Factors associated with length of stay in care homes: a systematic review of international literature

A protocol for the systematic review was prepared prior to conducting the review. A systematic search strategy was developed and reported using criteria established in the Preferred Reporting Items for Systematic Review and Meta-Analysis Protocols (PRISMA-P) 2015 statement [26]. The search strategy included a combination of free-text terms and subject indexing terms, such as MeSH and EMTREE (Table 1). The search strategy was developed through identification of key terms in the titles and abstracts of relevant studies identified in an initial scoping search of the literature.

The following electronic databases were searched for articles published in peer-reviewed journals, from database inception to September 2016, and were not limited by language or publication restrictions: MEDLINE, EMBASE, PsycINFO, Cumulative Index to Nursing and Allied Health Literature (CINAHL), Proquest, the Cochrane Library, including the Cochrane Methodology Register, Cochrane Central Register of Controlled Trials (CENTRAL), Cochrane Database of Systematic Reviews (CDSR), Database of Abstracts of Reviews of Effect (DARE), Health Technology Assessment (HTA) database and NHS Economic Evaluation Database (NHS EED), Web of Science, the Campbell Library, SCOPUS and Social Care Online.

In cases where a conference abstract or unpublished research met the inclusion criteria, the lead author was contacted where possible. Additional papers were identified through other sources, including reviewing the reference lists of publications that met the inclusion criteria, reverse citation searches and grey literature. Reverse citation searches were undertaken on included papers using the ISI Web of Science Citation Databases. Grey literature was searched using OpenGrey. Websites were searched using an abbreviated search strategy; these included the World Health Organization, European Association of Palliative Care and Age UK.

The review focused on older adults, defined as any adult or groups of adults aged 65 years or above. Participants were included in the review if they resided in a care home or long-term care facility. As previously stated, the term “long-term care facility” generally refers to a collective institutional setting where care is provided to older people, who live there, 24 hours a day, 7 days a week, as defined previously [8]. In the review, this definition was also applied to care homes. Studies in hospitals, assisted living facilities, sheltered housing and hospices were excluded.

The review had one outcome measure, length of stay, defined as length of stay within a care home until death. Length of stay could be measured in days, months, years or any other unit of time; and measured from any time point after admission. Studies exploring length of stay until discharge were excluded. The review was restricted to observational studies, including retrospective or prospective cohort studies and case-control studies.

In the first stage of screening, a decision on whether a paper met the inclusion criteria was based on the study title and abstract. In the second stage of screening, a final decision on inclusion was made based on reading the full paper.

In both stages, screening was conducted by one reviewer (DCM), and decisions checked by a blinded second reviewer (LD) on a subsample of 10% of the papers. Discrepancies were discussed, and a final decision was made by a third member of the research team (KF). Reasons for excluding full papers were recorded and reported (Fig. 1).

Data from the included studies were extracted using a data extraction form specifically developed for this review, informed by STROBE statement on the reporting of observational studies [27]. Data was extracted by one reviewer (DCM). The data collected from each study included: information on participants (number of residents, age of residents, gender of residents), the care home (type of care home, number of care homes), the methodology (study setting, type of study, dataset used, data collection period, follow up period, how length of stay was defined, variables included in model, statistical method used) and information on missing data. The results of the paper, including risk measure used and its value, confidence intervals and measure of significance were also extracted.

If a study contained more than one cohort, both cohorts were included as separate groups, such as Engle and Graney [28]. In development and validation studies, data from the development cohort was extracted. In studies where men and women were listed separately, such as Hedinger, Hamming and Bopp, [29], the cohort that reflected the sample majority was used. In cases where the results were significant, either through significance testing or confidence intervals, the result was classed as significant. In cases were no data on an included variable was reported in the results, the result was not significant or the confidence interval indicated none significance, the result was classed as not significant. If a study reported more than one result for the same outcome, i.e. age reported in groups, the result with either the most significant or the largest ratio was used to avoid duplication. Factors included in less than three studies were excluded from the analysis to avoid bias in the evidence synthesis, which required a minimum of three studies to apply the data synthesis.

There are numerous tools to assess the methodological quality and risk of bias in randomized controlled trials; however, few are designed specifically for observational studies [30]. To ensure that the quality assessment tool was appropriate for the topic area, the review used a modified version of the quality assessment tool used by Luppa et al., adapted for application to studies on factors associated with length of stay in care homes. The adapted tool has 14 items, scored as either 0 (not meeting the criteria) or 1 (meeting the criteria) (see Additional file 1). A score of 75% or more of the assessment criteria was defined as high-quality studies, between 50% and 75% was defined as moderate quality and less than 50% was defined as low quality.

Due to heterogeneity in the setting, sample and tools used, a meta-analysis was not appropriate to synthesize the data. Data synthesis was split into two stages. Firstly, the characteristics of the study were described, including the study design, sample size and variables reported. Numerical data was tabulated and presented for each factor, showing effect size and direction.

Secondly, in a similar approach used by Luppa et al., all factors identified across the studies were pooled and grouped, alongside the direction of the variables effect on length of stay and their significance (positive, negative and not significant). These findings were then cross-referenced with the quality assessment score for the study (see Additional files 2 and 3).

A judgment was made on the strength of the evidence for each risk factor from this cross tabulation. A factor was classed as being supported by strong evidence if there are consistent findings in at least 75% of studies in at least three high-quality studies. Moderate evidence was classed as consistent findings in at least 50% of studies in at least two high-quality studies. Weak evidence was classed as findings of one high-quality study and of at least two moderate- to low-quality studies or consistent findings (≥ 75%) in at least four or more moderate- to low-quality studies. If a risk factor was not classed as having strong, moderate or weak evidence, it was classed as inconclusive.

Fourteen of the 47 studies were judged to be of a high quality, 31 of a moderate quality and 2 of a low quality. The lowest scores on the methodological quality were on reporting the training and quality control methods for interviewers’ technique (reported in 8/47 papers) and the reliability and/or validity of study instruments (reported in 8/47 papers).

Factors associated with length of stay, minimum and maximum risk results and strength of evidence have been summarized in Table 2. Three factors had strong evidence to support their association with shorter lengths of stay: shortness of breath, receipt of oxygen therapy and admission to a nursing home. Nine factors had moderate evidence to support their association with shorter lengths of stay: cancer, increased contact with primary care, poor general health, poor mobility, low BMI or malnutrition, poor physical functioning, presence of pressure ulcers, older age and being male.

Weak evidence to support their association with shorter lengths of stay was identified for admission from hospital, behaviour problems, biochemical indicators, poor cognitive function, dementia or Alzheimer’s disease, depression, diabetes, poor appetite, presence of a feeding tube, help with feeding or diet, hallucinations, delusions, wandering or delirium, incontinence or catheter use, respiratory disorders or COPD, history of stroke, vision impairment, and being married. History of fractures or falls, being of white ethnicity and vaccinations decreased the risk of shorter lengths of stay.

A subgroup analysis was conducted on studies that investigated a follow-up period of 1 year or less (Table 3). Twenty-six papers were included, with the same criteria as applied to the full sample. Within 1 year of follow-up, oxygen therapy remained strongly associated with shorter stays, although residence in a nursing home and shortness of breath were associated with moderate evidence. Increased age, cancer, poor appetite, being male, poor general health, low BMI or malnutrition, poor physical functioning remained supported by moderate evidence, respiratory disorders or COPD increased from weak to moderate evidence.

To the authors’ knowledge, this is the first systematic review of international literature on factors associated with length of stay in care homes. The review had broad inclusion criteria and was not limited to one type of care home. Efforts to ensure that the studies identified were comparable limited the review to study designs that followed residents until death, meaning that residents who were discharged or moved to another care home were not included. The inclusion of variables related to resident and care home characteristics adds an additional dimension to the review.

The review is potentially limited by the heterogeneity in the terminology used in studies in care homes. Efforts were made to include a multitude of terms related to care homes in the search strategy used to identify studies from a variety of settings; however, it is possible that some studies that included less common terms for care homes were missed. The levels of care provided by different care homes, their admission criteria and the terminology used to define care homes vary between countries making the synthesis of data published on care homes problematic. For example, the definition of long-term care facilities applied in this review excluded studies in assisted living facilities and sheltered accommodation. Further discussion is needed to refine the terms used in this area and improve subsequent reporting. In addition, it is difficult to make meaningful comparisons between countries without considering the national policies and provision of care offered by care homes within a country’s wider health care system. The review also does not capture a resident’s living arrangement prior to admission that may affect the decision to enter a care home and their subsequent experience.

The focus of the review centered on studies that had explored a number of factors as their aim, without any prior hypothesis. Studies that explored the role of one factor on length of stay, such as depression or malnutrition, were excluded, for two reasons. Firstly, it would be impossible to develop a search strategy that could identify all factors associated with length of stay, without first having a basis on which to justify the inclusion of search terms for each factor. Secondly, the number of studies identified would be very large and difficult to synthesize.

The interpretation of the results is also limited by the limitations of the individual studies. Most studies only collected data at baseline, with follow-up restricted to the outcome measure of time until death. It is possible that changes in time-dependent characteristics, such as cognitive impairment, which may get progressively worse, were missed. There were numerous measures used to assess factors, for example, at least 6 tools were used to assess physical functioning and 12 for cognitive functioning; however, the analytical approach used in this review allowed these findings to be combined and weighted into a meaningful measure of association and methodological quality. Finally, the review did not perform study selection and study extraction in duplicate on the full dataset.

The study highlighted the varied trajectories of care home residents approaching end of life and the need for flexible, appropriate palliative care provision to accommodate different trajectories. This review has synthesized factors associated with variation in length of stay in care homes, and identified similar homogeneity within the care home population from admission to death, which is not yet fully understood.

Increasingly, care homes are taking on a complex role within the wider health system, catering for the diversities of an ageing population that can no longer live in the community. At one end of the spectrum, care homes are acting as proxy hospices for short-stay residents approaching end of life. At the other end, care homes are accommodating residents with cognitive impairments who may survive for many years post-admission. Supporting care homes in negotiating these two roles; delivering palliative care for short-stay residents while simultaneously providing a residential home for long-stay residents, in the same space, is imperative and requires further research. In addition, further thought should be given to the suitability of care homes in catering for such a wide variation in needs. The potential for other types of services, such as specialist dementia care units and assisted living facilities, in providing care for subgroups of care home residents could be examined, although there is debate as to whether such services provide better care [43, 44].

The findings of this review have identified numerous questions requiring further investigation. Firstly, further research is needed to explore the relationship between factors associated with care home admission and factors associated with length of stay. Longitudinal studies which follow community-dwelling older adults post-admission are required to fully understand this relationship. Secondly, in this review, characteristics related to the care home were only collected in eight studies, further data on variation in length of stay and care home-related factors could identify ways to improve the delivery of care. It is imperative that research on care home residents contextualizes the data within the long-term care setting to inform the generalizability of the findings internationally. Finally, inclusion and identification of care home residents in existing national datasets would allow comparisons within and between countries, and enable time-dependent variables to be monitored.