Background Although it is widely recognised that frail older persons need adaptation of healthcare services, it is unclear how hospital care in general can best be tailored to their frailty.
Objective To systematically review the evidence for hospital-wide interventions for older patients.
Methods PubMed, Cochrane CENTRAL, Cinahl and reference lists of included articles (1980–2009) were searched. Papers describing (1) randomised controlled trials, controlled clinical trials, controlled before–after studies or interrupted time series, (2) patients ≥65 years admitted to hospital, (3) hospital-wide organisational interventions, and (4) patient-related outcomes, quality of care, patient safety, resource use or costs were included. Two reviewers extracted data and assessed risk of bias independently, according to Cochrane Effective Practice and Organization of Care Review Group guidelines.
Results The authors included 20 articles out of 1175. The mean age of the study populations ranged from 74.2 to 85.8 years. Interventions included multidisciplinary (consultative) teams, nursing care models, structural changes in physical environment and/or changes in site of service delivery. Small or no effects were found on patient-related outcomes such as functional performance, length of stay, discharge destination, resource use and costs compared with usual care. Methodological quality evaluation showed data incompleteness and contamination as main sources of bias.
Conclusions No single best hospital-wide intervention could be identified using strict methodological criteria. However, several interventions had positive results, and may be used in hospital practice. Since strict methodological designs are not optimal for evaluating highly complex interventions and settings, the authors recommend studying hospital-wide interventions for older persons using adapted quality and research criteria.
- frail elderly
- quality of care
- systematic review
- healthcare quality
- healthcare quality improvement
- quality of care
Statistics from Altmetric.com
- frail elderly
- quality of care
- systematic review
- healthcare quality
- healthcare quality improvement
- quality of care
The quickly growing number of frail older surgical and non-surgical inpatients emphasises the need to develop hospital-wide interventions to improve outcomes of hospital care.1 Hospital-wide interventions are system interventions, not restricted to medical specialties or departments, that are available for all older hospitalised patients. Comprehensive geriatric assessment (CGA) has been introduced and further developed to maintain or improve functioning in frail older patients, and has been proven to be effective when implemented ward-based (as opposed to inpatient geriatric consultation service).2–4 CGA is a multidimensional, interdisciplinary diagnostic instrument designed to determine the medical, psychosocial and functional capabilities and limitations of older patients in order to develop a coordinated and integrated plan for treatment and long-term follow-up.3 However, having only one geriatric ward cannot improve care for all frail hospitalised older patients, since persons older than 65 years currently form the largest proportion of all inpatients. In addition, despite the development of CGA, there is still a high risk of poor functional outcomes and dependency during5 or after6 hospitalisation. Delirium and falls are examples of major and often preventable adverse events,7 8 which quickly increase with age.9
Thus, enforced by healthcare reforms, interest in effective and efficient care models for older patients, next to existing geriatric specialised wards, is increasing.10 Therefore, the primary objective of this article is to systematically review the evidence for hospital-wide interventions for frail older patients.
We performed a search of PubMed, Cochrane CENTRAL, and Cinahl, from 1 January 1980 to 15 May 2009, including only articles written in English. For PubMed a comprehensive search strategy was developed (appendix A), which was adapted for the other databases (appendices B and C). Methodological search filters for Medline (for PubMed) and Cinahl were used as described by the Cochrane Effective Practice and Organization of Care Group (EPOC) (http://www.epoc.cochrane.org). The snowball method was used to manually identify relevant references from the reference lists of included articles.
We explicitly searched for interventions that were developed to be implementable on a hospital-wide basis and therefore available for all hospitalised older patients. We defined hospital-wide interventions as integrated practices throughout the hospital system of care delivery for older patients, which are not restricted to medical departments or specialties (eg, geriatric departments as the only place providing special attention to older patients and therefore available only for the, clearly visible, frailest patients). The term ‘frailty’ was primarily used as a term to retrieve studies of interest, but not as an inclusion or exclusion criterion, since there is still much debate on its definition. Studies were considered for inclusion when they: (1) included patients 65 years or older and acutely admitted to hospital; (2) described an organisational intervention designed and piloted or implemented to improve hospital-wide quality, safety or effectiveness of care for (frail) inpatients ≥65 years; (3) reported outcomes related to quality of care, patient safety, patient-related outcomes, resource use or costs; and (4) were a randomised controlled trial (RCT), controlled clinical trial, controlled before–after study (CBA) or interrupted time series (ITT). Studies describing (1) medical specialty-, disease- or disability-specific interventions, (2) pre- or posthospital interventions (eg, improvement of transfers), (3) specialised hospitals (eg, rehabilitation, long-term, intermediate care) or (4) single-component interventions (eg, use of fall-prevention protocol) were excluded. The first and fourth exclusion criteria were chosen, as we are looking for interventions which serve, in concordance with CGA, all frail older patients with their complex and heterogeneous health problems.
Data extraction and quality assessment
Two researchers (FB and SR) conducted the initial search by independently examining each title and available abstract. Retrieved full-text studies were independently reassessed (FB and SR). A third researcher (MOR) was consulted in case of disagreement. Data were collected based on the checklist of the Cochrane EPOC Review Group, and abstracted using a modified version of the EPOC data extraction form (appendix D) (http://www.epoc.cochrane.org). Data collected included details of the intervention, patients and providers, setting and primary outcomes. Quality assessment was included by using the most recent 2009 EPOC form, which includes nine standard criteria to assess the risk of bias: randomisation, allocation concealment, baseline comparability, incomplete outcome data, blinding of participants, providers or outcome assessors, selective outcome reporting or other risks of bias. A consensus-based risk of bias table was constructed.
Data synthesis and analysis
It was not feasible to conduct a meta-analysis. Results of included studies were therefore analysed by making qualitative, descriptive summaries. We show results as presented by original studies. Additionally, effect sizes (Cohen d) were calculated (d of 0.20 judged as small, 0.50 as medium and 0.80 as large), when standard deviations and means were presented in the original article (for further details, see appendix E).
The search strategy identified 1175 citations, of which 11 articles could be included for analysis. The snowball method yielded an additional nine articles. Figure 1 details the results of the steps in the search strategy. The 20 included articles represent results of 17 studies (12 RCTs and five controlled clinical trials).
The characteristics of the studies included are shown in table 1. More detailed information is available in appendix table 1. The mean age of the population varied from 74.2 to 85.8 years across studies. Whereas most studies used age to select a frail population or selected frail patients during the intervention, seven of the included studies used additional criteria to select frail inpatients.17 19 22 23 25 28–30 As for the location of the intervention, one study described an intervention starting in the Emergency Department,30 whereas the other studies describe interventions initiated at general medical wards. All but two studies set up multidisciplinary teams; these two studies only made structural changes in physical environment and/or site of service delivery.23 30 Four studies initiated, in addition to a multidisciplinary team, an intervention including modifications of the physical environment.12 23 24 27 In seven studies, the main providers of the intervention were nurses.12 18 20 23 24 27 30 In one study, the main providers of care were rehabilitation staff.16 In the other studies (geriatric), physicians were the responsible professionals and/or main providers of the intervention. Interventions (I) were compared with controlled usual care (C) as provided throughout the hospital, prior or next to the interventions.
Risk of bias
On average, we found two main sources of potential bias (table 2). For 14 articles, it was unclear whether or not the incomplete outcome data had been addressed adequately (ie, it was not specified whether missing outcome measures potentially biased the results as presented in the article). Contamination was inadequately addressed or not described in 19 articles.
Effectiveness of interventions
Primary outcomes were functional performance, length of stay, mortality, discharge destination, readmission, complications, resource use and costs (table 1; for further details, see appendix table 2).
Fourteen studies presented results on functional patient outcomes. Of these, five studies (four Geriatric Consultation Teams (GCT),17 18 22 25 one dayroom23) showed significant effects for patients in the intervention group on mental health, emotional or cognitive status. Three studies (two GCT,19 28 one Acute Care for Elders unit24) demonstrated significant improvements in physical outcomes.
Length of stay
Of nine studies studying length of stay (LOS), one (primary nursing model of care12) had a significantly shorter LOS in one of the two experimental sites.
Eight studies focused on discharge destination. Of these, one (primary nursing model of care12) showed a statistically significantly higher nursing-homes admission rate, and one (GCT29) had a significantly lower number of nursing home admissions at 12 months.
Two studies studied in-hospital resource use, of which one (GCT17) showed a significantly higher rate of referral to rehabilitation services. Six studies measured postdischarge resource use. Three (two GCT,17 29 one geriatric-based ward26) showed a significantly lower average number of nursing home days per patients at 12 months, a higher mean number of referrals to community services or a higher number of outpatients visits per patients to a physical or occupational therapist up to 3 months' follow-up.
This systematic review assessing the effects of interventions to improve hospital-wide care for older inpatients showed that no single best evidence-based practice can be described that improves quality of care, safety and effectiveness. Different forms of geriatric consultation teams were partly effective in improving patient-related outcomes and process quality measures. Additionally, nursing models of care, wards admitting all older patients and environmental adaptations were found, with heterogeneous effects in different settings. The designs are methodologically not sufficiently strict, and the studies too heterogeneously described to allow summary statistics or a Cochrane high-quality evidence rating.
The heterogeneity in the studies can be explained in several ways. First, hospitals differ from site to site in catchment area and associated demographic and sociocultural setting, referral practice, specialisation, staff, and overall quality and safety of care. Consequently, care interventions differ greatly, even if they are based on a similar model of care.12 In addition, positive effects across studies were found on different outcomes, and positive outcomes showed only moderate or small effects (effect sizes ranged from 0.16 to 0.37). Ten studies introduced an intervention including the GCT's principles, of which four studies demonstrated no significant effects on their primary outcomes. Three showed small effects in mental status or mood.17 18 22 One of these 10 showed positive effects on the Barthel score and survival,19 and one in survival and readmission rates.21 Four studies which introduced an intervention with nurses as main providers found no or small effects, which they ascribed to limited availability of resources and thus limited intensity of the intervention. Of the two ACE unit studies which intended to be implemented hospital-wide eventually, one had positive results on functional outcomes.24 The other explains improvements in usual care as the main cause for the lack of significant results.27 It is also possible that usual care was contaminated by the intervention in the majority of studies, which may have influenced the ability to show positive effects.
However, although the effects are small, positive results are definitely important in such a frail population. About 22% of persons older than 80 years who are admitted to a hospital die within 1 year after discharge,31 and the average time for partial or full recovery after hospitalisation is 18 months.32 Therefore, each step forward is important in effectiveness of hospital care, such as stabilisation of functional performance, and is an important positive result. Studies showing no significant improvement in overall functional status, mortality or readmissions, but which do show a tendency towards less functional decline,15 17 21 22 25 27 mortality16 or readmission,19 26 are therefore very valuable.
Comparison with published literature
As far as we know, we are the first to review hospital-wide interventions, though there are articles describing intervention studies included here. Landefeld et al summarised lessons to be learnt from Geriatric Evaluation and Management (GEM) departments, ACE units and the HELP set-up.33 Similarly, Palmisano-Mills identified the implementation of different versions of four models of integrated care for older patients (including HELP, ACE units, NICHE and a Model of Transitional Care) in 24 hospitals in Connecticut. She found that few hospitals have implemented the original models but that the majority successfully implemented key components of the care models as well as their own innovative protocols.34 However, the success of these implementation projects was never substantiated in an RCT.
This review only included RCTs and controlled clinical trials, which has led to exclusion of studies with a lower methodological quality. However, as table 2 shows, none of the studies included still is without serious risk of bias, with only one study showing protection against contamination. This evokes the question of whether these studies are methodologically flawed but could have been performed better, or whether systematic review techniques applying strict methodological Cochrane criteria are less appropriate in selecting these complex evaluations of service delivery and organisation of care.35 As such, Harari and colleagues evaluated a hospital-wide intervention in which an Older Persons' Assessment and Liaison (OPAL) team improved processes of care. Although the study design did not meet our inclusion criteria, the results of this study are promising in terms of effectiveness and efficiency.36 The same conclusion may be drawn for the Older Adult Services Inpatient Strategies (OASIS) program, which aims for improvement in care for older patients throughout the hospital.37
Additionally, we found only one study on the hospital-wide Hospital Elder Life Program (HELP), which could be included in our review.20 The others were excluded due to the study design or, for example, a focus on delirium in a specific patient group. Not including such studies based on design criteria is debatable, as such studies seem to support the evidence-based practice of implementation of HELP and subsequently prevent cognitive and functional decline38 39(http://elderlife.med.yale.edu). This also applies for the Nurses Improving Care for Health System Elders (NICHE) programme, which has evolved into a national USA/Canadian geriatric nursing programme (http://hartfordign.org). Our Cochrane review criteria yielded only two studies implementing a programme based on NICHE.40 41 A third intervention for which we could include only two articles is the Acute Care for Elders Unit (ACE),24 27 which is mentioned as the state-of-the-art care model to improve hospital-wide care for older adults. Also, other studies support the evidence that development of ACE units can improve health and functioning of older persons, without increasing healthcare costs.42 43
It should be noted that our snowball method has favoured older studies. However, recently a non-RCT study of a proactive geriatrics consultation model was published,44 indicating that hospitals are still using similar models of care to improve care for frail older patients. The same accounts for the ACE unit, which was developed in the early 1990s, where efforts are still made to get (adapted versions of) this model of care disseminated throughout hospitals.45
The key message for hospital practice is that one should investigate what works best in a specific hospital, preferably by piloting an intervention that uses effective and innovative intervention components, and incorporates the barriers and facilitators of implementation as well (appendix table 3). This stepwise procedure is proposed by the Medical Research Council's framework for complex interventions.46 Dynamic and complex healthcare organisations, such as modern hospitals, require innovative interventions as well as innovative research designs that are flexible enough to allow changes to be made during the intervention (eg, time-series analyses, before–after studies).47 For innovative hospital reform interventions, this can be realised by transition management, which adapt interventions with regard to the facilitators and barriers met during the implementation process. For evaluation, apart from more flexible options than RCTs, we suggest using quality indicators (QIs) to monitor effects on the major health problems that are targeted. For example, the Assessing Care of Vulnerable Elders (ACOVE) indicators are objective and comprehensive measures, which are a useful starting-point for developing site-specific QIs.48 49 In addition, to be able to compare outcomes in older patients within and between studies, methods for incorporating key descriptors such as cognitive and physical functioning to adjust for different case mixes should be introduced into routine clinical practice.50 Another innovative and promising evaluation of healthcare reform by complex interventions is to follow the framework that has recently been proposed by Porter et al.51 This framework defines value measures as outcomes in evaluating healthcare practices. Porter provides a framework through which this value (or performance) of an intervention can be identified, using multilevel patient-oriented outcomes related to their full costs. For both scientific and societal evaluation, it would be an important step forward to be able to continuously monitor the value of an intervention for a specific inpatient group such as frail older patients.
The current ageing of the population and developments in hospital care explicitly call for comprehensive interventions aimed to improve care for all frail older patients throughout the hospital. While implementing evidence-based practices is stimulated, only a few hospital-wide intervention RCT studies could be identified. There is an urgent need to study alternative approaches and to set adjusted scientific standards to gain firm evidence-based improvements in hospital-wide care for frail older patients.
We thank M van Tulder and M Perry, for their support and comments on the review.
Funding The work was made possible by grant 60-6190-098-272 and grant 60-61900-98-129 of the National Programme for Elderly Care, coordinated and sponsored by ZonMw, The Netherlands, Organization for Health Research and Development.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.