PERFECTED enhanced recovery (PERFECT-ER) care versus standard acute care for patients admitted to acute settings with hip fracture identified as experiencing confusion: study protocol for a feasibility cluster randomized controlled trial

The aim of this study is to test trial and intervention feasibility to inform a definitive trial. The current study therefore has the following research objectives:

This multicentre, interventional, feasibility study is an open trial, which will be cluster randomised. In line with MRC guidance for complex interventions [20, 21], an integrated multimethod multi-perspective (patients, suitable informants (defined as individuals with regular contact who are prepared to complete proxy measures) and NHS professionals) process evaluation will be conducted. This will explore variations in implementation fidelity, dose and reach. Trial hospitals will be randomised to active or control arm within geographical region using a simple randomisation process. An ad hoc programme will be written in SAS to carry out this procedure (see Fig. 1 Randomization overview). The nature of PERFECT-ER means that patients, suitable informants and staff delivering treatments cannot be masked to the trial arm. Accordingly, this is an open unblinded trial. Statistical analysis will be undertaken by a subgroup-blind approach, blinded to trial arm and cluster.

The study will be conducted in acute hospital wards in ten NHS hospitals located in five different UK regions (England: four regions, eight hospitals; Scotland: one region, two hospitals) to which individuals who sustain proximal hip fracture are admitted. Each hospital will contribute one study ward. The unit of randomisation is the cluster, in this case, the hospital site. The following site inclusion and exclusion criteria have been identified.

Site inclusion criteria:

Site exclusion criteria: sites that have participated in the “PERFECTED WP2: Implementing optimised hospital care” research programme leading to the development and refinement of PERFECT-ER will be excluded.

Because of the cognitively vulnerable nature of the patient population, proxy measures will be used alongside patient-generated measures. Consent will therefore be sought from patients where possible and from “suitable informants” as defined subsequently. The eligibility criteria for patient participants and suitable informant participants are as follows.

Patient inclusion criteria:

Patient exclusion criteria:

Suitable informant inclusion criteria:

Suitable informant exclusion criteria: Individual not over 16 years of age.

PERFECT-ER is a multicomponent service improvement intervention with a systematic implementation process. It comprises of the PERFECT-ER checklist, which synthesizes best practice evidence of hospital-based dementia care delivery with current best practice evidence of the admission, preoperative, postoperative, rehabilitation and discharge management of hip fractures, agents of change (described below) and a model for change (plan-do-study-act). The checklist, consisting of patient-level and organizational-level care elements, is designed to overlay existing hip fracture pathways highlighting areas to improve care delivery to patients experiencing confusion. PERFECT-ER has been developed, piloted and refined in earlier PERFECTED work packages in partnership with a range of stakeholders in diverse development sites. As the current paper is reporting the protocol for a feasibility study the intervention is understandably not yet in the public domain.

Sites allocated to the active arm will be given PERFECT-ER 3 months before recruitment begins and resourced to support the implementation of PERFECT-ER using the NHS Service Improvement (plan-do-study-act) methodology [24]. The resource is in the form of an internal chance agent, known in this study as a Service Improvement Lead (SIL). Following the 3-month implementation period led by the SIL (and provided that individual active sites have reached predesignated PERFECT-ER adherence scores) sites will open to recruitment.

Control is treatment as usual. Local practices in each site will differ. We will collect relevant site profile data (please see “Site profile data” for details).

Recruitment will commence when individual active sites have reached predesignated PERFECT-ER adherence scores. Recruitment will take place over a 10-month period; participants will be followed for 6 months. A three-step recruitment process has been guided by our experiences from previous phases of the PERFECTED programme, other studies in this area [25, 26] and conversations with clinical and academic collaborators:

This three-step process will be closely monitored by the coordinating centre to identify trends that might be leading to over-recruitment or under-recruitment from specific groups. For example, if sites are consenting purely via personal consultees (England) or legal representatives (Scotland), monitoring will enable corrective actions and provide information to mitigate these recruitment trends in any future definitive trial. A more detailed overview of the recruitment process is shown in Fig. 2.

The aims of this trial are incompatible with only enrolling patients with minimal or mild confusion. It is important to ensure findings are broadly applicable and able to inform a future definitive trial. Participants who lack capacity to provide informed consent must be included. In this situation, the patient’s agreement to participate will still be obtained to their best level of understanding (in line with legislative frameworks in England and Scotland). Where patients in England are assessed as lacking capacity to make a decision about their initial or continued involvement with the study, we will seek personal or nominated consultee advice [27]. In Scottish trial sites where a patient is assessed not to have capacity, a legal representative will be sought and asked to consent in relation to the patient’s participation in the research [28]. In cases where written informed consent cannot be obtained, verbal consent can be taken (for example a patient with extremely poor eyesight or wrist/arm fractures). However this must be witnessed and fully documented in the patient’s notes [29]. For suitable informants in instances where written informed consent cannot be obtained, verbal consent can be taken over the telephone but must be witnessed and fully documented in the patient’s notes [29].

This feasibility study will not have a primary outcome. Instead, this study will inform the selection of the primary outcome for a potential definitive evaluation of PERFECT-ER. Because of the complexity and feasibility nature of this trial, different types of data will be collected from different sources (see Fig. 3). Additional information about selected outcome measures detailed in Fig. 3 can be found in Additional file 2.

All outcome assessment timeframes are taken from the point after a patient returns from surgery. Local research nurses will collect data from patients and their suitable informants at four time points. Assessment timeframes are baseline (0–5 days postoperative), follow up at time 1 (1 month ± 5 days), time 2 (3 months ± 5 days), and time 3 (6 months ± 5 days). Follow-up measures will be obtained by a local research nurse at the patient’s place of residence, with suitable informant follow-up data collected according to pragmatic arrangements.

We will also collect clinical measures and resource use data (Fig. 3). We will collect hospital care data (Accident and Emergency visits, inpatient and day-case admissions, outpatient appointments) from the clinical records of the participating NHS Trusts. Data on health and social care services and unpaid carers used by patients will be collected at each follow-up time point from the suitable informants (however only informants who are unpaid carers will be asked to complete questions on unpaid carer input). This will allow comparison of hospital-service use data from both sources. Findings from this collection and from the process evaluation workstream will also be used to explore the feasibility of collecting hospital service-use and cost data from routine sources. For patients cared for in English trusts, we will also collect data made available via the NHFD. Findings will also be used to explore the feasibility of economic data collection. We will interrogate these data to test for differences between participating hospitals to inform later generalisability. We will also draw on these results to make recommendations for cost data-collection methods for a definitive trial if indicated, taking into account research costs, data completeness and reporting burden experienced by participants.

To anticipate the expected wide variance in local practices, sites will complete a ward profiling assessment (comprising hospital characteristics, study ward characteristics, study ward staff profiling, and patient care profiles at organisational and patient level) at baseline, 4, 7 and 10 months following the trial opening. This will enable any drift from baseline status towards perceived improvements in service delivery to be monitored across control and active arms and provide rich contextual data for the process evaluation.

Following MRC guidance for complex interventions [21], an integrated multimethod, multi-perspective (patients, suitable informants and NHS professionals) process evaluation will be conducted. The process evaluation will aim to determine how, and under what circumstances, PERFECT-ER is implemented in each site and to examine how, and under what circumstances, implementing PERFECT-ER impacts on staff practices and perceptions, resource use, and patient and suitable informant outcomes. Learning from initial piloting of PERFECT-ER and its implementation mechanisms (PERFECTED WP2: Implementing the optimisation of hospital care delivery to older adults by NHS staff via action research methodology) will be used to inform procedures for collecting quantitative and qualitative data to inform the process evaluation. This data set will comprise PERFECT-ER checklist scores and SIL-generated reports and action plans. We will also undertake semi-structured interviews with patients (where appropriate) and suitable informants with experience of PERFECT-ER care delivery (25 patients and 25 suitable informants, 5 per population per active site). Interviews with patients will take place at their place of residence at the time (hospital, rehabilitation unit, care home or patient’s home). Locations and mode (face-to-face or telephone) of semi-structured interviews with suitable informants will be organised on a case-by-case basis sensitive to individual preference. Public patient involvement (PPI) is a central part of the PERFECTED programme. PPI colleagues are experts by experience [30]. They will receive training and assist in co-interviewing of suitable informants alongside academic researchers on a case-by-case basis.

Additionally, interviews and focus groups will be held with NHS professionals involved in ward-level and strategic implementation of PERFECT-ER. This research will also explore the economic aspects of caring for people who are confused and have hip fracture (for instance, to understand how time is used to provide appropriate care to this patient group). In addition, we will look at the feasibility of collecting detailed patient-level costing information (e.g. patient-specific nursing time, laboratory tests) from administrative systems in each site, mapping the availability of information based on correspondence with hospital administrators/finance officers and a documentary analysis.

The target sample size is 400 patient participants (200 per arm). This will comprise 40 patient participants’ in 10 different sites. As this is a feasibility study, this sample size was not decided on the basis of efficacy analyses; rather this is a purposive selection of groups with relevant experience of hip fracture and confusion and ability to access in the time available. Their data will be used to estimate the intra-class correlation coefficient for each outcome measure. Assuming a coefficient of no more than 0.1, 400 participants from 10 centres should provide a standard error no greater than 0.041, providing a basis for a definitive trial sample size.

Data from the trial will be analysed and reported in accordance with Consolidated Standards of Reporting Trials (CONSORT) [23]. An initial estimate of intervention efficacy will be provided using a multi-level model, with hospital as the highest “cluster” level and an appropriate error term depending on the nature of the outcome of interest. Between-treatment-arm effects will be estimated with 95% confidence intervals. Analyses will be on an intention-to-treat basis. There are no plans for imputing missing data and thus analyses will be on a complete-case basis only. No subgroup analyses are planned. In the light of new information, should it be decided during the course of the trial that a subgroup analysis is appropriate, this will be recorded in the statistical analysis plan prior to any data analyses. There will be no formal interim analysis. Deaths, falls, pressure ulcers, admissions to long-term residential care and safeguarding incidents are expected in this patient population. These rates will be collected at ward level throughout the study period. Analysis of recruitment rates, patient participant (including mortality) and suitable informant withdrawal rates will also be reviewed. Based on the data collected from this study, a formal sample size calculation will be carried out to inform a future, definitive trial.

Health and social care costs will be calculated drawing on the trial data, attaching unit costs from published sources to quantities of service use [31, 32]. The additional resource use associated with the intervention will be examined over the period of the trial by tracking potential indicators of impact, such as patient time spent in the anaesthetic room (i.e. this could change because relatives are encouraged to stay with the patient as part of the enhanced recovery checklist). SIL salary costs, on-costs and overheads will be estimated in order calculate their contribution to the costs of participants’ index hospital stays. Cost-effectiveness analyses will be from two perspectives - health and social care and societal. We will analyse outcomes including Bristol Activities for Daily Living Scale (BADLS) (patients) and quality-adjusted life-year (QALY) gain (for patients and carers). Incremental cost-effectiveness ratios will be computed and cost-effectiveness acceptability curves generated to examine the probability of cost-effectiveness over a range of willingness-to-pay thresholds for each outcome. Multivariate analyses of costs and outcomes appropriate for the clustered-randomised trial design will control for centre, baseline outcome measures or costs, and other confounders. We will compare results of analyses using societal, patient and carer weights for QALYs generated from DEMQOL-proxy, and different approaches to valuing unpaid care [32]. Variations in hospital service use within and between patients (e.g. hospital admissions, Accident and Emergency visits) will be explored through latent growth modelling. A sample of Participant Information Sheets and Consent Forms can be found in Additional file 3.

Data such as audio-files not already in a textual form will be transcribed verbatim and represented as “play script” transcripts. Once transcribed and formatted these documents will be imported into the NVivo qualitative software. Analysis will be conducted iteratively and draw out emergent themes from the data to explore variations in implementation fidelity, dose, adaptation and reach. The Consolidated Framework for Implementation Research (CFIR) will be used to further guide the analysis, including the identification of disconfirming data, points of implementation vulnerability will be examined to see how these may or may not lead to instances of breakdowns in the intervention and cases for critical learning that can inform the definitive trial will be taken forward [33]. Qualitative data collected pertaining to the association between costs and resource use, inputs and quality of care will also be explored. Data collected from mapping of patient-level hospital costing will be tabulated and summarised to draw out lessons for a definitive trial.