Frailty Index and functional level upon admission predict hospital outcomes: an interRAI-based cohort study of older patients in post-acute care hospitals

This study was a retrospective cohort study among patients aged 70 and older who were hospitalized in two geriatric post-acute care hospitals in Tampere (population base 232,000, of which 11% is aged 70 years or older), Finland, during the period of 1 February 2013 to 31 May 2016. These hospitals (230 and 190 beds) offered subacute care and rehabilitation for older patients who were first hospitalized in a tertiary or secondary care hospital (Fig. 1). In addition, one of the hospitals served as a supporting hospital for home care clients. Consequently, home care nurses or physicians in the emergency room could refer these patients directly to this hospital without hospitalization in an acute care setting. At the end of 2015, this hospital was closed due to organizational changes.

The results of the interRAI-PAC assessments (see below) were linked to hospital discharge records, which contained information about the patient’s usual residence, the place he/she was admitted from, dates of admission and discharge, discharge diagnosis and destination, and, when applicable, death during hospitalization. In patients with several hospitalizations during the observation period, the first to which interRAI data could be linked was included in this study. Information on the patient’s chronic diseases, functional ability, previous falls, smoking habits, and Body Mass Index (BMI) were collected from the interRAI-PAC. Some 2188 patients were included in the final analysis (Fig. 2).

There are several interRAI instruments with similar core items and divergent instrument-specific domains. The interRAI-PAC is designed for post-acute care and rehabilitation settings [26]. It contains information across domains, including functioning on the physical, cognitive and psycho-social levels as well as sociodemographic data, medical diagnoses, and current symptoms. Single items are combined to compose validated scales that measure different aspects of functional ability. InterRAI instruments have substantial interrater reliability [27, 28].

The use of interRAI-PAC instrument was started on 1 February 2013 in one post-acute care hospital and gradually in the other hospital. All the wards in this particular hospital had started to use interRAI-PAC by the beginning of the year 2016. Trained nurses performed the assessment within a few days of the patient’s admission to the ward. To obtain relevant information, they interviewed the patient and family members, observed the patient, and reviewed the medical records. The assessment consisted of 150 variables. The only missing variables were for weight or height (in 23 patients).

Based on previous findings of prognostic factors related to the outcomes of older inpatients [4‐7, 10, 13, 16, 18], associations of the interRAI scales measuring cognitive functions, ADLs, mood, and stability of health state were used in this study. The Cognitive Performance Scale (CPS) describes the cognitive status of the patient based on an algorithm [29]. The Activities of Daily Living Hierarchy Scale (ADLH) is an algorithm that considers a measure of ADL performance in locomotion, eating, toilet use, and personal hygiene [30]. The Depression Rating Scale (DRS) is based on existing symptoms of depression [31]. The Changes in Health, End-stage disease, and Signs and Symptoms Scale (CHESS) is a summary measure based on decline in cognition and ADL performance, certain symptoms (for example, weight loss, shortness of breath, and oedema), and ratings of a prognosis of less than six months, and it is designed to identify individuals at high risk for clinically significant decline [17].

The Frailty Index is a method to measure frailty in relation to the accumulation of health deficits [32], and it can be calculated from a variety of databases according to the standard procedure for selecting individual deficits [32]. The Frailty Index is the proportion of deficits present in an individual out of the total number of variables considered [32], and so higher scores are associated with adverse hospital outcomes – for example longer length of hospital stay, new discharge to a nursing home, and death [9, 10]. The Frailty Index from the interRAI Acute Care instrument (FI-AC) was previously derived and published by Hubbard et al. in 2015 [25]. The interRAI-AC instrument includes the same core items as the interRAI-PAC but has fewer items in total.

In this study, the Frailty Index (FI-PAC) was derived from the interRAI-PAC according to the standard procedure and the well-defined criteria created by Searle et al. [32], and leaning on the coding of variables in FI-AC. In short, all the items of the interRAI-PAC were evaluated against the Frailty Index criteria independently by two geriatricians. Secondly, eventual differences were negotiated to achieve a consensus of appropriate variables in post-acute care patient population. Finally, variables were compared with the coding of FI-AC [25]. There are several explanations for the differences between FI-PAC and FI-AC. First, some variables that were used in FI-PAC are not recorded in interRAI AC. Second, some differences are based on the differences in interpretation of the criteria for selecting appropriate variables to FI, mainly based on different characteristics of patient populations in post-acute and acute care settings. Finally, the Depression Rating Scale, Pain Scale, and Aggressive Behaviour Scale were included in the FI-PAC instead of using single variables, because the scales reflect both the patient’s situation and criteria for selecting variables to FI better than separate variables related to the issue. Of the variables considered, 57 variables were chosen for the FI-PAC [Additional file 1]. The FI-PAC was calculated for each patient by summing deficit points and dividing the sum by the total number of deficits considered. The only missing item was BMI (in 23 patients), and the denominator was adjusted to 56 items for these patients.

Prolonged hospital stay. Length of hospital stay was determined as the difference between the date of admission and the date of discharge. Length of stay in post-acute care hospital was recorded only for the patients who were discharged to their usual residency (own home or nursing home). It was not recorded for the patients who had emergency department admissions or who died during the hospital stay. In addition, length of hospital stay was not recorded for the patients who were admitted from home but were discharged to nursing home for long-term care (n = 69). This is because the delay of a new nursing home placement was most probably more dependent on the organizational factors than on patient’s condition. Length of hospital stay was dichotomously classified as less than 90 days and 90 days or more according to the usual cut-off for long-term care [33]. Hospitalization for 90 days or more was defined as a prolonged hospital stay.

In-hospital mortality was recorded from the discharge records and defined as death during the stay in the post-acute care hospital. In addition, deaths in patients who were referred to an acute care hospital because of an acute illness and who died there on the same day were also counted as in-hospital deaths (n = 4).

Patient characteristics were described using frequencies and percentages. The distribution of the FI-PAC was tested in all patients as well as in sex and age groups; the results are presented as means and standard deviations. The predictive ability of the FI-PAC on outcome measures was investigated using binary logistic regression analysis, adjusted for age and sex. Logistic regression analyses were also performed for sex and age subgroups. The receiver operating characteristic curve (ROC) and the area under the curve (AUC) with 95% confidence intervals (CIs) were calculated to clarify the discriminative ability of the FI-PAC for hospital outcomes. For each outcome measure, the optimal cut-off point of the FI-PAC for sensitivity and specificity was calculated using the Youden method, and positive and negative predictive values (PPV and NPV) were determined. To compare the predictive ability of the FI-PAC to that of existing interRAI scales, the ROC curve and the AUC with corresponding 95% CIs for hospital outcomes were also calculated for the ADLH, CHESS, CPS, and DRS scales. Data management and analysis were performed using IBM SPSS Statistics version 25.

Retrospective register-based studies in which the subjects are not contacted are not considered medical research by Finnish legislation (Medical Research Act 1999/488 § 2) [34] and, therefore, ethics committee approval was not required. Retrospectively collected health register data could be used for this study with permission of register owner without participants’ informed consent, based on current legislation (Data Protection Act 2018/2010, Act on the Publicity of Official Documents 1999/621, European Union General Data Protection Regulation) [35‐37]. Research plan was institutionally reviewed and permission to use the interRAI-PAC assessments and hospital discharge records was hence obtained from the administration of the City of Tampere (decision the Director of Hospital Services, in August 30, 2016).

The cohort included 2188 patients with a mean age (SD) of 84.7 (6.3) years. Most of the patients were female (n = 1499, 69%) (Table 1). Almost half of the patients (46%, n = 1004) had a memory disorder diagnosis. Only 12% of the patients (n = 255) were independent in all basic activities of daily living (BADLs) (bathing, personal hygiene, dressing, walking, locomotion, transfer to toilet, toilet use, bed mobility, and eating), while 18% (n = 395) were totally dependent on caregivers for all BADLs. Half of the patients came to hospital straight from home and the other half came from an acute care hospital.

Most of the patients (n = 1691, 77%) were discharged to their usual place of residence (own home or nursing home) (Table 1). The median length of stay in post-acute care was 35 days (interquartile range 18–87 days), and 409/1691 patients (24%) had a prolonged hospital stay. Some 204/2188 patients (9%) were admitted to the emergency department. The in-hospital mortality rate was 11% (n = 231/2188).

The FI-PAC was normally distributed, with a mean (SD) score of 0.34 (0.15), a minimum of 0.01 and a maximum of 0.76 (Fig. 3). There were no significant differences between age and sex groups.