Using the Primary care Academic CollaboraTive to explore the characteristics and healthcare use of older housebound patients in England: protocol for a retrospective observational study and clinician survey (the CHiP study)

Background Older housebound people are an under-researched group for whom achieving good primary health care can be resource intensive. Aims To describe the characteristics and healthcare use of older (≥65 years) housebound people; explore clinician views on delivery of care to housebound people; and assess the feasibility of using a new network of healthcare professionals to deliver high quality research. Design & setting Retrospective observational study of electronic GP records and clinician survey in England. Method Clinical members of a new UK research network called the Primary care Academic CollaboraTive (PACT) will collect the data. For part A, around 20 GP practices will be recruited and clinicians will identify 20 housebound and 20 non-housebound people, matched by age and gender (around 400 total in each group). Anonymised data will be collected on characteristics (age, gender, ethnicity, deprivation decile), long-term conditions, prescribed medicines, quality of healthcare (via Quality Outcomes Framework targets), and continuity of care. Reports with benchmarked practice-level data will be provided to practices to identify areas for quality improvement and to enhance engagement. For part B, 2–4 clinicians will be recruited from around 50 practices in England (around 150 clinicians) to complete a survey about delivery of healthcare for housebound people. For part C, data will be collected to assess the feasibility of using the PACT network to deliver primary care research. Conclusion Older housebound people are a neglected group both in terms of research and clinical care. Understanding the characteristics and use of primary healthcare of housebound people will help identify how to improve their care.


Background
Older housebound people are an under researched group for whom achieving good primary healthcare can be resource intensive.

Aims
To describe the characteristics and healthcare use of older (≥65 years) housebound people; explore clinician views on delivery of care to housebound people; and assess the feasibility of using a new network of healthcare professionals to deliver high quality research.

Design & Setting
Retrospective observational study of electronic GP records and clinician survey in England.

Method
Clinical members of a new UK research network called the Primary care Academic CollaboraTive (PACT) will collect the data. For part A, ̴ 20 GP practices will be recruited and clinicians will identify 20 housebound and 20 non-housebound people, matched by age and gender ( ̴ 400 total in each group). Anonymised data will be collected on characteristics (age, gender, ethnicity, deprivation decile), long-term conditions, prescribed medicines, quality of healthcare (via Quality Outcomes Framework targets) and continuity of care. Reports with benchmarked practice-level data will be provided to practices to identify areas for quality improvement and to enhance engagement. For part B, 2-4 clinicians will be recruited from 50 practices in England ( ̴ 150 clinicians) to complete a survey about delivery of healthcare for housebound people. For part C, data will be collected to assess the feasibility of using the PACT network to deliver primary care research.

Conclusion
Older housebound people are a neglected group both in terms of research and clinical care.
Understanding the characteristics and use of primary healthcare of housebound people will help identify how to improve their care.

Introduction
Housebound people (defined as those unable to attend the GP surgery, visited by a primary healthcare professional (HCP) at home) are an under researched group. 1 Research from the 1990s found that around 20% of people over 85 years were housebound 2 3 (now around 340,000 people within this age group in the UK). 4 Home visits are time consuming and, due to high workload pressures, acute problems are possibly being prioritised over proactive care of long-term conditions, such as optimising medicines and advance care planning.
Unmet healthcare needs (the difference between healthcare that is deemed necessary and the actual care received) 5 depend on the characteristics of patients seeking healthcare (e.g. sociodemographics) and the services available to them. 2 A study in France found that unmet healthcare needs were strongly associated with being housebound. 6 A qualitative study of marginalised UK communities found that housebound participants experienced significant problems accessing healthcare. 7 Problematic polypharmacy, where the potential harm from medicines outweighs the potential benefit, is one example of unmet healthcare need. 8 Medicine reviews offer an opportunity to improve prescribing safety by rationalising prescribing, simplifying regimens, and stopping or tapering down non-essential and potentially harmful medicines. 9 GPs and pharmacists report difficulty in stopping medicines without involving the patient. 10 GP home visits are generally focused on acute problems, and home-based pharmacist reviews are seldom provided, likely resulting in suboptimal medicine reviews for older housebound people. 11 Better understanding the healthcare needs of the UK housebound population is essential for improving healthcare services. Undertaking such research may benefit from novel distributed strategies to improve reach and recruitment.

Aims
The primary aim of the study is to describe the characteristics and primary healthcare use of older housebound people and to understand how they differ from non-housebound people.
The primary aim this study is broken down into the following specific objectives: 1. To examine the characteristics and long-term conditions of older (≥65 years) housebound and non-housebound patients, including: • socio-demographic characteristics (e.g. age, gender, ethnicity, deprivation decile); • long-term conditions, including morbidity count and Cambridge Multimorbidity Score; 12 • frailty (Rockwood score 13 ) 2. To compare healthcare utilisation of older (≥65 years) housebound and non-housebound patients, including: • prescribed medicines in the preceding 12-month period, including potentially problematic prescribing; • continuity of care; • quality of care; • number and type of in-hours and out-of-hours GP appointments; • and use of other healthcare services.
3. To explore clinician views on the delivery of care to housebound patients, including: • how care of housebound patients is organised and delivered; • and views on delivery of healthcare.
The secondary aim of the study is to assess the feasibility of using a new primary care network of trainees and HCPs (PACT) to carry out high quality research, including: • to describe the number of PACT members who are interested and, of those selected, the proportion that complete data collection; • to describe the characteristics of GP practices and PACT members; • to test whether valid high-quality data can be collected using the PACT model; • to assess whether the PACT model introduces active participation in research to previously research naïve practices and staff.

Method
What is PACT?
PACT stands for the Primary care Academic CollaboraTive -a new initiative, which aims to build the capacity of UK academic primary care through engagement of trainees, GPs and other HCPs (hereinafter PACT members) in high-quality research focused on improving patient care. 14 PACT members working in GP practices across the UK will collectively take part in projects. This innovative model of research will allow researchers to include patients who are difficult to identify by other research methods (e.g. via Clinical Practice Research Datalink), such as housebound patients, and to create bespoke datasets by combining data extracted from the GP records, patient surveys and practice workforce surveys. As PACT members are clinically trained, they can carry out detailed case note reviews of the GP records, extracting information that is not captured in routinely collected coded data.

Recruitment
The study will be advertised via the Clinical Research Network; PACT network (HCPs who have registered an interest in taking part in PACT projects); deaneries; social media; GP teaching newsletters; RCGP Associates in Training network; and the research team's local networks. We will purposefully reach out to those working in more deprived and rural areas. To register an interest, PACT members (who will collect data for the study) will complete an online expression of interest form, and a GP partner or practice manager will complete a practice agreement form.
For part A (observational study), ̴ 20 GP practices using EMIS (electronic GP records) will be invited to take part from six Clinical Research Networks (CRNs) (Box 1). For part B (clinician survey), ̴ 50 GP practices from across England will be invited to take part. We will endeavour to select practices serving a range of socio-demographic populations (e.g. urban/rural, deprived/affluent and different ethnic groups).

Box 1: Eligibility criteria for GP practices for observational study (part A)
-GP practice within the East Midlands, North East and North Cumbria, North West Coast, South West Peninsula, Yorkshire and Humber, and West of England CRNs -GP practice uses EMIS software (electronic GP record) -One or more foundation doctors, GP trainees or clinicians interested and available to collect data for the study (four half days of work) -Practice delivered home visits in April, May and June 2022 -Written consent from a GP partner or practice manager giving permission for the practice to take part in the study and for the PACT member to use four half-days of protected study time for the project -GP and/or nurse available to supervise the PACT member (two hours total, cost reimbursed via the CRN service support costs, Appendix 1) -Practice administration team available to support the PACT member with administrative tasks (three hours total, cost reimbursed via the CRN service support costs, Appendix 1)

Data collection Part A (observational study)
Data collection will be piloted in a small number of practices prior to scaling up. Detailed standard operating procedures (SOP) and 'Frequently Asked Questions' will be provided to PACT members.
The study team will troubleshoot via email, phone and videoconference calls. PACT members will complete online Good Clinical Practice training (three hours of online research training).
PACT members will identify 20 housebound (who have had at least one home visit) and 20 nonhousebound people (who have attended the GP surgery with no record of a home visit) aged ≥65 years with a face-to-face consultation in April to June 2022. Nursing home or residential home residents will be excluded. A pre-prepared electronic search will identify potentially eligible patients, using 'home visit requested', 'home visit' appointment slot and 'home visit' codes for housebound patients and 'GP surgery' codes for non-housebound patients (Box 2).

Power calculation for part A
To estimate the difference between housebound and non-housebound people in morbidity count, we used the results of a study in the United States (7603 adults ≥65 years): 15 morbidity count 2.5 in and 3.9 for non-housebound and housebound participants respectively (standard deviation 1.5). If we collect data on 200 housebound and 200 non-housebound people, this magnitude of difference could be detected with more than 90% power at the 5% significance level. If we collect data on 400 cases and 400 controls (our target), we will detect smaller differences with greater precision.

Part B (clinician survey)
PACT members will complete a short online survey about organisation of healthcare for housebound patients, including the workforce involved in home visits, the types of healthcare offered to housebound patients (urgent/non-urgent), initiatives for and coding of housebound (Supplementary

Part C (feasibility study)
We will collect data about the characteristics of PACT members who express an interest in taking . To test whether high-quality data can be collected using the PACT model, data quality exercises will be completed in ̴ 5 practices, whereby a second PACT member will independently check patient eligibility and manually extract healthcare use data.

Analysis plan Part A (observational study)
A descriptive analysis of the socio-demographic characteristics, long-term conditions and prescribed medicines will be completed for housebound and non-housebound patients.
Regression analyses will be used to compare the characteristics and healthcare utilisation of housebound and non-housebound patients, by: • examining the socio-demographic characteristics, long-term conditions and frailty of older housebound and non-housebound people (outcome: housebound status); and • comparing healthcare use of older housebound and non-housebound people using key outcome variables such as prescribed medicines in the preceding 12-month period, continuity of care and number of in-hours GP appointments (outcome: healthcare use).

Box 4: Summary of planned regression analyses for part A (observational study)
Aim: to examine the characteristics and long-term conditions of older (≥65 years) housebound and non-housebound patients.
Aim: To compare healthcare utilisation of older (≥65 years) housebound and non-housebound patients Key outcome variables: • prescribed medications in the preceding 12-month period (e.g. total number of medications, rates of polypharmacy (≥5 or ≥10 medications), anticholinergic burden score 17 and medication regimen complexity index); 18 • continuity of care; 19 • number of in-hours GP appointments (phone calls, home visits, video consultations, GP surgery visits); Exposure variables: • housebound or non-housebound Covariates: • age, gender, deprivation level • long-term conditions • frailty • GP practice

Part B (clinician survey)
A descriptive analysis will examine the clinician survey data. Mean and standard deviation will be used to describe distribution for normally distributed data; medium, IQR for non-normally distributed data. A subgroup analysis will be conducted for the following groups: rural/urban practices, deprived/affluent, practice list size.

Part C (feasibility study)
Using a flow diagram, we will examine the number of GP practices that expressed an interest, were invited and completed the project. A descriptive analysis will examine the characteristics of PACT members (e.g. age, gender, ethnicity, stage in training etc) and their GP practices (e.g. list size, deprivation decile, ethnicity estimates, urban/rural etc); the reasons they chose to participate; and their experience of taking part (Supplementary File 6). A subgroup analysis will be carried out for the following groups: research active/inactive practices, practice list size, urban/rural, practice deprivation level. Error rates will be calculated using data from the data quality exercise.

Summary
This study will use a novel primary care research collaborative to explore the characteristics and primary healthcare use of older housebound people, and clinician views on delivery of healthcare.

Strengths and limitations
This is one of two exemplar studies to test the feasibility of using a network of trainees and primary HCPs to conduct high quality research. 20 Using clinical staff to extract otherwise difficult to reach data is a key proposed benefit of this model and evaluating how this works in practice is a vital part of the data collection. Due to inconsistencies in coding of housebound status, housebound patients are difficult to identify using routinely collected data. This study will address this issue as clinicians will review the electronic GP records to check eligibility. A possible limitation is that some PACT members will have little research experience. Detailed instructions and support will be provided.
Furthermore, the data quality exercise will provide evidence about the reliability and reproducibility of the data. A final limitation is that, despite collecting a broad range of healthcare use data, we are unable to collect data about healthcare provided by the district nursing team.
Implications for research and/or practice At a local level, PACT members will be provided with practice-level data benchmarked against other practices and will be encouraged to discuss the data with the practice team to identify areas for quality improvement, such as interventions to optimise coding of housebound patients and to reduce high risk prescribing.
Nationally, this study will answer important questions about healthcare use for older housebound patients. Manually extracted data (e.g. from hospital letters) will be combined with coded data (e.g. long-term conditions and medicines), to create a rich and bespoke dataset. Data collected may inform changes to clinical practice and ideas for future research aimed at reducing unmet healthcare need in older housebound patients.