Using administrative health data for palliative and end of life care research in Ireland: potential and challenges

Setting

Ireland has a mixed public private health care system where publically funded health care is managed by the HSE and funded through the tax system. All residents are entitled to use the public health system. There are three private health insurance providers in Ireland and in 2018, 45% of the population had private health insurance²⁸. Privately insured patients in Ireland may be treated in public or private hospitals. The HSE National Clinical Programme for Palliative Care oversees the management and organisation of palliative care services in Ireland²⁹. Specialist palliative care is delivered by the HSE along with a number of voluntary service providers. Specialist palliative care teams provide care in acute hospitals, community settings and specialist inpatient units across the country.

Identifying potential datasets for PEoLC research

The Health Information and Quality Authority (HIQA) is an independent body that evaluates the quality of the information available on health and social care and makes recommendations to improve quality, minimise inconsistencies and fill gaps where data are not available^30,31. Quality is defined as data that are complete, valid, accurate, reliable, relevant, legible and available in a timely manner³². HIQA advocates eight guiding principles for organisations collecting data that include formalised governance arrangements, facilitating appropriate access to the data to optimise its benefits, continuous monitoring/improvements of data quality and effective information governance procedures. Standards for data quality include the use of data dictionaries, classification systems and clinical terminologies³⁰. A data dictionary is a descriptive list of names, definitions and attributes of data elements to be collected in an information system or database and aids in the standardisation of data definitions³³. Related to the concept of data dictionaries, data models describe how the data are organised and stored within an information system or database. This affects how relevant data from different systems data can be identified, extracted and compared. Data dictionaries and data models hold data about the data, also called metadata.

HIQA produces a catalogue of national health and social care data collections using a standardised template to describe existing data collections¹⁶. The 2017 catalogue was cross-referenced with a recognised list of diseases with associated palliative care needs (Table 1), based on a methodology by Murtagh et al.³⁴. Given Ireland’s aging population and identified future palliative care needs for cancer, neurodegenerative disease and dementia²⁰ particular focus was given to disease registry collections. Criteria used to assess the datasets included an examination of population coverage, the data collected, the availability of data dictionaries and data models and information on how the data can be accessed.

Death certificate data

Every death in Ireland is legally required to be notified to the state within three months of death, so death certificate data is population based at the national level. Death is a unique event so a person should only have one death certificate record. The Department of Social Protection, Central Statistics Office (CSO) and General Register Office collect and record date of death, address of residence of deceased, place of death, cause of death, occupation of deceased, age of deceased, sex of deceased, and marital status of deceased. Cause of death for all deaths registered from 2007 onwards are coded using ICD10 codes³⁸. Place of death is recorded as an address and is not classified e.g. into home, hospital, hospice, or long-term care facility³⁹. Information on how to access the data are available at https://www.cso.ie/en/aboutus/lgdp/csodatapolicies/dataforresearchers/.

Hospital Episode data

The Hospital In-Patient Enquiry system (HIPE) collects demographic, clinical and administrative data on discharges from, and deaths in, all acute public hospitals nationally. Details of each episode of care is recorded as a single record so that over time an individual can have multiple records within and across HIPE hospitals. HIPE is the only source of morbidity data available nationally for acute hospital services⁴¹. In 2016, 53 hospitals were contributing to HIPE. Data are not available in HIPE for emergency department attendances unless the patient is admitted to hospital. Data are also not available for 22 private hospitals. Clinical coders review the records of each patient and extract the relevant clinical data, and translate it into codes using the ICD-10-AM/ACHI/ACS 8th edition⁴². As well as a source of clinical information for many chronic diseases with associated palliative care needs (e.g. dementia, neurodegenerative diseases and cancer), diagnostic codes include ‘Z51.5 - Palliative care’ recorded when a patient has been seen by the palliative care team⁴². The guidance for recording palliative care in HIPE changed with the introduction of the 10^th edition ICD-10-AM/ACHI/ACS from January 2020. Palliative care should be recorded only where there is documented evidence that the patient has been provided with palliative care⁴³. Notwithstanding evidence of variation in how the code is used across hospitals, currently HIPE is the only available population level administrative dataset where a record of a patient being seen by a palliative care specialist can be identified⁴⁴. Additional relevant information for PEoLC research include admission type (elective/emergency) and patient destination on discharge with categories that include home, nursing home, transfer to another hospital, transfer to hospice and/or died. Information on accessing data and a data dictionary for HIPE data are available from http://www.hpo.ie/

Opportunities for data validation. The Minimum Data Set (MDS) is a national survey of demographic and patient activity data for specialist palliative care services in Ireland⁴⁵. Monthly aggregate data from specialist palliative care inpatient units, community (homecare) services, day care services and acute hospitals are returned to a national office. The Specialist Palliative Care MDS does not contain patient level data and is not listed in the HIQA catalogue. A summary analysis of MDS⁴⁵ for the period 2012 to June 2016 reported several metrics including

Data from all acute hospitals was incomplete at the time of analysis and excluded from the report, however MDS aggregated data of specialist palliative care activity in acute hospitals in 2016 has been used to validate HIPE coding of palliative care³⁷.

Primary care reimbursement service data

The HSE Primary Care Reimbursement Service (PCRS) is responsible for making payments to healthcare professionals including general practitioners (GPs), dentists and pharmacists, for the free or reduced costs services provided to the public under the General Medical Scheme and/or other schemes⁴⁶. Access to the schemes is means-tested on a rolling basis and/or determined by specified long-term disease. Qualifying individuals are given a medical card with a unique medical card number (MCN). Eligibility for a medical card can change with changing circumstances so that over time, one person can have had a number of medical cards. In 2018, 43.4% of the population (over 2 million people) were eligible for a medical or GP visit card⁴⁷. The PCRS dataset is one of the few national datasets that collects data in primary and community care settings. All expenditures around pharmaceuticals (drugs/medicine costs) are recorded against an MCN so that the data are transaction based. A data model is not currently available for PCRS so it is not clear how an individual is linked with medical card(s) within the PCRS database or whether an individual or a medical card is recorded more than once; neither is a data dictionary available. Further information on PCRS can be found at https://www.hse.ie/eng/staff/pcrs/.

The Irish Longitudinal Study on Ageing

The first wave of data collection for the Irish Longitudinal Study on Ageing (TILDA) surveyed a nationally representative sample of over 8500 people, aged 50 years and over, beginning in October 2009 with a further four waves of data collection in 2012, 2014, 2016 and 2018⁴⁸. Each individual within the TILDA dataset has a unique identifier and a wide range of data on the health, economic and social aspects of participants' lives are collected through personal interviews, self-completion questionnaires and health assessment measures⁴⁹. TILDA is unique in Ireland in that it contains detailed longitudinal data on education, income and occupation in this age group which is not readily available elsewhere.

TILDA data has been linked at the person level with death certificate data, matching on the basis of name, address and month and year of birth. Matching was performed for all individuals who died between Wave 1 (2009/2011) and March 2018, and of a total of 863 confirmed deaths among the TILDA sample, matching death records were obtained for 779 decedents (90.3%)⁵⁰. By law every death is registered and this is reflected in the high match rate achieved. Matching allows for detailed research around end of life care given the depth and breadth of information collected prospectively before death by TILDA. This work demonstrates person level data linkage with the other datasets described here is feasible. Anonymised data and documentation on TILDA are available for download from the Irish Social Science Data Archive⁵¹.

Cancer registry data

The NCRI collects data nationally for incident tumours recorded at the level of the patient, that is each patient should be recorded once only in the registry database. This relational data model simplifies data linkage where data are matched at the person level using demographic details. Over time a patient may have additional tumour and management data attached to their patient record. Information collected includes patient demographics (age and sex), type of cancer (site and staging), treatments and selected procedures, date and cause of death (from linked death certificate data). Clinical information is coded using international guidelines including international classification of diseases (ICD) codes^40,52. The focus of data collection is on the first year post-diagnosis with limited data collection thereafter. Completeness of case ascertainment is estimated to be 98.7%⁵³. Information on accessing data from the NCRI is available from the website https://www.ncri.ie/.

Other disease registries

Population based disease registries are a good starting point for PEoLC research because selection bias is reduced since the whole population with the disease are identified. The Irish Motor Neurone Disease (MND) register was established in 1995 and collects data on all known patients diagnosed with MND each year and currently it holds information on over 2,200 patients⁵⁴. Individual level demographic data are recorded so that linkage to HIPE, death certificate data and PCRS data should be feasible. Date of disease onset is also captured so that studies on the patient’s PEoLC needs throughout the disease trajectory are possible.

The Cystic Fibrosis registry requires patient consent for data collection. In 2017 it was estimated the registry coverage of the cystic fibrosis population was just over 90%⁵⁵. The characteristics of those patients not captured are unknown, so studies using the cystic fibrosis registry may be subject to selection bias. Detailed demographic information that includes name and address, date of birth and ethnicity are recorded. Additional information includes information on diagnostic tests, genotype, symptoms and method of diagnosis, age at diagnosis, number of hospitalisations between annual assessments, complications and other clinical data and social data. Data linkage to death certificate data, HIPE data and PCRS data should be feasible. The cystic fibrosis registry is unique among the datasets described here in that it collects ethnicity data⁵⁶.

The Irish Thoracic Society Interstitial Lung Disease (ITS-ILD) Registry began collecting data in 2016. Patient written consent is required for data collection. The first annual report (2018) on 154 patients found 46% of patients with idiopathic pulmonary fibrosis were referred through primary care, 12% of patients were referred for lung transplant assessment and 13% were referred to palliative care⁵⁷. A key finding is that most patients with idiopathic pulmonary fibrosis will need a lung transplant or palliative care. While only preliminary information on disease stage and no information on survival times were provided, the report demonstrates the value of disease registries in providing detailed information necessary to assess the need for palliative care services.

National renal transplant registry

The National Kidney Transplant Service for the Republic of Ireland was established in 1986 and is coordinated through Beaumont Hospital⁵⁸. The National Renal Transplant Registry collects data on parameters at time of transplant, renal disease and source of transplant. Patient data collected includes gender, area of residence and date of birth. The information is used to assess graft survival and patient survival, monitor factors affecting outcome and monitor performance¹⁶. Three further national transplant services exist in Ireland: the National Heart and Lung Transplant Service, the National Liver Transplant Service transplant and the National Pancreas Transplant Centre⁵⁹. Person level data from the transplant databases has been linked to cancer registry data to examine cancer incidence after organ transplantation⁶⁰.

Structural issues in healthcare organisation and delivery

The ad hoc development of the Irish health system has contributed to an information infrastructure that often does not link across service providers thus leading to duplication, fragmentation and increased costs. Patients cannot be easily tracked from hospital to community based care leading to large gaps and silos of under used data³⁰. Gaps exist particularly from the primary and community care sector as well as from outpatient clinics and emergency department attendances that don’t result in hospital admission. The lack of community and social care data is particularly relevant for PEoLC as a considerable amount of palliative care is delivered in the community.

Data governance

A second consequence of the ad hoc development of services means how data is managed and accessed across providers, many of whom are not part of the HSE, is not well defined. Private hospitals do not contribute to the HIPE national data collection so that studies based on HIPE data cannot be generalised to the whole population. Biases and omissions in the available data cannot be adequately assessed. Similarly inpatient hospice services in Ireland are mostly provided by charities, partially funded by the HSE but with separate and distinct governance structures⁶¹. Data models describing how the data are stored and organised are generally not available so that gauging the workload to manage and link data can be complex. Data dictionaries are generally not available so the datasets usefulness for PEoLC research cannot be evaluated in terms of the data items potentially available.

Individual health identifiers

The 2014 Health Identifiers Act⁶² mandated the creation of an individual health identifier (IHI) register so that all health service users can be uniquely identified. While work is ongoing to introduce IHIs across the Irish health system, they have not yet been widely incorporated into the national data collections described above. In the absence of unique identifiers, linking patient records across datasets requires probabilistic matching techniques⁶³, comprehensive strategies to guide the process including data cleaning and standardisation techniques⁶⁴ and detailed knowledge of the datasets to be linked. Address can be used in matching but over 35% of addresses in Ireland share their address with at least one other property. Eircode, Ireland’s postcode system, was launched in July 2015 where a unique postcode is assigned to each residential and business address. The integration of eircodes will facilitate probabilistic data matching of administrative datasets and allow geospatial analysis of the data.

Issues affecting quality or completeness of data within each dataset can affect the efficacy and accuracy of probabilistic matching. Issues can be systemic e.g. how the data are organised and stored. Temporal issues can include health service reconfigurations, changes to eligibility criteria (e.g. eligibility for medical cards) and/or changes in classifications systems over time.

Health Service Providers Identifiers

The 2014 Health Identifiers Act⁶² also legislated for the development of a national database to capture, maintain and publish quality assured and verified standard codes and identifiers for health related entities i.e. practitioners, organisations, services, locations, and information on the relationships between them⁶⁵. The repository will hold up-to-date information on health sites/locations, health care providers and services provided by the HSE and Private/Voluntary Organisations in Ireland. The introduction of health service provider identifiers will facilitate classification and enumeration of services that will benefit PEoLC research.

Place of death is an important outcome measure at a population level. In PEoLC research, place of death is commonly standardised to Own Residence, Hospital, Care Home and Hospice based on the place of death address,⁶⁶. There are no standards in use for Irish mortality data³⁹ so that categorising place of death based on the address of a healthcare facility can be difficult without local knowledge. Facilities range from specialised centres to large regional hospitals, general hospitals, community and district hospitals, public and private nursing homes. Some facilities provide different services on the same site e.g. nursing home and hospice services. Ambiguity around place of death could be reduced by requiring institutions to self-categorise the main services they provide from a standardised list.

Electronic Health records

The introduction of a national Electronic Health Record (EHR) in conjunction with IHI’s are a key part of the HSE’s strategic e-Health Programme⁶⁷. EHRs are the means by which data can be recorded and shared across organisations and care settings⁶⁸. Core functions will allow electronic prescribing and case management as well as the ability to aggregate data from these systems into a comprehensive national record, accessible to health and social care professionals, patients, service users and carers.

The 2020 HSE National Service Plan commits to progressing procurement of an electronic health record (EHR) solution in the National Children’s Hospital which will inform the procurement of an EHR solution for all health and social care services [69, p. 20]. To-date an electronic health record had been introduced at several maternity hospitals in Ireland⁷⁰. Several projects benefitting from the improving electronic health infrastructure have already been realised⁷¹.

In Scotland the availability of electronic medical records have been used to develop electronic palliative care summaries to improve patient care for those accessing out–of-hours services⁷². In England the impact of advance care planning (ACP) discussions have been evaluated in a hospice setting where that information has been recorded in the electronic patient record⁷³. A wider initiative that relies on the existence of electronic medical record has been in development for some time in England. The Electronic Palliative Care Coordination Systems aims to enable advance care planning, improve communication and coordination at end of life by providing up-to-date key information on patients believed to be in their last year of life⁷⁴. These studies from other countries demonstrate the opportunities for PEoLC research in Ireland as the electronic health infrastructure improves.

Health region

Several reconfigurations of the Irish health service have occurred since 2005, each of which can impact the continuity and quality of data collected. For example health boards have been replaced by HSE administrative areas and more recently by Community Health Organisation areas (CHO). In 2019, the Sláintecare report recommendation for a ‘common unit of geography’ for data collection and integration to increase capacity for cross-organisational research (Information and Research, page 24) has been initiated with the announcement of six integrated health regions to replace the CHOs⁷⁵. The data collection systems have not kept pace with these changes so that a patient cannot be accurately assigned to a CHO area using address data alone. Eircode postcodes could be used to assign every household to a distinct CHO and/or other geographical units. This would eliminate any ambiguity for both service providers and service users on where to seek health care in the first instance, help establish criteria for access to services and facilitate meaningful research around service provision by health region.

General Data Protection Regulations

In May 2018 the General Data Protection Regulation (GDPR) became law in the European Union^76,77. It regulates the processing of personal data relating to individuals in the EU so that personal data are

Included in GDPR is the principle of patient consent where by valid consent from individuals is required for the processing of their personal data. Consent must be a “freely given, specific, informed and unambiguous indication of the individual’s wishes”. GDPR force a stricter data governance regime on organisations so that data controllers i.e. the organisations collecting data, can be required to prove compliance with GDPR requirements.

The measures for data processing for health research are given more specific effect through Ireland’s Health Research Regulations Act (HRR)⁷⁸. Some of the specific measures enshrined by HRR Act were considered restrictive⁷⁹. Of particular concern were the requirements for explicit consent or approval from the Health Research Consent Declaration Committee (HRCDC)⁸⁰ for ongoing research involving retrospective chart reviews, use of biobank materials and research with individuals who lack capacity to consent. In 2021 the Department of Health amended the HRR to address these issues and clarify situations where a health research declaration consent application is required⁸¹.Under the new amendments, low risk retrospective chart reviews that have been approved by a research ethics committee and meet specified transparency requirements, no longer require a health research consent declaration⁸².

The requirements for compliance with GDPR and in particular the HRR are complicated by the fragmented health data infrastructure. Guidance notes are available to assist data controller organisations when making an application to the HRCDC for a consent declaration. In addition a public log of HRCDC applications provide an insight to the working of the committee. Information on the decision process for existing applications that include the decision outcome, any specific conditions attached and/or additional recommendations can alert researchers to potential requirements and/or shortcomings in their own application⁸⁰.

Underlying data

All data underlying the results are available as part of the article and no additional source data are required.