A Service Intervention to Reduce Falls in Hospital

Overview

The head of nursing at University Hospital Coventry and Warwickshire (UHCW) plans to roll out an intervention across groups of hospital wards over the next four months. The intervention is designed to reduce falls as part of quality improvement for the hospital. However, the head of nursing has asked the University of Warwick to help with the scientific evaluation of the intervention - to find out whether and to what extent falls on the wards are reduced by the intervention. The University of Warwick will have two functions: 1. To analyse data on falls to see if there has been a statistically significant drop in fall rates before and after the intervention has been implemented across these groups of wards 2. To determine a random order in which the groups of wards receive the intervention as this will make it easier to distinguish cause and effect.

Falls are the most commonly reported patient safety incident with 240,000 recorded in NHS hospitals in England each year. The national mean rate of falls per 1000 occupied bed day (OBDs) is 6.6. The rate of harm per 1000 OBD is 0.19. The head of nursing at UHCW is responsible for quality and safety in nursing and recognises that falls are a problem at UHCW. Consequently she plans to implement an intervention (an educational programme for ward staff) to reduce the potential harm caused by falls, and wishes to know whether the intervention is effective. For logistical reasons it is not possible to introduce the intervention to all wards at the same time, so it is necessary to roll the intervention out across clusters of wards over time. These clusters will be the unit of study. Human Subjects Protection Review will be exempt as the intervention is taking place at ward level, not the patient level. The hypothesis is that fall rates will tend to decline after the intervention is introduced. Standards for prevention of falls have been published by the Healthcare Quality Improvement Partnership (Royal College of Physicians. National Audit of Inpatient Falls: audit report 2015. London: RCP, 2015). The intervention will consist of education to improve compliance with these standards. NIHR CLAHRC West Midlands were approached to conduct an independent evaluation. However, the service imperative does not permit any delay in implementation of the intervention. Although the intervention must not be delayed, it cannot be implemented simultaneously across all hospital wards; it must be rolled out incrementally. The unit of roll-out is a cluster of wards. There are nine clusters of wards: Cluster A: * Ward 40 - Gerontology - Age * Ward 20 - Gerontology * Ward 21 - Gerontology Cluster B: * Ward 41 - Stroke * Ward 42 - Neurology * Ward 43 - Neurosurgery Cluster C: * Ward 30 - Respiratory * Ward 31 - Medical ward * Ward 34 - Clinical Haemotology * Ward 35 - Oncology Cluster D: * Ward 50 - Renal * Ward 52 - Orthopaedics * Ward 53 - Orthopaedics Cluster E: * Ward 10 - Cardiology * Ward 11 - Cardiothoracic Surgery Cluster F: * Cedar Ward - Orthopaedics * Hoskyn Ward * Mulberry Ward * Oak Ward - Rehab Cluster G: * Ward 32 - Head \& Neck * Ward 33 - Surgery * Ward 33 - Gastro * Ward 33 - Urology Cluster H: * Ward 21 - Short-stay - Gen Surgery * Ward 22 - ECU * Ward 22 - Surgical Assessment Unit * Ward 22A - Vascular Surgery * Ward 23 - Gynaecology Suite Cluster I: * Ward 12/CDU - AMU1 * Ward 3 (AMU3) * Ward 12 - Observation / Assessment Unit (ED) * Ward 1 * AMU 2 The phased introduction across wards evokes the possibility of step wedge cluster RCT (Hemming, et al. BMJ. 2015; 350:h391). However, clusters A and B must proceed first, and clusters H and I must proceed last as these are wards with very short stay. Five clusters (C to G) are therefore available for randomisation. Once clusters have been randomised to a given order, there are no foreseeable reasons to change the order. Accordingly, a list of the five eligible clusters was sent to the CLAHRC WM Director on 17/01/17 and were randomised independently (by Dr Mark Slater at the Dept of Physics, University of Birmingham) using Microsoft Excel (each cluster was assigned a random number, the numbers were then sorted from smallest to largest), as below: E - F - G - C - D. The primary outcome is fall rates, and we hypothesise that fall rates will decrease over the intervention period. Fall rates per 1000 bed days per month range from 0 to 34.48. Fall rates will be collated monthly over a one year period to provide a median of approximately six months pre- and six months post-intervention data points. The data will be expressed as falls per 1000 OBDs and will be harvested from the routine data system. The senior nurse on duty has a statutory requirement to collect data on falls and enter them on the hospital computer system. We will record the date at which the intervention team start working with a new cluster. The rate at which the intervention will be rolled out is uncertain at this stage. For analysis, the raw data from the system will be sent to Dr Karla Hemming at the University of Birmingham for statistical analysis, including adjustment for calendar time. Primary analysis will be restricted to the five randomised clusters, and secondary analysis will include the four non-randomised clusters. The evaluation will take account of correlation within clusters and auto-correlation over time.