Interrupted time series evaluation of the impact of a dementia wellbeing service on avoidable hospital admissions for people with dementia in Bristol,England

Abstract

Objectives

To determine whether a dementia wellbeing service (DWS) signposting people with dementia to community services decreases the rate of avoidable hospital admissions, in-hospital mortality, complexity of admissions (number of comorbidities) or length of stay.

Methods

Interrupted time series analysis to estimate the effects of the DWS on hospital outcomes. We included all unplanned admissions for ambulatory care sensitive conditions (‘avoidable hospital admissions’) with a dementia diagnosis recorded in the Hospital Episode Statistics. The intervention region was compared with a demographically similar control region in the 2 years before and 3 years after the implementation of the new service (October 2013 to September 2018).

Results

There was no strong evidence that admission rates reduced and only weak evidence that the trend in average length of stay reduced slowly over time. In-hospital mortality decreased immediately after the introduction of the dementia wellbeing service compared to comparator areas (x0.64, 95% CI 0.42, 0.97, p = 0.037) but attenuated over the following years. The rate of increase in comorbidities also appeared to slow after the service began; they were similar to comparator areas by September 2018.

Conclusions

We found no major impact of the DWS on avoidable hospital admissions, although there was weak evidence for slightly shorter length of stay and reduced complexity of hospital admissions. These findings may or may not reflect a true benefit of the service and require further investigation. The DWS was established to improve quality of dementia care; reducing hospital admissions was never its sole purpose. More targeted interventions may be required to reduce hospital admissions for people with dementia.

Keywords

health services evaluation dementia avoidable admissions

Introduction

In the United Kingdom (UK), around 850,000 people were estimated to be living with dementia in 2015.¹ This number is expected to increase to over 1 million by 2025.¹ People with dementia (PWD) have an increased risk of hospitalisation, including hospitalisations for ambulatory care sensitive conditions (ACSC, also referred to as ‘avoidable hospital admissions’),² conditions for which hospital admission could potentially be prevented by interventions in primary or community care. Once admitted to hospital, PWD also have poor outcomes when admitted to general hospitals, including longer length of stay and higher risk of death.³

The Bristol Dementia Wellbeing Service (DWS) is a partnership between the Alzheimer’s Society and the Devon Partnership NHS Trust (DPT) in the south-east of England, UK.⁴ The DWS has been providing support to people with dementia, their carers, GPs and other health professionals since April 2015. The service designates a dementia practitioner and a dementia navigator for each GP practice to aid early diagnosis, deliver therapeutic interventions and to signpost people with dementia and their families to resources that offer information, guidance, and practical and emotional support.⁴ They also help develop a wellbeing plan, which outlines medication regimes, support needs and activities and groups that can enhance quality of life.⁵ In its first year of operation, the DWS received 2068 referrals and delivered 2945 wellbeing plans.⁵

Assessments of the Bristol DWS have so far been limited to service user and carer feedback, which has been mostly positive, with those providing feedback through the ‘Friends and Family Test’⁶ (April 2015 to March 2016) saying they would be likely or extremely likely to recommend the service to other people.⁵ There have been no assessments of the impacts of the DWS on other outcomes for people with dementia. This study seeks to fill this gap by evaluating whether receiving support from the Bristol DWS decreases the rate of avoidable hospital admissions, in-hospital mortality, length of stay and complexity of admissions (number of comorbidities).

Methods

We conducted a longitudinal observational study using routinely collected data from patients admitted to hospitals in England between October 2013 and September 2018. We used a controlled natural experiment design, comparing data from Bristol before and after the implementation of the DWS to appropriate comparator areas, accounting for secular changes in health care utilisation. The study was developed and reported according to the RECORD extension to STROBE guidelines for observational studies using routinely collected data.⁷

Data sources

We used anonymised, individual episodes in the admitted patient care hospital episode statistics (HES-APC, October 2013 to September 2018). HES-APC is a routinely collected dataset that records all episodes of care provided to patients admitted (day case or inpatient) to NHS hospitals in England and NHS-funded patients treated in private hospitals.⁸ Each episode represents a period of care under one consultant team. Up to 20 diagnoses are recorded per episode using the International Classification of Diseases (ICD) version 10. HES also includes the registered GP practice for each patient, allowing linkage of hospital admissions to GP practice populations. GP practice list sizes and dementia populations were identified using the Quality Outcomes Framework (QOF) dementia indicators.^9-11 The QOF rewards GP practices based on indicators of quality of care, such as maintaining a dementia register.

Study population

We included patients with dementia registered at GP practices in Bristol Clinical Commissioning Group (CCG; clinically-led statutory NHS bodies in England responsible for the planning and commissioning of health care services for their local area between 2013 and 2022) (intervention area) and 10 CCGs considered most similar to Bristol CCG in terms of demographics and population density according to NHS Rightcare Commissioning for Value:¹² Brighton and Hove, Coventry and Rugby, Hull, Liverpool, Norwich, Portsmouth, Salford, Sheffield, Southampton and Sunderland.

Identifying dementia admissions (numerator)

Dementia hospital admissions were identified using ICD-10 dementia diagnosis codes (Online Supplement Table S1) in any of the 20 diagnosis fields. Episodes were combined into continuous inpatient spells (CIPS) using the NHS Digital method,¹³ which combines episodes within the same admission for the same patient even if this includes a transfer between hospitals or providers. We will refer to these CIPS as admissions; of these, 28% contained a single consultant episode, and 97% contained five or fewer episodes.

Dementia population at risk (denominator)

To determine the population at risk for a dementia admission, we extracted the number of dementia patients on the dementia register for each GP practice in the QOF dataset. Practices were excluded if they did not provide QOF data for all six financial years (2013/14-2018/19), so that analysis of trends would be consistent in terms of included practices over time. Four out of 46 practices (9%) were excluded from Bristol CCG, and 67 out of 512 practices (13%) were excluded from comparator CCGs.

Outcomes

We assessed four outcomes.

Hospital admission rate for ACSC among dementia patients

ACSC are conditions for which unplanned hospital admissions could potentially be prevented by interventions in primary or community care.¹⁴ We identified ACSCs for people with dementia from the diagnosis fields in HES using ICD-10 codes listed in the Directory of Ambulatory Emergency Care for Adults.¹⁵ We calculated hospital admission rates for ACSCs by taking the number of unplanned hospital admissions with a primary ACSC diagnosis for dementia patients registered at GP practices within an area (numerator) and dividing by the number of patients on the dementia register at those GP practices (denominator).

In-hospital mortality

In-hospital mortality is indicated in the HES data discharge method field (4 = died). We took the number of in-hospital deaths during ACSC admissions for people with dementia registered at GP practices within an area and divided by the number of ACSC hospital admissions for people with dementia at those GP practices, so it is conditional on being admitted.

Length of hospital stay

We used two measures of length of stay as this continuous variable is heavily skewed to the right with outliers who stay a long time: (a) proportion of ACSC hospital admissions over 30 days as this arbitrary cut-off indicates a long length of stay and is not overly influenced by outliers and (ii) average length of stay for ACSC hospital admissions when length of stay is 30 days or less.

Complexity of admission (Charlson comorbidity index)

We explored the complexity of ACSC hospital admissions in terms of the number of comorbid conditions listed on the same admission. The Charlson comorbidity index¹⁶ is a well-validated summary measure of comorbidity. It includes 17 categories of comorbid conditions and has been shown to be associated with mortality after 1 year. We used a Charlson index algorithm adapted for use with ICD-10 diagnosis codes¹⁷ to count the number of Charlson comorbidities recorded in each hospital admission.

Statistical analysis

For each outcome, we estimated values for each month and conducted controlled interrupted time series (ITS) analyses comparing the intervention area (Bristol CCG) to the comparator area (10 CCGs considered to be similar) as noted above. We assumed that the DWS in Bristol was fully operational at the start of October 2015, and the comparator areas did not introduce a similar service at that time. Interrupted time series analysis compares the trend before an intervention with the trend after the intervention using segmented regression models.^18,19 We compared the change in trend in the intervention area to any change in trend in the comparator area to reduce the chance that the change reflects some secular difference across all areas. We also controlled for annual seasonality in the data by including indicator variables for spring, summer and Autumn, and Winter as baseline (December, January, February).¹⁹ We adjusted for serial autocorrelation in the models by allowing for an autocorrelation structure using Newey-West standard errors^20-22 with maximum lag 2 months using the xtpoisson command to produce multiplicative models for count/rate data (admission rate, in-hospital mortality rate, proportion long stays) and the newey command for linear/additive models of averages (average length of stay and average number of comorbidities); in sensitivity analyses we allowed the maximum lag to be 1 or 5 months. We report four variables for the ITS analyses: ‘difference in Oct 2013’ refers to the difference between the intervention area Bristol and the comparator at October 2013 before the DWS was launched and reflects historical population or health care issues; ‘trend before’ is the difference in trend between Bristol and the comparator before the intervention; ‘change in level’ is the difference in the immediate level change at the time of the intervention (October 2015) between Bristol and the comparator; and ‘change in trend’ is the difference in the change in the slope of the trend after the intervention between Bristol and the comparator. For multiplicative models we used an ‘x’ before coefficients to indicate multiplication (e.g. x0.61); for additive models we used ‘+’ or ‘−’ before the coefficients.

All statistical analyses were conducted using Stata/MP version 16.1 for Windows. Stata code is available via github.²³

Results

Between October 2013 and September 2018, there were 12,400 dementia admissions (7.98 admissions per month per 100 PWD; 89% emergency) in Bristol CCG and 122,529 admissions (8.19 admissions per month per 100 PWD; 90% emergency) in the comparator area. There was a primary dementia diagnosis for 5% of admissions, and 65% of unplanned (emergency) admissions were for an ACSC, with little variation between areas. There were 7241 ACSC dementia admissions in Bristol (4.66 ACSC admissions per month per 100 PWD) and 65,488 in the similar CCGs (4.77 per month per 100 PWD).

We compared demographics for Bristol CCG and the comparator CCGs to those provided by the Bristol DWS for their service users during the study period. These were similar, with 61% of PWD being women in all cases. The proportion of white PWDs was 95% for Bristol CCG, 98% for the comparator CCGs and 93% for Bristol DWS. The majority was aged 75 years and older (at 83%, 84% and 84%), and 44%, 43% and 41% were aged 85 and older. Reasons for hospital admissions for people with dementia in each area were also similar, with the major reasons being urinary tract infection, acute respiratory tract infection, pneumonia and falls.

Hospital admission rate for ACSC among dementia patients

Interrupted time series analysis of ACSC hospital admission rates (per 100 dementia patients) for people with dementia showed that Bristol started with a slightly lower admission rate in October 2013 (x0.87; 95% CI 0.80, 0.95, p = 0.001) (Figure 1); rates converged over time, although there was no evidence for different trends in admission rates between Bristol and the comparator CCGs before or after the intervention (change in trend x1.00; 95% CI 0.99, 1.01, p = 0.771) (Table 1).

Figure 1.

Monthly dementia admission rates for ACSC conditions per 100 dementia patients for Bristol and 10 similar CCGs, October 2013 - September 2018.Note: vertical red line is our start of intervention (BDWS).

Table 1.

Results of segmented regression analyses for hospital admissions for ACSC among dementia patients, in-hospital mortality rates, length of hospital stay and complexity of admissions.

	Coefficient	p-value
Hospital admission rate for ACSC among dementia patients^*
difference, October 2013	0.87 (0.80–0.95)	0.001
trend before	1.00 (1.00–1.01)	0.378
change in level	0.98 (0.88–1.09)	0.727
change in trend	1.00 (0.99–1.01)	0.771
In-hospital mortality^*
difference, October 2013	0.83 (0.60–1.15)	0.259
trend before	1.01 (0.98–1.03)	0.596
change in level	0.64 (0.42–0.97)	0.037
change in trend	1.01 (0.98–1.04)	0.563
Proportion length of hospital stay >30 days^*
difference, October 2013	1.69 (1.39–2.07)	0.000
trend before	1.00 (0.98–1.01)	0.777
change in level	0.90 (0.69–1.16)	0.413
change in trend	1.01 (1.00–1.03)	0.137
Average length of hospital stay (≤ 30 days)^ꝉ
difference, October 2013	−0.16 (−0.77–0.44)	0.594
trend before	0.05 (0.01–0.09)	0.025
change in level	−0.38 (−1.29–0.54)	0.416
change in trend	−0.06 (−0.12–0.01)	0.031
Complexity of admissions^ꝉ
difference, October 2013	0.18 (0.06–0.30)	0.004
trend before	0.00 (0.00–0.01)	0.509
change in level	−0.02 (−0.14–0.11)	0.811
change in trend	−0.01 (−0.02–0.00)	0.037

Note: *Coefficients are from a Poisson model and multiplicative; ^ꝉ coefficients are from a linear regression model and additive.

In-hospital mortality rate

There was an immediate decrease in in-hospital mortality rates (per 100 dementia patients admitted for ACSC conditions) in Bristol after the introduction of the dementia wellbeing service (change in level x0.64; 95%CI 0.42, 0.97, p = 0.037) compared to the comparator CCGs, although this attenuated over time (Figure 2).

Figure 2.

Monthly in-hospital mortality rates per 100 dementia ACSC admissions for Bristol and 10 similar CCGs, October 2013 - September 2018.Note: vertical red line is our start of intervention (BDWS).

Length of hospital stay

Bristol had a higher proportion of ACSC hospital admissions with a length of stay of more than 30 days in October 2013 (x1.69, [95% CI 1.39, 2.07], p < 0.001) and this continued throughout the study period. There was some indication that the change in slope after the introduction of the DWS was fairly flat in Bristol CCG while the comparator CCGs showed a slight decline (change in trend x1.01, [95% CI 1.00, 1.03], p = 0.14).

Regarding the average length of hospital stay for stays that were 30 days or less, there was some evidence that Bristol had a more upward trend before October 2015 (+0.05 days per month, [95% CI 0.01, 1.09], p = 0.025) with a more downward change in slope than the similar CCGs after October 2015 (−0.06, [95% CI −0.12, −0.01], p = 0.031), although starting from a higher baseline level so the averages converged (Figure 3).

Figure 3.

Mean length of stay for dementia ACSC admissions ≤ 30 days for Bristol and 10 similar CCGs, October 2013 - September 2018.Note: vertical red line is our start of intervention (BDWS).

Complexity of admission

There was an upward trend in average monthly comorbidities for hospital admissions for dementia in both intervention and comparator areas (Figure 4). Bristol CCG had higher comorbidities for admissions in October 2013 (+0.18, [95% CI 0.06, 0.30], p = 0.004). There was some indication of a more downward change in slope in Bristol compared to the comparator areas after the introduction of the DWS (−0.01, [95% CI −0.02, 0.00], p = 0.037). This was a fairly small effect, with comorbidities in Bristol CCG reducing by 0.1 on average every year compared to the comparator CCGs.

Figure 4.

Average monthly comorbidities in terms of numbers of diagnoses listed on dementia admissions for Bristol and 10 similar CCGs, October 2013 - September 2018.Note: vertical red line is our start of intervention (BDWS).

Sensitivity analyses

Sensitivity analyses varying the maximum autocorrelation lag from 1 to 5 months showed the same pattern of significant results for all 5 outcomes, with only slight changes in confidence intervals and p-values, and therefore had the same interpretation. (See Online Supplement Tables S2 and S3).

Discussion

This study found little evidence that a new dementia wellbeing service in Bristol, England, had a major impact on avoidable hospital admissions for people with dementia. In-hospital mortality fell following the introduction of the service compared to control areas, although this appeared to attenuate over the following years. Results pointed to a greater number of very long stays (>30 days) in the intervention area, although average length of stay tended to be lower. People with dementia admitted to hospitals in the intervention area had more comorbidities in October 2013, but the rate of increase in comorbidities appeared to be lower following the introduction of the wellbeing service, with rates similar to those in the comparator areas at the end of the study period.

Our observations align with the findings of a systematic review of non-pharmacological interventions to reduce hospital admissions in people with dementia, which found limited evidence from randomised controlled trials.²⁴ Approaches such as care management by care manager, a nurse or community worker, whose job was to coordinate care of dementia patients were not found to reduce admissions (RR 1.00, 95% CI 0.76, 1.31), although a wide 95% confidence interval means that one cannot rule out a clinically important benefit. The review also did not find evidence of a reduction in mortality (0.91, 95% CI 0.59, 1.40) while there was a suggestion of a modestly reduced average length of stay (−0.16 days, 95% −0.32, 0.01).

Strengths and weaknesses

Our study used a national, longitudinal hospital administrative dataset (HES), which provides standardised information across areas. HES is linked to payments for hospitals, providing a strong incentive towards complete and accurate data for research. The controlled interrupted time series design, with a self-controlled before/after comparison, and a comparator area, reduces the likelihood of findings being due to other factors which would have to occur at the same time as our intervention, and differentially impact the intervention and comparator areas.

The study was an ecological analysis of a new service using non-randomised comparators. This makes causal inference problematic but is superior to a simple before and after comparison. The ITS approach enables stronger inference if the results are consistent with a priori assumptions. It is reasonable to hypothesise that the use of a personal navigator as part of the Bristol DWS might better link primary and secondary care services, allowing families to seek help earlier in the natural history of any illness and potentially avoiding a hospital admission or being admitted earlier with potential reductions in mortality and length of stay.

We observed a reduction in hospital mortality around the time of the intervention, although this might not be a true causal effect for the following reasons. First, it may be a chance finding (type 1 error) due to multiple testing although the p-value is 0.037. Second, if the observed effect was causal, it is likely that the benefits would not have been observed so soon after the introduction of the new service and it should have grown stronger over time as the service matured and was rolled out to more practices and patients. We observed that the reduction in mortality appeared to attenuate over time, which points to a ‘regression to the mean’ effect whereby 2015 happened to be a chance low rate. Third, if the reduction in mortality was due to earlier intervention or referral to primary or emergency care, we would have expected to also see some reduction in length of stay. There was weak evidence for a small reduction in average length of stay after the new service was introduced (change in trend, p = 0.031), but there was also weak evidence for increasing very long stays compared to comparator areas (p = 0.14). These two findings could be linked because very long stays were excluded from our calculation of average length of stay.

We further observed a reduction in the rate of increase in comorbidities over time in Bristol after the introduction of the DWS but not in the comparator CCGs. Comorbidities were assessed on what is recorded at hospital discharge and could be artefactually reduced if hospital doctors or coders based in Bristol CCG were less thorough in reporting comorbidities. While we cannot exclude this explanation, we believe this to be unlikely. It is not clear how the Bristol DWS could have reduced comorbidities as these are usually conditions not dealt with by the navigator other than advising families to seek primary care support. If anything, better access to primary care and navigator support should result in an increase in undiagnosed comorbidities being identified, and those with fewer comorbidities being admitted to hospital. We are not sure how to interpret this observation and it will be interesting to see if this trend continues or reverses with further follow-up.

Our intervention group included GP practices within Bristol CCG. Every GP practice was given a dementia navigator to help people with dementia access services. However, we do not know who accessed the DWS. As our study is an ecological analysis it could include some misallocation. Identification of people with dementia was based on the diagnoses recorded in hospital data, which may be imperfect. Denominator populations were based on GP practice dementia registers, which may also be imperfect, although DWS caseloads from each practice were similar to the numbers on the QOF dementia registers. We excluded GP practices that did not provide six years of dementia register information through the QOF for our denominators. Exclusions were relatively small (9% in Bristol and 13% in the similar CCGs) so would not be expected to have a large impact on the trend analyses.

Our comparator group consisted of ten CCGs with similar demographics to Bristol CCG. All areas have some form of dementia support services in place, most often signposting and support services run by charities such as the Dementia Support Service provided by Age UK in Salford.²⁵ The precise times for commencement of these services in comparator CCGs were difficult to ascertain from publicly-available information, but it is likely that this was before October 2015. For example, Liverpool CCG refer to a ‘dementia care navigator’ service that began in June 2012 and was planned to be extended in April 2015 and again in September 2015.²⁶ None of the services in the comparator CCGs appear to be identical to the Bristol DWS although it is possible that the comparator group had similar changes in service provision occurring at the same time. This could have masked any impact of the DWS in Bristol, but this seems unlikely when averaged over the ten areas.

There may be other factors influencing outcomes which we have not controlled for, for example, other changes in clinical practice that occurred around the time of initiation of the Bristol DWS. We have modelled a sudden transition point in October 2015. However, we are aware that the service began earlier and was rolled out over time until around July 2017; using October 2015 as a transition point appeared to provide an appropriate compromise. Further, our definition of comorbidities could have been impacted by coding changes, but the definitions are equivalent between the areas. There is a possibility that coding changes over time could be different in Bristol compared to other areas. Finally, our analyses do not consider cost effectiveness or the views of patients, carers and clinicians about Bristol DWS.

Study implications

The impact of the Bristol DWS on avoidable hospital admissions for people with dementia appears to be modest, but reducing hospital use is not the sole aim of the service. Families and carers of people with dementia value the service, with survey evidence finding positive feedback to quality reviews of the service, and the majority of those responding to surveys likely or extremely likely to recommend the service every year.^27,28 Qualitative studies in Northern Ireland and the United States have highlighted the benefits of some form of navigator service to help people understand and come to terms with their dementia diagnosis,²⁹ and to reduce caregiver burden for carers of people with dementia.³⁰ Future evaluations of the DWS should consider qualitative work involving people with dementia, their families and carers, as well as those providing the service to better understand the overall value of the service.

Conclusions

Our study did not find a major impact of the Bristol DWS on avoidable hospital admissions among people with dementia. The initial reduction in hospital mortality is not likely to be caused by the service and we cannot explain the reduced complexity of avoidable hospital admissions following the introduction of the service. It is possible that these benefits only occur during a longer follow-up period when the service has been operating for some time. It is also possible that the DWS has improved other aspects of care and patient and carer quality of life, which we have not assessed in this study. More targeted interventions may be required to reduce hospital admissions for people with dementia.

Supplemental Material

Supplemental Material - Interrupted time series evaluation of the impact of a dementia wellbeing service on avoidable hospital admissions for people with dementia in Bristol, England

Supplemental Material for Interrupted time series evaluation of the impact of a dementia wellbeing service on avoidable hospital admissions for people with dementia in Bristol, England by Tim Jones, Maria Theresa Redaniel and Yoav Ben-Shlomo in Journal of Health Services Research & Policy

Footnotes

Acknowledgements

The evaluation reported here would not have been possible without information provided by the Bristol Dementia Wellbeing Service, and the use of the Hospital Episode Statistics via a licence with NHS Digital (NIC 17875): Hospital Episode Statistics, copyright © 2020, re-used with the permission of the Health and Social Care Information Centre (‘NHS Digital’). All rights reserved.

Declaration of conflicting interests

The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: All authors have completed the ICMJE uniform disclosure form at coi_disclosure.pdf and declare: TJ and MTR had financial support from NIHR ARC West for the submitted work; no financial relationships with any organisations that might have an interest in the submitted work in the previous 3 years; no other relationships or activities that could appear to have influenced the submitted work.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institute for Health Research Applied Research Collaborative West (NIHR ARC West; NIHR200181). The views expressed in this article are those of the author(s) and not necessarily those of the NHS, the NIHR, the Department of Health and Social Care, or the Bristol Dementia Wellbeing Service.

Ethical approval

We were provided with routinely collected Hospital Episode Statistics data under licence from NHS Digital (DARS-NIC-17,875-X7K1V). The licence allows us to use the information under Section 261 of the Health and Social Care Act 2012, 2(b) (ii): ‘after taking into account the public interest as well as the interests of the relevant person, considers that it is appropriate for the information to be disseminated’.

ORCID iD

Tim Jones

Supplemental Material

Supplemental material for this article is available online.

References

Prince

Knapp

Guerchet

, et al. Dementia UK: update. London: Alzheimer’s Society, 2014.

Phelan

Borson

Grothaus

, et al. Association of incident dementia with hospitalizations. J Am Med Assoc 2012; 307: 165–172.

Sampson

Blanchard

Jones

, et al. Dementia in the acute hospital: prospective cohort study of prevalence and mortality. Br J Psychiatry 2009; 195: 61–66.

Dementia Wellbeing Service . What is the dementia wellbeing service? https://www.bristoldementiawellbeing.org/Use-our-service/What-is-the-Dementia-Wellbeing-Service, (2017, accessed 10 February 2023).

Bristol Dementia Partnership . Bristol dementia wellbeing service quality account 2015/16 Bristol, https://www.bristoldementiawellbeing.org/SiteUploads/86/Uploads/DWS_Quality_Accounts_final.pdf (2016, accessed 10 February 2023).

NHS England . Friends and family test 2020, https://www.nhs.uk/using-the-nhs/about-the-nhs/friends-and-family-test-fft/ (2020, accessed 10 February 2023).

Benchimol

Smeeth

Guttmann

, et al. The REporting of studies conducted using observational routinely-collected health data (RECORD) statement. Plos Med 2015; 12: e1001885.

Herbert

Wijlaars

Zylbersztejn

, et al. Data resource profile: hospital episode statistics admitted patient care (HES APC). Int J Epidemiol 2017; 46: 1093i-1093i.

NHS Digital . QOF 2016-17 Recorded disease prevalence, achievements and exceptions Mental health and neurology group, GP practice level, https://digital.nhs.uk/data-and-information/publications/statistical/quality-and-outcomes-framework-achievement-prevalence-and-exceptions-data/quality-and-outcomes-framework-qof-2016-17 (2017, accessed 10 February 2023).

10.

NHS Digital . QOF 2015-16 Recorded disease prevalence, achievements and exceptions Mental health and neurology group, GP practice level, https://digital.nhs.uk/data-and-information/publications/statistical/quality-and-outcomes-framework-achievement-prevalence-and-exceptions-data/quality-and-outcomes-framework-qof-2015-16 (2016, accessed 10 February 2023).

11.

NHS Digital . QOF 2014-15 Recorded disease prevalence, achievements and exceptions Mental health and neurology group, GP practice level, https://digital.nhs.uk/data-and-information/publications/statistical/quality-and-outcomes-framework-achievement-prevalence-and-exceptions-data/quality-and-outcomes-framework-qof-2014-15 (2015, accessed 10 February 2023).

12.

NHS England . Similar CCG explorer tool, https://www.england.nhs.uk/publication/similar-10-ccg-explorer-tool/ (2016, accessed 10 February 2023).

13.

Arabadzhyan

Castelli

Gaughan

, et al. A comparison of NHS Digital and CHE algorithms for spells and CIPS construction, https://www.york.ac.uk/media/che/documents/papers/researchpapers/CHERP182_NHS_update2018_2019_supplementary.pdf ( 2021, accessed 10 February 2023 ).

14.

Purdy

Griffin

Salisbury

, et al. Ambulatory care sensitive conditions: terminology and disease coding need to be more specific to aid policy makers and clinicians. Public Health 2009; 123: 169–173.

15.

NHS Ambulatory Emergency Care Network . Directory of emergency ambulatory care for Adults: NHS elect, https://www.ambulatoryemergencycare.org.uk/uploads/files/1/AEC-Directory_6th_edition_February_2018.pdf (2018, accessed 10 February 2023).

16.

Charlson

Pompei

Ales

, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987; 40: 373–383.

17.

Quan

Sundararajan

Halfon

, et al. Coding algorithms for defining comorbidities in ICD-9-CM and ICD-10 administrative data. Med Care 2005; 43: 1130–1139.

18.

Bernal

Cummins

Gasparrini

. Interrupted time series regression for the evaluation of public health interventions: a tutorial. Int J Epidemiol 2017; 46: 348–355.

19.

Bhaskaran

Gasparrini

Hajat

, et al. Time series regression studies in environmental epidemiology. Int J Epidemiol 2013; 42: 1187–1195.

20.

Bottomley

Scott

JAG

Isham

. Analysing interrupted time series with a control. Epidemiol Methods 2019; 8: 20180010.

21.

Dorward

Khubone

Gate

, et al. The impact of the COVID-19 lockdown on HIV care in 65 South African primary care clinics: an interrupted time series analysis. Lancet HIV 2021;8:e158-e165.

22.

Linden

. Combining synthetic controls and interrupted time series analysis to improve causal inference in program evaluation. J Eval Clin Pract 2018; 24: 447–453.

23.

Jones

. Dementia Wellbeing github archive, https://github.com/jonestim2002/dementia_wellbeing (2022, accessed 10 February 2023).

24.

Packer

Ben Shlomo

Whiting

. Can non-pharmacological interventions reduce hospital admissions in people with dementia? A systematic review. Plos One 2019; 14: e0223717.

25.

Age UK . Dementia support service, https://www.ageuk.org.uk/salford/our-services/dementia-support-service-4c1978bd-46d8-ec11-bea2-00155d806b13/ (2022, accessed 10 February 2023).

26.

Liverpool Clinical Commissioning Group . Liverpool commissioning strategy for dementia (FINAL). Liverpool: Liverpool Clinical Commissioning Group and Liverpool City Council, 2014.

27.

Devon partnership Trust, Alzheimer’s Society . Bristol Dementia Wellbeing Service Quality Account 2019/20, https://www.bristoldementiawellbeing.org/SiteUploads/86/Uploads/[sm]_DWS_Quality_Accounts_2019_20_final.pdf (2020, accessed 10 February 2023).

28.

Devon Partnership TrustSociety As . Bristol dementia wellbeing service quality account 2020/21, 2021, https://bristoldementiawellbeing.org/SiteUploads/86/Uploads/DWS_Quality_Account_2020-21_v2.pdf (2021, accessed 10 February 2023).

29.

Hagan

. What next? Experiences of social support and signposting after a diagnosis of dementia. Health Soc Care Community 2020; 28: 1170–1179.

30.

Bernstein

Merrilees

Dulaney

, et al. Using care navigation to address caregiver burden in dementia: a qualitative case study analysis. Alzheimers Dement 2020; 6: e12010.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

0.70 MB