Background
Geriatric syndromes are common clinical conditions in older adults [
1]. They are often connected to each other with multiple shared underlying aetiological factors that involve different organ systems [
1]. Frailty is a geriatric syndrome in which the patient’s ability to resist stressful events is reduced as a result of age-related cumulative decline in many physiological systems [
2]. At least in its early stages, frailty is a potentially reversible condition [
3].
Frail older patients [
4,
5] and those suffering from other geriatric syndromes [
6,
7] are vulnerable to adverse outcomes. Frailty predicts prolonged hospital stay [
8‐
10] and in-hospital mortality [
10‐
12]. Impaired functional ability in activities of daily living (ADLs) and impaired cognition predict all-cause mortality among hospitalized patients [
13,
14]. Symptoms of depression associate with in-hospital mortality, all-cause mortality, and length of hospital stay [
15,
16]. In addition, stability in health state, measured by combining different instability symptoms with functional ability, declined cognition, and poor prognosis, predicts all-cause mortality among institutionalized patients and patients with neurological conditions [
17,
18], but studies among hospitalized patients are lacking.
Even though geriatric syndromes are highly prevalent among acutely ill hospitalized patients [
6,
19], the recognition rate of these conditions is low [
6]. However, hospitalization offers opportunities to identify and act on geriatric syndromes and undiagnosed diseases [
20]. The Comprehensive Geriatric Assessment (CGA) was developed to improve the identification of older patients with geriatric syndromes [
19]. The CGA includes an assessment of the patient’s medical, psychological, cognitive and functional problems, as well as environmental and social factors. The assessment leads to a treatment plan, rehabilitation, and follow-up [
19]. Performing the CGA during a stay in acute care increases the patient’s likelihood of being alive and living at home one year later [
19].
There is currently no clear consensus about the contents of the CGA, and several different CGA approaches have been developed. One example is the interRAI assessment system, which can be used as a CGA tool [
21]. Similarly, frailty does not yet have an internationally recognized standard definition, nor is there a gold standard for detecting it [
22]. Instead, there are multiple frailty instruments that are based on one of two widely used frailty models: the phenotypic model [
23] and the cumulative deficit model [
24]. The phenotypic model defines frailty as the presence of three or more of five factors in an individual [
23]. In the cumulative deficit model, frailty is defined as the cumulative effect of individual deficits [
24]. The Frailty Index is based on this latter model [
24]. Although the interRAI instrument is lacking a frailty scale, it can be derived from the database [
25].
To the best of our knowledge, no previous studies have dealt with the prognostic effects of the Frailty Index and different interRAI scales in post-acute care. The aims of this study were 1) to derive a Frailty Index (FI-PAC) from the interRAI Post-Acute Care instrument (interRAI-PAC), 2) to determine how the FI-PAC associates with hospital outcomes (in-hospital mortality, prolonged hospital stay, and emergency department admission), and 3) to clarify how the other scales of the interRAI-PAC compare in the prediction of hospital outcomes.
Discussion
In this large retrospective cohort study of older patients in a post-acute care setting, we derived a Frailty Index (FI-PAC) from the interRAI Post-Acute Care instrument (interRAI-PAC) to summarize the results of the comprehensive assessment. A Frailty Index has previously been derived from the interRAI Acute Care instrument [
25], and it has been shown to predict multiple adverse outcomes in hospitalized older patients [
10], but the interRAI-PAC has not been previously used for that purpose. Most variables are the same in the FI-PAC as in the Frailty Index derived from the interRAI assessment system for Acute Care (FI-AC), but one difference is that instead of using single variables, we included the Depression Rating Scale (DRS), Pain Scale (PAIN), and Aggressive Behaviour Scale (ABS) in the FI-PAC. Another difference is that we did not include the number of medications in the FI-PAC. In addition, we included walking speed.
We succeeded in deriving a Frailty Index from the interRAI-PAC with the expected normal distribution in this study population [
25,
39]. The distribution of the Frailty Index is usually skewed in population-based samples, but it tends to change to a normal distribution in more morbid and unwell groups of older people [
41]. However, a skewed distribution was also found in hospitalized older patients in a study by Cesari et al. [
11]. This discrepancy could be attributed to the better functional ability of the patients in their study. The mean score for the FI-PAC was 0.34, which was close to the mean score of 0.32 for the FI-AC [
25]. There were no significant differences between age and sex groups, and this finding is consistent with the finding of Hubbard et al. [
25].
It transpired that the FI-PAC was associated with both prolonged hospital stay and in-hospital mortality, and it had a good discriminative ability (both AUCs over 0.70). Previous studies have not dealt with length of hospital stay in the post-acute care setting, but the results from acute care showed an association between the Frailty Index and prolonged length of stay [
8,
9]. In accordance with our results, Hubbard et al. found an association between the FI-AC and in-hospital mortality [
10]. This finding is also consistent with previous studies that have examined the predictive ability of the Frailty Index [
11] and the Clinical Frailty Scale [
40,
41] for in-hospital mortality in the acute care setting.
It was noted also that the FI-PAC associated with emergency department admission, but the predictive ability was only modest. This result may be explained by the fact that most short-term readmissions to acute care hospitals are due to medical issues [
42,
43] – for example, acute and chronic diseases – and the impact of these diseases on admission to acute care is greater than that of frailty status.
Interestingly, the FI-PAC was equal but not superior to ADLH in predicting prolonged hospital stay and in-hospital mortality. However, having a high Frailty Index significantly increased the odds for adverse hospital outcomes in patients with ADL impairments or cognitive decline compared to the effects of these conditions alone. In their analysis based on the FI-AC, Hubbard et al. did not compare the predictive ability of the FI-AC to the standard interRAI scales [
10]. Although several studies have shown that ADL impairment upon admission to acute hospital is a strong predictor of prolonged hospital stay and mortality in older patients [
14,
43,
45], it was surprising that functional impairment, measured by the short ADLH scale, was as good a prognostic instrument as the multicomponent Frailty Index. These results are, however, in agreement with Chen’s findings, which showed that frailty and functional dependence were comparable in predicting short-term outcomes after gastrointestinal surgery [
46]. A possible explanation might be that frailty is a complex phenomenon and different instruments – for example, the Frailty Index – can measure only some aspects of it [
3]. Although the Frailty Index consists of a variety of different health-related items, it more or less represents a sum of comorbidities and disabilities rather than a measure of the biological aspects of frailty [
47]. If measuring biological (phenotypic) frailty had been possible in our study, the results might be different.
It can thus be suggested that, in clinical practice, calculating the Frailty Index for the purpose of identifying patients with poor outcomes does not bring additional value over assessment of functional ability. Instead, the detection of functional impairment can be used to define frailty [
48]. From a clinical point of view, assessment of the patient’s functional ability is simple, quick, and inexpensive, and it is usually already part of the nurses’ assessment protocol. Owing to the multifactorial basis of functional impairment [
49], factors underlying each person’s functional decline are probably different regardless of similar scores on the Frailty Index. Thus, the detection of functional impairment should in turn lead to the comprehensive clinical and interprofessional evaluation of the patient in order to clarify underlying factors and make a plan for proper treatment and rehabilitation.
For clinical decision making, cut-off points with approximate discrimination between robust, prefrail and frail individuals have been developed. In older adults with functional decline, the cut-off point is about 0.25 between robust and prefrail and about 0.40 between prefrail and frail [
10,
38]. We considered it important to clarify the clinically relevant cut-off points for the FI-PAC that can be used to differentiate persons who are likely to experience adverse outcomes during their hospitalization from those who are likely to survive without complications. Optimal cut-off points, based on the ROC curves, varied from 0.30 to 0.35 in our study population. The problem with the Frailty Index in this patient population is that by using the cut-off point of 0.35, half of the patients are classified as being at risk for adverse outcomes. However, scores that were lower than the cut-off points ruled out most patients who did not face adverse outcomes during hospitalization.
The strengths of our study are the representative sample size and quite homogenous patient population, the complete records, and the representation of real-life patients due to the retrospective nature of the study. However, a note of caution is due here since our materials did not include all patients that had a treatment period in a post-acute care hospital during the study period, because the interRAI assessment was not made for everybody. There are many possible reasons for missing assessments. One reason is that the introduction of interRAI-PAC was gradual in different wards, but hospital discharge records were collected the same period of time from both hospitals. In addition, the assessment was not done for the patients who were in a terminal care phase and to the patients with suspected hospital stay for less than seven days. Another reason may be related to the fact that the completion of an interRAI assessment is time and resource demanding [
50], which may lead to a substantial number of the missing assessments in real-life clinical context [
51]. However, this is unlikely to cause systematic bias in our analysis.
Another source of uncertainty is our lack of knowledge of incidents occurring during the whole hospital treatment period of the patient – for example, the length of stay in an acute care hospital, diagnoses of acute diseases, or treatments given. The predictive ability of the FI-PAC probably varies between different patient groups, for instance between patients whose reason for hospitalization is acute disease versus patients whose reason for the hospital stay is postoperative rehabilitation. Therefore, caution must be applied when applying our results to diverse patient groups. In addition, although our materials cover all post-acute care in our city and although the patients represent unselected population (in terms of social or insurance status), it is acknowledged that in international context, the current patient numbers are modest and the results may not be fully generalizable to other health care systems.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.