Comprehensive Geriatric Assessment in hospital and hospital-at-home settings: a mixed-methods study

Types of intervention and comparison

We used the definition of the CGA that was developed by Rubenstein et al. 8 to identify studies that were eligible for inclusion in the review ‘CGA is a multidisciplinary diagnostic process intended to determine a frail elderly person’s medical, psychosocial, and functional capabilities and limitations in order to develop an overall plan for treatment and long-term follow-up’. We included RCTs that compared the CGA delivered on a specialist ward or across several wards by a mobile team with usual care on a general medical ward without the CGA. We excluded studies that did not evaluate the CGA in an inpatient setting and studies of condition-specific interventions (e.g. stroke units, geriatric orthopaedic rehabilitation),19,20 as these condition-specific interventions require specialist skills for assessment, acute management and rehabilitation.

Outcomes

The primary outcome was living at home (the inverse of death or admission to a nursing home combined) at follow-up. Secondary outcomes were death, admission to a nursing home, dependence, activities of daily living (ADL), cognitive function, length of stay, re-admission, cost and cost-effectiveness.

Inclusion criteria are detailed in Table 2.

Identification of studies

We searched the following databases with no restrictions (language or date) on 5 October 2016:

• Cochrane Central Register of Controlled Trials (CENTRAL; 2016, Issue 9) in The Cochrane Library

• MEDLINE (Including Epub Ahead of Print, In-Process & Other Non-Indexed Citations) via OvidSP (from 1946)

• EMBASE via OvidSP (from 1974)

• Cumulative Index to Nursing and Allied Health Literature (CINAHL EBSCOhost; from 1982)

• Database of Abstracts of Reviews of Effects (DARE; 2015, Issue 2) in The Cochrane Library

• Health Technology Assessment (HTA database; 2016, Issue 3) in The Cochrane Library.

We also searched the following clinical trials registers on 5 October 2016:

• ClinicalTrials.gov (https://clinicaltrials.gov)

• World Health Organization’s International Clinical Trials Registry Platform (ICTRP) (http://apps.who.int/trials/search/Default.aspx).

In Appendix 1 we detail the search strategies for MEDLINE, EMBASE, The Cochrane Library and CINAHL. We also checked the reference lists of studies included in the review, along with the reference lists of related systematic reviews and meta-analyses. 9,10,14,21–23 When we contacted triallists to request IPD, we also asked whether or not they had identified any published or unpublished data.

The titles and abstracts of all the studies identified by electronic searches were screened for inclusion by one review author (MG), full-text papers were assessed by two authors working independently (MG and GE) and disagreements were resolved by discussion with a third reviewer (SS).

Data extraction

We designed a data extraction form based on a modified version of the Cochrane good practice extraction form. 24 Sections of the data extraction form included:

• population and setting – inclusion/exclusion criteria

• methods – aim, characteristics and conduct of the randomised trials (e.g. unit and method of allocation), year and method of recruitment

• Cochrane Effective Practice and Organisation of Care (EPOC) risk of bias criteria (classified as low/high/unclear) and justification of each judgement

• participants – number of participants, mean age, male-to-female ratio, clinical details

• interventions – team members and team organisation for intervention and control groups, location, team/ward

• outcomes – outcome definition, time points measured

• results – number of events, mean, standard deviation (or other variance) and number of participants for the intervention and comparison.

Two review authors (MG and GE) independently extracted the data on all the studies that fulfilled the inclusion criteria and disagreements were resolved by discussion with a third reviewer (SS).

Survey of triallists

We sent a survey to the triallists of the 29 trials included in the review to obtain a detailed description of the CGA models evaluated in the RCTs (see Appendix 2). We contacted the investigators by e-mail or telephone and each triallist was sent a minimum of three reminders. The survey included questions on (1) the population using the service (including mean age of the population, location and inclusion/exclusion criteria), (2) the intervention characteristics (including details of the core team members, the processes of care and clinical leadership) and (3) the control group characteristics (e.g. whether or not standard assessment tools were used).

Prior to this update of the Cochrane review, we developed a logic model (Figure 1) to help explain the relationship between the key components of the CGA intervention that were intended to achieve the desired outcomes. We refined the logic model with the findings from the survey and revisited the pathways that constitute the possible mechanism of action of CGA as researched.

FIGURE 1.

Logic model. QoL, quality of life.

Risk of bias of included studies

The risk of bias of the included studies was independently assessed by three review authors (MG, GE and SS) using the Cochrane EPOC risk of bias criteria. 25 We resolved areas of uncertainty or disagreement by discussion.

Statistical analysis plan

The aims of the statistical analysis were to:

1. update the meta-analyses with published data and IPD

2. estimate the impact of the CGA on health-care utilisation costs and determine the cost-effectiveness of the CGA

3. examine whether or not the treatment effects and costs differed for patients with different levels of frailty, and to investigate any differences in care home costs between the compared services (the analysis of care home costs was not possible because of a lack of data on care home length of stay)

4. conduct a metaregression analysis to assess the effects of trial covariates on the primary outcome (living at home).

Missing data

We contacted authors of the included studies for missing data and for missing information from the trial survey.

Data synthesis

We combined published data using fixed-effect meta-analysis for living at home, death, admission to a nursing home, dependence, ADLs, cognitive function, re-admission to hospital and length of stay. We grouped trials by ward or by team for all outcomes. We calculated relative risks (RRs) for binary outcomes, standardised mean differences (SMDs) for continuous measures that used different scales to measure ADLs and cognitive function, and mean difference for continuous outcomes, such as length of stay. We analysed dependence by combining a binary definition of dependence (as defined by trials) with deterioration in ADLs. Tests of heterogeneity were undertaken using Cochran’s Q26 and the I² statistic27 and we did not retain a pooled analysis if the values of I² were > 70%.

We conducted a metaregression analysis by using a fixed-effect model to assess the effects of trial covariates on living at home at the end of follow-up period (3–12 months). 28 Trial covariates consisted of team or ward intervention, age or frailty as a criterion for targeting the delivery of the CGA (frailty typically included criteria such as geriatric syndromes, risk of nursing home admission and functional or cognitive impairment), timing of admission from emergency department directly or after 72 hours (stepdown) and outpatient follow-up. We used post-estimation Wald tests to derive F-ratios and p-values. We used metaregression to test for an interaction between the covariates (e.g. team and ward) for the outcomes of living at home at the end of follow-up (3–12 months), mortality at the end of follow-up (3–12 months), admission to a nursing home at the end of follow-up (3–12 months) and dependence.

We used Stata^® version 13 (StataCorp LP, College Station, TX, USA) and RevMan version 5 (The Cochrane Collaboration, The Nordic Cochrane Centre, Copenhagen, Denmark) when performing all analyses.

Cost-effectiveness

We conducted a cost-effectiveness analysis to examine whether or not costs and health outcomes differed between those receiving inpatient CGA and those not receiving the CGA.

We used the cost of length of stay in days from 17 trials as the main driver of resource use. We derived the costs of providing the CGA from IPD provided by one trial29,30 that evaluated a version of the CGA that included an attending geriatrician and outpatient follow-up. We valued relative costs using English unit cost prices for 2013/14,31 and a NHS health service perspective was taken. 32 We compared incremental health outcomes of the CGA versus usual care. For trials that reported the cost of the CGA we used the following measure of cost-effectiveness:

• We calculated quality-adjusted life-years (QALYs) using IPD from three trials29,33,34 that assessed patient ADL with the Barthel Index. We converted the Barthel Index to EQ-5D-3L (EuroQoL-5 Dimensions, three-level version) UK scores, based on methods described by Kaambwa et al. 35 to calculate QALYs. We selected studies with mean Barthel scores at baseline that were similar to the population in the Kaambwa et al. 35 study (Barthel score ranged from 14.8 to 16.5, on a scale of 0 to 20). We used the IPD provided by Edmans et al. 29 to validate the mapping exercise by comparing the QALYs calculated using the Barthel Index with QALYs based on EQ-5D-3L using IPD from Edmans et al. ,29 as this study provided data for the EQ-5D (EuroQol-5 Dimensions) and the Barthel Index. A meta-analysis using a fixed-effect model was performed to estimate incremental QALYs.

• We estimated life-years (LYs) using the IPD from four trials29,33,34,36 by calculating the time to death from recruitment expressed as a fraction of a year.

• Using the IPD, we created a variable ‘life-years living at home’ (LYLAHs) after discharge from hospital, as a measure of independence and well-being in an older population, based on IPD from two trials. 29,36,37

We constructed a decision model to estimate an incremental cost-effectiveness ratio (ICER) of inpatient CGA compared with inpatient care without the CGA. The ICER was expressed as a cost per QALY gained, cost per LY gained and cost per LYLAH gained from a health service perspective. We used the RR of living at home at the end of follow-up in the decision model and multiplied this by the incremental LYLAH to adjust LYLAH with the probability of living at home. The input parameters used in these models are presented in Table 3. Uncertainty in the input parameters of the model was addressed by performing 10,000 draws of all incremental cost and incremental health outcome parameters using prespecified distributions and recording incremental costs, incremental QALYs, incremental LYs and incremental LYLAHs from each draw. These results were plotted on cost-effectiveness planes and cost-effectiveness acceptability curves to display the uncertainty in the estimated ICERs.

Analysis using individual patient data

We requested IPD from the investigators of the trials included in the update of the Cochrane review of the CGA (including the original review). We contacted the investigators by e-mail or telephone and each triallist was sent a minimum of three reminders. We performed a two-stage meta-analysis of IPD with each model initially run within each trial. 40

We used fixed-effects logistic meta-analyses for two outcomes: living at home and death. 40 For a third outcome (time to death), we used fixed-effect time-to-event meta-analysis and used Cox regression models to calculate the log hazard ratio and its standard error; the pooled effect was expressed as the hazard ratio for inpatient CGA compared with general medical care. All three meta-analyses were adjusted for the participant’s age, sex and baseline index by applying a threshold score of ≤ 15, out of a maximum score of 20. 41

Sensitivity analysis

We ran random-effects meta-analyses using the DerSimonian and Laird method42 in a sensitivity analysis and compared the results with those of the fixed-effects meta-analyses used in the analyses. 40 We assessed the impact of excluding three trials43–45 that included participants who were admitted to hospital from a nursing home for the outcomes of living at home and admitted to a nursing home. We also assessed the impact of using data at 6 months’ follow-up, rather than 12 months’ follow-up, for three trials. 34,46,47

Reporting bias

We assessed reporting bias by creating a funnel plot for the main outcome (living at home) at 3–12 months’ follow-up, recognising that, if there are a small number of trials, these plots are not necessarily indicative of publication bias.

Certainty of evidence

We used the GRADE (Grades of Recommendation, Assessment, Development and Evaluation) framework to assess the certainty of the evidence by creating a summary-of-findings table, and followed the approach of the GRADE working group48 and guidance developed by EPOC. 49 We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and risk of bias) to assess the certainty of evidence as it relates to the main outcomes,48 and graded the evidence as being of very low, low, moderate or high certainty. We included the most important outcomes of living at home, mortality, admission to a nursing home, dependence, cognitive function, hospital length of stay and cost-effectiveness. Three review authors (MG, GE and SS) independently assessed the certainty of evidence.

Patient and public involvement

We established a SSC to ensure delivery, governance and advice, which met five times over the course of the project. A lay member of the public was on the committee and gave valuable feedback on various aspects of the project, including the review. We were also in contact with, and received feedback from, the Oxford-based DeNDRoN and Age UK.

Study selection

For this update, we screened 7147 titles and abstracts for eligibility in the update of the review and we excluded 7131 records. The flow of studies through the search process is outlined in Figure 2.

FIGURE 2.

The PRISMA flow diagram for identification of published studies included in review.

Characteristics of included studies

Full details of the included studies (population, intervention, comparison, outcomes and risk of bias) are described in Appendix 3. We retrieved the full text of 16 articles and identified seven eligible trials (from eight publications) to include in this update. 29,36,38,50–53 Twenty-nine RCTs (n = 13,766 participants) were included in this review (seven studies of these were from the update) from nine countries (Australia, Canada, China, France, Germany, Norway, Sweden, the UK and the USA). We received IPD from five trials;29,33,34,36,38 this limited this aspect of the analysis to a subgroup of trials (n = 1692).

Eleven trials targeted the CGA to the frailest or most at-risk participants,29,33,34,36,45,46,54–58 and 11 targeted CGA on the basis of age. 38,43,44,47,50,53,59–63 The CGA was delivered in a dedicated geriatric ward in the majority (n = 20) of trials34,36,38,43,45–47,50,51,53–57,59–62,64,65 and by using a team approach that covered more than one ward/unit in eight trials. 29,33,44,58,63,66–68 The process of the CGA in the two models is described in more detail in Figure 3.

FIGURE 3.

Components of in-hospital CGA. ●, present or carried out; ○, recommendation made or staff accessed from general pool.

The intervention was case management by a geriatrician at the point of discharge from an acute medical unit in the AMIGOS (acute medical unit comprehensive geriatric assessment intervention study) trial,29 and, in another study, the CGA intervention was care in a specialist medical and mental health unit. 36 Most trials described the control group as usual care, but in three trials the control group received enhanced care. 29,36,51 For one trial,36 the control group was a mixture of care on geriatric medical wards (70%) and general medical wards (30%). Outpatient follow-up was provided in nine trials and duration of follow-up ranged from 3 to 12 months.

Survey of triallists

Thirteen of the 29 triallists completed the survey29,33,34,36,38,43,45,46,53,57,66–68 and reported that the elements of the CGA that they considered to be critical to success were tailoring treatment plans to the individual (13 out of 13 trials), MDT meetings (12 out of 13 trials), clinical leadership (10 out of 13 trials), specialty knowledge, experience and competence (11 out of 13 trials) and involving patients and carers in goal-setting (10 out of 13 trials) (Figure 4). Interestingly, triallists reported similar staff profiles in the control group and the CGA intervention group (see Figures 3 and 5). The main exception was that only four trials reported that a specialist trained geriatrician was part of the control group (Figure 5). 33,36,38,68 MDT meetings took place in the majority of trials in the CGA intervention group (12 out of 13 trials), but in only 6 out of 13 trials in the control group. 36,38,46,57,67,68

FIGURE 4.

Key components of the CGA reported by triallists.

FIGURE 5.

Components of in-hospital control group: processes of care and staff profiles.

Risk of bias within included studies

Risk-of-bias assessments of the included studies are reported in Table 4. Two trials were available only as abstracts52,67 and these were assessed as having unclear risk of bias for each of the domains. The majority of the trials were assessed as being at low risk of selection bias (judged by sequence generation and allocation concealment), and two trials that used an open allocation schedule were assessed as having a high risk of bias. 53,61 We classified all trials as having a high risk of performance bias, as it was not possible to blind participants or personnel to the allocated intervention, and we assessed detection bias as low risk for objective measures of outcome. We assessed subjective measures of outcome as having low or unclear risk of bias in 26 trials and a high risk of bias in one trial,53 as the outcome assessors were not blinded to functional status. We assessed attrition bias as being low or unclear in 24 trials and as high in three trials,43,59,63 with one of these trials59 reporting attrition for functional outcomes of > 25%. Twenty-five trials did not publish a protocol and, therefore, we assessed them as having unclear risk of selective reporting bias; four trials did publish protocols29,33,36,67 and we assessed these as having low risk of selective reporting bias. In 21 trials, there was low or unclear risk of contamination of the control group as there was little evidence that the control group received the CGA intervention. However, in six trials it is likely that the control group received the intervention33,36,51,53,56,67 and, therefore, these trials were classified as having a high risk of bias.

Synthesis of results

Living at home

The CGA increases the likelihood of patients being alive and in their own homes at hospital discharge [RR 1.05, 95% confidence interval (CI) 1.01 to 1.10; 11 trials, n = 4346 participants, high-certainty evidence; I² = 43%],44,45,50,53,55,57–60,62,63 and at 3–12 months’ follow-up (RR 1.06, 95% CI 1.01 to 1.10; 16 trials, n = 6799 participants, high-certainty evidence; I² = 13%). 29,33,34,36,38,43–47,54,56,58,62 (Figure 6). There was little evidence of an interaction between ward and team (F = 3.54, p = 0.08; meta-regression).

FIGURE 6.

Living at home, RR (end of 3–12 months’ follow-up). ESD, early supported discharge; GEMC, Geriatric Evaluation and Management Centre; UCOP, usual care outpatients.

Mortality

The CGA results in little or no difference in mortality at hospital discharge (RR 1.04, 95% CI 0.82 to 1.32; 11 trials, n = 4346 participants, high-certainty evidence; I² = 16%),44,45,50,53,55,57–60,62,63 or at 3–12 months’ follow-up (RR 1.00, 95% CI 0.93 to 1.07; 21 trials, n = 10,023 participants, high-certainty evidence; I² = 0%). 29,33,34,36,38,43–47,54,56,58,60–62,65,67,68 (Figure 7). There was no evidence of an interaction between ward and team (F = 0.07, p = 0.80; meta-regression).

FIGURE 7.

Mortality, RR (end of 3–12 months’ follow-up).

Admitted to a nursing home during follow-up

The CGA decreases the likelihood of patients being admitted to a nursing home at discharge (RR 0.89, 95% CI 0.81 to 0.98; 12 trials, n = 4459 participants, high-certainty evidence; I² = 31%)44,45,50,53,55,57–60,62,63,66 and at 3–12 months’ follow-up (RR 0.80, 95% CI 0.72 to 0.89; 14 trials, n = 6285 participants, high-certainty evidence; I² = 3%). 29,33,34,36,43–47,54,56,58,62 (Figure 8). There was evidence of an interaction between ward and team (F = 7.64, p = 0.02; meta-regression).

FIGURE 8.

Admission to a nursing home, RR (end of 3–12 months’ follow-up).

Dependence

The CGA results in little or no difference in dependence (RR 0.97, 95% CI 0.89 to 1.04; 14 trials, n = 6551 participants, high-certainty evidence; I² = 0%). 29,34,38,43–45,47,50,56,59,60,62,68 (Figure 9). There was no evidence of an interaction between ward and team (F = 0.61, p = 0.45; meta-regression).

FIGURE 9.

Dependence, RR.

Cognitive function

Five trials reported cognitive function at follow-up, but because of the high level of statistical heterogeneity we did not retain the meta-analysis (n = 3534 participants; low-certainty evidence; I² = 73%). 33,36,43,58,67 The SMD ranged from –0.23 to 0.35 (Figure 10).

FIGURE 10.

Cognitive function, SMDs.

Length of stay

There was a high level of statistical heterogeneity in the 17 trials that reported length of stay; therefore, we did not retain the meta-analysis (n = 5309 participants; low-certainty evidence; I² = 80%). 29,34,36,38,43,44,53,54,56,58,60,61,63,66,68 Mean hospital length of stay ranged from 3.4 days to 40.7 days in the CGA group and from 3.1 days to 42.8 days in the control group (Figure 11).

FIGURE 11.

Length of stay, mean differences. WMD, weighted mean difference.

Cost-effectiveness

We used the meta-analysis of published data from 17 trials, and IPD from five trials,29,33,34,36,38 to estimate the incremental cost and incremental health outcomes of the CGA versus usual care. Results from the main cost-effectiveness analysis are detailed in Table 5. Health-care costs per participant in the CGA group were estimated to be £234 (95% CI –£144 to £605; 17 trials, low-certainty evidence) higher than in the usual-care group. Furthermore, the CGA may lead to a slight increase in QALYs of 0.012 (95% CI –0.024 to 0.048), a slight increase in LYs of 0.037 (95% CI 0.001 to 0.073) and a slight increase in LYLAHs of 0.019 (95% CI –0.019 to 0.155) (see Table 5). The ICER in terms of QALYs was £19,802, which is close to the threshold suggested by NICE as a ceiling value for a QALY;32 the cost for a LY gained was £6305, and, for a LYLAH gained, the cost was £12,568. The probability that the CGA will be cost-effective at a £20,000 ceiling ratio for QALYs, LYs and LYLAHs was 0.50, 0.89 and 0.47, respectively (see Table 5). We have plotted the cost-effectiveness planes with ICERs expressed as cost per QALY gained (Figure 12), per LY gained (Figure 13) and per LYLAH gained (Figure 14); these give the distribution of each draw of all incremental cost and incremental health outcome parameters.

TABLE 5 - Results from the main cost-effectiveness analysis

Incremental outcomes	ICER (£)	Probability of the CGA being cost-effective at a £20,000 ceiling ratio	Estimate (95% CI)
QALY (cost–utility analysis)	19,802	0.50	0.012 (–0.024 to 0.048)
LY (cost-effectiveness analysis)	6305	0.89	0.037 (0.001 to 0.073)
LYLAH (cost-effectiveness analysis)	12,568	0.47	0.019 (–0.019 to 0.155)

FIGURE 12.

Cost-effectiveness plane with ICERs expressed as cost per QALY gained.

FIGURE 13.

Cost-effectiveness plane with ICERs expressed as cost per LY gained.

FIGURE 14.

Cost-effectiveness plane with ICERs expressed as cost per LYLAH gained.

Results from metaregression

Differences in effectiveness of the CGA delivery between wards and teams for each outcome were uncertain, and these analyses were underpowered (discharged home: F = 1.91, p = 0.20, n = 8 trials ward, n = 3 trials team; living at home at the 3–12 months’ follow-up: F = 3.54, p = 0.08, n = 12 trials ward, n = 4 trials team). There was also uncertainty about using age versus frailty as a criterion for targeting the delivery of the CGA on the main outcome of living at home [at discharge: F = 0.18, p = 0.68, n = 7 trials for age, n = 4 trials for frailty; end of follow-up (3–12 months): F = 0.98, p = 0.34, n = 5 trials for age, n = 11 trials for frailty], delivering the CGA on admission to hospital versus 72 hours after admission [at discharge: F = 0.51, p = 0.49, n = 6 trials for the CGA on admission to hospital, n = 4 trials for the CGA delivered 72 hours after admission; end of follow-up (3–12 months): F = 0.45, p = 0.51, n = 4 trials for the CGA on admission, n = 7 trials for the CGA delivered 72 hours after admission] and between outpatient follow-up versus no outpatient follow-up (at end of follow-up: F = 0.17, p = 0.69, n = 5 trials outpatient follow-up, n = 7 trials no outpatient follow-up).

Subgroup analysis using individual patient data

Results of subgroup analysis using IPD indicate that, in the five trials providing IPD (n = 1692, 12% of the total number of participants, adjusted for age, sex and frailty), there was little or no difference in the odds of living at home at the end of follow-up for participants in the intervention group versus the control group [odds ratio (OR) 0.95, 95% CI 0.74 to 1.24; I² = 0%] (see Appendix 5). 29,33,34,36,38 Similarly, results on mortality indicate little or no difference in the odds of mortality at the end of follow-up (OR 0.92, 95% CI 0.70 to 1.21; I² = 0%) (see Appendix 5). Time-to-event analysis showed little or no difference in the time to death [hazard ratio (HR) 0.88, 95% CI 0.72 to 1.08] (see Appendix 5).

Sensitivity analysis

Rerunning the analyses using random-effects rather than fixed-effect models, or removing trials that did not exclude participants who were admitted to hospital from a nursing home, had little effect on the associations (data not shown).

Publication bias

The Harbord test (bias = 0.87, p = 0.18) and Egger’s test (bias = 0.87, p = 0.17) show little evidence of small trial bias for the main outcome of living at home at the end of follow-up (3–12 months).

Grades of Recommendation, Assessment, Development and Evaluation

The summary of findings is given in Table 6.

Logic model

We refined the protocol logic model to map out the chain of events, or pathways, that constitute a mechanism of action of the CGA as researched, using the information we received from the survey of triallists (Figure 15). We have highlighted these additions (specialty knowledge, experience and competence, tailoring treatment plans to the individual and involving patients and carers in goal-setting) in italics (see Figure 15).

FIGURE 15.

Logic model: the CGA in hospital from Cochrane review and survey of triallists. Italicised text indicates additions to the protocol logic map, as a result of information received from the survey of triallists.

Survey development and piloting

Response rate

Of the 41 community trusts or health boards contacted, 14 were excluded as they did not provide an out-of-hospital community-based service, leaving a sample of 27. We did not identify a Health and Social Care Trust in Northern Ireland that provided the CGA in the community at the time we ran the survey (2014–15). Of the 27 community trusts or health boards, 19 completed the survey. Three community trusts or health boards completed the survey but did not describe the type of service they provided. Eight respondents were interviewed, including five consultant geriatricians, two consultant nurses and a community psychiatric nurse team leader.

Survey findings

Of the 24 services described, 11 were based in Scotland, six were based in England and six were based in Wales; one respondent did not name the trust/health board (Table 8). We received two responses for Ayrshire and Arran, Greater Glasgow and Clyde, Lothian, Tayside, and Solent, and each described two services.

Of the 19 community trusts/health boards detailed in Table 8, 11 (57.9%) reported that the acute trusts provided a similar admission avoidance hospital-at-home service.

Interview findings

We interviewed lead clinicians from eight of the trusts to obtain more details about the services that they provided. Two of the leads described admission avoidance hospital-at-home services (Fife and West Lothian) and two described intermediate care services (Ayrshire and Arran, and Dorset). Ayrshire and Arran intermediate care service was led by a senior specialist nurse in geriatric medicine and provided step-up and step-down beds. A summary of the information provided by these four trusts is reported in Table 14. The service in Dorset was described as intermediate care and provided admission avoidance hospital at home and early supported discharge with the same staff working in both services (including a consultant geriatrician).

Four other trusts generally provided a lower level of care that was not an alternative to inpatient care. These were:

1. Betsi Cadwaladr Community NHS Trust (Wrexham) Home Enhanced Care Services, which was supplementing primary care

2. Solent Community NHS Trust admission avoidance hospital at home/early discharge and integrated care management, which was case management of frail older people in the community

3. Liverpool Community NHS Trust/Liverpool Frailty Service, which was a combination of a community and acute care package (a frailty pathway), and involved working across three NHS trusts

4. Cwm Taf Community NHS Trust/Community Mental Health, which was community care incorporating parts of admission avoidance.

The detailed summary table for these four trusts is included in Appendix 8. Those interviewed described both perceived successes and difficulties in implementing the CGA service, and these are described below.

Service evaluations

We received service evaluations for Fife, Lanarkshire and West Lothian admission avoidance services, Ayrshire and Arran, and Dorset intermediate care services and a community mental health service for older people (Cwm Taf). The content of the service evaluations varied from trust to trust.

Setting and data

We used patient-level data collected by three of the 14 Scottish health boards of all patients aged ≥ 64 years who were admitted (referred to as the index admission) to either admission avoidance hospital at home or inpatient hospital between August 2014 and December 2015 (17 months) in sites 1 and 2, and between January 2015 and December 2016 (24 months) in site 3. These three health boards cover a population of almost 1.5 million people in urban and rural areas. The ISD, part of NHS Scotland, de-identified and cleaned the data. We obtained signed release forms from each health board’s Caldicott guardian.

Data were available for each person for 2 years prior to their index admission and from the point of their index admission to 6 months after index discharge from hospital at home or hospital. Box 2 presents a full list of all variables included in the data set. Figure 16 provides schematic examples of the differing calendar time periods studied before and after index admission for people admitted between August 2014 and December 2015 to hospital at home (patients A and B) or hospital (patients C and D) in site 1. As this illustrates, the follow-up period for each patient included the period between index admission and index discharge, and 6 months after index discharge.

FIGURE 16.

Illustration of obtained data from site 1.

Selection of patients in the hospital-at-home and control cohorts

We included patients aged ≥ 65 years who were classified as an unscheduled admission to general or geriatric medicine. In the control cohort, we excluded those with a diagnosis that would not be eligible for management through hospital at home; these exclusions included acute intracerebral crisis (intracerebral infections, trauma or haemorrhage), stroke and related codes, acute coronary syndromes and myocardial infarction, surgical emergencies including vascular, urological, gynaecological and general surgical presentations, orthopaedic diagnosis of fractures and trauma, cardiothoracic diagnoses, poisoning and complications of surgery. We also excluded from the control group those who had a diagnosis [i.e. primary and secondary ICD-10 (International Classification of Diseases, Tenth Edition) codes] that was not observed in any of the hospital-at-home admissions in each setting (1081 patients in site 1, 1405 in site 2 and 451 in site 3) (Figure 17). Each patient was counted as a single episode of health care.

FIGURE 17.

Flow chart of study population. A&E, accident and emergency; ACS, acute coronary syndrome; HAH, hospital at home; MI, myocardial infarction; TIA, transient ischaemic attack.

Intervention and intervention costs

All service models of hospital-at-home in the three geographical areas provided an admission avoidance function and had similar structures, but with differences in the type of professionals included in the service team, the capacity of the service and the assessment process for patients referred to the service.

We collected data on the costs of hospital-at-home using a template derived from the Cost-It tool of the World Health Organization. 73 The cost categories included staff, training, transport, information and communication, clinical materials/equipment, support services, laboratory services, diagnostics, overheads and other costs. Clinician managers supported by finance staff in the three health boards completed this template based on the actual spending for the hospital-at-home service for the time periods covered by the ISD data. The cost per hospital-at-home admission was calculated by dividing the total costs of the hospital-at-home service by the total number of hospital-at-home admissions during the same period.

Characteristics of the admission avoidance hospital-at-home services and the comparison inpatient hospital care services are detailed in Table 15.

Statistical analysis

We used an iterative approach to the analysis, starting with a description of the two cohorts (i.e. those admitted to hospital at home and those admitted to hospital) for each health board. We calculated means, standard errors and frequencies to describe differences in patient characteristics at index admission and tested differences using a Mann–Whitney test for continuous variables and a chi-squared test for categorical variables. We also estimated the mean differences in resource utilisation costs (with bootstrapped standard errors) and the unadjusted RR of mortality between the two cohorts for each health board. In addition, we investigated the association of being admitted to hospital at home or hospital with mortality and cost over a minimum follow-up period of 6 months. To do this, we followed the Medical Research Council guidelines on performing natural experiments. 74 In line with the literature, we adopted a step-wise strategy to select the propensity score matching (PSM) technique that most reduced observed confounding between the two cohorts in each health board. 75–77 First, we matched the two cohorts in each setting using a range of the most commonly used PSM techniques, including Mahalanobis, 1 : 1, K-to-1, kernel, local linear regression, spline and inverse probability weighting techniques. Second, the performance of each PSM technique on covariate balancing was assessed based on the mean and median percentage standardised bias as well as Rubin’s B [the absolute standardised difference of the means of the linear index of the propensity score in the treated and (matched) non-treated group] and Rubin’s R [the ratio of treated to (matched) non-treated variances of the propensity score index]. Following Rubin’s78 recommendation, we considered B < 25 and R between 0.5 and 2 to indicate sufficient balance. Third, we chose the PSM technique that had the lowest values on these performance indicators in each of the three health boards. We matched the two cohorts in each health board by sociodemographic characteristics (i.e. age, gender, socioeconomic status), diagnosis code (i.e. primary and secondary ICD-10 code) of index admission, morbidity (i.e. type of long-term condition), mortality during follow-up (for the analysis of cost), 2-year costs prior to the index admission (by cost category as listed in Box 2) and date of index admission (to account for seasonal trends).

We performed a doubly robust estimation to further reduce confounding by using a regression analysis after performing the most suitable PSM technique and including the confounding variables listed above as covariates. 79 In the regression, we used generalised linear regression models (GLMs) with gamma distribution and log link to investigate the association of hospital at home with total costs during the follow-up period, and total costs in 6 months following index discharge. We also used GLMs with Poisson distribution and log link to estimate the RR of mortality. Robust standard errors were specified in all regression models. We calculated Kaplan–Meier survival curves, with and without using the weights from the PSM, and used log-rank tests to test the equality of the survival functions.

Subgroup analysis

We conducted a subgroup analysis, running the same regression models used in the main analysis, to investigate the association of hospital-at-home services with costs and mortality for the population that had a diagnosis of dementia. We considered this population to be important because of their complex health-care needs and the increasing prevalence of dementia. 80,81 In a second subgroup analysis, we excluded patients who died during the follow-up period and investigated the association of hospital at home with costs. In both subgroup analyses, PSM was performed to match subcohorts in each setting.

Sensitivity analysis

In a univariate sensitivity analysis, we reduced and increased the intervention cost of admission avoidance hospital at home by 50%, as there are no standard unit costs to benchmark these types of services and we were concerned that costs for these services may vary because of economies of scale, size, experience, setting, human resource capacity and error. This sensitivity analysis was expected to affect the costs during index admission and the costs of admission to hospital at home in the 6 months after discharge.

Characteristics of the population cohorts

Between August 2014 and December 2015 (17 months), 1771 patients were admitted to hospital at home in site 1; between January 2015 and December 2016 (24 months), 1547 patients were admitted to hospital at home in site 2; and between August 2014 and December 2015 (17 months), 443 patients were admitted to hospital at home in site 3 (see Figure 17). In the same period, there were 14,220 patients admitted to hospital in site 1, 5399 patients admitted to hospital in site 2 and 2295 patients admitted to hospital in site 3.

There were few differences between each of the hospital-at-home cohorts, with the main difference being that a larger proportion of the population in site 3 lived in a more affluent area (i.e. scored ≥ 5 on the Scottish Index of Multiple Deprivation). Patients admitted to hospital at home were, on average, 3–4 years older than those admitted to hospital and were more likely to be female (ranging from 5% in site 3 to 9% in site 2) and a higher proportion had more than four long-term conditions (approximately 7%) than patients admitted to hospital (Table 16). The largest difference between those admitted to hospital at home and those admitted to hospital in sites 1 and 3 was in the proportion of patients with dementia (10% higher in the hospital-at-home cohorts than in the hospital cohorts), whereas, in site 2, the largest difference was the proportion of patients with renal failure (also 10% higher in the hospital-at-home cohort than in the hospital cohort).

We compared the two cohorts in each site from index admission to 6 months post discharge from hospital at home or hospital (Table 17). In all sites, there was, on average, a higher percentage of deaths in those receiving health care in hospital than in those receiving health care in hospital at home (6% vs. 1% in site 1; 6% vs. 3% in site 2; 4% vs. 1% in site 3). There was a lower percentage of deaths in the group of patients who had received health care in hospital at the end of the follow-up period (i.e. during index admission and 6 months after discharge) (21% vs. 28% in site 1; 22% vs. 32% in site 2; 17% vs. 27% in site 3). Patients in the ‘hospital at home’ cohort lived an average of 8 (site 1), 10 (site 2) and 12 (site 3) fewer days during the whole follow-up and their index admission was, on average, shorter in site 1 (mean difference –2.64, 95% CI –2.97 to –2.31) and site 3 (mean difference –2.02, 95% CI –2.66 to –1.37) and longer in site 2 (mean difference 1.25, 95% CI 0.86 to 1.64).

The cost during a hospital-at-home admission was, on average, lower than the cost of hospital admission in site 1 (mean difference –£2318, 95% CI –£2420 to –£2217) and site 3 (mean difference –£1096, 95% CI –£1398 to –£793), and slightly lower (mean difference–£153, 95% CI –£277 to –£29) in site 2. In the hospital-at-home cohort, these costs included the intervention costs of delivering the service at home, which were £628 per admission and £113 per day in site 1, £2928 per admission and £398 per day in site 2 and £865 per admission and £118 per day in site 3. In each health board, staff were the major driver of the intervention (i.e. hospital at home) cost (site 1 95% of health-care cost, site 2 87% of health-care cost, site 3 94% of health-care cost). Detailed information on the interventions costs in each site is presented in Appendix 9.

In the 2 years prior to the index admission, each of the three hospital-at-home cohorts incurred higher average health-care costs than the cohorts that were admitted to hospital. In site 1, this was 40% (mean difference £3219, 95% CI £2513 to £3925) higher, driven primarily by higher costs of non-elective hospitalisation. We observed a similar pattern in sites 2 and 3, where the mean costs in the hospital-at-home cohort were on average 56% (mean difference £5064, 95% CI £3984 to £6143) and 57% higher (mean difference £4115, 95% CI £2467 to £5764), respectively, and, again, this was because of non-elective hospitalisation. In the 6 months following discharge, and excluding the costs of the index admission, costs were on average 43% higher (mean difference £1839, 95% CI £1423 to £2255) in site 1 for those who had been admitted to hospital at home, 16% higher (mean difference £875, 95% CI £156 to £1595) in site 2 and 92% higher (mean difference £3068, 95% CI £2178 to £3958) in site 3. The larger increase in costs in all settings was because of higher non-elective hospitalisation costs in the group of patients who had received hospital-at-home care (mean difference £1517, 95% CI £1134 to £1899 in site 1; mean difference £529, 95% CI –£77 to 1135 in site 2; mean difference £2618, 95% CI £1779 to £3458 in site 3) during the 6-month follow-up.

When the cost of the index admission was included in the analysis, the cost during follow-up (i.e. including the index admission and 6-month health-care resource use after index discharge) was 6% lower (mean difference –£480, 95% CI –£996 to £36) in the hospital-at-home cohort than in the control cohort in site 1, whereas in site 2 these costs were 8% higher (mean difference £722, 95% CI £32 to £1413) and in site 3 they were 35% higher (mean difference £1973, 95% CI £1019 to £2927).

Compared with the control cohort, the mean costs per lived day were 13% lower (mean difference –£12, 95% CI –17 to –£6) in the hospital-at-home cohort in site 1 but 34% higher (mean difference £37, 95% CI £18 to £56) and 66% higher (mean difference £36, 95% CI £18 to £53) in sites 2 and 3, respectively.

Selection of propensity score matching technique

In the propensity score matched analysis, there were 1696 (site 1), 925 (site 2) and 427 (site 3) patients in the hospital-at-home cohort and 11,571 (site 1), 3849 (site 2) and 1683 (site 3) patients in the hospital cohort (see Figure 17). Local linear regression matching was the best PSM technique to match the cohorts in sites 1 and 3 for costs and mortality because it resulted in a lower mean (1.5 and 1.8, respectively) and median (1.2 and 1.6, respectively) percentage standardised bias, as well as the lowest Rubin’s B (9.4 and 9.6, respectively). Based on the same criteria, kernel matching was selected to match the cohorts in site 2. Rubin’s R was within the suggested range (i.e. from 0.5 to 2.0) in the selected techniques. These results are presented in Appendix 10.

Main propensity score matched analysis

The results of the main analysis are presented in panel A in Table 18. After PSM and regression analysis, the health-care cost during index admission in hospital at home and over 6 months after index discharge was, on average, 18% lower (ratio of means 0.82, 95% CI 0.76 to 0.89) than admission to hospital in site 1. Excluding the cost of the index admission (hospital at home or hospital), the costs during the 6 months following discharge for those who had been admitted to hospital at home were, on average, 27% higher (ratio of means 1.27, 95% CI 1.14 to 1.41) than for patients who had been admitted to hospital. In site 2, the difference in costs between the cohorts was close to zero (ratio of means 1.00, 95% CI 0.92 to 1.09) during the index admission and 6-month follow-up period and 9% (ratio of means 1.09, 95% CI 0.95 to 1.24) more costly in the 6 months after index discharge (i.e. excluding the index admission). In site 3, patients admitted to hospital at home had on average 15% higher costs during the entire follow-up period (ratio of means 1.15, 95% CI 0.99 to 1.33) and 70% higher costs during the 6 months after discharge (ratio of means 1.70, 95% CI 1.40 to 2.07) than patients admitted to hospital.

There may be an increased risk of mortality in all three hospital-at-home cohorts (site 1 RR 1.09, 95% CI 1.00 to 1.19; site 2 RR 1.29, 95% CI 1.15 to 1.44; site 3 RR 1.27, 95% CI 1.06 to 1.54) compared with the hospital cohort after PSM and regression analysis to adjust for confounding. The Kaplan–Meier survival curves presented in Figure 18 show higher survival rates in the inpatient control cohorts in all three health boards and, after weighting with the propensity score, the control cohort in site 2 still had a higher survival rate than the hospital-at-home cohort. The difference in statistical significance of survival in site 3 between the results reported in Table 18 and the survival curve after weighting is explained by the fact that Kaplan–Meier curves are weighted only with the propensity score without performing an additional regression analysis. Full results of the main analysis are presented in Appendix 11.

FIGURE 18.

Survival curves before and after PSM. HAH, hospital at home. (a) Kaplan–Meier survival estimates unweighted (site 1) log-rank test χ² (1) = 40.73, pr > χ² = 0.0000; (b) Kaplan–Meier survival estimates weighted (site 1) log-rank test χ² (1) = 1.06, pr > χ² = 0.3026; (c) Kaplan–Meier survival estimates unweighted (site 2) log-rank test χ² (1) = 60.13, pr > χ² = 0.0000; (d) Kaplan–Meier survival estimates weighted (site 2) log-rank test χ² (1) = 11.18, pr > χ² = 0.0008; (e) Kaplan–Meier survival estimates unweighted (site 3) log-rank test χ² (1) = 21.81, pr > χ² = 0.0000; (f) Kaplan–Meier survival estimates weighted (site 3) log-rank test χ² (1) = 3.33, pr > χ² = 0.0680. Note: the cohorts in each setting were matched on age, gender, socioeconomic status, primary and secondary ICD-10 codes of index admission, type of long-term condition and 2-year costs prior to the index admission (by cost category as listed in Box 2). Weighted refers to weighting the observation of each patient based on the propensity score to be in the hospital-at-home cohort as described in the PSM section (see Chapter 4, Statistical analysis).

Results of the subgroup analysis

Patients with dementia (see panel B in Table 18) admitted to hospital-at-home services in sites 1 and 2 had ≈25% lower costs (site 1 ratio of means 0.75, 95% CI 0.65 to 0.87; site 2 ratio of means 0.76, 95% CI 0.66 to 0.88) during the index admission and 6 months post index discharge. After excluding the index admission period, the same difference in mean costs remained in site 2. We found that the population who were admitted to hospital at home and had a diagnosis of dementia may have an increased risk of death (site 1 RR 1.05, 95% CI 0.89 to 1.24; site 2 RR 1.41, 95% CI 1.19 to 1.67; site 3 RR 1.65, 95% CI 1.12 to 2.41) compared with those who had a diagnosis of dementia and who were admitted to hospital.

When we excluded people who died during follow-up (i.e. during index admission and 6 months after discharge), patients admitted to hospital at home in site 1 had lower costs (ratio of means 0.85, 95% CI 0.77 to 0.94), whereas there was 11% increase in costs in site 2 (ratio of means 1.11, 95% CI 1.00 to 1.25) and a 20% increase in site 3 (ratio of means 1.20, 95% CI 1.00 to 1.43). The mean costs were higher in the hospital-at-home cohort when the costs during the index admission were excluded (ratio of means 1.23, 95%CI 1.08 to 1.40 in site 1; ratio of means 1.17, 95% CI 0.99 to 1.38 site 2; ratio of means 1.71; 95% CI 1.36 to 2.15 in site 3) than for patients admitted to hospital (see panel C in Table 18). Full results of the subgroup analysis are presented in Appendix 11.

Results of the sensitivity analyses

The results from the sensitivity analysis (see panel D in Table 18) showed that patients in the hospital-at-home cohort had 13% lower costs (ratio of means 0.87, 95% CI 0.81 to 0.94) during the follow-up period (i.e. during index admission and 6 months after index discharge) when the hospital-at-home service costs were assumed to be 50% higher than in the main analysis. In site 2, the results from the sensitivity analysis showed that the uncertainty in hospital-at-home service costs led to increased costs or cost savings by about 18% during the whole follow-up period. In site 3, the sensitivity analysis showed a 23% cost increase (ratio of means 1.23, 95% CI 1.07 to 1.42) if the intervention costs of hospital at home were 50% higher.

Overview of methods

We performed a comparative case study using qualitative methods90 to examine the meaning and delivery of the CGA in different health-care settings (Figure 19). The analytical focus was the CGA within the context of the health care delivered and received, and it recognised that service models evolve over time (illustrated in changes to participating services over the course of the study). At the outset, we had discussions with staff at each site to understand the history, purpose and set-up of each service, and supplemented these with reference to local policy documents, assessment and care planning documentation. We followed these discussions with focus groups of patients and carers, before directly observing the delivery of health care and conducting interviews with staff who were involved in delivering the CGA and with patients and their caregivers. 91–93 We obtained permission to conduct the study from the Oxford C National Research Ethics Service (NRES) Committee South Central (Health Research Authority) (reference number 15/SC/0266).

FIGURE 19.

The sequence of research activities.

Focus groups

We conducted four focus groups with older people who had recent experience of admission to hospital or hospital at home, either as a patient or as a caregiver (usually a partner). We asked participants to describe the event leading up to the acute admission, how they felt, what was communicated to them and what happened at each stage. We encouraged them to discuss other dimensions of delivery and the priority they attached to these. The findings from the focus groups informed the interview topic guides. 94

Selection of sites for observation and interviews

We initially selected two locations and had to identify an additional site as there was a substantial delay in the set-up of the research in one of the sites as a result of the reorganisation and eventual merge of two trusts. One site was an Older People’s Assessment Team that was established in 2013 in a trust in southern England that provided health care to a population of around 460,000 in a mainly rural area. The team covered an Older People’s Assessment Unit and also the accident and emergency department. The second site (hospital at home 1) was a hospital-at-home service provided by a city teaching hospital in northern England, covering a population of 500,000; the organisational hub for the service was located a short distance from the main hospital site and received referrals from the Older People’s Assessment Unit and GPs. The third site (hospital at home 2) was in Wales, and here we recruited from the rapid response element of a hospital-at-home service that is part of a Frailty Programme and covers a population of close to 150,000. This was of interest in that the Frailty Programme offered an integrated model of care that spanned acute, community, primary and social care services. We also recruited participants from a rehabilitation ward in a community hospital that provides inpatient and outpatient care services, accommodates patients waiting for placement in nursing or residential care homes and accepts direct admission from home by the Frailty team. The rehabilitation ward is under the care of a consultant geriatrician.

These sites were selected because they represented both urban and rural settings, the size of the geriatrician-led hospital-at-home services differed (one service had up to 30 beds; the second service had 50 beds) (Table 19) and the acute assessment focus of the Older People’s Assessment Unit in southern England contrasted with the rehabilitation focus of the community hospital ward in Wales.

Observation and shadowing

We scheduled a short period of general observation (an average of 6 hours in each setting) prior to conducting the interviews. This included attending the virtual ward rounds for hospital at home, meeting with the MDT and attending the MDT meetings in all sites. Our interest was on how the work of the MDT was organised and delivered to patients, and the mechanisms and systems in place to enable co-ordination of assessment, care planning and discharge.

Interviews with staff, patients and carers

We conducted semistructured interviews with a purposive sample of staff from different disciplines (nursing, medical, care, therapy, social worker and voluntary sector worker) who were directly involved in the assessment of need and delivery of treatment, therapy and care; we pursued lines of inquiry that were unanticipated at the outset, but that emerged through reflection of the conversation with the participant. We sought to explore the concept of the CGA and understand how it was delivered in each service model. We intended to interview approximately four or five team members in each study site.

Interviews with patients and caregivers

Staff at each site referred eligible participants to the research team, that is those who were not severely ill, receiving palliative care or close to death, and who were typical of those receiving the CGA, for example older people who had symptoms or signs of frailty and who had experienced an acute medical event. We conducted individual interviews with patients and/or their caregivers at, or near the point of, discharge from the service. We included the option of interviewing patients and their caregivers together, as this can provide insight into how problems associated with poor health are managed. 95,96 Our aim was to interview five or six patients who had received the CGA hospital at home and the CGA hospital care, and their caregivers, from each of the included service settings. We left open the possibility of undertaking a small number of additional interviews to pursue promising lines of inquiry not anticipated in advance. Although we used a topic guide, the interviews had a conversational style to enable people to talk about the areas that were important to them. We were interested in their perceptions and experiences of their health event and the health care received.

Interviews were audio-recorded and subsequently transcribed. The data set for each patient included biographical information (e.g. age, household composition, informal and formal support and health problems), the interview context and the interviewer’s observations.

Consenting process

Potential participants (patients and caregivers) who were referred to the study were provided with an information leaflet and the opportunity to discuss the study with the staff who provided care and the researcher. The consent procedure adhered to the Mental Capacity Act 2005. 97 Staff invited to be interviewed were also provided with an information sheet and were asked to sign a form to consent to the interview.

Analysis

The analysis considered factors that contribute to the CGA as a multidimensional diagnostic and therapeutic process that intends to determine an older person’s capabilities and limitations (medical, functional, mental and social) to ensure that problems are identified and managed appropriately. 8 As a starting point, we considered the components that were considered critical to the success of the CGA, as reported by 13 triallists (Box 3). 12 These concepts were expanded for the purpose of this analysis, for both inpatient and hospital-at-home models of the CGA. The intention was to develop a narrative that explored the assumptions within the existing logic model for the CGA (Figures 1 and 15), whereby professionals’ understandings of activities, outputs, outcomes and impact might be refined and informed by understandings of their significance for patients and caregivers.

Data were analysed using a framework approach to facilitate a comparative analysis:98,99 (1) reading transcripts to gain familiarisation with all data, noting key issues in relation to the research questions; (2) generating a thematic framework that allowed integration of core components from the Cochrane review of the CGA together with key issues identified in stage 1; (3) applying the framework systematically to the transcripts, coding the data in accordance with the framework within NVivo 11 (QSR International, Warrington, UK) and adding additional themes where appropriate; (4) creating a chart for each main theme, with a row for each health-care service and columns for themes; and (5) review of the chart for patterns and associations, to develop the final analysis. The framework method was selected as it facilitates management of a large data set in a way that supports answering the research questions and comparing data across cases, while situating each perspective in the context of other aspects of each individual’s account. 20 The research team critically reviewed the analysis at various stages of progression, allowing for an iterative construction of findings, ensuring comprehensiveness and explicitly incorporating a range of perspectives. 100

Description of the services delivering Comprehensive Geriatric Assessment

One function of the services was to support the effective use of in-hospital acute beds for older people, albeit by providing support at different stages of an episode of health care. The hospital-at-home services provided an alternative to admission to hospital or facilitated early discharge from hospital, the medical assessment unit (MAU) provided a type of early triage to support transfer to another health-care setting and the community hospital provided rehabilitation before determining the discharge setting. Although each service differed in terms of organisation, location in relation to the acute hospital and function within the wider health system, these CGA models are examples of the dynamic nature of the CGA as developed in a UK NHS context and they also represent the challenge of delivering integrated care to frail older people, whose needs transcend the boundaries of acute care.

One model, an Older People’s Assessment team, is physically located within a MAU. The team proactively reaches into accident and emergency and the MAU to provide a timely comprehensive assessment of need and to plan the care for older patients.

Two models provide a hospital-at-home service for older people. These deliver a type of rapid response that provides acute medical and nursing care in the person’s home as one component of an integrated system of nursing, therapy and enabling support that spans acute, community and social care for frail older people. The provision of health care might include physiotherapy, occupational therapy, reablement assistance to enable people to resume their daily life routines and social care support with personal care and tasks of everyday living. Criteria for admission of a patient are similar for both hospital-at-home services and include medical or physical deterioration, for example from an acute respiratory infection, that requires comprehensive management and intravenous administration of antibiotics in the community setting, or for an acute exacerbation of a chronic condition, such as heart failure, or following a fall. The rationale is that the service is most effective for those patients who have not spent time in hospital, or who have been in hospital for a short period of time but are still potentially medically unstable and need additional monitoring, with recovery time facilitated by clinical and therapy staff. Whatever the pathway, suitability is dependent on the need for support for up to four times daily with regular clinical review.

A fourth service is a community hospital that provides a more traditional model of the CGA on a rehabilitation ward. Features of the models are summarised in Table 19. It is important to note that, in each case, the services are not static; features, such as team complement and bed capacity, can vary over time.

Participants

A total of 20 professionals were interviewed: five from the Older Person’s Assessment team, six from one of the hospital-at-home services, five from the second hospital-at-home service and four from the community hospital inpatient ward.

The findings are organised around three overarching, albeit related, organisational and service delivery categories of the CGA: (1) assessment, (2) care and treatment planning and (3) team working.

Assessment

Box 4 provides an overview of the focus for each service, their approaches to CGA documentation and temporal considerations for their assessments.

The Older People’s Assessment team provides a comprehensive multidisciplinary assessment, close to the point of acute admission of older people. Patients are assessed within 12 hours of arriving in accident and emergency and an initial treatment plan and appropriate support is mobilised. Its distinctive feature is the provision of the CGA to those who require acute health care: ‘the frailest and most complex’:

. . . within the first 12 hours of getting to the hospital . . . to do the right thing for the patient and for their family . . . whether to optimise the acute admission by making sure patients get to the right wards . . . avoid admission and turn the person around within a few hours . . . or shorten their length of stay to within a day or . . . an overnight stay . . . with specialist multidisciplinary input.

Geriatrician

Any member of the team may carry out the initial CGA assessment, although specific disciplinary expertise might be drawn upon depending on patient need. A nurse explained it in the following terms:

It is a bit like a jigsaw . . . So we can actually work quite independently as practitioners, the therapy team, Age UK, so we all go off and do our own priorities in our own direction but then bring it back to the team.

Nurse practitioner

A feature of the Older People’s Assessment team is the involvement of an Age UK worker, who provides time-limited emotional, practical and social support to patients being discharged home and who require some help to manage the transition to home. This might include, for example, advice and support with accessing benefits, information about community resources and befriending, and referral for a carer assessment.

For patients referred to hospital at home (service 1) by the Older People’s Acute Assessment unit, the CGA starts in the hospital and is continued through observation of the person in their home environment by a doctor, nurse or therapist:

At home you can address most of it because it is the ideal environment.

Geriatrician

. . . we go on the basis of the assessment that follows CGA principles and the people that we tend to typically refer on . . . are those . . . who require a home-based assessment following ‘discharge to assess’ principles . . . anyone who we would feel requires an assessment in a home environment to determine . . . functional capabilities . . . or with . . . a relatively broad range of medical conditions . . . it’s difficult to apply fixed criteria or protocol to referring people into the service because of the complexity of the patients.

Geriatrician

Hospital-at-home staff estimated that the initial assessment takes approximately 1 hour to complete. They record data directly onto a standardised CGA template, saved within an electronic system that is also used by primary care. The content of the assessment can be the same irrespective of the disciplinary expertise of the professional conducting it:

. . . we take the patients through a full assessment – we check all the mobility, which obviously you can do that with a patient anyway. So we check to make sure they can do basic transfers, that they can get on and off the bed, that they’re able to make their own meals and we follow them up with like an enablement process anyway, so it is really patient-centred, but obviously if we need any information then we will contact the family if they’re not present, for any concerns of theirs.

Senior nurse

There was some variation in the way that hospital-at-home services were organised: therapists were not always familiar with the term CGA and, at times, separate assessments were conducted by different members of the MDT. This might reflect how the services were established and the availability of resources to support a shared assessment:

It would possibly be helpful to work more with the medical team . . . make one assessment rather than sort of three separate assessments, which happens.

Physiotherapist

In describing the process, the majority of staff interviewed made explicit reference to the concept of the CGA. Within hospital-at-home services, they also reported being able to give greater attention to a patient’s social circumstances, finding it easier to establish goals relating to a patient’s functional status prior to their recent health event and to engage with families:

A full health and social MOT [Ministry of Transport test] and takes about 1 hour to conduct. It is a full systemic enquiry from a medical point of view and then we look at social needs and mobility. So we get a full overview of that patient’s functionality within their home environment.

Senior nurse

If we didn’t use this CGA then it would be quite easy to miss parts of an assessment . . . we could potentially create more visits or miss diagnoses.

Occupational therapist

For us, CGA is a way of finding other illnesses besides what they were referred with . . . when we go and do the CGA, then they get a full examination.

Nurse

The conception of the assessment process appeared to be based on obtaining an understanding of the health event that required admission to one of the services, the social situation and the home environment. Although the CGA could inform decision-making around where a person was discharged to, it was not always possible to also consider other aspects of living with long-term conditions that contribute to patient well-being.

Care and treatment planning

When formulating a treatment and care plan, the content of the assessment is considered in the multidisciplinary ‘board round’ meeting, held on weekday mornings within each of the hospital-at-home services, and at a similar type of meeting held several times during the day in the case of the Older People’s Assessment inpatient team and weekly in the community hospital. Within each of the hospital-at-home services, the MDT includes senior experienced clinical staff and there is provision for additional reablement support with functional and personal care needs. In the context of one of the hospital-at-home services, the therapy and enabling support is provided for up to 6 weeks, depending on availability, to optimise function and determine longer-term formal care needs. However, organising the delivery of health care from the various disciplines can be challenging because of the different time scales:

For us the problem . . . is the different timescales in that the medical and nursing response is acute and rapid . . . whereas the physio[therapist] and OT [occupational therapist] on the care side of the service doesn’t have that same limitation on them . . . so sometimes it can be difficult if you identify an urgent need that requires their input because their work is more scheduled and sometimes it is more difficult to pull someone immediately to address it but . . . we almost always manage it. The thing that probably lets us down more often than not is availability of care, if all the care slots are used up.

Senior nurse

One hospital-at-home team can provide a ‘bridging’ package of care, delivered by their own staff, until formal carer input can be arranged and initiated. It was notable that members of the MDT who were interviewed referred to ‘goal-setting’ when describing the process of care planning. Thus, part of decision-making around treatment was establishing goals based on the knowledge acquired of the person and how they were functioning prior to the event that brought them into the service. It was less clear how far it involved negotiation, and this is considered below through patients’ and caregivers’ perspectives.

Interviewees acknowledged that, although older people may wish to remain in their own homes and not be admitted to hospital, they and their families also require reassurance that they will be ‘safe’ and ‘looked after’ in the absence of visible support mechanisms of a hospital ward. Generating reassurance was seen to require clear communication about the service purpose, detailed information about the plan of care and the delivery of the treatment and responsiveness to a change or deterioration in the patient’s condition:

In terms of looking at caring for someone at home you have to have that comprehensive overview. You can’t possibly manage someone without knowing as much about them as possible. That obviously isn’t just medical, it is social aspects as well.

Senior nurse

Two features of how multidisciplinary work was conveyed present questions about the level of engagement of patients and caregivers in decision-making. First, in describing the encounter with the patient for the CGA assessment, the presence of a relative/caregiver was perceived as desirable in order to extend knowledge of the patient. For those patients living with dementia, the caregiver was seen to be a critical source of information about the person, especially in instances in which the patient was unable to provide it. Second, care plans and records of staff members’ actions are typically left in the patient’s home and are seen as one means through which communication may occur, yet these require a level of engagement on the part of patients and caregivers that may not be feasible in the context of acute illness.

The sharing of information between professionals was supported by the use of a structured pro forma:

We have a CGA document that incorporates all the relevant assessments required from the medical, nursing and therapy perspective.

Geriatrician

The flow of information could be interrupted by patients being moved to different wards or other health-care settings and this could challenge the continuity of care:

Sometimes . . . if someone needs an overnight stay, we strive to keep that person on the acute medical unit but with bed pressures, not infrequently, patients get moved to base wards and the momentum and pace . . . is lost because new teams have to pick up the patient.

Geriatrician

We pull it [the assessment] together but then when people go up to the wards I am not sure how cohesive that would stay.

Nurse practitioner

Capacity to respond is affected by the availability of resources beyond the hospital, particularly for those who require multiple types of support (domiciliary, therapy and nursing care), ‘when it all gets very difficult’.

Team working

Interviews with staff in each of these CGA service models suggested a reflexive and creative approach to practice and professional development. It was a style of team working that built on the varied disciplinary skills and expertise, and, in some instances, used a more distributed form of leadership and responsibility than the traditional conception of the CGA.

The importance of ad hoc discussions that involved different team members during the day, to review action plans and progress in respect of individual patients, was emphasised by different staff in the multidisciplinary teams. These types of discussions facilitated the flow of information and were supported by the team sharing a work space. Although there is access to a common set of patient notes, verbal communication was reported as key and the importance of informal communication and exchange of personal knowledge by the different members of the MDT was highlighted:

We have a lot of respect for each other and that’s what makes it work . . . and as professionals we have learned lots . . . Coming into this role I was experienced but I’ve learned so much about older people’s mental health [from other team members] . . . it’s always a learning curve but everyone’s clinical skills have been so much more improved by being part of the MDT.

Physiotherapist

The development of the hospital-at-home services provided an opportunity for individual professionals to develop their expertise and roles:

Yes, the [hospital at home] nurses are . . . well they’re diagnosticians. They’re examining patients; they’re taking full control of the patients. I think it’s an immensely rewarding job but you have to be very confident as a nurse. So, the senior nurses that are involved who have come from senior positions often in hospital environments such as a ward sister on surgery . . . But you need your staff to be trained and competent and know the limits of their skill base. You’ve got to trust them and you’ve got to let them grow as individuals . . . trust is absolute because you can’t duplicate what the nursing staff and therapy staff do.

Geriatrician

At the same time, the nature of the work environment is seen to require, from all staff, a level of preparedness and role flexibility to respond to the unexpected:

If you go into someone’s house and whoever’s going in has to be prepared to either . . . you might be making a meal, or you might be giving somebody an enema or you might be deciding what medicines need adjusting. And I think what really has impressed me is that everybody, whoever they are, whether you’re a therapy staff, whether you’re . . . even me as a doctor, have to be prepared to do anything when you go in, and that flexing of roles is absolutely essential.

Geriatrician

In summary, the forging of multidisciplinary ways of working are seen as enhancing disciplinary expertise and contributing to the development of new models of service appropriate to settings that lack the support structures of a hospital ward.

Participants

An overview of patients and family caregivers who participated in the interviews is shown in Tables 20 and 21. The caregivers had been providing a range of informal care and support, had varied relationships to the patient and included those who were living together with, or separately from, the patient. Interviews lasted, on average, around 45 minutes and were audio-recorded. Those involved in the study included patients who presented with delirium, functional decline, dependence, falls, immobility and memory problems or dementia. A common feature of all was the co-occurrence of multiple and interacting chronic health problems, acute exacerbations of existing conditions, and infections. Often, when living with spouses/partners, the relationship was of mutual and reciprocal interdependence.

The findings are organised around the main themes that relate to the delivery of the CGA, as described in accounts from patients and caregivers. These include assessment, the location, perceptions of MDT input and team working, making progress, experiences of discharge and follow-up arrangements, own ways of coping and managing risk. To protect anonymity, all participants were allocated pseudonyms and these are used below.

Perceptions of assessment

Patients’ and caregivers’ accounts suggested that approaches to the CGA varied between, and also within, services in both inpatient and hospital-at-home settings. Key aspects related to components of the assessment experienced and whether or not caregivers had been included within processes. In both inpatient and hospital-at-home settings, caregivers and patients described tasks of assessment as those aspects ‘done to’ the patient. Patients’ accounts suggested that technical aspects of assessment, and later monitoring of their stability, were welcomed as reassurance or had generated confidence in professionals through contact achieved. For some patients, accounts of a medically focused assessment suggested that they valued this as an approach that addressed their expectations of the specialist service:

You know that there is always blood pressure, every I don’t know, it seemed like every 5 minutes actually. So, there was plenty of company.

Nancy, patient

I was told I would have X-rays, I would have blood taken, I would have examinations . . . to discover what problems I have. Which is what it is all about, really . . . to throw up something that might provide a clue, to sort of make the situation better.

William, patient

However, caregivers often perceived that a physical investigation focus had dominated over broader aspects of assessment that they would have valued:

I guess what I was expecting was for someone to see if . . . there’s any additional help that can be offered. This was more on the medical side, just a couple of ‘obs’ [observations], yeah, blood pressure and checking the blood readings, sugar levels, that’s it.

Anne, caregiver

In some cases, caregivers described that they had prompted professionals to undertake additional functional aspects of assessments, for example of their relative’s mobility at home, that they felt may otherwise have been overlooked. Others (including through the same service) had perceived a broader, personally tailored approach from professionals, giving consideration to needs both sensitively and holistically. For some, discrepancies had become apparent between the views of family members and the professionals’ portrayal of the patient’s situation, or caregivers felt that problems and challenges were insufficiently acknowledged in professionals’ assessments:

I find that the nursing notes don’t reflect what I’m seeing . . . There’s no mention of any confusion or muddled stuff going on.

Connie, caregiver

Social services thought she was Benjamin Button, for heaven’s sake.

Connie, caregiver

Inclusion of caregivers in the assessment

An aspect of assessment frequently raised by caregivers related to their own contribution to their relative’s care and challenges that they were encountering, particularly valuing when professionals gave consideration to these issues. Some caregivers conveyed a sense of feeling excluded from assessments, or that their own health needs had not been taken into account:

They’re not dealing with me, that’s the only thing, they’re only dealing with her situation, like . . . They were to deal with her leg, they don’t deal with other things . . . They noticed I wasn’t well and everything.

Alan, caregiver

Caregivers described concerns when they had not received an approach from the health-care team during their relative’s assessment in inpatient settings, including instances in which a decision and plan had already been made for discharge home. In addition to not being asked to provide information about their relative’s health, function or home environment, some family caregivers described ways that their ‘hidden’ support was key to enabling their relative to live at home, but was not necessarily acknowledged by staff:

They were happy that [my mother] could cook for herself and clean and get herself up, which is all fine, except she can’t because she relies pretty much on me . . . she assumes I’m not included in what you call ‘help’.

Bill, caregiver

In addition, caregivers striving to support their relative at home had not always felt able to raise the topic or ask about additional assistance if an opportunity for discussion had not been created by professionals within interactions:

Well, at the moment I’m struggling. [The hospital-at-home team] haven’t asked about any of that.

Alan, caregiver

Although some caregivers felt that their own knowledge about their relative and the family’s ways of coping had been overlooked during assessments, when caregivers’ concerns or opinions had been taken into account and acted on by professionals, their confidence had been enhanced:

So the specialist came out and she rang me back about the assessment. She said, ‘You were right. There are problems there, you did have reason to be concerned’.

Lisa, caregiver

Impact of location: inpatient setting

Patients’ and caregivers’ accounts highlighted challenges experienced as a result of the environment and time constraints in the inpatient facilities. The business of hospitals was often perceived to have contributed to problems in establishing rapport with staff. In some cases, caregivers felt that processes that seemed to be part of standard practice had hindered their opportunity for discussion with the team:

I don’t know why it’s so important to move people into a discharge lounge area then, basically, when you ask a question you’ve got 10 other people in the room . . . there’s no privacy, and you can’t take anyone to task over it.

Bill, caregiver

Patients had also experienced ways that their own usual routines, such as self-managing their medications, had become disrupted on admission to wards where they were required to relinquish control to hospital processes:

When you go into hospital they . . . take the dosette box, and then they won’t give me the tablets out of it. And until the doctor prescribes them, you don’t get them. I’ve been awake at 1 o’clock at night, waiting for my tablets.

Rhys, patient

However, some patients and caregivers perceived benefits of the inpatient setting; for example, it could seem more restful than being at home when unwell. For others, the shared space of the ward had also facilitated valued interactions with ‘peer’ patients during the admission:

At least if she’s in the hospital we can go at visiting time; it’s less stress on us and on her as well . . . you know how [. . .] communities that come in [to the home] and it’s hectic, they don’t realise when one person, they need a rest.

Alishah, caregiver

Social interaction, it is very important. We [patients] go into the day room and watch TV . . . watching the programme ‘Jerry Springer’, it stimulates an argument!

Lilian, patient

Other patients had felt cut off from desired interactions with others because of the ward’s physical layout or because of equipment that they had found challenging to use:

The space between the beds . . . we [neighbouring patients] can’t make a conversation . . . that’s been awkward, not being able to see the people, hearing their voices but not being able to answer to them, because I can’t hear what they’re saying.

June, patient

You have a television with a telephone attachment, but the trouble was I couldn’t really reach it. It is really quite cumbersome.

Nancy, patient

Impact of location: hospital at home

For patients and caregivers who had experienced hospital at home, a key consideration was the potential for disruption to the rhythm and routines of their home life. This was minimised when staff clearly communicated the anticipated times of their visits and when they were reliable in following arrangements through, in comparison with accounts of when patients and caregivers felt that this had not been achieved:

I think it should have been set that it were a time for them to come, ‘cos you couldn’t do ‘owt really; you were waiting in for them.

Joyce, patient

For some patients, concern about the stability of their condition led to feelings of anxiety about hospital-at-home care, particularly when thinking about lack of rapid access to clinicians overnight:

I think it is better if you get treated [in the hospital] because at home, even if someone is doing everything for you, but if you have a problem at night, have trouble breathing, who can you turn to? . . . In hospital they are checking on you and the doctor, everything is there.

Laila, patient

Challenges were also described by caregivers in managing to support their relative within the home environment; these were often linked to whether or not the caregiver had felt included in the team’s assessments and planning:

Somebody’s got to be there . . . this business of one size fits all . . . not everybody’s got a bed and a toilet that’s very convenient.

Peter, caregiver

In addition, some patients found that support provided by formal carers at home did not always meet their needs or expectations in managing their difficulties with daily tasks:

I can’t say the carers really know what you wanted. I know they haven’t got a lot of time but they don’t see to the bed or anything like that anymore. My other carers used to make the bed, you know, but [the current carers] haven’t got the time.

Jean, patient

When caregivers had experienced a flexible approach from hospital-at-home staff in responding to their relative’s extended care needs, especially during the challenging time following discharge from hospital, this had supported their own ability to cope:

Once [my husband] did have a bad bout and the bed was absolutely swamped again, so [the hospital-at-home nurses] cleaned, you know, they had a really long job to clean him up and change the bed and get him tidied up again . . . I couldn’t have done it without them.

Penny, caregiver

Other caregivers gave examples of hospital-at-home staff providing equipment, which they felt had facilitated ways of managing for their relative at home. Concerns were often raised by caregivers about managing stairs and other risks within the home, as considered further below.

Perceptions of interactions with multidisciplinary team

Aspects of MDT input described in caregivers’ and patients’ accounts included interventions experienced, relational aspects of communication, perceptions of teamwork, monitoring progress and planning discharge.

Accounts from inpatient and hospital-at-home settings demonstrated ways that holistic assessments had been perceived to lead to direct interventions being taken by the team. These encompassed management of medical problems, including rationalisation of medications, as well as practical aspects of managing. In other cases, explanations of findings of assessments had resulted in clarification of what to expect:

The doctor that came out with the nurses explained that they couldn’t physically find anything wrong with her. That they thought it was just obviously the dementia progressing.

Pauline, caregiver

Patients’ and caregivers’ perceptions of communication approaches by teams varied in both inpatient and hospital-at-home contexts (including experiences within the same service). For some, lack of continuity had disrupted rapport-building when different team members had come to the home and could be compounded by an approach of ‘being informed’, rather than ‘being included’, within discussions:

One’s telling you one thing one day and then somebody else coming along and saying no we don’t think that should happen, we think this should happen. I think, well, you know, that’s not what they said yesterday.

Pauline, caregiver

By contrast, continuity through staff members taking time to understand the particular challenges for both patient and caregiver through sequential visits was valued and was perceived to enable professionals’ meaningful monitoring of changes over time. Relational aspects of care through personal interactions with staff, and the role of rapport and humour, were appreciated across settings. For those who experienced inpatient care, the regular contact and interactions with non-clinical team members was also highlighted:

A very nice lady came along with the trolley and tea or coffee . . . somebody came along with the water jugs . . . and then the cleaner came along . . . It was constant, you know, loads of company.

Nancy, patient

Many patients and caregivers across the case studies did not differentiate between multidisciplinary staff roles and responsibilities. When distinction was described between staff members’ focus of input, this seemed to contribute to a perception of thoroughness and personalisation in addressing needs of the patient and, in some cases, the caregiver also:

The physio[therapy] girls . . . walked him round the ward and took him to the stairs and things like that, so yes, they were very thorough with him. And I did see an Age Concern lady.

Grace, caregiver

Teamwork appeared to be most visible within accounts from patients who had received care from the Older People’s Acute Assessment team. In this case, some had also perceived distributed elements of responsibility for decision-making and planning within the MDT:

Yeah, that is what the doctor said, ‘We will see what the physio[therapist] says and you might have to stay another night’.

Joyce, patient

Own perceptions of making progress

Although none of the participants used the CGA terminology of ‘goal-setting’, some caregivers from hospital-at-home contexts talked about the ways this concept was enacted within their own everyday tasks and interactions with their relative. In addition, patients did not give accounts of targets that they had worked towards with MDT members, but some accounts included descriptions of patients’ own markers of meaningful progress:

I’m sort of walking more steady, being on me feet. I managed to take myself by taxi to get my new hearing aid yesterday, and I felt it such a big step.

Olive, patient

Patients also conveyed a sense of enhanced confidence when noticing their own achievements, and these could be reinforced through supportive interactions with team members in inpatient and hospital-at-home contexts, for example:

Speech therapy . . . I am practising, and the nurses on the ward they say, I can make myself clear to them . . . sheer persistence on my part, determination.

Lillian, patient

Involvement in discharge planning

A key concern raised by caregivers, which was raised in relation to inpatient and hospital-at-home services, was insufficient involvement in determining discharge arrangements. For some patients also, conflicting communication from the team about discharge plans and lack of family involvement had raised anxiety:

[The consultant] said we would have a meeting that would include my family and me as well, but that is not going to happen . . . I thought of a load of other things I needed to ask, and I didn’t realise discharge was imminent.

Theresa, patient

For one hospital-at-home service, the team removed their folder of patient notes from the home at their last visit and because of this, some patients and caregivers deduced that their support would no longer be available to them. This demonstrated unintended consequences that may arise from a team’s routine way of working and individuals’ own sense-making of observed actions of professionals:

No, they didn’t tell me [about discharge], but I knew they’d come and got the folder.

Joyce, patient

The positive contribution of documentation, when explicitly shared between patients, families and professionals, was illustrated through one caregiver’s account of having an advanced care plan. This lady valued the way the inpatient team acknowledged her husband’s preferences, as had previously been documented, and felt that this facilitated discussions and her own reassurance in determining discharge arrangements.

For other caregivers, accounts of difficulties with discharge arrangements included their own rationalisations that actions taken had been unavoidable or were perceived to have been beyond the professionals’ ‘control’, instead situating these challenges in larger narratives about the health-care system in which staff are required to act:

I realise that it’s terrible to send someone out [late at night], but there’s nothing you can really do about it because we all know we’re struggling and beds are at a premium . . . it’s not really the nurses’ fault, it’s the system’s fault.

Susan, caregiver

Addressing needs following discharge

Participants’ experiences of continuity following discharge included instances of contact with staff who had a remit crossing transitions (the Age Concern worker’s role, based in the Older People’s Acute Assessment team) or when inpatient staff had provided them with information about onward referrals to community services, which they had already actioned. However, patients and caregivers commonly described that ongoing continuity of care would depend upon their engagement with the GP, with variable perceptions that were often linked to previous experiences with their GP practices. Some caregivers valued that the GP could address gaps they experienced in information sharing from an inpatient service, and some patients also felt confident to continue to manage by accessing their GP after discharge, as and when they felt necessary:

The only way to find out whether it is all right [to stop medication] is to go and see your GP . . . I think it was up to me, really, to find out whether I needed to take them. You can’t expect people to go on looking at you as if you were about 3 [years old].

Nancy, patient

However, a cause of concern for some patients and caregivers came from lack of clarity over which health-care service would be involved or available for any further problems after discharge from hospital-at-home service:

I’ve still got this chest problem, which is what I was going to ask: I go back to my GP do I, now? I don’t get involved with the virtual people?

Hector, patient

Although accounts about provision of information resources at the time of discharge were limited from patients and caregivers, experience suggested that this could be a useful means of raising awareness on support options later, such as through an Age Concern booklet. Similarly, few patients or caregivers talked about receiving a copy of the discharge letter from the service, although, for some, it had provided clarification on the outcome of assessments that had given confidence for moving forward. Others felt that they would not find it necessary to have their own or their relative’s discharge summary if the GP had a copy:

You get your discharge papers . . . they go in your paper bin or something like that, you know, because your doctor’s got one, and you know what you’ve got, and what tablets you’re taking.

Elizabeth, caregiver

Own ways of coping and managing risk

For patients living alone, the importance of broader support networks through friends and neighbours, and support from formal carers, was apparent. In addition, patients and caregivers demonstrated a range of their own strategies for managing and moving forward:

Down the road there is a masseuse who has a practice, she came in and massaged my legs. They had got very stiff, you know, they needed to be made a bit more mobile. So that worked alright . . . and we had a chat about how lambing was going on!

Nancy, patient

For caregivers living with their relative, their unique knowledge and previous experience of providing support were often linked to accounts of risk and safety awareness at home, including their established ways of coping from before the episode of health care and anticipation of what would be likely to follow:

The physiotherapist came yesterday and they suggested a seat for [my Dad], so that he sits down and washes his hands . . . I was trying to explain that it would put him more at risk . . . he’s not going to use it. I know my Dad, because obviously I’ve been looking after him for about 7, 8 years now.

Hamza, caregiver

Families’ considerations about coping with needs and risks were frequently tied to their own availability to provide support, associated with anxieties about the level of input that their relative would require, which they may not have had opportunity to discuss with professionals prior to the discharge. In some cases, concern was expressed for another family member’s ability to cope, particularly where a spouse was living with the patient and the caregiver was providing support to both parents. These situations highlight caregivers’ work in family communication and their knowledge of the sometimes fragile dynamics of support available for their relative, combining complex and interdependent needs. Management of risks of a sudden decline, falls or ensuring safety on stairs was considered in terms of flexible family availability to provide supervision or assistance. Some caregivers also described experiences of trying technological measures as a form of extended support for their relative:

She had a nasty fall and she had a terrible black eye, and I had arranged for my Mum to have the call button . . . I’m so worried about her being home, because she didn’t even call the call button when she had that fall.

Sandra, caregiver

When families were no longer able to sustain the level of practical support that they felt was essential for their relative to continue living safely in their home, especially overnight, they began to envisage alternatives to their relative remaining there.

Data management

Data were stored securely on servers within the Nuffield Department of Population Health, University of Oxford, and managed as per standard operating protocols.

Participants

Of the 78 people who registered to participate, 53 (68%) completed round 1 of the CGA Delphi exercise. Thirty participants were medical doctors with experience of older people’s medicine (56.6%), five were nurses (9.4%), four were allied health professionals (two occupational therapists and two physiotherapists) (7.5%) and one was a policy-maker/senior administrator for older people’s health care (1.9%). There were 13 patients, caregivers or members of a PPI group (24.5%). These participants included one patient, six caregivers and six members of a CGA PPI group (the last of which included a lay person, a member of the public, a volunteer and a social worker).

Of the 78 people who registered to take part in the Delphi exercise, 59 (76%) completed round 2. Thirty-three participants were medical doctors (56%): 25 consultant geriatricians, one community geriatrician, six junior doctors/registrars and one psychogeriatrician. There were eight nurses (14%), four allied health-care professionals (one occupational therapist, two physiotherapists and one pharmacist) (7%) and one person was a worker for an older people’s organisation. There were 13 patients, caregivers or members of a PPI group (22%). These participants included nine PPI research volunteers, two participants providing a carer perspective and two participants providing a patient perspective.

Results from round 1

Results from round 2

The statements, together with the per cent rating ‘very important’ from round 2 are detailed in Table 24. In the final version of the CGA Delphi, we included statements that ≥ 69% of participants from both panels scored as 7–9 and < 15% scored as 1–3. 119 We excluded the following criteria from the final version of the CGA Delphi: mild or moderate frailty, having more than one long-term condition, social isolation or care home residents as criteria to target the delivery of the CGA, hospital length of stay as an outcome, and access to a recreational therapist or spiritual support. In all but one area of disagreement, judged as a difference of ≥ 20% in the score between the two groups (health-care professionals, and patients and caregivers), the health professionals rated a statement as less important than patients and caregivers. These included people having more than one long-term condition being targeted for the CGA (49% vs. 93%), death as an outcome (57% vs. 77%), the well-being of the carer (57% vs. 77%), access to a dietitian (39% vs. 62%), access to a care co-ordinator (60% vs. 85%) and being provided with the opportunity to be present at a MDT meeting (64% vs. 100%). The one exception was the rating for access to a social worker, with health-care professionals rating this as more important (93% vs. 62%) than patients and caregivers.

Changes from the protocol

We initially planned to recruit a group of 20 clinicians (including members of the MDT) to participate in the online Delphi exercise and obtain views from patients and caregivers through focus groups. Following advice from the SSC, we invited patients and caregivers to participate in the Delphi exercise and increased the number of participants to at least 50. We planned to use a cut-off point of ≥ 70% participants scoring 7–9 (really important) and < 15% scoring 1–3 (not important) to determine the statements to include in the final version of the Delphi. 119,120 We reduced this to ≥ 69% because of a large number of statements receiving a score of very important from 69% of participants; it is not unusual for this type of post hoc pragmatic decision to be taken in Delphi exercises. 120

The effectiveness of implementing Comprehensive Geriatric Assessment in hospital

In our update of the Cochrane Review of hospital-based CGA,12 we included 29 trials recruiting 13,766 participants across nine mostly high-income countries. The findings provide a good indication that the CGA increases the likelihood that patients will be alive and in their own homes after discharge and decreases the likelihood that patients will be admitted to a nursing home at 3–12 months’ follow-up; there is little or no difference in mortality at follow-up. There was low-certainty evidence on the impact of the CGA on cognitive functioning and on the difference in effect between wards and dedicated teams as this analysis lacked power owing to the relatively small number of studies.

The cost-effectiveness of implementing the Comprehensive Geriatric Assessment in hospital

Comprehensive Geriatric Assessment delivered in hospital may lead to a slight increase in costs (data from 15 trials29,43,45–47,50,53,54,56,57,59,60,62,63,66), albeit with some uncertainty because of variation in the measurement of resource utilisation in the trials that were included in the calculation of the costs and in the unit costs. Hospital length of stay, the main driver of resource use, varied among the trials with an average of 1.63 days to 40.70 days in the intervention group, and from 1.80 days to 42.80 days in the comparison group. This variation could be partly explained by the different populations recruited. The trial that reported the longest length of stay was conducted in 1999;56 it recruited older people who had multiple chronic conditions or functional decline and who were at risk of a nursing home placement. The CGA intervention in the trial that reported the shortest length of stay was delivered in an Acute Elderly Assessment Unit. 53 We found that the CGA may lead to a very slight increase in QALYs resulting in ICER close to the National Institute of Health and Care Excellence ceiling of £20,000. There was also a very slight increase in LYs gained and in LYLAH. The evidence for the cost-effectiveness analysis is of low certainty because of imprecision and inconsistency among studies.

A non-randomised comparison of the populations who received Comprehensive Geriatric Assessment delivered in a hospital-at-home setting versus those who received it in hospital

We analysed administrative data supplied by the ISD (NHS Scotland)131 to compare the characteristics of the populations who had received health care from three geriatrician-led hospital-at-home services with the populations who were admitted to hospital with similar diagnoses. We assessed mortality and cost at 6 months’ follow-up and used PSM in combination with regression analysis to reduce observed confounding. In this study, there was a greater than four-fold difference between the three sites in the cost of providing geriatrician-led hospital at home, and the populations that had received this type of home-based health care may incur an increase in health-care costs in the 6-month period after their admission to hospital at home, compared with the population that had received their health care in a hospital.

Determining the populations most likely to benefit from the Comprehensive Geriatric Assessment

Eleven of the trials included in the Cochrane Review delivered the CGA to patients on the basis of age alone (the age cut-off point used was between 65 and 75 years), and the remaining 18 trials used needs-based eligibility criteria (such as falls, reduced mobility, confusion). A small minority of the participants who attended the consensus meeting endorsed age or frailty as the sole criterion to determine eligibility for the CGA, and the majority agreed that age should be used with other criteria. Those who responded to the Delphi reported that the criteria that are most important to determine eligibility for the CGA are a high level of frailty, problems with cognitive functioning, delirium, a fall, assistance with ADL, functional decline and recurrent hospital admissions. Patients and caregivers placed more importance than health-care professionals on the importance of having more than one long-term condition and the role of social isolation in a CGA assessment.

The analysis of administrative data provided an opportunity to examine the characteristics of the populations receiving hospital-level care outside a research setting (‘real-world data’). We used a pragmatic approach to identify the two cohorts, by applying the criteria for admission employed by three hospital-at-home services. In this study, we found that the population that received health care from the three geriatrician-led hospital-at-home services was older, more socioeconomically disadvantaged, had higher morbidity, higher rates of previous hospitalisation and may be more likely to die during the 6-month follow-up period.

The content and process of delivering the Comprehensive Geriatric Assessment in NHS community settings and the barriers to implementing the Comprehensive Geriatric Assessment

Methodological strengths

This programme of research used a range of methods to assess the effectiveness, cost and organisational features of the CGA. We also explored the assumptions that underpin the CGA, using the theory of change, and conducted a narrative analysis of professionals’ understandings of the CGA activities, outcomes and impact and the significance placed on these by patients and caregivers.

The comparative analytic focus of the provision of the CGA in different settings increased our confidence in the findings, particularly when similar themes were identified across settings and there was coherence between the findings from the interviews and the Delphi exercise. By applying the theory of change to the logic model of the CGA we identified the possible mediating factors, and potential consequences, if key assumptions (the exchange of knowledge, a comprehensive focus to the assessment and patient and caregiver engagement) are not satisfied. 17,133 The logic model provided an additional method to integrate the findings from the different studies and to assess the coherence of the findings. However, without the theory of change, the linearity of the logic model would have limited our understanding of the contextual and mediating factors that affect outcome.

The inclusion of randomised trials from nine countries in the update of the Cochrane review of the CGA strengthens the applicability of the findings to different settings in countries that have developed health systems. The problems with adjusting a LY for health-related quality of life in an older population has been debated, and well-being-based outcomes have been suggested as alternatives. 69 In response to these concerns, we used LYLAHs as an indicator of independence and well-being. 37 We used IPD to calculate the cost-effectiveness of inpatient CGA compared with inpatient care without the CGA and used LYLAHs after discharge from hospital as a measure of independence and well-being in an older population. We also expressed the ICER as a cost per QALY and cost per LY gained from the NHS perspective (i.e. including only hospitalisation costs and costs of the CGA delivery). Our analysis of administrative data followed Medical Research Council guidelines74 and provided real-world evidence on the impact of hospital at home as currently delivered in three Scottish health boards, and the sensitivity analyses helped to address uncertainty in the results.

Limitations

It emerged through the interviews with staff, patients and caregivers that for the CGA to work it is necessary for a range of support services to be in place that provide social as well as health care. The research plan did not include an investigation of social care, and instead we relied on the interviews with health-care professionals, patients and caregivers to identify social care needs. Furthermore, our analysis of cost did not include the cost of home or social or residential care. A second perspective that was missing was primary care and, in particular, how GPs could be involved with the CGA process. This was an aspect that was mentioned in the Delphi, with respondents placing importance on the background information GPs could provide. Related to this is how the CGA process is communicated to the GP at the end of an episode of health care.

A second limitation was that the number of trials that supplied IPD was smaller than expected, despite receiving assurances from several triallists that data would be available and sending several reminders to the triallists to request data. Various reasons for not providing data were provided, with the most common being a move to a different location and no longer having access the data, or the data being stored on a device that was no longer accessible.

The Delphi exercise included a relatively small number of allied health professionals; this might be because of our approach to identify participants, but might also reflect the fact that the term CGA might not have resonance with this group. A further limitation is the small sample of patients and caregivers who contributed to the Delphi; although there was little variation in their responses, this could be because of the sources that we contacted to identify people to contribute.

A limitation of the analysis of the administrative data from the three health boards in Scotland was the non-randomised comparison and risk of residual confounding. Although matching individuals and performing regression analysis reduces this risk, it is possible that the two groups still differed. If clinical decision-making by GPs and geriatricians to admit patients to either hospital-at-home or hospital relied on variables that were not included in our analysis, our findings might be biased because of confounding by indication. This type of confounding cannot be measured directly because it is not based on established criteria or available diagnostic codes. If clinicians did not consider hospital at home as a substitute service to hospitalisation, but rather as a service that supplemented existing services, then confounding by indication might increase the risk of residual confounding in our analysis. Data were not available to match and adjust for differences in the use of community and social services prior to index admission or to include the cost of community and social care services in the analysis. Although residual confounding might be present, the results of the survivors’ subgroup analysis were very similar to the results of the main cost analysis. Another limitation of this study is the lack of data on the cost of informal care, the quality of care in the two services and patient health status, which precludes a complete estimate of the cost of hospital-at-home services compared with inpatient hospital services. A further limitation is the lack of information on the degree to which the CGA was delivered in each of the three hospitals and the extent to which we compared the CGA in hospital at home with the CGA in hospital.

Search strategy

1. Geriatric Assessment/

2. Health Services for the Aged/

3. Needs Assessment/

4. Risk Assessment/

5. exp Diagnostic Services/

6. ‘Health Services Needs and Demand’/

7. exp Health Services/

8. exp ‘Delivery of Health Care’/

9. exp ‘Outcome and Process Assessment (Health Care)’/

10. ((multidisciplinary or multi-disciplinary) adj5 assess*).tw.

11. 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10

12. geriatrics/

13. 11 and 12

14. 1 or 2 or 13

15. ((geriatric or elderly or old age) adj5 consultation).tw.

16. ((geriatric or elderly or old age) adj5 evaluation).tw.

17. ((geriatric or elderly or old age) adj5 assess*).tw.

18. (gemu or gemus).tw.

19. 14 or 15 or 16 or 17 or 18

20. randomized controlled trial.pt.

21. controlled clinical trial.pt.

22. randomized.ab.

23. placebo.ab.

24. drug therapy.fs.

25. randomly.ab.

26. trial.ab.

27. groups.ab.

28. 20 or 21 or 22 or 23 or 24 or 25 or 26 or 27

29. exp animals/not humans.sh.

30. 28 not 29

31. 19 and 30

32. (2015* or 2016*).dc,dp,ed,ep,yr.

33. 31 and 32

Search strategy

1. Geriatric Assessment/

2. Health Services for the Aged/

3. Needs Assessment/

4. Risk Assessment/

5. exp Diagnostic Services/

6. ‘Health Services Needs and Demand’/

7. exp Health Services/

8. exp ‘Delivery of Health Care’/

9. exp ‘Outcome and Process Assessment (Health Care)’/

10. ((multidisciplinary or multi-disciplinary) adj5 assess*).tw.

11. 3 or 4 or 5 or 6 or 7 or 8 or 9 or 10

12. geriatrics/

13. 11 and 12

14. 1 or 2 or 13

15. ((geriatric or elderly or old age) adj5 consultation).tw.

16. ((geriatric or elderly or old age) adj5 evaluation).tw.

17. ((geriatric or elderly or old age) adj5 assess*).tw.

18. (gemu or gemus).tw.

19. 14 or 15 or 16 or 17 or 18

20. crossover procedure/

21. double blind procedure/

22. single blind procedure/

23. randomized controlled trial/

24. (random* or trial or placebo* or crossover or ‘cross over’ or ((singl* or doubl*) adj1 (blind* or mask*)) or assign* or allocat* or volunteer*).tw.

25. 20 or 21 or 22 or 23 or 24

26. (exp animals/ or nonhuman/) not human/

27. 25 not 26

28. 19 and 27

29. (2015* or 2016*).dp,dd,yr,em.

30. 28 and 29

Search strategy

#1 [mh ‘geriatric assessment’]

#2 [mh ‘health services for the aged’]

#3 [mh ‘needs assessment’]

#4 [mh ‘risk assessment’]

#5 [mh ‘diagnostic services’]

#6 [mh ‘health services needs and demand’]

#7 [mh ‘health services’]

#8 [mh ‘delivery of health care’]

#9 [mh ‘outcome and process assessment (health care)’]

#10 ((multidisciplinary or multi-disciplinary) near assess*):ti,ab,kw

#11 (or #3–#10)

#12 [mh geriatrics]

#13 [mh aged]

#14 #12 or #13

#15 #11 and #14

#16 ((geriatric or elderly or old age) near consultation):ti,ab,kw

#17 ((geriatric or elderly or old age) near evaluation):ti,ab,kw

#18 ((geriatric or elderly or old age) near assess*):ti,ab,kw

#19 (or #1–#2. #15–#18) Publication Year from 2015 to 2016

Search strategy

S1 (MH ‘geriatric assessment+’)

S2 (MH ‘health services for the aged’)

S3 (MH ‘needs assessment’)

S4 (MH ‘patient assessment’)

S5 (MH ‘nursing assessment’)

S6 (MH ‘diagnostic services+’)

S7 (MH ‘risk assessment’)

S8 (MH ‘diagnostic services+’)

S9 (MH ‘health services needs and demand’)

S10 (MH ‘health services+’)

S11 (MH ‘health care delivery, integrated’)

S12 (MH ‘health care delivery’)

S13 (MH ‘outcome assessment’)

S14 (MH ‘process assessment (health care)’)

S15 TI (((multidisciplinary or multi-disciplinary) n5 assess)) or AB (((multidisciplinary or multi-disciplinary) n5 assess))

S16 S3 OR S4 OR S5 OR S6 OR S7 OR S8 OR S9 OR S10 OR S11 OR S12 OR S13 OR S14 OR S15

S17 (MH ‘geriatrics’)

S18 S16 AND S17

S19 TI (((geriatric or elderly or old age) n5 consultation)) or AB (((geriatric or elderly or old age) n5 consultation))

S20 TI (((geriatric or elderly or old age) n5 evaluation)) or AB (((geriatric or elderly or old age) n5 evaluation))

S21 TI (((geriatric or elderly or old age) n5 assess*)) or AB (((geriatric or elderly or old age) n5 assess*))

S22 TI (gemu or gemus) or AB (gemu or gemus)

S23 S1 OR S2 OR S18 OR S19 OR S20 OR S21 OR S22

S24 PT randomized controlled trial

S25 PT clinical trial

S26 TI ( randomis* or randomiz* or randomly) OR AB ( randomis* or randomiz* or randomly)

S27 (MH ‘Clinical Trials+’)

S28 (MH ‘Random Assignment’)

S29 S24 OR S25 OR S26 OR S27 OR S28

S30 S23 AND S29

S31 S30 Limiters - Published Date: 20100101-20161231; Exclude MEDLINE records