HEALTH REGISTERS. How, why and for whom?

Transcription

1 HEALTH REGISTERS How, why and for whom? Prepared by the Royal Australasian College of Surgeons for the Commonwealth Department of Health and Aged Care December 2000

2 TABLE OF CONTENTS EXECUTIVE SUMMARY 3 1 INTRODUCTION Definitions Use of Health Registers Register Types Disease Screening Registers Disease Prevention Registers Acute Health Episode Registers Chronic Disease Registers Health Intervention Registers Vital Event Registers Other Registers 14 2 REGISTERS: HOW THEY WORK Utilisation of Information Positive Outcomes Critical Success Factors Audit, Research and Quality Activities Strengths and Weaknesses Gaps and Opportunities 30 3 RECOMMENDATIONS Data Element Definitions Survey of Existing Registers Data Linkages Funding 37 4 CENTRAL OVERSIGHT BODY Establishment of a Central Oversight Body Remit of Central Oversight Body Membership of Central Oversight Body Data Collection, Auditing and Reporting Scope and Boundaries Confidentiality Funding 40

3 5 SURGICAL OPPORTUNITIES Surgical Register Development Opportunities Priorities Performance Indicators Clinical Effectiveness 42 REFERENCE LIST 43 3

4 EXECUTIVE SUMMARY Health registers have been in existence for many years and have been used for a variety of different purposes from disease screening through to intervention recording. Information held by registers is of importance to health professionals, researchers and patients. Register data specifications must meet national and international standards to allow for meaningful comparisons to be made. To this end, the National Health Information Agreement will ensure that registries have more efficient data collection processes, reduced data duplication, and are established in accordance with government health policy directives. A major factor currently limiting the effective use of health register data is the difficulty associated with establishing linkages between registries. Data linkage is essential in managing and maintaining diverse national data registries of health information that can be pooled together to address specific research questions. Registers play an important role in audit, research and quality activities. Recommendations for extending these roles include: all Australian health registers should adopt the definitions associated with the generic elements outlined in the National Health Data Dictionary; there should be an audit of all existing health registers; data linkage between health registers should be pursued; funding for register development should be the responsibility of the Commonwealth Department of Health and Aged Care; and a Central Oversight Body (COB) should be established to coordinate the development of all Australian health registries. This body would offer support 4

5 Health registers can provide an excellent model for monitoring surgical outcomes, especially in the areas of audit, performance indicators, and clinical guideline development. Thus, surgical registers would enhance clinical effectiveness. In addition, the role of the Australian Safety and Efficacy Register of New Interventional Procedures Surgical (ASERNIP-S) could be further developed to include development, coordination and dissemination of these quality activities. 5

6 INTRODUCTION DEFINITIONS The development of registries or registers can be traced as far back as the Domesday Book prepared in 1089 in England. 1 Despite the longevity of the register concept, a systematic review of the literature by the Michigan Department of Health in 1989 found no clear consensus concerning the operational definition of a register. The only consistent factor that distinguished registers from other databases in the literature was that the data collected in registers related to specific, identifiable persons. On this basis, a register was defined "as a database of identifiable persons containing a clearly defined set of health and demographic data collected for a specific public health purpose". 1 Similarly, the Australian National Health Information Management Group defined a health register as a collection of records containing data about aspects of the health of individual persons. The cases are typically patients or clients of a health service or health program, from which the data is collected USE OF HEALTH REGISTERS Health registers are established for varying reasons. The major reason is to provide accurate population-based information to assist with the development of health policy, health services prioritisation and/or resource allocation. This includes providing (1) epidemiological data (i.e. the incidence or prevalence of specific diseases or conditions), and (2) information on health service delivery and health interventions. The other main reason for establishing a register is to facilitate research on a health 6

7 topic, either through record linkage to other existing data collections or by using the register as a sampling frame for more detailed research and/or sub-group analysis. 2 The information collected in health registers has many uses, including: 3 estimating the magnitude of a health problem or condition; determining risk measures including the incidence or prevalence of disease or health outcomes; examining trends of disease or health events over time; assessing service delivery and identifying groups or sub-populations at highest risk of the disease or health outcome; documenting the types of persons or patients serviced by a health provider; conducting research, particularly survival analysis, evaluating health effects of specific exposures, and investigating etiologic or causation hypotheses; serving as a source of potential donors e.g. bone marrow or organ donor registries; serving as a source of potential participants in clinical trials; and serving as a comprehensive source of information on research projects, trials, interventions or organisations in a particular health or medical area. 1.3 REGISTER TYPES Health registers that currently exist in Australia fall into seven main categories: 2 Disease Screening e.g. State and Territory Breast Screening and Pap Smear Registries; Disease Prevention e.g. Australian Childhood Immunisation Register; 7

8 Acute Health Episode e.g. Cardiac Arrest Register, Neonatal Intensive Care Register; Chronic disease e.g. State and Territory Cancer Registries, National Diabetes Register, Australian Cystic Fibrosis Data Registry; Health Intervention e.g. Coronary Angioplasty Registry; Vital Events e.g. National Death Index; and Genetic disease or events e.g. the West Australian Birth Defects Registry, the NSW and ACT Hereditary Bowel Cancer Registers. 4,5 The Disease Registers Unit within the Health Division of the Australian Institute of Health and Welfare (AIHW) has responsibility for several core databases including The National Cancer Statistics Clearing House, the National Death Index, and the National Mortality Database. It has also recently started development on the National Diabetes Register and the Breast and Cervical Cancer Screening Database. The aim of the Disease Registers Unit is to maintain and link these national health databases in addition to developing standards for data input. The ultimate goals of the Disease Registers Unit are to facilitate effective monitoring and investigation of disease patterns, identify potential risk factors and provide a valuable information resource for epidemiological studies. 6 The following outline is not exhaustive but represents some of the major registries and databases in existence in Australia at this time. Many of the databases have been 8

9 the subject of moves to amalgamate and house the various registries under one government health agency Disease Screening Registers The National Cancer Statistics Clearing House is supported by the State and Territory cancer registries and is a database of all cancers diagnosed between 1982 and Its aim is to enable computation and publication of national statistics on cancer; allow tracking of cancer patients that move interstate; facilitate exchange of information between cancer registries; and promote nationally standardised data sets for the individual registries. The Clearing House produces reports of national incidence and mortality data. In addition, it periodically produces a more detailed analysis of the data that includes information on histology, cancer site location, and geographical incidence profiles. The database is available to clinicians, patients, pharmaceutical companies, and governments once they have successfully passed a strict scientific and ethical review process. 6 The Breast and Cervical Cancer Screening Database is a collection of information gathered from the State and Territory Breast Screening and Pap Smear Registries to examine patterns of cancer screening in Australian women. This database also provides information to the Health Registers and Monitoring Unit at the AIHW, which is responsible for monitoring and reporting on the performance of two structured national screening programs, namely the BreastScreen Australia Program and the National Cervical Screening Program. The Disease Registers Unit is currently developing new data standards that will increase the level of comparability between the 9

10 7, Disease Prevention Registers The Australian Childhood and Immunisation Register (ACIR) was established as part of a national strategy to increase levels of age appropriate childhood immunisation in Australia, and thereby reduce the incidence of preventable childhood diseases. The Health Insurance Commission collects data on the immunisation status of all children under seven years of age living in Australia. The bulk of the data is retrieved from Medicare enrolment information. However, a child may also be admitted to the database from immunisation details that are supplied by an immunisation provider when the child is not enrolled with Medicare. The ACIR also enables the government to link the immunisation status of a child to the eligibility criteria for the payment of the Commonwealth Childcare Rebate, Childcare Assistance and Maternity Immunisation Allowance. Registered immunisation providers and approved Internet clients can obtain access to the ACIR. They can request information, record immunisation services, view the immunisation history of an individual child, and monitor claims lodged under their provider number. In addition, it serves as an optional reminder/recall service that informs parents when the next immunisation is due; provides information for determining the payment of linked Commonwealth rebates; provides feedback and incentive payments to encourage registered general practitioners to provide appropriate childhood immunisation services; and provides an effective management tool for government monitoring of immunisation coverage and delivery. 9 10

11 1.3.3 Acute Health Episode Registers The Australian and New Zealand Neonatal Network (ANZNN) is a voluntary collaboration of all level III neonatal intensive care units (NICUs) in the region. This prospective data collection commenced for babies born from the 1st January 1994 and provides an ongoing audit of the care and outcomes of infants admitted to NICUs. Data are collected in hospitals either via ANZNN forms or by incorporation of the ANZNN data set into the local neonatal intensive care unit audit. Data include items such as maternal age, pregnancy history, and ethnicity as well as information about the infants including gestational age, birth weight, sex, etc. 10 The National Notifiable Diseases Surveillance System (NNDSS) was established in 1990 under the auspices of the Communicable Diseases Network of Australia and New Zealand. It is responsible for coordinating the national surveillance of more than 40 communicable diseases that are endorsed by the National Health and Medical Research Council (NHMRC). Notifications made to the State and Territory health authorities are computerised, deidentified and supplied to the Network secretariat for collation, analysis and publication in the Communicable Diseases Intelligence Chronic Disease Registers The National Diabetes Register provides a record of Australian residents diagnosed with diabetes after 1 January 1999 who are using insulin. Presently, the prevalence of diabetes can only be derived or extrapolated from localised studies. Therefore, the goal of the National Diabetes Register is to fill a void in 11

12 7 The Australian Cystic Fibrosis Data Registry is maintained by Cystic Fibrosis Australia Inc. and is currently being adapted to the specifications of the Victorian Cystic Fibrosis Register. The Data Registry currently consists of two main modules, the National Register and the Centre Database(s). The former consists of information derived from input sent by the State and Territory Centres responsible for data collection on cystic fibrosis. Each Centre Database is extracted from the National Register and is a generic database intended to serve the needs of all Centres. The Centre Database aims to provide a means for individual Centres to maintain their own client data and to enable the Centres to collect and store the information required by the National Register. Additionally, this allows for efficient updating of the National Register data set Health Intervention Registers The Australian Safety and Efficacy Register of New Interventional Procedures Surgical (ASERNIP-S) is administered by the Royal Australasian College of Surgeons. The register contains safety and efficacy information on selected new surgical procedures. It contains reports on, for 12

13 Vital Event Registers The National Death Index is a database that contains records of all deaths occurring in Australia since It is maintained in cooperation with the State and Territory Registrars of Births, Deaths and Marriages. The database has data variables such as name, date of birth, age, sex, etc. This data can be cross-referenced with the National Mortality Database to provide information on underlying cause of death. The National Death Index is designed for use in epidemiological studies and access to the database is only permitted once a researcher has formally applied to the AIHW, and their study has met certain eligibility criteria. Current users of the National Death Index include government departments, the Red Cross, universities, industry and clinicians. The data forms a useful adjunct in directing policy and resource allocation decisions, and in identifying disease patterns and health outcomes

14 1.3.7 Other Registers The following databases cannot be easily categorised because they cover a range of different health issues within their focus, such as acute health episode, health intervention, and vital events. The National Cardiovascular Monitoring System comprises the National Centre for Monitoring Cardiovascular Disease, an Advisory Committee and a number of collaborative projects with other centres that monitor progress towards national goals and targets for cardiovascular disease, as well as trends and inequalities in mortality, morbidity and risk factors for the disease. The National Cardiovascular Disease Database (NCDD) provides a conduit for easy access to the data currently held by the National Centre for Monitoring Cardiovascular Disease. It contains information on deaths from cardiovascular diseases, prevalence of traditional risk factors, and data on cardiovascular procedures and operations conducted in Australia. The database is accessible to all people including epidemiologists, policy makers, researchers and patients can access the database. 15 The Perinatal Data Collection contains national data on all births and perinatal deaths in Australia since The database receives notifications from State and Territory perinatal data collections. Data for the Congenital Malformation Monitoring System is also derived from this source. The notification forms for the Perinatal Data Collection feature a minimum data set and are usually completed by midwives or medical practitioners. These forms are completed for all babies of 20 weeks gestational age or a birth weight of 400 g. The main purpose of the database is to describe the demographic, medical and pregnancy characteristics of mothers and the outcomes of their 14

15 16 The Assisted Conception Database contains the results of infertility treatment by assisted conception in Australia and New Zealand since It includes data on In-Vitro Fertilisation, Gamete Intrafallopian Transfer and IntraCytoplasmic Sperm Injection. Each infertility centre records data on a standard form, which is submitted to the database managers. Data collected include maternal and paternal age, cause of infertility, details of the infertility treatment, pregnancy complications, plurality and outcome, as well as infant details

16 REGISTERS: HOW THEY WORK UTILISATION OF INFORMATION A survey of representatives of the International Association of Cancer Registries indicated that most registries provide descriptions of cancer patterns, patient care, and outcomes. They monitor these factors with respect to cancer control or prevention initiatives, and provide a research database that is often utilised by others. Some registries, however, have an expanded role that includes direct participation in epidemiologic research as well as the implementation of control programs, particularly screening. 18 Health registers are operated by a range of different agencies, including: federal or state governments, universities, groups of hospitals, non-profit organisations and private groups. 3 The diverse roles and activities of Australian health registers are evident in the three examples below: The NSW and ACT Hereditary Bowel Cancer Registers identify people with, or at high genetic risk of, colorectal cancer and assist with identifying, tracing and alerting family members that are at risk. The Registers also operate a screening reminder service for high risk individuals and their clinicians, and provide information regarding the use and availability of genetic testing, counselling and support groups. 5 The role of the NSW Pap Test Register is to document women's Pap tests and cervical histology results. The Register assists clinicians to follow-up women with abnormalities who have not had a subsequent Pap test or cervical biopsy within 16

17 19 The Australian Childhood Immunisation Register has recorded the details of children's immunisations in Australia from January 1, All children (<6 years) enrolled with Medicare are included on the Register. The Register's role is to collect useful information on immunisation coverage in Australian children to inform health policy, for example the recent pre-school measles mass immunisation program. The Register also provides a reminder service for when vaccinations are due or overdue. 20 Access to information in registers is important to health professionals, researchers, and patients. Systems should be developed in accordance with basic informatics principles including adherence to standards, usability considerations, and iterative testing and evaluation. 21 Linkage is also essential in managing and maintaining diverse national data registries of health information. To ensure that the information collected is of value, it is important to be able to pool a range of health related databases to address specific research questions. When establishing data collection systems it is also important to ensure that the data specifications match national and international standards. This will allow meaningful data comparisons to be made and amalgamated, when required, between different population databases. 22 Consequently, the National Health Information Agreement, as implemented by the National Health Information Management Group, was developed in Australia to 17

18 ensure that the collection, compilation, and interpretation of nationally relevant health information is both appropriate and efficient. To achieve this, it was considered essential to have national consensus on definitions, standards and rules of collection of information, and on guidelines for the access, interpretation and publication of national health information. Such measures will make data collection more efficient, reduce duplication of effort by standardising core data sets and data definitions, and also ensure that the data collected is specific to purpose and is in accordance with government requirements with respect to health policy directives. 22 To this end, the Australian Institute of Health and Welfare has established one of the few metadata registries in the world. This National Health Information Knowledgebase is currently based on elements in the Australian National Health Data Dictionary and was created on behalf of the National Health Information Management Group. The registry is an electronic repository and query tool for health metadata and provides information about the use of particular data elements, such as their definition and permitted values, as well as information on related data types, other data collections that are available, and who is responsible for them. The Knowledgebase has been constructed according to ISO/IEC principles. Thus, the Knowledgebase can be used to find out what data collections are available on a particular health related topic, and any related official national agreements, definitions, standards and work programs. It is intended that future incarnations of the Knowledgebase will provide access to actual data through its searching and querying tools. National Minimum Data Sets and prescribed indicators of performance for registries are also included in the Knowledgebase, which will ensure consistency with national and international standards for data collection and database assessment. Thus 18

19 the Knowledgebase provides direct integrated access to the major elements of health information design in Australia which comprise:23 The National Health Information Model; The National Health Data Dictionary; The National Health Information Work Program; and The National Health Information Agreement. 2.2 Positive Outcomes One of the major functions of health registers is to monitor the impact of populationbased health interventions or changes in health policy. However, the effect of these factors on population health is often not seen for decades. This is particularly the case when the diseases are chronic such as cancer or cardiovascular disease - for example, the inability to determine the impact of smoking cessation programs on lung cancer incidence rates and myocardial infarction mortality rates. Currently in Australia, Cancer Registries have documented the impact of mammographic and cervical screening programs as initial increases in incidence rates (due to increased casefinding) followed by indications of reductions in mortality (due to earlier diagnosis and treatment) Information from the central cancer registry in Illinois was used to identify target populations for breast and cervical cancer screening programs. The success of breast and cervical cancer screening programs were also evaluated using specific surveillance indicators.27 The more immediate impact of health policy or health interventions can be seen using some registers. For example, determining the impact of immunisation schedules and coverage (as recorded by the Australian Childhood Immunisation Register) on the 19

20 incidence of communicable diseases (as recorded by the National Communicable Diseases Surveillance Network). Two studies in Finland have investigated the feasibility of using routinely collected health-register data and data linkages to follow-up children's health, and compared this with data obtained through cohort studies. Data collection using health registers was found to be a feasible method, which saved both time and financial resources when compared with cohort studies.28,29in Slovakia, highly detailed information on cancer incidence collected by the National Cancer Registry of Slovakia between 1968 and 1998 contributed largely to research on the relations or association between the environment and the occurrence of cancer. 30 The Metropolitan Atlanta Congenital Defects Registry, operated by the American Centers for Disease Control, monitors the occurrence of congenital defects in a defined population for changes in trends or patterns that may suggest avoidable risk factors. This register has provided data for a large number of seminal epidemiologic studies, including determination of the protective effect of peri-conceptual consumption of folic acid on the risk of infant neural tube defects. 3 Health interventions registers often provide very accurate information of the treatments that patients receive in hospital. However, they may under-report a substantial amount of information about treatments received on an outpatient basis. Therefore registers that ascertain data from multiple sources are more likely to be more accurate

21 2.3 CRITICAL SUCCESS FACTORS Critical factors for the development of a successful health register were determined through the conduct of a systematic literature review and survey by the Michigan Department of Public Health in In this review and other research studies, nine factors were identified as critical: An implementation plan, including a pilot phase. 1 This plan includes: - a clear statement of the purpose, use and scope of the register, 2 together with development of a timeline; - consultation with relevant health care professionals; - staff identification and training; - an estimate of the size of the register at establishment and projected to the next 3-5 years; - operation of the register at a specified geographic level. This should be aligned with known populations at risk of the disease or health condition; 2 - coverage of the register directly proportional to the prevalence of the disease or health condition. For example, rare health events require a wider reporting base to ensure the statistical validity of the results 2 derived from the register; - identification of data sources for case ascertainment; - development of case-finding processes; - development of data collection instruments as well as data collection and entry procedures; - selection, development and implementation of hardware, software and data processing methods; 21

22 - reporting and dissemination processes so that relevant data is available and accessible in a timely manner; and - development of a quality control system. Adequate documentation of: the procedures and personnel used to administer the register; inclusion and exclusion criteria; definitions of data sources, data elements, data collection and data entry procedures; protocols for matching or linking to other data sources; data processing and analysis procedures that are routinely conducted; confidentiality guidelines; and research access procedures. 1 Quality control procedures. Three factors influence the quality of health register data: (1) completeness - that is, the proportion of cases in the target population that appear in the register, (2) validity - the proportion of cases in the register that are attributed to have a particular characteristic and actually do have that characteristic and, (3) timeliness. Quality assurance procedures should be undertaken at regular intervals to maintain a high quality of data. This can range from simple ongoing data entry checks to a formal audit comparison of register data and source data. 1,2,32-37 Case definition and ascertainment procedures. Inclusion and exclusion criteria should be clearly delineated so that the persons involved in case ascertainment will make consistent decisions as to the relevance of a potential case to the register. Case ascertainment procedures should be piloted in a new health register so that inclusion and exclusion criteria can be adequately and comprehensively defined. Three factors must be addressed when a new health register is established: (1) will case ascertainment be active, passive or a 22

23 1 Determination and definition of data elements. Data elements should be kept to a minimum with only the most useful or essential, in terms of the purpose of the register, being collected. Ideally data elements should not be changed or modified over time as this limits the usefulness of the register. Basic demographic information should be collected to facilitate the generalisation of results and the matching and merging of data with other registers. A minimum data set should be established and data definitions should follow a standardised format that is consistent with other health registers. 1,2 Data collection and processing procedures. The choice of data collection and entry methods will impact on the quality of data obtained. Clear and unambiguous data collection forms or measurement tools are required and should be designed, piloted, carefully reviewed and modified during the development phase of the health register. Further, data verification and editing procedures should be in place to reduce error associated with case definition and data entry. 1,2 Sustainability of the health register. The register should be sustainable and adequately resourced over the long term. The prime motivation for establishing registers is to provide ongoing data collection. Therefore, a stable funding base, adequate staffing, and appropriate information technology resources are critical. The financial viability, suitability and appropriateness of the data collection should also be periodically reviewed. 23

24 Data access policy. Information held in a health register must be able to be accessed in an efficient, flexible and timely manner. However, a data access policy must be developed by the health register 1 so that confidentiality or privacy agreements associated with the data are not breached in the pursuit of information for research or other purposes. Framework for dissemination of health register data. The value of a health register is dependent upon the extent to which the information it collects is fully analysed and disseminated to relevant agencies. New registers should detail a dissemination framework and a mechanism for responding to requests for data. Further, the routine reporting required to fulfill the purpose of the health register should be defined and delineated AUDIT, RESEARCH AND QUALITY ACTIVITIES There is a significant body of evidence that illustrates the relationship between registers and audit activities. Some registers used for this purpose are national and collect information on surgical interventions within a clinical speciality while others are institution and intervention/disease specific. The following are some pertinent examples: The United Kingdom Thoracic Surgical Register and the Cardiac Surgical Register were established by The Society of Cardiothoracic Surgeons of Great Britain & Ireland to provide anonymous activity and mortality data from all National Health Service (NHS) cardiothoracic/thoracic units. This data includes surgeon specific information, which allows the Society to investigate sub-standard performance by individual surgeons. The data is statistically analysed, summarised and aggregated into an annual audit. This report is 24

25 38 Sweden has established National health care Quality registers that contain individual based data on diagnoses, interventions and outcomes. Aggregated data is made available on a departmental and national level. The purpose of the registers is to facilitate continuous quality improvement in the delivery of health care services. An example is the Hip Fracture register that has approximately 75% coverage throughout the country. Hip fractures were chosen because they represent 25% of all orthopaedic inpatient activity and are thus considered a high volume group. The register has been successful in providing a model for quality control from an outcome perspective. 39 The Department of Surgery at the Wellington School of Medicine conducted a retrospective data collection on patients treated for rectal cancer at Wellington Hospital between 1986 and The results indicate, using a comparison with accepted best practice, a need for improvement in restorative resection, pelvic recurrence and anastomotic leak. 40 Data collection on all vascular surgical activity from nine metropolitan hospitals in Melbourne was commenced in The information is used for clinical auditing and, if necessary, the head of the Auditing Monitoring Committee will interview a surgeon whose practice performance is outside two standard deviations of the mean. 41 Data from cancer and other health registers are widely utilised for research purposes. Since 1976, cancer registers in the United States have voluntarily contributed data for annual patient care studies undertaken under the auspices of the National Cancer Data Committee of the Commission on Cancer. These 25

26 42 The need for medical staff to demonstrate systems of self-assessment and peer review is becoming increasingly necessary, particularly in the area of patient safety and quality. It has been recommended that the use of information and information technology should be maximised in the pursuit of quality and safe standards of care. The use of health registers is instrumental in achieving this STRENGTHS AND WEAKNESSES The National Health Priority Areas initiative, a collaboration between the Commonwealth, State and Territory Governments, has identified five health areas that impose a high social and financial cost on Australians and, with the implementation of targeted interventions, offers the opportunity for significant health gain. The five priority areas are - cardiovascular health, cancer control, injury prevention and control, mental health and diabetes mellitus Information from various national health registers has been used to identify these priority areas - particularly the National Mortality database and the National Cancer Statistics Clearinghouse. 46 An investigation of these health priority areas has uncovered deficiencies and strengths in the national data collection systems. With respect to Injury Prevention and Control, Mental Health and Cardiovascular Disease, certain deficiencies have been identified in the National Mortality Database including incomplete information 26

27 (particularly indigenous information), and information of variable quality and reliability, i.e. affected by under-reporting and misclassification. 46 Some of these problems have been ameliorated by supplementing data from registers derived from national surveys. Efforts to promote standardisation of data elements and data linkage are underway. The coverage of cancer registers appears to be wide and includes all the critical data elements. However, one major weakness of cancer registers is the time taken to process information is often several years. Furthermore, long-term indicators such as 5-year survival rates take a long time to collate. Prevalence and incidence rates on diabetes mellitus could not be calculated due to a lack of detailed data. As a consequence, a National Diabetes Register has been set up to monitor national diabetes indicators and provide information for epidemiological and clinical studies. 45 Data linkage between registers and other data collections, such as hospital audits or surveys, is essential for maximising the benefit of information collected. It facilitates sub-group analyses and the pursuit of more detailed research hypotheses, along with reducing the need for conducting new studies or establishing new data collections. However data linkage is often problematic because of ethical and privacy requirements concerning the use of identified data. 2 However, it is possible to link same-person records across multiple databases using deterministic or probabilistic decision criteria or computer algorithms, despite the absence of personal names. This can be achieved with acceptable sensitivity and positive predictivity if the data is uniform and of a sufficiently high quality The Western Australian Health Services Research Linked Database project is an excellent example of the benefits 27

28 that can be accrued using this technique. 50 Potential problems with data linkage studies relate primarily to the variation in the content or definition of data elements, and in the data quality of different registers, as well as data protection issues. 28 Audits generally occur at selected sites on a sample of the population at risk, have a local and short-term focus and capture more detailed information. Audits are often reliant on information that is routinely collected by hospitals. The limitations of hospital data for research purposes include: (1) the selectivity of hospital admissions according to personal characteristics, severity of disease, associated conditions and hospital admission policies, (2) hospital records are not prospectively designed for research purposes and, therefore, may be incomplete, missing or unreadable and may vary in terms of the diagnostic quality and definitions used, which makes it difficult to compare or combine across hospitals, and (3) prevalence or incidence rates often cannot be calculated as the population(s) at risk (the denominator) is usually not defined. 51 Thus, the information obtained from hospital audits, unlike health registers, is not necessarily generalisable to other sites or circumstances, and risk measures cannot be calculated. The decision to utilise a health register or audit depends entirely on the purpose of the data collection, whether short-term or long-term results are required, and the financial resources available. In Britain there has been a directive issued from the General Medical Council that patient data should not be submitted to a cancer registry without patient permission. While there is a possibility that patient data can be abused once logged in the registry, there are tough protection procedures in place to prevent this occurring. However, it is important to consider that even a small percentage of refusals would bias the data and 28

29 make it virtually worthless. Currently the data is transferred automatically from oncology and pathology records to the cancer registries. The data cannot be deidentified because it must be linked to the individual to enable statisticians to follow up survival rates and avoid counting cases twice, as well as enabling linkage with occupational information for risk factors, etc. Therefore, it is important to consider the data input setup of registries to ensure that they are fulfilling their function without admitting bias that would make conclusions based on such data essentially worthless. 52 There is little information available on how registers are compiled, the factors that predict inclusion, or the relationship between registration and the level of care received by the patient. Wide variations in levels of registration have been shown to exist between practices, particularly in voluntary systems. 53 However, it is important to be aware of and assess the completeness and accuracy of surveillance data. One method of doing this would be to compare locally ascertained data with recorded registry sources in selected diagnoses. Such a comparison between the Northern Region Cancer Registry and the district of South West Durham checked patients who had been diagnosed with lung, skin, and malignant cervical cancer during the calendar period from 1989 to Of 544 cases of cancer identified from all sources, 448 (95.8%) were registered, while of these 448, 11.8% showed disagreements between local and registry data involving classification of site and timing of registration. 54 Data sources can be an important issue for the veracity of registration data when designing or assessing a health registry. For example, inaccuracies in the assessment of pathology data have been shown to cause problems when used as criteria for cancer 29

30 registration in a national registry. In Scotland during 1992, 71 apparently missed cancer cases were considered ineligible for registration as independent primary malignant neoplasms, on the basis of their pathology assessment, out of a total of 218 missed cases of cancer. Thus, it was concluded that unverified computerised pathology data should not be used as the sole independent method for registration, and should only be used with critical examination as a supplement to other methods of determining registration GAPS AND OPPORTUNITIES Health registers are best suited for problem detection and identification, as well as for tracing new, rare or rapidly developing health problems. Registers are also essential for monitoring the delivery of public health services, and documenting events such as births and deaths. Other reasons to establish a register include: Information is not available or is under-reported elsewhere; Long-term/follow-up data is required; Population-based information is required to inform and develop health policy or interventions that will have the greatest effect on the largest section of the population with the health problem, particularly those at greatest risk or need; and When policy-makers need to see the effect of health policy and broad-based health interventions on population health to provide a foundation for longrange planning of resource allocation, health service delivery and health funding. 30

31 On the basis of a systematic literature review and survey of registers, the Michigan Department of Public Health commissioned the development of a set of criteria for determining when new health registers should be created. Six critical issues were identified to determine whether a register is warranted: 1 An evaluation of the stated purpose of the register. Is there a clear statement of goal or rationale for how the information collected will assist with the accomplishment of that goal? The evaluation of a proposal for a register should focus on whether the data collected can solve or facilitate solving the health problem or issue in question. Is the health problem or issue in a priority area? A review of the function, duration and scope of the register. A health register should only be established when there is a need to collect information over the long term. Registries are usually an expensive method of obtaining health information due to the fixed costs associated with developing an administrative structure and staff. Short-term projects are therefore better suited to a focused research study that is able to collect the appropriate health information, identify health service requirements and evaluate health service delivery systems at a much lower cost. Sampling is not, however, appropriate for all situations particularly for the maintenance of vital statistics, health service tracking and maintenance of referral information. Consideration of existing alternative data sources. Prior to the development of a health register there should be a careful review of existing sources of information to determine whether this information is already available. 31

32 Assessment of the practical feasibility of the register. A register should only be established if there is a reasonable expectation that it can achieve its stated purpose or goals. This relates to issues of: - reliable and sustainable methods of case ascertainment; - the use of restrictive inclusion criteria and its impact on the generalisability of results; - maintaining adequate levels of reporting compliance; - confidentiality and privacy issues; - the timely collection and processing of data; - the availability of reliable and valid instruments to measure data elements; and - the costs associated with all these issues. Likelihood of sufficient start-up and long-term funding. The major problem in the establishment and maintenance of a health register is cost and a long-term commitment to funding. In the Michigan survey, managers of health registers consistently complained about the lack of funding and resources to fully utilise register data. Data collection and processing exhausted budgets, leaving few resources for quality control, data analysis, interpretation and dissemination of findings. 1,2 Research is often only conducted on register data by outside agencies or personnel in an ad hoc manner. It often takes several years before the full benefits or impact of health register data can be realised, particularly for large scale registers on chronic health problems. An evaluation of the cost-effectiveness of the health register. The establishment of a health register should be evaluated in terms of the likely costs necessary to develop and maintain it. Some factors that should be 32

33 1,51 - Total number of cases, as well as risk measurements such as incidence and prevalence of the disease, condition or health event; - Indices of severity such as case fatality ratio and quality-adjusted life years; - Overall and cause-specific mortality rates; - Indices of lost productivity such as bed-disability days; - Index of premature mortality such as years of potential life lost; - Medical costs; and - Preventability and expected benefit of health register information in reducing morbidity and mortality. Guidelines and standards for the establishment of health registers, once approved and funded, have been developed by both the Michigan Department of Public Health, and the Australian National Health Information Management Group. 1,2 33

34 RECOMMENDATIONS DATA ELEMENT DEFINITIONS It is clear that well-documented, precise and unambiguous definitions of the data elements to be included in a health register are crucial to the accuracy and success of the register. These definitions must be readily accessible and routinely utilised by those persons providing information to the register. All Australian health registers should adopt the definitions associated with the generic data elements outlined in the National Health Data Dictionary (version 9, Australian Institute of Health and Welfare). This would ensure uniformity across health registers, facilitate the combination or comparison of data, and reduce unnecessary duplication of data. 2 This appears to be occurring for newly established registers but cannot easily be undertaken for existing registers due to the potential loss of long-term data or information on trends (i.e. re-classification of data elements would mean comparisons or integration with historical data could not occur). It is also clear that a register should only be established when this method of data collection is the most resource efficient method of collating information on the specified health issue. The register should not duplicate other data collections SURVEY OF EXISTING REGISTERS A national survey should be undertaken to identify all existing Australian registers. As part of the survey the oversight body should carry out the following analysis for each existing register: 34

35 define the purpose, use and scope; describe the measurement tools utilised; identify the data elements and data definitions utilised; and determine whether definitions are based on the NHDD or could be adapted to the NHDD. A survey similar to the one proposed was undertaken by the Michigan Department of Public Health, although it was restricted to registers funded by that Department. It included face-to-face interviews with directors of health registers, researchers who used registry data and key public health officials. 1 As recommended by the 1999 National Public Health Information Development Plan, the appropriateness, utility, feasibility and cost-effectiveness of implementing record linkage between health registers and other existing data collections, such as hospital or surgical audits and national surveys, should be systematically investigated. 56 This investigation should also examine the feasibility of a unique patient identifier for the Australian population. This would increase the potential and ease of establishing useful data linkages thus reducing some of the current barriers to effective use of register data. The example of the Swedish Fractured Hip Audit is an ideal example of the possible benefit DATA LINKAGES The linking of data has other documented advantages that contribute to the effectiveness of registers. 57 Administrative data sets do not record the outcome of health care interventions, which makes it extremely difficult to evaluate the efficiency and effectiveness of health interventions. In addition, measures of utilisation that are based on aggregated data are confounded by other factors (for example, changes in 35

36 the incidence of the condition, the availability of new technology or behavioural changes). Only when a comprehensive health record management system becomes available can the value of administrative data sets be fully exploited. The National Public Health Information Development Plan 1999 also acknowledged the need to improve the information available on public health in Australia, particularly with respect to record linkages and the need to investigate all aspects of this activity. 58 Record linkage contributed to the successful implementation of the Quality of Surgical Care Project 1998 in Western Australia. The linked databases have access to 8 million health records from a variety of sources including hospital morbidity data, births, deaths, cancer registrations, 59 and provide information on: epidemiology of selected diseases requiring surgical care; trends in utilisation; comparisons of results with national and international standards of best practice; the outcomes of new and established surgical procedures; and the establishment of benchmark standards as well as recommending and evaluating the implementation of appropriate changes in surgical practice. To enable the capture of all the life long patient data through linked registers, a unique identifier would need to be used on all health care access events for each patient to ensure capture of all related information. The benefits of a unique patient identifier include allowing timely confidential access to health information by different health service providers, providing the opportunity for data to be aggregated, supporting 36