Research methods

Overview

In this chapter we provide an overview of our study design, and we outline the quantitative and qualitative data and methods used (note that detailed methods are provided in Chapters 3–6).

Study design

This study used a mixed-methods design, with findings from planned qualitative and quantitative analyses informing further investigation and analysis throughout the study (see Chapter 1, Structure of the report and Figure 2).45 The case study method permits development and testing of theories on how efforts to bring about change interact with the context in which they take place. A multiple case study approach, in this case, in which each HASU is a case study, allows the development and testing of theories in several organisational contexts.46–48 Our large qualitative data set was derived from interviews with HASU staff and patients, and observations on the ward and elsewhere in the eight hospitals. After gaining access to the SSNAP data, we looked at temporal variation (objectives 1 and 2) in the processes and clinical outcomes of HASU care for a 1-year period (from 1 January 2014 to 31 December 2014) (see Chapter 3). Chapter 4 is an analysis of the reasons for the patterns of variation found in our SSNAP data set, derived from dialogues between the quantitative and the qualitative research teams in the study (objective 3). Our third study was informed, in part, by qualitative data, with interviews and observations suggesting that nursing seniority and variation in providing key roles (such as nurse in charge) was perceived to be more important than the number of nurses on the ward (objectives 1–3). Therefore, we examined how nursing roles were provided through different time periods, and the impact on clinical outcomes and processes (see Chapter 5). Chapter 6 reflects the development of our conceptual thinking as a mixed-methods team throughout this project (objectives 3 and 4). We have recognised the limitations of using the SSNAP to assess temporal variation, leading to a qualitative analysis of the forces or structures driving temporal variation, and how this affects measurement of variation in stroke research. Our methods are reported in accordance with the COnsolidated criteria for REporting Qualitative studies (COREQ) (see Appendix 1) and STrengthening the Reporting of OBservational studies in Epidemiology (STROBE) reporting guidelines.49

FIGURE 2 Interlinkages of the mixed-methods analysis, including deviation from the protocol. (Figure not reproduced here; see the original article.)

Perspective

We approached this research with a commitment to the study of the HASU system from both a quantitative and a qualitative perspective. Although not an ethnographic study, we adopted some principles of multisited ethnography, such as following people and objects through systems.50 For example, we based some of our observations around the journey being followed by a stroke patient all the way from the emergency room, through scans and tests, and up to the HASU ward itself. We followed members of staff, such as consultants on their ward rounds and nurses around the hospital. We also focused on relative cultural meanings, such as what 24/7 means from a variety of professional and patient perspectives. We committed to be reflexive and iterative in our approach to data collection, regularly meeting to debrief and make decisions about future plans for data collection, and quantitative and qualitative analysis.51 Finally, we sought to understand both our qualitative and our quantitative data through our immersion in the services. This led to new additional quantitative analyses being undertaken, such as our analysis of HASU nursing data.

Setting

Our study was set in all eight of the London HASUs. These are dedicated wards, often sited in close proximity to relevant specialties, such as neurology, stroke wards, gerontology, etc. The wards range in size from 18 to 50 beds. We did HASU-specific profiles, noting bed numbers, shift hours, access to scans (e.g. magnetic resonance imaging) (see Tables 11 and 18). In addition to the ward, the HASU teams work in the ED to assess and treat incoming patients. This also includes moving them to scanning facilities and working with radiologists to interpret CT scans. Some patients are either assessed or temporarily situated in other wards when the HASU has inadequate capacity; these are known as ‘outliers’. Therefore, our observational work occurred in all of these locations, but not outside the hospital.

The London HASUs are in the following hospitals:

• Charing Cross Hospital
• The Royal London Hospital
• King’s College Hospital
• St George’s Hospital
• Northwick Park Hospital
• Princess Royal University Hospital
• Queen’s Hospital
• University College London Hospital.

Ethics approval

The study received approval from the National Research Ethics Service Committee London, Westminster Branch (reference number 14/LO/1355) and local approvals at each participating trust.

Quantitative data collection

Sentinel Stroke National Audit Programme

We obtained anonymised patient-level data from the SSNAP for all patients treated in London HASUs with a primary diagnosis of stroke (ischaemic stroke or primary intracerebral haemorrhage) between 1 January 2014 and 31 December 2014. The SSNAP collects data on clinical characteristics, care quality (from the time of admission up to 6 months after stroke) and outcomes for all stroke patients admitted to acute care hospitals in England. Data are entered into the SSNAP database by local clinical teams via a secure web interface. During our study period, the case ascertainment in the SSNAP, which is calculated as the proportion of all acute stroke patients admitted to hospitals, was estimated to be 90% for England.52

Primary outcomes: health outcome measures

Health outcome measures were (1) whether or not the patient died within 3 days of admission to the hospital and (2) functional outcome, measured by a mRS53 score of 3–6 (moderate, moderately severe, severe disability or death) at the end of the inpatient spell. Mortality data in the SSNAP are not available after hospital discharge; therefore, we measured mortality up to 3 days after admission to minimise the number of missed deaths. LOS was measured in the HASU and total LOS in hospital.

Secondary outcomes: quality-of-care measures

Quality-of-care indicators were measured from time of hospital admission or onset of stroke symptoms (the latter used for those who were already in hospital when they had their stroke) (Box 1).

Brain imaging within 1 houra,b and within 12 hours.a,b

Admission to a stroke unit within 4 hours.b

Dysphagia screen within 4 hours.a,b

Assessment by a nurse trained in stroke management within 24 hours.a,b

Administration of intravenous thrombolysis to eligible patients.a,b

Door-to-needle time < 1 hour in patients receiving thrombolysis.a,b

Assessment by a stroke specialist consultant physician within 12 hours and within 24 hours.b

Assessments by a physiotherapist, occupational therapist and SLT within 24 hours and within 72 hours.b

We also included measures with more stringent time constraints to reflect the time-critical nature of acute stroke care.

The denominators used for each measure were consistent with the SSNAP key indicators.54 Most outcomes were measured for all patients, but there were exceptions: patients who were medically unwell or refused to be screened were excluded from the dysphagia screen measure; patients with only ischaemic stroke who met the Royal College of Physicians’ guideline minimum threshold for thrombolysis were included in the thrombolysis rate; door-to-needle times included only those patients who received thrombolysis; patients who were medically unwell, refused to be assessed or had no relevant deficit were excluded from the therapist measures.

To inform our cost analysis we also created alternative versions of some of these variables that were not time limited, that is whether or not the patient:

• had brain imaging
• was admitted to a stroke unit
• had dysphagia screening
• was assessed by a nurse trained in stroke management
• was assessed by a stroke specialist consultant physician
• was assessed by a physiotherapist, occupational therapist or speech and language therapist (SLT).

Independent variables

In-hours and out-of-hours measures

To examine variations across the week, we initially used a flexible specification of time of admission, measured in six 4-hour periods, 00.00 (midnight) to 03.59, 04.00 to 07.59, 08.00 to 11.59, 12.00 (noon) to 15.59, 16.00 to 19.59 and 20.00 to 23.59, every day of the week (42 periods). We also created a more restrictive measure to examine broad trends across the week: Monday to Friday 08.00 to 19.59, Monday to Friday 20.00 to 07.59, Saturday and Sunday 08.00 to 19.59, and Saturday and Sunday 20.00 to 07.59 (four periods). These specifications were also used by Bray et al.22

Patient characteristics

In every model we controlled for sex, age (continuous variable), ethnic group (six categories), type of stroke (infarction or primary intracerebral haemorrhage), comorbidities prior to admission, mRS before stroke (0–2, 3–5), level of consciousness on arrival at the hospital (four categories), method of admission to the hospital (three categories), time from onset of stroke symptoms to admission (four categories), month of admission (12 categories) and stroke unit. When analysing mRS scores of 0–2 compared with scores of 3–5 at the end of the inpatient spell, we additionally controlled for the number of days after admission at which the mRS score was measured.

We also computed data from the SSNAP on the number of patients arriving at the hospital in each time period.

Nurse staffing data

Comprehensive data on nurse staffing were collected from nursing rotas from each of the London HASUs for every time of the day across the 12-month study period (between 1 January and 31 December 2014). Nurse staffing measures comprised the band of the nurse in charge [bands 5–8 and registered bank (or agency) nurse] and whether the nurse in charge was also responsible for administering intravenous thrombolysis or whether there was an additional nurse to perform this activity. We also collected data on the total number of specialised stroke nurse staffing on duty in the HASU by band (as above), the number of beds in each HASU and the times of the nursing shift handover period. These data were combined with the SSNAP data set based on the time of arrival at the hospital, so that for every patient who was admitted to a London HASU during the study period we also knew the characteristics of the nurse staffing at the time that they were admitted to the hospital. Note that we were unable to collect medical staffing data and so did account for this in our analyses (see Chapter 5).

Qualitative data collection

Sampling

Staff were sampled purposively to ensure coverage of a range of professional roles found in the HASUs (Table 2), including medical, nursing, therapy, administrative and managerial staff. Patients were also recruited purposively to reflect variations in sex, age, stroke type, discharge destination and carer perspectives (Table 3). The recruitment procedure was as follows: HASU staff provided researchers with a list of eligible patients in the HASU or stroke unit previously admitted to the HASU (see Recruitment and consent). Patients were introduced by clinical staff and given study information to consider. They were told that the interviewers were not members of current clinical staff and that their participation would not affect their care. Full and informed consent was obtained at the time of the interview. Interviews were conducted at participants’ homes, public places or clinical settings, according to their preference. The interviews were conducted between 6 and 99 days after admission to the HASU (average 38 days, median 33 days).

Recruitment and consent

To analyse the organisation of all eight London HASUs during the day and during the night, on weekdays and at weekends, authors GB, ABL, JE and PX conducted 76 interviews with HASU staff and 31 interviews with stroke patients and carers, and undertook 41 non-participant observations of HASU activity over approximately 85 hours. JE also gathered data through informal discussion on staff numbers and seniority on the ward, the composition of staff attending suspected stroke calls in the emergency room and the availability of scans across the week from each site. All interviewers were female and employed as research staff on the project. GB, ABL and PX are postdoctoral researchers with significant interviewing experience and no prior experience of working in stroke. JE was a stroke co-ordinator in a HASU prior to joining the project, with an occupational therapy background. She received research interview and analysis training and feedback from the rest of the team. Qualitative work was overseen by NJF and AIGR, who are experienced in both qualitative research and the London stroke system through previous work.

Patients were approached if they had a provisional or tentative stroke diagnosis and experience of receiving care in the HASU. We excluded patients and carers unable to sufficiently understand spoken or written English, and those who were unable to give informed consent by written or oral methods. Carers were interviewed alone or alongside patients, if patients requested it, and when the patient suffered from severe cognitive impairment or speech difficulty as well as when a patient had died during, or after, admission to a HASU.

The staff interviews were conducted according to a semistructured topic guide. The guide comprised questions about a typical day (see Report Supplementary Material 1), the perceived differences in and attitudes towards working in and out of hours, team working and the HASU environment. The patient interviews followed a broadly biographical topic guide, asking about their journey up to, during and after their HASU stay. We also prompted them to reflect on their experiences in and out of hours. All participants were interviewed only once and interviews were recorded digitally.

Observations

Global written consent for observations was gained initially from service leads and from a selection of staff at team meetings. Thereafter, verbal consent was given by staff members for researchers to make observations of staff meetings and of the stroke care pathway.

Non-participant observations were conducted at least four times at each HASU site, two visits on weekdays, one in the evening during the week and one at the weekend, with some extra visits to certain HASUs to confirm or add to our findings. We aimed to collect information about two dimensions of HASU service organisation and delivery, how (1) a patient progresses through the stroke pathway, assessment in the ED, care on the HASU and HASU repatriation; and (2) HASU staff organise themselves to meet the London specifications in day-to-day delivery operations [e.g. thrombolysis for at least 90% of eligible patients, timely brain scans, direct admission to HASU within 4 hours, daily multidisciplinary team (MDT) meetings]. This research design drew on ongoing research to identify likely service elements and factors influencing the design and implementation of HASU care (Table 4).

Modes of analysis

Quantitative analysis

We investigated variations in quality of acute stroke care and outcomes by day and time of admission in London HASUs and evaluated whether or not nurse staffing affected 24/7 access to high-quality acute stroke care in eight London HASUs. For the former, we ran patient-level regressions, regressing each measure against time period of admission. To inform our cost analysis we also tabulated the non-time-dependent measures described above against time period of admission. For the latter, we also included nurse staffing measures in models for London HASUs. In every model we controlled for baseline patient characteristics (sex, age and ethnic group) and case mix (type of stroke, comorbidities prior to admission, mRS before stroke, level of consciousness on arrival at the hospital, method of admission to the hospital, time from onset of stroke symptoms to admission and month of admission). We tested for statistically significant variations using Wald tests and likelihood ratio tests.

Qualitative analysis

The investigation was driven by understanding how London HASUs organise and deliver 24/7 stroke care in practice, how these service patterns have been influenced by workforce and organisational factors, and how stakeholders, including staff, patients and carers, experience these services.

The qualitative analysis drew together interview and observation data using a composite, thematic analysis across HASUs, as well as a comparative case study approach. The thematic approach55 allowed for developing an understanding of the central issues HASUs, as a particular type of service delivery model, face when delivering care in acute trust settings across 24/7 time scales. The comparative case study approach46–48 illuminated similarities and differences between HASUs and across professional groupings.

We used an iterative approach to analysis, following a five-step process to generate thematic and case study insights.

Step 1: data familiarisation

Data were compressed into workable, standardised forms (i.e. transcription formatting) using several initial mapping methods, allowing team members to become familiar with the types of data being collected (e.g. producing tables showing distribution of participants, making lists of information gathered through observations, key components of the service). These were adapted and refined as we became more familiar with the data. Simultaneously, we filled some of the gaps through ongoing data collection.

Step 2: initial framework coding guided by protocol

Three large groupings based on the central objectives of the study were developed as a way to initially sift through data focused by the protocol questions. A handful of patient, carer and staff transcripts were read by each of the four researcher team members (approximately five per researcher), with the groupings then applied and relevant data extracted.

Step 3: initial coding and framework development

The qualitative team met and discussed central issues emerging under the three-capture category rubric. Together, we developed a preliminary framework of codes reflecting central ideas, grouped into seven categories. Each category was defined along with the assigned codes. Codes then guided the extraction of relevant pieces of data in the interviews and observations. Code extracts were sorted by HASU to begin to identify differences by HASU and professional grouping.

Step 4: refining codes and coding framework

The research team met regularly to discuss extracts and codes to both refine and deepen our understanding of ideas emerging from the data. Some ideas led to new or redefined codes and the collapsing of multiple similar codes into one code. It also began the process of prioritising the most important of the codes, getting to the heart of the study’s driving research questions.

The team regularly met with the overall team (principal investigator, advisors and quantitative team) to feed back learning and solicit insights to further develop analytical threads.

Step 5: ‘headlines’ and written memos

The research team moved from the refined codes in iterative fashion to linking them into arguments, combining pieces of evidence to generate convincing answers to the driving research questions. In this process, we elaborated on the comparative case studies by professional groups and HASUs.

Synthesis of approaches

We used our monthly project meetings to bring together observations, queries and suggestions between the qualitative and the quantitative researchers. Interactions centred on key ideas such as:

• Can we explain why certain variables do and do not show temporal variation?
• What are the best ways to capture temporal variation?
• What factors would we expect to influence temporal variation?
• What further data can we gather or analyse to expand our findings?

Patient and public involvement

Two stroke patient representatives contributed to development of our study protocol and research questions. Over the course of the study patient representatives participated in three Study Steering Committee meetings (in June 2014, June 2015 and July 2016), in which they participated in discussions of interim findings and ongoing development of the study. Lay summaries of our work were produced for each meeting and representatives were offered individual meetings to discuss matters alone before the main Study Steering Committee meeting. Our patient representatives raised issues related to staff handover and confirmed the importance of the interface between hospital and social services, which we incorporated into our analysis.