Frailty status at admission to hospital predicts multiple adverse outcomes

Abstract

Aims

frailty is proposed as a summative measure of health status and marker of individual vulnerability. We aimed to investigate the discriminative capacity of a frailty index (FI) derived from interRAI Comprehensive Geriatric Assessment for Acute Care (AC) in relation to multiple adverse inpatient outcomes.

Methods

in this prospective cohort study, an FI was derived for 1,418 patients ≥70 years across 11 hospitals in Australia. The interRAI-AC was administered at admission and discharge by trained nurses, who also screened patients daily for geriatric syndromes.

Results

in adjusted logistic regression models an increase of 0.1 in FI was significantly associated with increased likelihood of length of stay >28 days (odds ratio [OR]: 1.29 [1.10–1.52]), new discharge to residential aged care (OR: 1.31 [1.10–1.57]), in-hospital falls (OR: 1.29 [1.10–1.50]), delirium (OR: 2.34 [2.08–2.63]), pressure ulcer incidence (OR: 1.51 [1.23–1.87]) and inpatient mortality (OR: 2.01 [1.66–2.42]). For each of these adverse outcomes, the cut-point at which optimal sensitivity and specificity occurred was for an FI > 0.40. Specificity was higher than sensitivity with positive predictive values of 7–52% and negative predictive values of 88–98%. FI-AC was not significantly associated with readmissions to hospital.

Conclusions

the interRAI-AC can be used to derive a single score that predicts multiple adverse outcomes in older inpatients. A score of ≤0.40 can well discriminate patients who are unlikely to die or experience a geriatric syndrome. Whether the FI-AC can result in management decisions that improve outcomes requires further study.

Introduction

Frailty is proposed as a summative measure of health status and marker of individual vulnerability. Older people who are frail have a diminished capacity to compensate effectively for external stressors and thus are vulnerable to adverse outcomes [1]. Identification of inpatients who are frail has the potential to target their care more appropriately. While frailty status is attractive as a risk stratification tool, the evaluation of frailty in the acute care (AC) setting has proven problematic. Many inpatients are unable to complete the performance-based tests integral to some frailty measures [2] and discriminative ability may be lacking for tools that categorise the majority of cohorts as ‘frail’ [3]. Comparison of multiple frailty instruments has reported generally poor predictive properties [4] and low sensitivity for adverse events (AE) [5, 6].

The quantification of frailty by counting accumulated deficits to determine a frailty index (FI) score has the potential to overcome some of these challenges. A FI is feasible for even bed-bound older inpatients [7] and as a continuous rather than dichotomous measure may harness greater granularity across the health spectrum. An electronic FI (eFI) has recently been validated using health record data in primary care [8]. An existing hospital-based assessment system, which serves other clinical and administrative purposes, can also be used to derive a FI. A FI derived from the interRAI-Acute Care Instrument (FI-AC) [9] increases with chronological age and reaches a 99% limit of 0.69, below the theoretical maximum of 1.0 [10]. The strong association between the FI-AC and death has been reported [10], but a valid frailty measure should be more than a mortality-prediction tool. Geriatric syndromes such as delirium, pressure ulcers and falls exert a high toll in personal suffering as well as greater healthcare costs [11, 12]. Functional decline and institutionalisation should also be considered since these outcomes are of primary importance to older people themselves [13].

Here, we evaluate the predictive validity of the FI-AC in older inpatients. We consider how admission FI-AC score relates to discharge destination and explore its association with other clinically important adverse outcomes.

Methods

Setting and sample

Previous studies have described recruitment of the study cohort [10, 14–16] and the relationship of polypharmacy with adverse outcomes in hospitalised older people [17]. In brief, 1,418 patients aged 70 and older admitted to 11 AC hospitals across Australia were recruited. These hospitals included small secondary care centres, rural hospitals, metropolitan teaching centres and major tertiary referral facilities. The majority of patients (N = 1,220) were recruited from general medical wards with 198 in orthopaedic or surgical wards. Patients admitted to coronary or intensive care units, for palliative care or for less than 24 h were excluded.

Measures

Frailty

The interRAI-AC tool, now known as the interRAI-AC-Comprehensive Geriatric Assessment (AC-CGA), was used for data collection. This instrument was specifically developed for use in the acute setting to support CGA of older inpatients [9, 18]. It collates information across a large number of domains including sociodemographic data, physical, cognitive and psycho-social functioning, medications, medical diagnoses, advance directives and discharge destination. Nurse assessors who were trained to use the interRAI-AC instrument gathered data at admission (within 24 h in the ward) and discharge. To obtain information for each item in the interRAI instrument, patient and family interviews, direct observations, staff interview and medical records were used. A number of scales embedded in the interRAI instruments combine single items belonging to domains such as activities of daily living (ADL), instrumental ADL and cognition; these are used to describe the presence and extent of deficits in these domains [18].

A frailty index (FI-AC) was calculated for each patient using candidate variables recorded in the interRAI-AC. Variables were chosen and coded as deficits according to well-defined criteria [10]. Deficits included co-morbidities, incontinence and pain as well as impairments in function, sensorium, mood and cognition. Each individual's deficit points were summed and divided by the total number of deficits considered (here = 56). For example, an individual with 12 deficits out of 56 counted had an FI-AC of 0.21. The FI has been contextualised against clinical descriptors: 0.25 has been proposed as the cut-off between ‘fit’ and ‘frail’ in community-dwelling older people [19] and scores of 0.4 and above describe older people who are dependent on others for ADL and have a significantly higher risk of death [20].

Outcome measures

Discharge destination

Discharge destination after the AC episode was classified on an ordinal scale as community, continuing inpatient care (including rehabilitation, palliative or extended care), residential aged care (RAC) or death.

Newly discharged to RAC

Those not in RAC at admission and who were discharged directly from AC to RAC, were classified as newly discharged to RAC. Those who died in hospital were excluded.

In-hospital mortality

In-hospital mortality was recorded for those patients who died during the hospital episode.

Fall in hospital

In-hospital fall was defined as having at least one fall during the period of hospitalisation. These data were collected prospectively by the research nurses using all available sources of information including interviewing the patient and medical staff, daily ward visits to review medical records and checking the forms or systems for recording AEs.

Functional decline

This was assessed using change in the ADL short form scale that consists of four items (personal hygiene, walking, toilet use and eating). Scores on the ADL scale range from 0 to 16, with higher scores indicating greater impairment [12]. In hospital functional decline was defined as having a worse (higher) ADL score on discharge compared to admission. Those who died in hospital were excluded.

Delirium

As part of the interRAI-AC, varying mental function and acute changes in mental status from baseline were evaluated by the nurse assessors at admission and discharge. The two items were combined to screen for delirium. This screener has been validated in a prospective observational study with good positive predictive value (PPV) of delirium [21]. Delirium in hospital was recorded if the interRAI delirium screen was positive at the admission or discharge assessments or if delirium and/or any acute change in cognitive function was noted in the hospital records on daily ward visits by the nurse assessor.

Pressure ulcer incidence

As previously described [22], new cases of pressure ulcer were assessed by the research nurse who visited the wards daily and viewed medical charts to record AEs including onset of any pressure ulcer that developed during the AC stay. Those with pressure ulcer present at admission were excluded.

Readmissions or deaths within 28 days

Patients were followed up at 28 days post discharge from AC by telephone and/or viewing medical records to determine if the patient had been readmitted to hospital or died during that period.

Composite AE in hospital

This included inpatient fall, delirium or pressure ulcer.

Composite adverse outcome

Patients discharged to a RAC facility or who died in hospital or within 28 days post discharge were combined in this category.

Analysis

Frequency distributions were used to describe cohort characteristics. Percentages are given as the proportion of available data. To investigate the predictive ability of the FI-AC on various AEs, we performed ordinal regression for discharge outcome and logistic regression between each of the other AE and FI-AC adjusting for age and gender. We categorised the FI into nine groups such as 0–0.1000, 0.1001–0.2000 and so on in order to compute an odds ratio (OR) and 95% confidence interval (CI) of OR. To summarise discrimination we prepared the receiver operating characteristic (ROC) curve and computed the area under the curve (AUC), also known as c-statistics. The ROC curve is seen as a better means of assessing a binary logistic regression model's ability to accurately classify observations, with AUC values of 0.70 and higher considered acceptable discrimination [23].

For each adverse outcome, the FI cut-point at which optimal sensitivity and specificity occurred was calculated as well as PPV and negative predictive value (NPV) related to this FI score.

Ethics

Ethical approval for secondary analysis of the interRAI data set was granted from the University of Queensland medical research ethics committee. The University of Queensland has access to de-identified clinical and research data on interRAI-AC assessments held in data repositories managed by the University of Queensland under a licensing agreement with the interRAI organisation, Michigan, USA.

Results

In the cohort of 1,418 patients, mean age was 81 (SD 6.8) years, and 55% were female. Prior to admission, 85% were living independently in the community and median length of stay in AC was 7 days (interquartile range 4–11 days). Principal diagnoses (ICD codes) were recorded for 1,411 participants (99.5% of the sample). Sociodemographic and clinical characteristics of the study population are shown in Table 1.

An FI-AC could be derived for 100% of the cohort and was normally distributed around a mean (SD) of 0.32 (0.14). In ordinal regression models, those discharged to the community (n = 917) were the least frail (0.28 ± 0.12), while those discharged to other inpatient care (n = 237), RAC (n = 207) or who died (n = 57) were progressively frailer (0.39 ± 0.13; 0.41 ± 0.13; 0.47 ± 0.16 respectively) (OR: 1.93 [95% CI:1.77–2.12]).

In logistic regression models, adjusting for age and gender, each increase of 0.1 in FI was significantly associated with an increased likelihood of each adverse outcome considered, with the exception of readmissions within 28 days (Table 2).

For each adverse outcome, the cut-point at which optimal sensitivity and specificity occurred was for an FI > 0.4. At this FI value, the specificity was higher than its sensitivity. The PPV varied from 7% for inpatient pressure ulcer to 52% for delirium with consistently high NPVs (88–98%).

Discussion

In this cohort of older inpatients, a FI derived from an existing assessment system was significantly associated with multiple adverse outcomes. While frailty was measured within 24 h of admission to hospital and may have been impacted by acute illness, our findings confirm the importance of patients’ current status in prognostication. Older, complex patients are now scattered throughout general medicine and surgical wards and many doctors feel inadequately trained to meet their individual needs [24]. Identification of older patients most vulnerable to AEs is a critical first step to optimise management and, ultimately, improve outcomes. The generation of a FI could help all doctors, not just those with training and expertise in geriatric medicine, to stratify risk status.

While the continuous nature of the FI is one of its strengths, in this study we were motivated to explore a cut-point that might inform decision-making. Our findings are clinically coherent. The FI of 0.40 lies between the descriptors moderately frail (mean FI = 0.36) and severely frail (mean FI = 0.43) on the Clinical Frailty Scale [20]. Results are also consistent with previous studies. In 2305 community dwellers >65 years, no one with FI > 0.45 was free of both disability and co-morbidity [25]. Here, the NPV of FI > 0.40 was consistently high, indicating that geriatric syndromes are uncommon in older inpatients who are more robust. Integrated, multi-component interventions which address common shared risk factors can significantly reduce the incidence of geriatric syndromes, including falls, delirium and functional decline [26]. Components include encouraging early mobilisation, optimising nutrition and hydration and providing reorientation and engagement. Whether systematic application of these strategies can reduce adverse outcomes for frail older people on non-geriatric wards is a focus of current studies by our group [27].

The links between baseline frailty and both delirium [28–30] and falls [30] in older inpatients have previously been recognised. This is congruent with the conceptualisation of a frail older person as a complex system on the threshold of failure. When redundancy has been lost, even minor triggers can result in perturbations of higher order functions such as divided thinking or upright ambulation [1, 31]. On the other hand, we found no association between frailty status and readmission to hospital. Again, this is consistent with previous studies [4]. Readmissions to the acute sector are notoriously difficult to predict [32] but medical instability and socio-environmental vulnerability [33] seem to be key contributors, neither of which is captured by current single-point frailty measures. Incorporation of assets, such as the presence of a supportive care-giver and affluence, may improve prediction of this outcome. Neighbourhood deprivation, for example, can be reflected in patients’ residential postcodes [34] which are already routinely included in electronic admission documentation. The identification of patients most at risk could trigger implementation of interventions proven to be effective at reducing hospital readmissions, such as follow-up phone calls and home visits [35].

This study has certain strengths. The study population is a large cohort of patients recruited from secondary and tertiary care settings with detailed assessment of patients’ functional and cognitive status. Data collection was comprehensive and complete with less than 2% missing data in the final analysis models. We also acknowledge methodological weaknesses. The instrument used, the interRAI-AC, does not yet have widespread international uptake. Yet, it does have more uptake than any other omnibus assessment system. The derivation of an FI-AC from this instrument does not depend on the further collection of data. Importantly, it can be calculated for even bed-bound and dependent older people, as exemplified by the 100% completion in this cohort.

Electronic medical records have considerable potential to individualise patient care [36]. For example, they have already been used to improve the risk stratification of patients presenting with suicidal behaviour [37]. Comparison of the eFI derived from community-based assessment [8] to the FI at admission to hospital could track health across care settings, providing a frailty trajectory which may be more critical to prognosis than either measure in isolation [38]. The identification of high-risk patients would enable us to move forward with the next step, of whether knowing such information with precision can aid the clinical judgement now employed daily to proceed with usual care, or to modify it based on the vulnerability of the person to whom it is aimed. Whether such modification will target healthcare resources more effectively and appropriately, and result in improved outcomes for older inpatients, should be the focus of future enquiries.

Conflicts of interest

Through the Dalhousie Industry Liaison Office K.R. has asserted copyright of the Clinical Frailty Scale (not mentioned here). K.M. founded and is Chief Scientific Officer of DGI Clinical, which has contracts with pharma in the areas of dementia, and haematological disorders.

Funding

This study was supported by the Australian National Health and Medical Research Council (http://dx.doi.org/10.13039/501100000925) Project Grant (APP 1065352) on Stratification of Risk for Older People in Hospital. The funding source had no involvement in the design of the study, analysis or writing of the paper.

References