Hospitals and wards

We aimed to recruit 16 wards in eight NHS hospitals in England and Wales: one elderly care and one orthopaedic trauma ward in each hospital.

The inclusion criteria for ward participation were the demonstration of adequate ward nurse staffing, as assessed against national guidance,^56,57 and the agreed involvement of a named ward manager, a senior nurse and a VSM (if local staff intended to use voluntary services as a component of the intervention). In addition, there was a requirement for the equivalent of 1 day per week per ward of dedicated time for 3–4 months from an experienced senior nurse to lead the implementation.

Wards were excluded if they had previously participated in the development of the POD system of care or if they intended to implement other delirium prevention initiatives during the trial.

Participants

As the POD system of care is a whole-ward intervention, all patients in the ward (regardless of eligibility and consent status) would have the potential to be exposed to the intervention. Patients were therefore recruited for completion of individual outcome assessments only.

Patients aged ≥ 65 years who were admitted to the study wards during the study period were eligible for participation.

Patients were excluded if they had prevalent delirium on admission to the ward; if discharge was planned within 48 hours of admission; if a delirium assessment had not been performed by a research assistant (RA) within 24 hours of admission (elderly care patients) or preoperatively (orthopaedic trauma patients); if consent had not been obtained with 48 hours of admission to the ward; if end-of-life care was being provided; and if they had transferred from another ward or were not under the care of the ward team. Participants were recruited for 6 months following the 6-month intervention implementation period.

Primary outcome

We assessed for differences in new-onset delirium within 10 days of recruitment between patients in the intervention group (POD programme) and patients in the control group (usual care), as this is the expected primary outcome for a definitive trial. Delirium was assessed using the four-item Confusion Assessment Method (CAM)^58,59 (Table 8).

TABLE 8

Structured assessment process to complete the CAM

Secondary outcomes

Physical and social independence were measured by the RAs at baseline and at 3 months (postal questionnaire) using the Nottingham Extended Activities of Daily Living (NEADL) scale;⁶² anxiety and depression were measured by the RA using the Clinical Anxiety Scale⁶³ and the Geriatric Depression Scale Short Form,⁶⁴ respectively, at 30 days.

Research assistant Confusion Assessment Method training

A CAM training and monitoring process was developed that followed recommended practices.⁶⁵ We developed a three-stage training process. Stage 1 was central or local classroom teaching about delirium, trial research procedures and the administration of outcome measurement instruments, including the CAM. Stage 2 was specific to the CAM and involved local experiential learning consisting of (1) one-to-one practice sessions, (2) pilot interviews with patients and (3) within-site inter-rater reliability assessments. Stage 3 was a further within-site inter-rater CAM reliability performance check conducted at the local sites.⁵⁹

Data collection

Data collection was undertaken by locally based RAs who were trained in study procedures and outcome measures.

Primary outcome

The RAs performed cognitive assessments [AMTS and months of the year backwards (MotYB) test] and the CAM daily for up to 10 days post recruitment (or until discharge, if sooner) to detect the presence of new delirium. Each CAM item was assessed and recorded on a clinical research form that was dated and signed. This document was stored digitally and accessed at a later date to investigate achievement of the completeness of the CAM assessments.

Discharge assessment

At the point of discharge, the RAs recorded the date of discharge (or date of death), episodes of falls in hospital and discharge destination (living alone, living with another person, residential care home, nursing home, other).

Thirty-day assessment

At 30 days post recruitment, the RAs performed a cognitive assessment (AMTS and MotYB test) and the CAM, and asked the patient to complete a questionnaire about health-related quality of life using the EQ-5D,⁶⁸ about anxiety using the Clinical Anxiety Scale,⁶³ about depression using the Geriatric Depression Scale Short Form⁶⁴ and about patient experience using selected questions from the patient-reported experience measure from the National Audit of Intermediate Care.⁶⁹

Three-month assessment

At 3 months post recruitment, postal questionnaires were used to provide information on physical and social independence (NEADL scale),⁶² health-related quality of life (EQ-5D),⁶⁸ and health and social care resource use and living arrangements. Proxy completion of the questionnaires was permitted.

Estimation of recruitment rates

We recorded the number of sites expressing an interest in participating in the trial and reasons for non-progression. To assess the feasibility of recruiting participants for a definitive trial, we calculated the number of patients screened, eligible, assessed for delirium, with prevalent delirium, with capacity to consent and for whom consent for trial participation was obtained. Baseline descriptors of the screened and recruited populations were obtained (see Characteristics of the screened and registered participants: generalisability) and compared between the groups to assess for imbalance.

Estimates of completeness of data collection and follow-up rates

The reliable calculation of delirium incidence rates requires close adherence to a predetermined delirium detection process. We therefore recorded the number of in-hospital delirium assessments and 30-day delirium assessments conducted, and calculated the missing observations. We also recorded the number and timing of participant withdrawals from follow-up data collection and the reasons for withdrawal (including deaths), and the number of participants with missing self-reported outcome questionnaires at each time point.

Assessment of between-ward intervention contamination

The number of staff moving on and off study wards within sites was collected for a sample period of 1 week. The number of participants moving wards during their hospital stay was tabulated. Incidence rates of new-onset delirium at the sites that were randomised at the hospital level were calculated and compared with those from sites that were randomised at the ward level, to assess possible between-ward contamination. Incidence rates of new-onset delirium were calculated for participants recruited within the first 3 months of sites opening to recruitment, and for participants recruited between 3 and 6 months of sites opening to recruitment to determine if the intervention delivery was sustained when more time had elapsed since initial training. Service improvements introduced on participating wards during the study period were recorded.

Estimation of a sample size for a future definitive randomised controlled trial

To inform the sample size calculation for a possible definitive trial, we calculated the incidence of new-onset delirium within 10 days of admission by ward type, by study arm and overall, together with corresponding 95% CIs. We used multilevel logistic regression that adjusted for demographic characteristics (age and sex), delirium risk factors (medications associated with delirium, e.g. benzodiazepines, opiates, H₁ antihistamines),⁷⁵ sensory impairment (hearing impaired, use of hearing aid or sight impaired), cognitive impairment and/or dementia, Charlson Comorbidity Index,⁶⁶ NEWS⁶⁷ category and ward type. In the regression model, ward type was fitted as a random effect. The number of new patients admitted per ward during the recruitment period was used to estimate cluster size. The intracluster correlation coefficient (ICC) and associated 95% CI were calculated using the covariance parameter estimate from a multilevel logistic regression without adjustment for participant or ward characteristics.

Criteria for continuation to a future definitive randomised controlled trial

A priori criteria for progression to a definitive randomised controlled trial were defined as a minimum of six of the eight wards (75%) completing the POD manual milestone checklist (to provide assurance that the POD implementation was successful) and an overall recruitment rate of at least 10% of the potential recruitment pool. The criteria did not include thresholds for projected clinical effectiveness or cost-effectiveness.

Model development

To estimate the cost-effectiveness of an integrated delirium prevention intervention in the context of the trial, a decision-analytic model was used. The model was developed after consultation with the research team (including clinical experts) and the parameter values were identified following targeted searches of the literature. The working report describing the model development and results is included in Appendix 4.

Cost-effectiveness study

The aim of the economic study was to establish the feasibility of conducting an economic evaluation of the POD system of care and to determine preliminary estimates of its cost-effectiveness. Specific objectives were to:

• determine the feasibility of collecting the assessments needed (quality of life and health-care resource use) for an economic evaluation in this patient group
• determine the number of missing data in assessments
• determine the validity and responsiveness of quality-of-life assessments in this group
• determine the feasibility of collecting and of using/interpreting proxy-completed assessments
• estimate the cost of the POD intervention
• provide estimates of the cost-effectiveness of POD, compared with usual care
• compare these estimates with those from the earlier evaluation based on decision modelling.

Quality of life was assessed using the EQ-5D⁶⁸ at baseline and at 1 and 3 months. Health-care resource use was captured using a specially designed questionnaire completed by patients (and/or proxies) at 3 months. Costs were calculated from the perspective of health services and Personal Social Services. The cost of the POD intervention was estimated to include material costs (e.g. printing of manuals), the time to deliver and receive the training and also the time to provide support during POD delivery. This information was provided by the POD research team, which kept a contemporaneous diary of visits and travel. The feasibility of data collection was determined by observing the extent of missing data.

The primary economic evaluation adopted the NICE-preferred approach of a cost–utility analysis comparing the costs and benefits of POD and usual care.⁷⁶ The analysis time horizon was 3 months, based on the trial follow-up. The main analysis result was the incremental cost-effectiveness ratio (ICER) per quality-adjusted life-year (QALY). ICERs below the range of £20,000–30,000 indicate that POD would be considered cost-effective. Non-parametric bootstrapping was employed to determine the level of sampling uncertainty. No discounting of costs or effects was conducted.

See Appendix 5 for full details of the methods.

Hospitals and wards

Twenty hospitals expressed interest. Twelve hospitals subsequently completed and returned site survey forms, and eight of these were recruited. Among the four hospitals not recruited, two withdrew (one because of poor staffing and one was unable to identify a suitable ward), one did not respond to the request for a site visit to progress towards recruitment and, for one, regulatory approvals came after the other requisite number of POD sites had gained approval. Of the 16 recruited wards, nine were elderly care medicine and seven were surgical/trauma orthopaedic. Seven of the eight hospitals registered had one elderly care ward and one orthopaedic trauma ward. The remaining hospital site had two elderly care wards registered.

None of the wards/hospitals withdrew from the study.

Patient screening and recruitment rates

Screening and recruitment took place between August 2014 and February 2015. A total of 4449 patients admitted to the 16 wards were screened for eligibility, 3274 (73.6%) of whom were considered eligible. The most common reasons for exclusion at screening of the remaining 1175 patients were as follows: 538 (45.8%; 12.1% of those screened) had a recorded diagnosis of delirium on admission, 352 (30.0%; 7.9% of those screened) had an expected duration of stay of < 48 hours, 139 (11.8%; 3.1% of those screened) were not under the care of the ward medical team and 105 (8.9%; 2.4% of those screened) were receiving end-of-life care (Figure 4).

FIGURE 4

The Consolidated Standards of Reporting Trials flow diagram. a, Withdrawals here are from researcher questionnaires; b, withdrawals here are from postal questionnaires; one patient in the POD arm withdrew from researcher visits at 30 days, but not from (more...)

Delirium assessment was performed by the RAs on 1537 (34.5% of screened) of the 3274 eligible patients. The remaining 1737 (39.0% of screened) patients did not have a delirium assessment. The most common reasons for not performing a CAM assessment were as follows: research staff missed patient [691 (39.8% of those excluded)], ward staff advised not to approach [374 (21.5%)], patient unavailable [283 (16.3%)] and patient refused [246 (14.2%)].

Of the 1537 patients who were screened for delirium, 1418 (31.9% of screened) were assessed as not having prevalent delirium. The remaining 119 were excluded: 113 had prevalent delirium, and the reasons are unknown for six.

Of the 1418 patients assessed as not having delirium, 1340 (30.1% screened) either had capacity (n = 1182) or a consultee had been identified (n = 158); these were approached for participation. The remaining 78 patients were excluded as they were without capacity and a consultee had not been identified. Of the 1340 patients approached for participation, 626 were excluded. The most common reasons for exclusion were as follows: 459 (73.3% of those excluded) patients refused; 59 (9.4%) were discharged before consent was obtained; 29 (4.6%) were missed by the researcher; and, for 26 (4.2%), the consultee refused. Consent was obtained for 714 (16.0% of screened) patients, and 713 were registered to the study (one patient was discharged before registration): 343 were registered to the POD system of care and 370 were registered to control (see Figure 4). Figure 5 is the recruitment graph. Patient accruals to the study between the sites ranged from 65 to 105 (see Appendix 6, Table 50).

FIGURE 5

Recruitment graph.

Eligibility violations

A total of 13 (1.8%) participants were identified as not fulfilling the eligibility criteria: five (1.5%) in the POD arm and eight (2.2%) in the control arm. The main criteria violated were as follows: the delirium assessment (CAM) was not performed (n = 1 in the POD arm and n = 6 in the control arm), and the participant had prevalent delirium on admission (n = 3 in POD and the POD arm and n = 2 in the control arm) (see Appendix 6, Table 51). Patients identified as breaching an eligibility criterion were included in all analyses.

Final follow-up

Thirty-three (4.6%) participants withdrew during the study period [15 (4.4%) from the POD arm and 18 (4.9%) from the control arm] (see Appendix 6, Table 52). Of these, 19 (57.6%) withdrew within 10 days of recruitment.

Characteristics of the screened and registered participants: generalisability

The characteristics of the screened and registered participants were similar with respect to age, sex and ethnicity (Table 9). In the screened and registered populations, the mean overall ages were 83.1 years [standard deviation (SD) 8.05 years] and 82.7 years (SD 7.84 years) respectively; female patients accounted, respectively, for 67.2% and 68.3% of the populations; and 89.3% and 91.7%, respectively, of the populations were of white ethnicity. Participant age and ethnicity were broadly similar across sites: the mean overall age varied between 79.0 years (SD 7.34 years) and 85.1 years (SD 7.58 years), and between 83.8% and 97.6% of participants were of white ethnicity (small differences due to more missing data for some sites). Some differences between sites were noted with respect to sex; female participants accounted for between 49.2% and 100.0% of participants (see Appendix 6, Table 53).

TABLE 9

Characteristics of the screened and registered participants by arm

Baseline characteristics

The two arms were well balanced with respect to residence, hearing aid use, benzodiazepines use and comorbidities, although some imbalance between groups was evident for all other characteristics, namely reason for admission, ward type, cognitive impairment and/or dementia, highest NEWS category within 48 hours of admission, visual impairment, hearing impairment, and opiates and antihistamines prescribed (Table 10). Residence and hearing impairment were balanced across sites; however, some imbalance between sites was apparent for all other characteristics (see Appendix 6, Table 54).

TABLE 10

Demographic characteristics and NICE risk factors by arm

Intervention fidelity

Ten health-care professionals with experience in older people’s care assessed fidelity. The mean score for each domain was as follows: installation, 4.5 points (range 3.5–5.0 points); delivery, 32.6 points (range 27.3–38.3 points); coverage, 7.9 points (range 4.2–10.1 points); and duration, 0.38 points (range 0–1.0 points).⁷³ Of the 10 delirium risk factors, infection, nutrition, hypoxia and pain were the most consistently addressed, and cognitive impairment, sensory impairment and multiple medications were the least consistently addressed.⁷³ Overall fidelity to the intervention was assessed as being high (≥ 80%) in two wards, medium (51–79%) in five wards and low (≤ 50%) in one ward.

Staff

Contamination data were received from all sites [two sites (sites 2 and 4) provided data by telephone or e-mail only]. Site 2 stated that staff were never moved to other wards from the trauma ward and no staff were brought onto the ward. No data were provided for the elderly care ward at this site. Site 4 stated that wards did not keep any records of staff movement.

During the 1-week data collection period, there were 216 reports of staff moving into the 12 study wards that were providing data: 115 on elderly care wards and 101 on orthopaedic trauma wards, most commonly health-care assistants (51.9%) (see Appendix 6, Table 56).

Only 13 staff moves off the ward were recorded during the 1-week data collection period: four from elderly care wards and nine from orthopaedic trauma wards (see Appendix 6, Table 56).

Delirium incidence rates

The incidence rates of new-onset delirium were similar between arms at the sites randomised at the hospital level and at those randomised at the ward level (see Appendix 6, Table 57). However, incidence rates were lower when ward-level randomisation was used.

Incidence rates of new-onset delirium were similar between arms for participants recruited during the first 3 months of sites opening to recruitment and for participants recruited between 3 and 6 months of sites opening to recruitment, although incidence rates were lower during the later 3 months of recruitment (see Appendix 6, Table 58).

Patients

During their hospital stays, 135 (18.9%) participants moved wards: 58 (16.9%) in the POD group and 77 (20.8%) in the control group (see Appendix 6, Table 59). Among the sites, the percentage of participants moving wards ranged from 10.0% to 25.8% (see Appendix 6, Table 60).

Wards

No ward reported any new multicomponent delirium prevention measures during the study period. Four POD wards introduced service improvements consisting of dementia training; observations using electronic handheld devices; attempts to decrease noise levels; and identification of patients at risk of delirium with education on hearing aids, nutrition and hydration. One control ward introduced a new rounding chart.

Confusion Assessment Method assessments

Taking into account the length of stay and excluding the assessments not expected as a result of death, withdrawal or discharge, of an expected 5645 CAM assessments, 5065 (89.7%) were completed during the first 10 days of recruitment (Table 12).

TABLE 12

Number of in-hospital CAMs performed out of number expected (based on length of stay) by randomised arm

Non-completion rates in sites ranged from 3.5% to 14.8%.⁵⁹ The main reasons for non-completion of the CAM were participants were too ill [n = 186 (32.1%)] or participant refusal [n = 163 (28.1%)] (see Appendix 6, Table 61).

Of the 5065 CAM assessments, six (0.1%) had missing responses to CAM questions, two (0.04%) omitted the AMTS and 25 (0.5%) omitted the MotYB test (see Appendix 6, Table 62).

At 30 days, out of an expected 629 CAM assessments, 513 (81.6%) were completed (Table 13).

TABLE 13

Number of 30-day CAMs performed out of number expected (excluding deaths and withdrawals) by randomised arm

Non-completion rates of the 30-day CAM in sites ranged from 2.8% to 30.6% (see Appendix 6, Table 63). More 30-day CAM assessments were not performed for participants in the POD group (n = 78) than for those in the control group (n = 38). Similar reasons for non-completion of the CAM were evident across the groups, although some differences between groups were noted for participant refusal (POD arm, 29.5%; control arm, 15.8%) and participants moving out of the area (POD arm, 12.8%; control arm, 21.1%) (see Appendix 6, Table 64).

Questionnaire return rates

Return rates of the questionnaire booklets to the CTRU were as follows: at baseline, 699 (98.0% of 713 registered participants); at 30 days, 511 (81.8% of 625 expected); and, at 3 months, 400 (70.5% of 567 expected) (see Appendix 6, Table 65). Participant age, sex, ethnicity and residence were broadly similar between those who did and those who did not complete the 30-day researcher questionnaire (see Appendix 6, Table 66).

The most common reasons for non-completion of the 30-day and 3-month questionnaire booklets were as follows: could not contact participant [27 (23.7%) and 92 (55.1%) for the 30-day and 3-month questionnaire booklets, respectively] and participant refused to complete them [26 (22.8%) and 31 (18.6%) for the 30-day and 3-month questionnaire booklets, respectively] (see Appendix 6, Tables 67 and 68).

Questionnaire compliance

Of the 511 30-day follow-ups, 313 (61.3%) were undertaken within ± 2 days of the due date [mean 28.9 (SD 4.93) days]. Of the 400 3-month follow-ups, 259 (64.8%) were undertaken within ± 2 weeks of the due date [mean 102.4 (SD 19.39) days].

Primary outcome

Fifty-seven (8.0%) of the 713 participants developed new-onset delirium within 10 days of recruitment: 24 (7.0%) of the 343 participants registered to wards delivering the POD system of care and 33 (8.9%) of the 370 participants registered to the control wards. New-onset delirium was slightly higher in orthopaedic trauma wards than in elderly care wards (10.0% vs. 6.4%).

Multilevel logistic regression analysis (adjusting for participant characteristics collected at registration and for ward type) was used to explore the between-group differences in delirium incidence. Although there was evidence that participants in the POD arm had lower odds of developing delirium, this result was not statistically significant (odds ratio 0.68, 95% CI 0.37 to 1.26; p-value = 0.2225) (Table 14). An unadjusted analysis confirmed this finding (odds ratio 0.77, 95% CI 0.44 to 1.33).

TABLE 14

Multilevel logistic regression analysis (adjusted for participant characteristics collected at registration and for ward type) for the effect of POD on delirium occurrence (within 10 days of recruitment)

Missing data

The return rate of the EQ-5D was 98.6%, 77.5% and 65.3% at baseline, 1 month and 3 months, respectively (94–98% fully completed) (see Appendix 5, Table 34). The completion rate of the resource use questionnaire was lower at 48.7%. Participants with cognitive impairment at baseline were less likely to return the questionnaire than individuals with no cognitive impairment (see Appendix 5, Table 35).

Baseline imbalance

There was some baseline imbalance between the groups, and adjustment was required. QALYs were adjusted using baseline EQ-5D, age, ward type (orthopaedic vs. general), sex and cognitive impairment status (yes vs. no).

Validity of patient outcome assessments

A significant, positive correlation existed between the EQ-5D and NEADL scale scores at 3 months (r = 0.66), indicating that they measure similar constructs in this patient group (see Appendix 5, Figure 18). The trial sample had lower EQ-5D scores at baseline than UK age-matched population norm averages (reported in Kind et al.⁷⁷) (see Appendix 5, Table 37). Patients who experienced delirium had a lower average baseline EQ-5D score than those who did not (0.09, compared with 0.26), and this difference was maintained across the different time points (1 month: 0.28, compared with 0.50; and 3 months: 0.15, compared with 0.43) (see Appendix 5, Table 38).

Validity of proxy outcomes assessments

At baseline, proxy-completed EQ-5D values were similar to self-completed (by participants) EQ-5D values, but, at 1 and 3 months, proxy-completed (or aided) EQ-5D completion underestimated quality of life (see Appendix 5, Table 39).

Costs

The POD group participants had higher average resource use for every health-care resource except general practitioner (GP) surgery visits and psychiatrist, psychologist or counsellor visits (see Appendix 5, Table 41). Participants in the POD group had, on average, 2.2 more overnight days in hospital and 1 more day in nursing/residential homes. Overall, the hospital inpatient stay appeared to be driving costs: mean costs were £4965 for the POD group and £4365 for the control group (see Appendix 5, Table 42). The average cost of the POD intervention was estimated as £10.98 per patient (see Appendix 5, Table 40).

Quality-adjusted life-years

The EQ-5D scores at baseline were slightly higher for the POD group than for the control group [mean 0.261 (SD 0.393) for the POD group vs. 0.234 (SD 0.347) for the control group]. Despite the fewer cases of delirium in the POD group, there were negligible between-group differences in QALYs, although, in all analyses, these were in favour of the control arm.

Cost-effectiveness

The trial-based ICER was dominated by standard care (see Appendix 5, Table 43). That is, the POD intervention resulted in higher costs and lower QALYs, albeit the QALY differential was negligible. The difference in cost varied from £920 in the complete-case group to £1127 for the complete-case and imputed items group. The difference in QALY varied from –0.01 in both imputation groups to –0.02 in the complete-case analysis (see Appendix 5, Table 43). NHS total cost and QALYs were replicated 10,000 times in a Monte Carlo simulation (see Appendix 5, Figures 20–22). Using a £20,000 per QALY threshold, the probability that the POD intervention was cost-effective was 0.01 (1% chance) in a simulation using adjusted QALYs and complete-case and imputed items. This chance increased to 10% when using unadjusted QALYs and complete-case data only. The findings were robust to sensitivity analyses.

The health economics model (see Model development) was updated using information from the trial. There were differences between some parameter values used in the original model and those observed in the trial: lower delirium incidence, lower delirium rate reduction and much lower utility values. As POD appeared to result in additional resource use (a difference of £419), a sensitivity analysis was run in which this was added to the POD cost; see Appendix 5, Table 47, for the updated model parameters and assumptions. The updated model showed that the POD system of care had an incremental cost and QALY of £1775 and 0.11, respectively, resulting in an ICER of £16,133, which indicated that the POD intervention was cost-effective (see Appendix 5, Table 48). The probabilistic sensitivity analyses yielded mean incremental costs and QALYs of £1774 and 0.11, respectively, and an ICER of £15,454. The cost-effectiveness acceptability curve with a ‘willingness to pay’ of £20,000 showed that POD had a 100% chance of being cost-effective (see Appendix 5, Figure 23). The results of the trial and model-based analyses were conflicted, and thus limit the confidence we can place on the economic evaluation results. It is unclear why this divergence occurred. Clearly, the model time horizon was much greater than that for the trial analysis and it is possible that the (albeit small) differential in delirium occurrence, when extrapolated over a lifetime, fully reflected the costs and benefits of the intervention, and thus explains to some degree the contrast between trial and model outcomes. The model also fixes the assumed relationship between delirium incidence and outcomes (length of stay, mortality and health-related quality of life), whereas, in the trial, we assessed the relationships directly; it could be that there was an unexpected or lower relationship between these factors than anticipated or that data quality led to bias.

There were significant issues relating to data quality (e.g. missing data and reliance on proxy reports); future research should seek to identify the optimal strategy for data collection in this population.

Delirium

Severity of delirium episodes, duration of delirium episodes and time to first episode of delirium (including persistent delirium) were similar between the two groups (see Appendix 6, Tables 69–74).

Falls

Nineteen falls were reported in wards receiving the POD system of care and 20 falls were reported in wards receiving control, with mean falls rates of 1.6 (SD 1.00) for the intervention group and 1.3 (SD 0.39) for the control group.

Length of hospital stay

The length of stay for patients who were discharged or who died while in hospital was similar between the groups: a mean of 9.7 days (SD 7.12 days) among those patients registered to POD and a mean of 9.8 days (SD 6.91 days) among those patients registered to the control.

Deaths

A total of 104 patient deaths were reported within 3 months of recruitment (14.6% of all registered patients): 56 (16.3% of patients registered) in the POD group and 48 (13.0% of patients registered) in the control group. Of these, 28 (26.9%) deaths occurred within 10 days of patient recruitment (see Appendix 6, Table 75). The number of deaths between the sites ranged from 5 (7.7%) to 22 (21.2%).

Discharge destination

Of those patients discharged, a larger proportion of patients registered to the POD group [176/248 (71.0%)] than patients registered to the control group [194/288 (67.4%)] were discharged home (Table 17).

TABLE 17

Discharge location by randomised arm

Of the 536 patients discharged, 118 (22.0%) had a change in discharge destination from independent to institutionalised accommodation: 47 out of 248 (19.0%) in the intervention arm and 71 out of 288 (24.7%) in the control arm (Table 18).

TABLE 18

Change between baseline and discharge location by randomised arm

Patient-reported outcomes

Raw (unadjusted) scores for the NEADL scale, Clinical Anxiety Scale and the Geriatric Depression Scale score showed little between-group differences, although control scores appeared very slightly higher for the NEADL at both baseline and 3 months (see Appendix 6, Tables 76–79).

Poor outcome

Poor outcome (defined as death, persistent delirium or change in accommodation at hospital discharge from home to residential care/nursing home or from residential home to nursing home) was similar: 80 out of 343 (23.3%) in the POD group and 72 out of 370 (19.5%) in the control group (see Appendix 6, Table 80).

Safety

There were no unexpected serious adverse events reported that were clearly attributable to the POD intervention.