Cohort profile: The SAIL long-term conditions e-cohort (SLTC cohort) investigating area-level changes in healthcare resource use in Wales

Abstract

Introduction
The prioritisation of acute cases of coronavirus during the pandemic caused significant disruption to non-urgent healthcare services, creating a backlog of undiagnosed and untreated individuals with long-term conditions. Previous research has explored the impact of the pandemic on long-term conditions in Wales, but not the geographic variation or underlying area-level characteristics associated with these changes.

Objectives
We created the SAIL long-term conditions e-cohort (SLTC cohort) within the Secure Anonymised Information Linkage (SAIL) Databank to describe changes in healthcare service use of individuals living with long-term conditions during the COVID-19 pandemic, and to facilitate future investigations into the underlying reasons for these changes.

Methods
Individuals were included in the cohort if they interacted with health services with a long-term condition between January 2017 and December 2022. Interactions were identified using primary and secondary care datasets within the SAIL Databank. We linked this interaction level data with individual, residence, and area-level demographic data. We calculated area-level age-sex-standardised rates of interactions, based on an individual's address at the time of interaction, for the 3 years pre-COVID-19 (2017-2019) and during-COVID-19 (2020-2022). Percentage changes in rates between these time periods were calculated, and we investigated the underlying area-level characteristics associated with these differences.

Results
The SLTC cohort contains 1,277,532 individuals. Age-sex standardised interaction rates varied by Welsh Index of Multiple Deprivation (WIMD) quintiles and Rural-Urban Classification. Areas in the most deprived WIMD quintile had the greatest median percentage decrease (23.5%) in primary care rates of interactions from pre- to during-COVID-19, and the least deprived overall WIMD quintile had the smallest (16.9%). Areas classified as 'Urban city & town in a sparse setting' had the greatest decrease in primary care interactions (29.7%), and `Rural village' areas had the smallest decrease (17.1%). Secondary care rates of interactions showed less variation in rates of interactions between the two time periods.

Conclusion
We have created a cohort that links area-level characteristics and measures of healthcare resource use, in a study period that covers pre- and during-COVID-19, which will allow researchers to investigate geographic variation of changes in healthcare resource use over this time period and the underlying influences. This cohort can also be further linked to other area-level characteristics of interest, such as travel times to general practices, or access to green space measures.

Introduction

The COVID-19 pandemic caused worldwide disruption to healthcare services. While demand for healthcare surged amid pressures to treat acute cases of the virus, research has shown that attendances to general practices, hospitals and accident and emergency departments for non-COVID-19 related healthcare decreased during the COVID-19 pandemic [1, 2]. In Wales, the first case of COVID-19 was reported on 28^th February 2020, leading to significant changes in healthcare delivery. Diagnosis rates of long-term conditions decreased in 2020 and 2021 compared to previous years in Wales [3] creating a potential backlog of patients living with undiagnosed, and therefore unmanaged, conditions. It has been estimated that a typical general practice – where long-term conditions are usually diagnosed and managed – serving a population of 10,000 patients may have over 400 undiagnosed long-term conditions [3].

This backlog of undiagnosed and untreated patients was reflected worldwide, with millions of people left with delayed or postponed treatments and procedures, while services were directed towards tackling COVID-19 related illnesses [4, 5].

This is a concern for individuals’ worsening prognoses and subsequent presentations to healthcare settings with more severe conditions which are harder to manage and treat effectively. Delayed diagnosis not only impacts an individual’s quality of life and health outcomes but also places greater strain on secondary healthcare services, as patients are more likely to require emergency or specialised care. These challenges can lead to increased healthcare costs and exacerbate pressures on already stretched systems. In order to direct resources effectively, understanding where services were most impacted can help inform strategies to target support for the worst hit communities.

The SAIL long-term conditions e-cohort (SLTC cohort) was set up to investigate how the COVID-19 pandemic impacted non-COVID-19 healthcare service use and provision in Wales, UK, with a particular focus on how this varies at the local community level. We have used Electronic Healthcare Records (EHRs), linked to residence and area-level data to create a cohort of individuals with long-term conditions, along with the characteristics of the areas in which they live. This will allow us to identify ‘hot’ and ‘cold’ spots of healthcare resource utilisation across Wales, comparing pre- and during-COVID-19 rates of healthcare service interactions. This cohort will then facilitate investigations into which communities and subgroups of the population were at the highest risk of the indirect negative effects of COVID-19. In subsequent work, we will use the cohort to examine the underlying individual, environmental, residential, and area-level characteristics associated with the greatest impact on healthcare services.

Our research, using the SLTC cohort, will provide evidence on where policies, interventions and resources could be targeted, to address the ‘undiagnosed’ populations of Wales, or to identify areas where long-term conditions are more prevalent.

Methods

Data sources/datasets

The cohort was created using data accessed and analysed within the Secure Anonymised Information Linkage (SAIL) Databank (www.saildatabank.com) [6–8]. The SAIL Databank is a trusted research environment, containing anonymised individual level EHRs from primary and secondary care healthcare settings, as well as demographic data, birth, and death data, among other datasets for the entire resident population of Wales over time.

The Welsh Longitudinal General Practice (WLGP) dataset [9] and the Patient Episode Database for Wales (PEDW) dataset [9] were used to identify diagnoses of long-term conditions. These contain patient records from primary care (attendances and clinical information from General Practices (GPs)) and secondary care (hospital admissions, day case activity) health services respectively. The WLGP dataset covers 86.1% of the population and 82.4% of GP practices in Wales and records are coded using Read v2 codes. The PEDW dataset contains all inpatient and day case activity undertaken in the National Health Service (NHS) Wales, and records are coded using the International Classification of Diseases version 10 (ICD-10). PEDW contains information on patients treated by the NHS, including where treatment occurs in private care facilities but is funded through the NHS. However it does not contain information on privately funded treatments in private hospitals.

The Welsh Demographic Service Dataset (WDSD) which contains information on all individuals registered to a Welsh GP [9] was used to obtain individual-level demographic data. EHRs were linked to the WDSD using Anonymised Linking Fields (ALFs).

The WDSD also contains details of residential address history. Individual addresses are replaced with a Residential Anonymous Linking Field (RALF) [10, 11], that allows for further linkage to geospatial characteristics at the household level.

RALFs are mapped to Lower Layer Super Output Areas (LSOAs) [12] which are small area statistical geographies and comprise between 400 and 1200 households and have a resident population between 1000 and 3000 persons.

The Welsh Index of Multiple Deprivation (WIMD) is the Welsh Government’s official measure of relative deprivation for small areas in Wales [13]. It contains measures across the following domains: Income, Employment, Health, Education, Access to Services, Community Safety, Physical Environment, and Housing. Each of the 1909 LSOA’s in Wales has domain-specific measures of deprivation and are then ranked from 1 (most deprived) to 1909 (least deprived) within each domain, and across overall deprivation. They can be further grouped into deciles and quintiles.

The Rural/Urban Classification is the UK Government’s categorisation of geographic areas by population density, subdivided into six settlement types; Urban city and town, Urban city and town in a sparse setting, Rural town and fringe, Rural town and fringe in a sparse setting, Rural village and dispersed, Rural village and dispersed in a sparse setting [14].

The matching process and assignment of unique ALFs to person-based records has been previously reported [8], demonstrating high quality and enabling accurate linkage.

Study population

The SLTC cohort is population level containing all individuals living in Wales and diagnosed with at least one of 17 long-term conditions any time up until December 2022, and a long-term condition related interaction with healthcare services between January 2017 and December 2022. These interactions could be the initial diagnosis of the long-term condition, or a subsequent interaction with health services where the long-term condition was recorded.

The time periods were selected based on data completeness and availability. The pre- and during-COVID-19 periods were designed to be of equal length to allow for comparisons. The entirety of 2020 was included in the during-COVID-19 period to account for potential behavioural and service utilisations changes prompted by awareness of the pandemic, prior to the first confirmed case in Wales.

The long-term conditions were defined using the general practice ‘Quality and Outcomes Framework’ [15]. The selection of conditions was informed by collaboration with NHS clinical colleagues to ensure a focus on conditions that are generally managed rather than cured, and to ensure high data quality. Records of interactions were extracted from routinely collected EHRs from the NHS, UK.

The long-term conditions included were: anxiety disorders, asthma, atrial fibrillation, coronary heart disease, chronic kidney disease, chronic obstructive pulmonary disease, dementia, depression, diabetes mellitus, epilepsy, heart failure, hypertension, inflammatory bowel disease, osteoporosis, peripheral vascular disease, rheumatoid arthritis, and stroke and transient ischaemic attack.

Variables

Individual health records for primary care were extracted from WLGP where the Read code matched with a Read code for long-term conditions. We extracted all records recorded between 1st January 2017 and 31st December 2022. Multiple records for an individual (indicated by variable ALF) were retained. Records related to the same long-term condition with the same event date were counted as a single interaction with primary care. The list of Read codes used to define long-term conditions can be found in Supplementary Appendix 1.

Individual health records for secondary care were extracted from PEDW where the ICD-10 code matched with an ICD-10 code for long-term conditions. We extracted all records where the hospital spell had a start date between 1st January 2017 and 31st December 2022. Multiple records for an individual (indicated by variable ALF) were retained. Records related to the same long-term condition and the same hospital spell were counted as a single interaction with secondary care. The list of ICD-10 codes used to define long-term conditions can be found in Supplementary Appendix 1.

The WDSD contains variables shown in Table 1. The variables Anonymised Linking Field (ALF), Address Start Date, and Address End Date were used for data linkage with interaction level data. Individuals were linked to an address based on the date of interaction with healthcare services falling between the address start date and address end date in WDSD. LSOA Codes were used for data linkage with area-level characteristic datasets. The WIMD dataset was used to extract area-level measures of deprivation for the cohort including: access to services and physical environment domains, as well as overall deprivation. Linkage to Office for National Statistics Rural Urban Classification data enabled area-level grouping of the cohort by rural-urban classes.

Table 1: SLTC Cohort linked datasets, variables included at both individual level and area-level.

Dataset	Variables	Individual/area-level
Welsh Longitudinal General Practice	Anonymised Linking Field (ALF)	Individual
	Event Date	Individual
	Read Code	Individual
	Anonymised Linking Field Status Code	Individual
Patient Episode Database for Wales	Anonymised Linking Field (ALF)	Individual
	Spell Start Date	Individual
	Spell End Date	Individual
	ICD-10 Code	Individual
	Anonymised Linking Field Status Code	Individual
Welsh Demographic Service Dataset	Anonymised Linking Field (ALF)	Individual
	Address Start Date	Individual
	Address End Date	Individual
	Residential Anonymised Linking Field (RALF)	Individual
	LSOA Code	Individual
	Week of Birth	Individual
	Date of Death	Individual
	Sex	Individual
Welsh Index of Multiple Deprivation	Lower Layer Super Output Area (LSOA) Code	Area
	WIMD Quintile	Area
	Access to Services Quintile	Area
	Physical Environment Quintile	Area
	Average travel time to pharmacy	Area
	Average travel time to GP	Area
	Proximity to accessible natural green space score	Area
	Ambient green space score	Area
Rural-Urban Classification	Lower Layer Super Output Area (LSOA) Code	Area
	Rural-Urban Classification	Area
Office for National Statistics Population Estimates	Year	Area
	Mid-year Population Estimates	Area

Linkage to individual level demographic data provided further information on individuals and allowed us to carry out exclusions of invalid records.

We calculated rates of healthcare interactions – age-sex standardised using the Welsh standard population (see Supplementary Appendix 2) – per 100 000 persons, for every LSOA in Wales, from 2017-2019 and 2020-2022.

We used these to calculate the percentage change in rates pre- and during-COVID-19, to identify which areas of Wales were most affected by the COVID-19 pandemic in terms of healthcare service use.

We further linked these area-level counts and rates to area-level socio-demographic and environmental datasets to understand which factors were associated with variation in the pre-during-COVID-19 changes between small areas across Wales. Median changes in age-sex standardised rates were also summarised by WIMD quintiles and Rural-Urban classifications to explore how healthcare utilisation changes varied across communities with different characteristics.

Analyses were performed using R 4.1.3, with the dsr package used to compute standardised rates.

Exclusions

The following exclusions were applied for individuals within the SLTC cohort:

• Non-Welsh address at the time of interaction (Already excluded as unable to link to WDSD data).
• Interactions occurring before week of birth.
• Interactions occurring after date of death.
• Anonymised linking field status not valid.
• Sex missing, or unknown.

Patient and public involvement

Public involvement was incorporated through the SAIL Consumer Panel. The panel provided insights into healthcare access during the COVID-19 pandemic, identifying long-term conditions as a priority for research, and highlighting regional differences in healthcare service access and use across Wales.

Results

Records relating to 1 375 055 persons were extracted from the data, 97 501 individuals could not be linked to the WDSD, and were excluded resulting in 92.9% of individuals as linkable to individual and area-level demographic level data. A further 22 individuals were excluded as shown in Figure 1.

Figure 1: Cohort inclusion flow diagram.

We linked the remaining person level EHRs with individual and area-level demographic data.

The final number of individuals in the SLTC cohort was 1 277 532, of these 463 218 individuals entered the cohort in the first year, then 262 485 in 2018, 197 658 in 2019, 117 406 in 2020, 118 755 in 2021, and 118 010 in 2022.

The numbers of individuals at each stage are shown in Figure 1.

A total of 7 502 485 interactions with healthcare services in Wales with a record of a long-term condition between January 2017 and December 2022 were identified.

Within the SLTC cohort, 74.1% (946 679) of individuals had at least one primary care interaction, with 2 555 732 primary care interactions in total. For secondary care, 59.5% (760 704) of individuals had at least one interaction, with 4 946 753 secondary care interactions in total.

More than half of the cohort (55.1%) had interactions relating to more than one long-term condition in the 6-year timeframe. In total, 175 798 individuals in the cohort died between January 2017 and December 2022.

Across the 6 years of coverage, the most deprived areas had the greatest number of individuals with long-term conditions who interacted with healthcare services, followed by the 2nd most deprived, 3rd most deprived, 4th most deprived, then the least deprived quintile (see Table 2).

Table 2: Number of distinct individuals by WIMD deprivation quintile and Rural-Urban Classification with a long-term condition health service interaction by year of interaction. Note: Yearly counts per WIMD Quintile/Rural-Urban Classification will not sum to total yearly counts, as individuals who moved between WIMD quintiles/Rural-Urban Classifications within a year and had interactions with healthcare services in each are counted more than once.

	Year
	2017	2018	2019	2020	2021	2022
WIMD Quintile
1 (most deprived)	104 794	106 010	106 747	80 719	85 633	92 542
2	97 808	98 738	100 624	75 457	81 143	90 045
3	92 021	92 544	93 278	70 199	76 980	85 517
4	89 406	90 068	91 083	69 537	76 318	83 619
5 (least deprived)	84 029	85 058	86 130	67 647	73 893	80 546
Rural-Urban Classification
Urban city and town	315 749	319 592	323 174	247 746	265 077	290 578
Urban city and town in a sparse setting	9714	9755	9836	7546	8029	8675
Rural town and fringe	62 705	63 052	63 433	48 094	53 722	58 658
Rural town and fringe in a sparse setting	17 053	17 259	17 420	12 916	14 041	15 609
Rural village and dispersed	27 458	27 827	28 534	21 540	24 091	26 308
Rural village and dispersed in a sparse setting	32 403	32 147	32 807	24 236	26 930	30 297
Total	463 218	467 762	473 525	360 916	390 411	428 572

We calculated age-sex standardised rates of interactions recorded between 2017-2019 (pre-COVID-19) and between 2020-2022 (during-COVID-19) for each LSOA and compared the differences between pre- and during-COVID-19 rates by quintiles of overall WIMD, and the domain specific indicators Access to Services, and Physical Environment (Table 3a), which were selected due to their relevance to area-level factors such as access to healthcare services, proximity to green spaces, and levels of air pollution. We also compared the differences between pre- and during-COVID-19 age-sex standardised rates by Rural-Urban groups.

Table 3: Median Percentage difference in age-sex standardised rates of primary and secondary care interactions from 2017-2019 to 2020-2022 by a) WIMD quintiles (Overall, Access to Services, and Physical Environment) - 1(most deprived) to 5(least deprived) and b) Rural-Urban Classification.

(a)
WIMD Domain	Quintile	Median percentage difference age-sex standardised rates (pre to during-COVID-19) (%)
		Primary care	Secondary care
Overall	1	–23.5	–12.3
	2	–21.7	–13.4
	3	–22.4	–13.5
	4	–20.6	–12.6
	5	–16.9	–11.2
Access to Services	1	–20.8	–12.8
	2	–20.6	–13.2
	3	–22.3	–13.5
	4	–20.7	–11.3
	5	–22.3	–12.8
Physical Environment	1	–24.1	–10.9
	2	–20.6	–12.2
	3	–21.7	–14.3
	4	–19.6	–13.6
	5	–20.9	–11.3
(b)
Rural-Urban Classification	Median percentage difference age-sex standardised rates (pre to during-COVID-19) (%)
	Primary care	Secondary care
Urban city & town	–20.5	–11.5
Urban city & town in a sparse setting	–29.7	–8.4
Rural town and fringe	–22.6	–14.4
Rural town and fringe in a sparse setting	–28.6	–18.1
Rural village	–17.1	–11.4
Rural village in a sparse setting	–24.7	–15.0

Both primary care and secondary care interactions decreased during-COVID-19 across all deprivation quintiles. The most deprived overall WIMD quintile had the greatest median percentage decrease (23.5%) in primary care rates of interactions from pre- to during-COVID-19, and the least deprived overall WIMD quintile had the smallest median percentage decrease (16.9%). The pattern of decreases in primary care interactions across the Access to Services domain quintiles was not linearly correlated with deprivation (from 20.6% in the 2^nd most deprived and 22.3% in the 3^rd most deprived). There was less variation in the percentage change of secondary care age-sex standardised rates, across the WIMD quintiles (between 10.9% and 14.3% decrease).

Areas classified as ‘Urban city & town in a sparse setting’ had the greatest decrease in primary care interactions (29.7%) in the during-COVID-19 period compared to the pre-COVID-19 period, and ‘Rural village’ areas had the smallest decrease (17.1%). However, the ‘Urban city & town in a sparse setting’ group had the smallest decrease in secondary care age-sex standardised rates of interactions (8.4%), and the ‘Rural town and fringe in a sparse setting’ group had the greatest decrease (18.1%). Changes in primary and secondary care age-sex standardised rates by Rural-Urban Classification are shown in Table 3B and visualised in Figure 2.

Figure 2: Primary and secondary care % change in age-sex standardised rates by Rural-Urban Classification.

Discussion

We have created the SLTC cohort to: enable quantification of the variation in non-COVID-19 healthcare resource use geographically across Wales; identify the areas of Wales that experienced the greatest changes; and, facilitate understanding the underlying reasons.

We have found the biggest changes in primary care interactions occurred within the most deprived WIMD quintile and areas classified as ‘Urban city & town in a sparse setting’, while the biggest changes in secondary care interactions occurred within areas classified as ‘Rural town and fringe in a sparse setting’, but there was less variation between WIMD deprivation quintiles.

The results of this study build on the work of Moynihan et al [16], who conducted a systematic review on the impact of the COVID-19 pandemic on healthcare utilisation. While their review highlighted trends in healthcare service utilisation, this study and cohort contribute to the existing literature by addressing a key gap, examining area-level changes in healthcare use and considering the underlying characteristics of the areas most affected. For example, we found that "Urban city & town in a sparse setting" areas experienced the greatest decrease in primary care interactions but the smallest decrease in secondary care interactions, indicating a potential shift in healthcare service utilisation from primary to secondary care. This shift is concerning, as ideally long-term conditions are mostly treated and managed in primary care. The observation that areas in the most deprived WIMD quintile experienced the greatest reduction in primary care interactions highlights a potentially disproportionate impact on these services for the most deprived populations in Wales.

In contrast to the study by Qi et al [3] which focused on diagnosis rates and found limited differences across deprivation levels, this study examined both initial diagnoses and ongoing management. While deprivation may not have significantly impacted the initial diagnosis of long-term conditions, it is likely to have influenced access to follow-up care.

These findings have implications for ongoing healthcare catch-up initiatives and provide useful insight for preparedness for potential future health crises, emphasising the importance of ensuring equitable access to primary care services during times of disruption.

Future work will use the SLTC cohort to apply spatial auto-correlation methods and identify clustering of similar values of age-sex standardised rates. Geospatial analyses will enable a better understanding of the local relationships between area-level socio-demographic and environmental factors (such as distances to general practices, proximity to greenspace, Enhanced Vegetation Index (EVI)) and various healthcare resource use outcome measures. We aim to investigate which area-level characteristics had the greatest impact on the effects of the pandemic on healthcare service use. There is also potential for the cohort to be used to investigate condition-specific changes in healthcare-service use, to explore whether individuals with specific conditions were disproportionately impacted by the pandemic.

Strengths and limitations

The main strength of the SLTC cohort is the use of longitudinal nationwide healthcare interaction records, demographic and area-level characteristic data with population-scale coverage in Wales. The use of anonymised linked datasets within SAIL Databank, a secure repository of large-scale datasets regarding the population of Wales, allows us to consider the geospatial characteristics associated with variation in healthcare resource use at a small area-level.

A further strength is the investigation of both primary care and secondary care interactions, which will help us to understand how individuals living in different parts of Wales have changed which services they use more or less frequently during-COVID-19.

This cohort has complete population coverage for secondary care and demographic records, while primary care records cover 86.1% of the population. This provides a more representative sample of the population of Wales compared to other types of cohorts, so studies using the SLTC can inform public health strategies aimed at improving the health of the whole population, and relieve pressures on the National Health Service.

This study included healthcare interactions for individuals diagnosed with a broad range of long-term conditions. These conditions defined using the Quality and Outcomes Framework criteria were selected for comparison with previous work [3] investigating temporal trends in diagnosis rates. A limitation of using these selected conditions is that the variation in healthcare use across areas may not be representative of other healthcare conditions (e.g. cancer).

The data used to create the SLTC cohort is routinely recorded administrative data and while it is highly valuable, it does not fully capture human behaviour. We therefore cannot account for individuals’ decisions not to attend primary care appointments or go to hospital for emergency or planned treatment nor understand the reasons why. As the study period of the SLTC cohort spans the peak of the COVID-19 pandemic, there may be further underlying reasons – such as fear of contracting the virus - why individuals may have avoided using healthcare services, and therefore not appear in routine data. It doesn’t necessarily imply the services were not available to them.

We are currently limited to statistical geographies when creating small area estimates of clustering. In rural areas where the statistical geographies are larger, we may not be able to detect intra-small area variations in disease distribution. Area-level measures of deprivation are used to make inferences about individuals’ socio-economic status, which would inevitably vary within small areas. This ecological fallacy of area-level characteristics limits the conclusions that can be drawn at an individual level; however, the purpose of this study is to help direct scarce resources to the area’s most in need rather than identify individual persons most affected.

A limitation of this study is that it does not account for the effects of specific lockdown periods on healthcare utilisation. This decision reflects the study’s primary focus on overall changes in service use patterns across pre- and during-COVID-19 periods, rather than a time-series analysis. Future studies could utilise this cohort to explore these temporal trends and underlying causes in greater detail.

Data availability statement

Anonymised individual-level data was extracted and linked within the SAIL Databank. Aggregated outcome data was requested out of the secure e-research platform and linked to area-level datasets.

Applications to access data within the SAIL Databank can be submitted here https://saildatabank.com/data/apply-to-work-with-the-data/, and will be reviewed by an independent Information Governance Review Panel.

Conflict of interest

None declared.

Ethics statement

This research was conducted within the Secure Anonymised Information Linkage (SAIL) Databank, with the permission and approval of the SAIL Independent Information Governance Review Panel (IGRP) project number 1542.

The cohort was created using anonymised data within the SAIL Databank, which is exempt from requiring individual patient consent.

Acknowledgements

This study makes use of anonymised data held in the Secure Anonymised Information Linkage (SAIL) Databank. We would like to acknowledge all the data providers who make anonymised data available for research.

Funding

This work was supported by Health and Care Research Wales [HRG-20-1755(P)]

References

1. Carr A, Smith JA, Camaradou J, Prieto-Alhambra D. Growing backlog of planned surgery due to covid-19. BMJ. 2021 Feb 9;372(8280):n339. 10.1136/bmj.n339

   10.1136/bmj.n339

2. Macdonald N, Clements C, Sobti A, Rossiter D, Unnithan A, Bosanquet N. Tackling the Elective Case Backlog Generated by Covid-19: the Scale of the Problem and Solutions. Journal of Public Health. 2020 Sep 15;42(4). 10.1093/pubmed/fdaa155

   10.1093/pubmed/fdaa155

3. Qi C, Osborne T, Bailey R, Cooper A, Hollinghurst J, Akbari A, et al. Impact of COVID-19 pandemic on incidence of long-term conditions in Wales: a population data linkage study using primary and secondary care health records. British Journal of General Practice. 2023 Jan 23;73(730):e332–9. 10.3399/bjgp.2022.0353

   10.3399/bjgp.2022.0353

4. World Health Organization (WHO). Addressing backlogs and managing waiting lists during and beyond the COVID-19 pandemic [Internet]. 2022. Available from: https://eurohealthobservatory.who.int/publications/i/addressing-backlogs-and-managing-waiting-lists-during-and-beyond-the-covid-19-pandemic.

5. Wang J, Vahid S, Eberg M, Milroy S, Milkovich J, Wright FC, et al. Clearing the surgical backlog caused by COVID-19 in Ontario: a time series modelling study. Canadian Medical Association Journal. 2020 Sep 1;192(44):cmaj.201521. 10.1503/cmaj.201521

   10.1503/cmaj.201521

6. Jones KH, Ford DV, Thompson S, Lyons R. A Profile of the SAIL Databank on the UK Secure Research Platform. International Journal of Population Data Science. 2020 Oct 19;4(2). 10.23889/ijpds.v4i2.1134

   10.23889/ijpds.v4i2.1134

7. Ford DV, Jones KH, Verplancke JP, Lyons RA, John G, Brown G, et al. The SAIL Databank: building a national architecture for e-health research and evaluation. BMC Health Services Research. 2009 Sep 4;9(1). 10.1186/1472-6963-9-157

   10.1186/1472-6963-9-157

8. Lyons RA, Jones KH, John G, Brooks CJ, Verplancke JP, Ford DV, et al. The SAIL databank: linking multiple health and social care datasets. BMC Medical Informatics and Decision Making. 2009 Jan 16;9(1). 10.1186/1472-6947-9-3

   10.1186/1472-6947-9-3

9. Health Data Research UK. HDRUK Innovation Gateway [Internet]. [cited 2024 Feb 26]. Available from: https://www.healthdatagateway.org/.

10. Rodgers SE, Demmler JC, Dsilva R, Lyons RA. Protecting health data privacy while using residence-based environment and demographic data. Health & Place. 2012 Mar;18(2):209–17. 10.1016/j.healthplace.2011.09.006

    10.1016/j.healthplace.2011.09.006

11. Rodgers SE, Lyons RA, Dsilva R, Jones KH, Brooks CJ, Ford DV, et al. Residential Anonymous Linking Fields (RALFs): a novel information infrastructure to study the interaction between the environment and individuals’ health. Journal of Public Health. 2009 May 15;31(4):582–8. 10.1093/pubmed/fdp041

    10.1093/pubmed/fdp041

12. Office for National Statistics (ONS). 2011 Census geographies [Internet]. [cited 2024 Feb 26]. Available from: https://www.ons.gov.uk/methodology/geography/ukgeographies/censusgeographies/2011censusgeographies.

13. Welsh Government. Welsh Index of Multiple Deprivation 2019 [Internet]. 2019 [cited 2024 Feb 26]. Available from: https://www.gov.wales/welsh-index-multiple-deprivation-full-index-update-ranks-2019.

14. Department for Environment, Food & Rural Affairs. Rural Urban Classification [Internet]. 2016 [cited 2024 Feb 26]. Available from: https://www.gov.uk/government/collections/rural-urban-classification.

15. Quality and Outcomes Framework [Internet]. statswales.gov.wales. [cited 2024 Jun 3]. Available from: https://statswales.gov.wales/Catalogue/Health-and-Social-Care/NHS-Primary-and-Community-Activity/GMS-Contract/quality-and-outcomes-framework.

16. Moynihan R, Sanders S, Michaleff ZA, et al. Impact of COVID-19 pandemic on utilisation of healthcare services: a systematic review. BMJ Open 2021;11:e045343. 10.1136/bmjopen-2020-045343

    10.1136/bmjopen-2020-045343