by up to 210% and procedure rates by up to 70 percentage points. Conclusions: Cross-jurisdictional linked hospital data is essential to understand patient journeys of NSW residents who live in border areas and to evaluate adherence to treatment guidelines for STEMI. MBS data are useful where hospital data are not available and for procedures that may be conducted in out-patient settings.

Enhanced reporting of deaths among Aboriginal and Torres Strait Islander peoples using linked administrative health datasets.

BACKGROUND: Aboriginal and Torres Strait Islander peoples are under-reported in administrative health datasets in NSW, Australia. Correct reporting of Aboriginal and Torres Strait Islander peoples is essential to measure the effectiveness of policies and programmes aimed at reducing the health disadvantage experienced by Aboriginal and Torres Strait Islander peoples. This study investigates the potential of record linkage to enhance reporting of deaths among Aboriginal and Torres Strait Islander peoples in NSW, Australia. METHODS: Australian Bureau of Statistics death registration data for 2007 were linked with four population health datasets relating to hospitalisations, emergency department attendances and births. Reporting of deaths was enhanced from linked records using two methods, and effects on patterns of demographic characteristics and mortality indicators were examined. RESULTS: Reporting of deaths increased by 34.5% using an algorithm based on a weight of evidence of a person being Aboriginal or Torres Strait Islander, and by 56.6% using an approach based on 'at least one report' of a person being Aboriginal or Torres Strait Islander. The increase was relatively greater in older persons and those living in less geographically remote areas. Enhancement resulted in a reduction in the urban-remote differential in median age at death and increases in standardised mortality ratios particularly for chronic conditions. CONCLUSIONS: Record linkage creates a statistical construct that helps to correct under-reporting of deaths and potential bias in mortality statistics for Aboriginal and Torres Strait Islander peoples.

Investigating linkage rates among probabilistically linked birth and hospitalization records.

BACKGROUND: With the increasing use of probabilistically linked administrative data in health research, it is important to understand whether systematic differences occur between the populations with linked and unlinked records. While probabilistic linkage involves combining records for individuals, population perinatal health research requires a combination of information from both the mother and her infant(s). The aims of this study were to (i) describe probabilistic linkage for perinatal records in New South Wales (NSW) Australia, (ii) determine linkage proportions for these perinatal records, and (iii) assess records with linked mother and infant hospital-birth record, and unlinked records for systematic differences. METHODS: This is a population-based study of probabilistically linked statutory birth and hospital records from New South Wales, Australia, 2001-2008. Linkage groups were created where the birth record had complete linkage with hospital admission records for both the mother and infant(s), partial linkage (the mother only or the infant(s) only) or neither. Unlinked hospital records for mothers and infants were also examined. Rates of linkage as a percentage of birth records and descriptive statistics for maternal and infant characteristics by linkage groups were determined. RESULTS: Complete linkage (mother hospital record - birth record - infant hospital record) was available for 95.9% of birth records, partial linkage for 3.6%, and 0.5% with no linked hospital records (unlinked). Among live born singletons (complete linkage = 96.5%) the mothers without linked infant records (1.6%) had slightly higher proportions of young, non-Australian born, socially disadvantaged women with adverse pregnancy outcomes. The unlinked birth records (0.4%) had slightly higher proportions of nulliparous, older, Australian born women giving birth in private hospitals by caesarean section. Stillbirths had the highest rate of unlinked records (3-4%). CONCLUSIONS: This study shows that probabilistic linkage of perinatal records can achieve high, representative levels of complete linkage. Records for mother's that did not link to infant records and unlinked records had slightly different characteristics to fully linked records. However, these groups were small and unlikely to bias results and conclusions in a substantive way. Stillbirths present additional challenges to the linkage process due to lower rates of linkage for lower gestational ages, where most stillbirths occur.

Newborn feeding practices at the time of discharge from hospital in NSW in 2007: a descriptive study.

OBJECTIVES: To describe the pattern of infant feeding at discharge from care after birth and the characteristics of mothers who are at risk of low rates of breastfeeding. METHODS: Data were obtained from the NSW Midwives Data Collection for 2007. Information on infant feeding was obtained for babies who were alive at discharge from care after birth. Of 96 030 births reported, 93 505 (97.4%) were eligible for analysis. A descriptive analysis of factors associated with variations in breastfeeding was carried out. RESULTS: In 2007, 80% of babies were fully breastfed, 7% were partially breastfed, and 13% were not breastfed. Babies born to mothers with the following characteristics had relatively low rates of full breastfeeding: teenage mothers (69%); Aboriginal mothers (64%); mothers born in South-East Asia (71%), North-East Asia (72%) and Melanesia, Micronesia and Polynesia (74%); mothers who commenced antenatal care later than 20 weeks gestation (74%); mothers who smoked (67%); mothers who received general anaesthetic during delivery (67%); mothers who gave birth by caesarean section (76%); mothers living in the most socially disadvantaged areas (73%); mothers living in remote and very remote areas (73% and 76% respectively); and mothers of preterm infants (70%). CONCLUSION: There is a need to improve overall rates of breastfeeding initiation in NSW. Particular attention and support needs to be given to the groups of mothers identified in this study as having relatively low rates of full breastfeeding.

Process trumps potential public good: better vaccine safety through linked cross-jurisdictional immunisation data in Australia.

OBJECTIVE: To provide insights into complexities of seeking access to state and federal cross-jurisdictional data for linkage with the Australian Childhood Immunisation Register (ACIR). We provide recommendations for improving access and receipt of linked datasets involving Australian Government-administered data. METHODS: We describe requirements for linking eleven federal and state data sources to establish a national linked dataset for safety evaluation of vaccines. The required data linkage methodology for integrating cross-jurisdictional data sources is also described. RESULTS: Extensive negotiation was required with 18 different agencies for 21 separate authorisations and 12 ethics approvals. Three variations of the 'best practice' linkage model were implemented. Australian Government approval requests spanned nearly four years from initial request for data, with a further year before ACIR data transfer to the linkage agency. CONCLUSIONS: Integration of immunisation registers with other data collections is achievable in Australia but infeasible for routine and rapid identification of vaccine safety concerns. Lengthy authorisation requirements, convoluted disparate application processes and inconsistencies in data supplied all contribute to delayed data availability. Implications for public health: Delayed data access for safety surveillance prevents timely epidemiological reviews. Poor responsiveness to safety concerns may erode public confidence, compromising effectiveness of vaccination programs through reduced participation.

The definition of opioid-related deaths in Australia: implications for surveillance and policy.

The reported number of deaths caused by opioid use depends on the definition of an opioid-related death. In this study, we used Australian Bureau of Statistics (ABS) mortality data to illustrate how choice of classification codes used to record cause of death can impact on the statistics reported for national surveillance of opioid deaths. Using International Classification of Diseases version 10 (ICD-10) codes from ABS mortality data 1997-2002, we examined all deaths where opioids were reported as a contributing or underlying cause. For the 6-year period there was a total of 5,839 deaths where opioids were reported. Three possible surveillance definitions of accidental opioid-related deaths were examined, and compared to the total number of deaths where opioids were reported for each year. Age restrictions, often placed on surveillance definitions, were also examined. As expected, the number of deaths was higher with the more inclusive definitions. Trends in deaths were found to be similar regardless of the definition used; however, a comparison between Australian states revealed up to a twofold difference in the absolute numbers of accidental opioid-related deaths, depending on the definition. Any interpretation of reported numbers of opioid deaths should specify any restrictions placed on the data, and describe the implications of definitions used.

Diagnosis of fetal growth restriction in perinatal deaths using brain to liver weight ratios.

We determined brain to liver weight ratio (BLWR) thresholds for fetal growth restriction (FGR) using autopsy information on 395 perinatal deaths comprising stillborn babies who died during labour and neonatal deaths. FGR was defined using two methods: (1) birth weight for gestational age (WGA) less than the 10th percentile; and (2) WGA less than the 10th percentile or discordant birth weight/length. The association between BLWR and FGR was investigated using odds ratios, and classification statistics were calculated for a range of BLWR thresholds. Using WGA, 84 cases (21.3%) were FGR and a further 15 cases (n = 99, 25%) had discordant birth weight/length. The BLWR ranged from 1.02 to 7.30 and was positively associated with FGR. BLWR was not associated with FGR for babies with congenital central nervous system or chromosomal abnormalities. Excluding these, for FGR defined using WGA and discordant birth weight/length, a BLWR threshold of 5.0 was 100% predictive of FGR. A BLWR threshold of 3.0 for babies over 28 weeks gestation and 3.7 for more preterm babies optimised case detection while minimising missed and false positive cases. Additional evidence of FGR should be sought for babies with a BLWR of less than 5.0 to confirm FGR.

Reporting postpartum haemorrhage with transfusion: a comparison of NSW birth and hospital data.

AIM: Postpartum haemorrhage rates have been increasing in NSW and internationally, and blood transfusion is required in severe cases. Using routinely collected administrative data provides a convenient method with which to monitor trends in both postpartum haemorrhage and associated transfusion use. In order for this to be feasible however, the reliability of reporting of the conditions needs to be assessed. METHODS: This study used linked data to compare the reporting of postpartum haemorrhage with transfusion as reported in the NSW Admitted Patient Data Collection (hospital data), with the same information obtained from the Perinatal Data Collection (birth data), for births in NSW from 2007 to 2010. RESULTS: The rate of postpartum haemorrhage requiring blood transfusion was 1.0% based on the hospital data and 1.1% based on the birth data, with a rate of 1.7% if identifying cases from either source. Agreement between the two sources improved from fair to moderate over the time period. CONCLUSION: Postpartum haemorrhage requiring transfusion recorded in the birth data shows only moderate agreement with hospital data, so caution is recommended when using this variable for analysis. Linkage of both datasets is recommended to identify birth information from birth data and postpartum haemorrhage with transfusion from hospital data until further validation work has been undertaken.

Are pregnancy outcomes associated with risk factor reporting in routinely collected perinatal data?

AIM: To assess reporting characteristics of commonly dichotomised pregnancy outcomes (e.g. preterm/term birth); and to investigate whether behaviours (e.g. smoking), medical conditions (e.g. diabetes) or interventions (e.g. induction) were reported differently by pregnancy outcomes. METHODS: Further analysis of a previous validation study was undertaken, in which 1680 perinatal records were compared with data extracted from medical records. Continuous and polytomous variables were dichotomised, and risk factor reporting was assessed within the dichotomised outcome groups. Agreement, kappa, sensitivity and positive predictive value calculations were undertaken. RESULTS: Gestational age, birthweight, Apgar scores, perineal trauma, regional analgesia and baby discharge status (live birth/stillbirth) were reported with high accuracy and reliability when dichotomised (kappa values 0.95-1.00, sensitivities 94.7-100.0%). Although not statistically significant, there were trends for hypertension, infant resuscitation and instrumental birth to be more accurately reported among births with adverse outcomes. In contrast, smoking ascertainment tended to be poorer among preterm births and when babies were <2500 g. CONCLUSION: Dichotomising variables collected as continuous or polytomous variables in birth data results in accurate and well ascertained data items. There is no evidence of systematic differential reporting of risk factors.

Reporting of Aboriginal and Torres Strait Islander peoples on the NSW Admitted Patient Data Collection: the 2010 data quality survey.

The reporting of Aboriginal and Torres Strait Islander peoples on the NSW Admitted Patient Data Collection was ascertained using a stratified purposive sample of NSW public hospital patients in 2010. Information was collected by interviewing patients and compared with patient information obtained on admission. The study used the methods used in the national survey by the AIHW in 2007 and the study results were compared to the AIHW survey results. The level of correct reporting was 90.7% (95% CI 84.6-94.2). These results, while indicative, should be interpreted with caution as some people may not have identified themselves as Aboriginal or Torres Strait Islander either on hospital admission or in the survey, and non-random sampling can produce non-representative samples.

Using linkage between hospital and ABS mortality data to enhance reporting of deaths among Aboriginal and Torres Strait Islander peoples.

OBJECTIVE: To investigate the potential of record linkage between the Australian Bureau of Statistics (ABS) mortality data and the NSW Admitted Patient Data Collection (APDC) to improve reporting of deaths among Aboriginal and Torres Strait Islander peoples. METHODS: ABS mortality data for 2002 to 2006 were linked with APDC records for 2001 to 2006. Six algorithms were developed to enumerate deaths. Possible biases by age, sex and geographic remoteness were investigated. RESULTS: Levels of reporting ranged from baseline reporting on the ABS mortality data to the largest enhancement with the 'ever reported as Aboriginal or Torres Strait Islander' algorithm. Enhancement was more likely in females, older people and residents of major cities. CONCLUSIONS: Data linkage substantially improved reporting of Aboriginal and Torres Strait Islander deaths. An algorithm that includes both the number of APDC records and the number of facilities reporting a person as Aboriginal or Torres Strait Islander was considered most promising. IMPLICATIONS: Inclusion of other datasets in the enhancement process is warranted to further improve reporting and address possible bias produced by using APDC records only. Further work should take into account the possibility that a person may be falsely reported as Aboriginal or Torres Strait Islander or not reported in either hospital or death records.

Cross-sectional reporting of previous Cesarean birth was validated using longitudinal linked data.

OBJECTIVE: The aim of this study was to demonstrate the feasibility of using linked health records to assess data quality in population health data. STUDY DESIGN AND SETTING: Reproductive histories of 155,897 women were constructed by longitudinal linkage of the New South Wales (Australia) birth records in 1998-2005, and 127,952 birth and hospital discharge records in 2000-2005 were cross-sectionally linked. History of Cesarean section (CS) derived from the longitudinal linkage ("gold standard") was used to validate the CS history fields (i.e., "Was the last birth by Cesarean section?" and "Total number of previous Cesarean sections?") in birth records and to validate "vaginal birth after previous Cesarean (VBAC)" and "maternal care for uterine scar" in hospital records. RESULTS: The reporting of CS at last birth was reliable with sensitivity, specificity, positive predictive value (PPV), and negative predictive value all >95% as was the number of previous CS (weighted kappa=0.97). For the hospital data, sensitivity and PPV were 46% and 99% for VBAC, 92% and 99% for maternal care of uterine scar, and 85% and 99%, respectively, for any prior CS. CONCLUSION: Assessing data quality by record linkage is feasible and can be done more quickly and cheaply than by any traditional validation study.

Hospital costs of older people in New South Wales in the last year of life.

OBJECTIVE: To estimate hospital inpatient costs by age, time to death and cause of death among older people in the last year of life. DESIGN AND SETTING: Cross-sectional analytical study of deaths and hospitalisations in New South Wales from linked population databases. PARTICIPANTS: 70,384 people aged 65 years and over who died in 2002 and 2003. MAIN OUTCOME MEASURES: Hospital costs in the year before death. RESULTS: Care of people aged 65 years and over in their last year of life accounted for 8.9% of all hospital inpatient costs. Hospital costs fell with age, with people aged 95 years or over incurring less than half the average costs per person of those who died aged 65-74 years ($7028 versus $17,927). Average inpatient costs increased greatly in the 6 months before death, from $646 per person in the sixth month to $5545 in the last month before death. Cardiovascular diseases (43.1% of deaths) were associated with an average of $11,069 in inpatient costs, while cancer (25.0% of deaths) accounted for $16,853. The highest average costs in the last year of life were for people who died of genitourinary system diseases ($18,948), and the highest average costs in the last month of life were for people who died of injuries ($8913). CONCLUSION: Population ageing is likely to result in a shift of the economic burden of end-of-life care from the hospital sector to the long-term care sector, with consequences for the supply, organisation and funding of both sectors.

Characteristics of unmatched maternal and baby records in linked birth records and hospital discharge data.

Linkage of routinely collected health data collections is increasingly being used to investigate maternal and infant morbidity and mortality. Such data have the advantage of being population based and readily available. However, in using such data it is important to understand the data linkage process, the proportions of unmatched records and the characteristics of these records so that potential bias can be recognised. This article describes the differences in characteristics of matched and unmatched mothers' and babies' records generated in the linkage of birth records with hospital discharge data and explores some of the reasons for these differences. The study population included over 250,000 women and their babies discharged from hospital following delivery in New South Wales, Australia between 1 January 2000 and 31 December 2002. Hospital discharge and birth data were linked using probabilistic linkage methods for both mothers and babies. Matching rates were 98.5% and 99.0% for maternal birth and hospital discharge records, respectively, and 98.8% and 99.4% for baby records. Unmatched maternal records had higher proportions of Australian-born women, private hospital births and stillbirths compared with matched records. Unmatched baby records had higher proportions of low-birthweight babies, preterm births and in-hospital deaths than matched records. With the possible exception of stillbirths, these differences are unlikely to cause important bias in studies relying on matched records only. Our results suggest studies using linked data should generally examine and report on the characteristics of unmatched records, and recognise them as a potential source of bias.