A detailed description of the APEP trial protocol has been presented previously . The trial received ethical approval from the local clinical research ethics committee.
The study was a prospective, sham-intervention controlled, randomised trial, with blinded randomisation and outcome measurement. It was completed between March 2015 and January 2017.
Patient Selection and Setting
Recruitment took place in one 350-bedded general teaching hospital. All wards admitted older medical inpatients, including one small geriatric ward. At the time of the trial, 17 medical consultants and four geriatricians were based in the hospital. Rehabilitation and general staffing levels were comparable across all wards. Irrespective of ward allocation, medical inpatients aged 65 and over, needing an aid and/or assistance to walk on admission, and admitted from and planned for discharge home (rather than for institutional care), with an anticipated hospital stay ≥3 days were recruited. The following patients were excluded: inpatients ˃48 hours prior to screening; unable to follow simple commands in the English language; admitted with an acute psychiatric condition, or requiring end-of-life or critical care; ordered bedrest, or contraindications to walking (e.g. hip fracture or high ventricular rate atrial fibrillation); baseline Short Physical Performance Battery (SPPB) score 0/1; participated in the trial within the previous 12-months.
Recruitment to the trial was completely independent from routine physiotherapy referrals and services. Using the electronic hospital management system, the principal investigator (RMC) identified suitable patients. Patients were not recruited on Fridays as no exercises sessions were delivered over the weekend. The medical team confirmed their suitability prior to the patient being approached. Patients were verbally informed about the study, and questions were answered. Relatives were contacted by phone if requested. Cognitive impairment is highly prevalent in patients aged ≥70 years of age in acute care . Therefore, we aimed to include patients who could exercise with one-to-one guidance. Those with significant cognitive impairment, who were unable to follow instructions and guidance were excluded. If cognition appeared or was reported as poor, their next-of-kin assented to their inclusion. All participants gave written informed consent. Patients were recruited as resources allowed; a maximum of five patients participated in the study simultaneously.
Both groups received usual care with additional exercises. The augmented prescribed exercise programme (APEP) was delivered to the intervention group, and a sham programme to the control group. The APEP programme aimed to improve strength, balance and walking, while the sham intervention was mainly breathing and stretching exercises. For a full description, please see Appendix 1: Description of APEP and Sham Exercise Programmes. These additional exercises were performed twice-daily, Monday-Friday, and the sessions lasted up to thirty minutes, (depending upon the patient’s exercise tolerance). At each session, contraindications were reviewed and verbal consent sought. Routine physiotherapy was not affected by the APEP trial and was delivered based on physiotherapy’s assessment of the patient and competing caseload, averaging three sessions per week. Patients in the acute stroke unit fitting the inclusion criteria were invited to participate. Patients with a major stroke met the exclusion criteria. They continued with their usual care, which was similar to other general medical patients until they were transferred to an offsite rehabilitation stroke unit.
On admission, the patients’ demographics and medical history were noted. Their home situation, medication use, co-morbidity (Cumulative Illness Rating Scale-Geriatrics, CIRS-G ), cognition (Six Item Cognitive Impairment Test, 6CIT)  frailty, including grip strength (Survey of Health, Ageing and Retirement in Europe, SHARE-FI) , falls history over the previous six months, and falls efficacy (Falls Efficacy Scale International, FES-I) , were measured on initial assessment.
The assessment schedule is described in Table 1.
The effects of the intervention on healthcare utilisation (length of stay and readmission rates), physical performance (SPPB)  and QoL (EuroQol 5 Domain 5 Level Scale, EQ5D5L) were measured. In addition, the effects on functional independence (Nottingham Extended Activities of Living, N-EADL), functional ambulation (Functional Ambulatory Classification, FAC), and falls rate, were also measured.
The primary outcome measure was length of stay (bed nights) (LOS). The number of readmissions over the subsequent three months was also recorded; both data were readily available from the electronic hospital information system.
Physical Performance and Daily Activity
Functional Independence was measured using N-EADL , premorbidly, on admission, and at the three-month follow-up.
The SPPB  was used to measure physical performance on admission, at discharge and at follow-up. Walking speed was assessed over four metres and only in those who could walk without assistance (standby assistance without contact was provided as required).
The FAC was used to measure their functional ambulation. Patients’ walking was observed on admission, at discharge and at follow-up. On admission, the patients were asked to self-report their premorbid ambulatory level. Self-report was also used at the follow-up when the patient couldn’t attend in person. While the FAC has not been validated as a self-reported tool, it did provide some information about their ambulatory level when observation was impossible.
Patients’ walking was continually measured using the Stepwatch Activity Monitor (SAM).
Falls and QoL
Number of falls over the previous six months was self-reported on admission. In-hospital falls were recorded from hospital notes, while post discharge falls were self-reported at follow-up.
QoL was measured using the EQ5D5L  on admission, discharge and at follow-up. The next-of-kin was asked to complete this if the patient was unable. The reliability of proxy reports has been debated with evidence suggesting that proxy reports are poorer than self-reports . However, other studies have found little or no difference between self and proxy reports in older adults , patients with traumatic brain injury and Parkinson’s Disease , therefore, the decision to include proxy reports was made.
Changes in living arrangements (change in accommodation, support or home adaptations) were recorded at discharge and follow-up.
Procedure for Data Collection
Patients were assessed within forty-eight hours of admission, and within twenty-four hours of planned discharge, and followed-up between two and three months following discharge home, at their medical check-up appointment, or by phone. After initial assessment, patients were randomly allocated to the intervention or control group using concealed allocation. A blinded research assistant assigned and recorded the patients using a computer-generated randomisation sequence, in varying block size. While patients were informed that they would be allocated to either the APEP or control group and upon allocation, they were neither explicitly informed nor encouraged to ask about their allocation.
Patients who had not begun the exercise sessions before withdrawal, transfer or discharge, were replaced, using the same process as above. Patients who began the exercise sessions before withdrawal from the study were not replaced. To prevent contamination, it was planned that patients who were in the same room but allocated to different exercise groups, would complete their exercise sessions in different locations separately. However, this event never occurred. Therefore, all patients were treated by their bedside. The discharge and follow-up assessments were completed by a blinded research physiotherapist.
An adverse event included a fall, cardiac ischaemia or pulmonary embolism during exercise, or an exacerbation of a condition as a result of the intervention (e.g., exacerbation of painful joints). Death or admission to intensive or critical care were considered as Serious Adverse Events. In the occurrence of adverse events, the Sponsor’s Clinical Research Supporting Officer, the Hospital Risk Manager and the treating consultant were informed. All the necessary hospital procedures and documentation were completed.
Deviations from the published protocol
There were four significant deviations from the previously published protocol . First, accelerometer-recorded walking activity was collected on a considerably lower number of patients than planned. Second, the trial was terminated early, due to a change in discharge procedures, with 190 patients of the planned 220 patients included. Third, in order to detect a deterioration in physical performance, we only recruited patients with an SPPB score of ≥2 on admission. And finally, we introduced a phone follow-up assessment for patients unable to attend a face-to-face assessment. For further details, please see Appendix 2: Deviations from the published protocol.
All of the descriptive information is presented in Table 1. Throughout the results, means (±SD) are presented for normally distributed data and medians [IQR] are used for non-normally distributed data. Normality of their distribution was determined using histograms.
Intention-to-treat analysis was employed on the length of stay, death and readmission rates as full data was available irrespective of drop-outs. Time-to-event analysis using the Cox proportional hazards model was used to measure the effect of the APEP on length of stay (time to discharge) i.e. discharge being the event. The effects of the APEP on walking activity in hospital, and physical performance and QoL, both at discharge and at follow-up, was estimated using linear regression. Logistic regression was used to estimate the effects on falls, readmissions and deaths at follow-up, and post hoc were grouped to report the combined negative effects. These models estimated the effects of the intervention on the absolute scores, rather than the changes in scores.
For the adjusted models, the most important covariates were selected post hoc based on the results from the preceding observation study  subject matter expertise and clinical judgement, with each model included their corresponding baseline score. The effects of APEP on time to discharge and step-count were adjusted for age and frailty only. For all other adjusted models, similar covariates were used.