|
nif:isString
|
-
The protocol for this trial and supporting CONSORT checklist are available as supporting information; see Checklist S1 and Protocol S1. This cluster randomised trial was approved by the Multi-Centre Research Ethics Committee for Wales (08/MRE09/12), who sanctioned post-recruitment contact with a vulnerable population in an emergency setting, and inclusion of all patients who did not opt out of the study. Participants therefore did not have to give oral or written consent to participate in the trial during their emergency episode, but consented to follow up in response to information about the trial sent by participating ambulance services 7–10 days after their index event. Following processes agreed by the Ethics Committee, these services passed non-dissenting participants' contact details to the research team, and kept records of dissent in hard copy and electronically. Paramedics gave informed consent to participate. We registered the trial at: http://www.controlled-trials.com/ISRCTN10538608.
Cluster trials are appropriate to evaluate interventions targeted at health professionals. Thus Support and Assessment for Fall Emergency Referrals 1 (SAFER 1) was a cluster trial with paramedics as the unit of randomisation. [24] We recruited patients at two UK study sites from November 2009 until October 2010. Delays in implementing a national information technology programme [25] reduced these from three to two: Site one, an urban centre where we recruited paramedics from four ambulance stations; and Site two, where we recruited paramedics from nine stations across a mixed urban and rural area.
Paramedics were eligible to participate in SAFER 1 if they worked at any of 13 ambulance stations with a falls referral pathway in place; they continued to be eligible if they moved from one of these stations to another. In practice such a pathway requires a community-based falls service to accept direct referral of older people who fall by paramedics at the scene of their fall. Within an agreed space of time (typically within 1 week) falls services contact the patient and arrange a home visit to assess clinical and social needs and to arrange ongoing community based support [26]. The chief investigator invited all eligible paramedics to participate in SAFER 1 using local media to support recruitment. The trial team consented volunteers and passed anonymous details to the West Wales Organisation for Rigorous Trials in Health (WWORTH) for randomisation stratified by current ambulance station. Patients were eligible for SAFER 1 if they were: aged 65 or over; living in the catchment area of a participating falls service; and attended by a study paramedic following their first emergency call categorised by the call-taker as a fall during the study period. We excluded those living in nursing homes as they were not eligible for care from participating falls services.
The health technology evaluated in the experimental arm was CCDS on hand-held Tablet computers for use by paramedics to decide whether to take patients who had fallen to an Emergency Department or leave them at home with referral to a community-based falls service. Site one implemented the CCDS simultaneously with a system for electronic patient data capture; while Site two, where a different electronic data capture system was already in place, added CCDS software to the existing system. However neither site fully integrated CCDS with the electronic software; in particular Site one experienced many teething problems including loss of network signal and hardware failures. Control paramedics at both sites provided usual care, with paper-based protocols to assess patients and make decisions about their care, including patients who had fallen. Usual care comprised assessment, treatment on scene as required and default conveyance to the Emergency Department unless the patient refused to travel to hospital. Although we know that practice is variable, we did not attempt to standardise care in the control arm as there is little evidence about what is best for patients. Both groups could refer older people who had suffered a fall to community-based falls services.
Effectiveness – proportion of participants left at scene without conveyance to an Emergency Department and proportion referred to falls servicesSafety – proportion of participants with adverse events (harm) up to one month (999 call, Emergency Department attendance, emergency admission to hospital, or death);Cost-effectiveness – comparison of costs of implementation of CCDS for paramedics and its benefits in the form of patient utility modelled over 12 months.
Self-reported falls; fall-related self-efficacy (‘fear of falling’) [27]; health-related quality of life (SF12) [28] and patient satisfaction (Quality of Care Monitor) [29] were gathered through postal questionnaires completed by patients or their carers. Operational indicators – ambulance service job cycle time, length of episode of emergency care and costs of care – were gathered from routine NHS sources. Though we had planned to include quality of clinical documentation, internal validation showed that the adoption of CCDS led to double data entry and risk of intervention bias. We explored implementation and adoption issues through focus groups and semi-structured interviews with practitioners, and reported the results elsewhere [30].
After redesigning SAFER 1 following delays in implementing the intervention, we powered it to detect clinically important changes in the proportion of participants who make another emergency call for a fall within a month (or die) – the ‘safety’ criterion. We calculated that a simple random sample of 622 participants would yield 80% power when using a 5% significance level to detect a fall in that proportion from 30% to 20%. To adjust for clustering by the 42 paramedics recruited (rather than the 13 ambulance stations at which they worked), we assumed that the intra-paramedic correlation coefficient (IPCC) was 0.02, and applied Donner's formula [31] to yield a target sample size of 865, namely 42 paramedics each recruiting an average of 20.6 participants or 622×[1+(20.6–1)×0.02].
The West Wales Organisation for Rigorous Trials in Health (WWORTH) independently used random number tables to allocate paramedics, consented and stratified by current ambulance station, between intervention and control arms. It was possible to blind analysts to these allocations, but not paramedics or patients.
Patient recruitment and data retrieval: Attending paramedics consented patients to treatment, but not trial participation owing to the emergency nature of the contact. Ambulance service staff identified potential participants from electronic records completed by control room staff, then confirmed eligibility from records completed by attending paramedics. They contacted participants by post within 10 days of the index call to give them the opportunity to opt out of follow up. At both sites we retrieved identifiable data about subsequent emergency calls and referrals to falls services and their outcomes from the ambulance services. Site one retrieved anonymised linked data about Emergency Department attendances, emergency hospital admissions and deaths from a central databank [32] although this process delayed analysis and reporting. At Site two we retrieved identifiable data about Emergency Department attendances and emergency admissions from individual National Health Service care providers; and about deaths from the Office of National Statistics. The flow of paramedics and patients through the trial is shown in Figure 1.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0106436.g001 Flow of paramedics and patients through the trial. In pre-specified analyses we used SPSS version 19 to fit multi-level logistic, linear and negative binomial regression models to, respectively, binary, measurement and count data available at one month on referrals to falls services, the hierarchy of ‘harms’ in Figure 2, and related outcomes, adjusting for statistically significant confounders, but not for multiple testing. Potential confounders included: ambulance service (site); patient's age, gender and distance to nearest Emergency Department; date of recruitment and whether call was out of hours. For secondary outcomes we again used multi-level models, adjusted for significant confounders, and imputed missing data, by published rules when available. Specifically, missing responses to individual SF12 questions were imputed using Expectation Maximisation methods [33] missing SF12-related scores were imputed using regression-based methods and set to zero on participant death. Similar regression-based methods impute missing ‘fear of falling’ and participant satisfaction scores.
Figure data removed from full text. Figure identifier and caption: 10.1371/journal.pone.0106436.g002 Hierarchy of harms. To extend the outcomes of SAFER 1 to one year, we adopted a modelling approach similar to that used by Goitein to extend the outcomes of CT scans of gastric carcinoma to survival over time [34]. Specifically we inferred that participants referred to falls services in SAFER 1 would achieve the clinical outcomes reported for intervention participants with similar characteristics in the Community Falls Prevention Trial of referral to falls services [22]; and that participants not so referred would achieve the clinical outcomes reported for control participants with similar characteristics in the Community Falls Prevention Trial. In doing so we exploited the similarity of populations and outcomes between SAFER 1 and the Community Falls Prevention Trial, in particular by standardising by age and referral rates using Site two, where electronic data capture was already in use as the standard.
We undertook a cost-effectiveness analysis from the perspective of the UK NHS and personal social services and used quality of life measured by the SF12, referrals to falls services, 999 calls, Emergency Department attendances and inpatient stays as outcomes over the next 30 days. To extend our time horizon to one year, we integrated the effects of CCDS on referrals to falls services as estimated by SAFER 1 with the effects of referrals to falls services on events over one year, especially patient utilities, as estimated by the earlier trial [22]. We used probabilistic sensitivity analysis to assess the extent to which the intervention gave value for money relative to using the same resources elsewhere. We estimated the implementation costs of CCDS, taking account of whether electronic data capture was already in place. We estimated the costs of staff, equipment and consumables from the ambulance services with and without existing electronic data capture; and the costs of healthcare use by multiplying that use by published unit costs (Table 1).
Table data removed from full text. Table identifier and caption: 10.1371/journal.pone.0106436.t001 NHS unit costs. aDepending on level of treatment receivedbDerived from discussions with Age Concern; equivalent to the unit cost of referral to ‘Hospital at Home or Early Discharge Schemes’ Deviations from the original study protocol: (1) We reduced paramedic training in consultation with participating ambulance services from two days to half a day including assessment of competence (2) We were unable to analyse some outcomes that varied between hospitals, for example categorisation of falls; we therefore analysed only the generic outcome ‘emergency admission to hospital’ (3) To reduce questionnaire length and maximise response rates, we did not collect costs incurred by participants, not least because in the UK they do not contribute financially to care provided by the falls services. (4) We did not measure outcomes at six months as planned, owing to delays in implementing the intervention.
|
![[This page as RDF]](http://data.gesis.org/somesci/static/rdf-icon.gif){kind=icon}