Design and participants
This prospective longitudinal study is part of a larger prospective observational study, Hospitalization Process Effects on Mobility Outcomes and Recovery (HoPE-MOR) [37]. Older adults admitted from February 2019 through February 2020 to one of three internal medicine wards in a medical center in northern Israel, were screened for participation in the current study. Patients recruited for the study had an unplanned admission due to a non-disabling, acute medical illness (e.g., pneumonia and acute bronchitis) and met the following inclusion criteria: (1) age 65 years and above, (2) able to speak, read and write Hebrew, (3) able to sign an informed consent form as determined by a health professional, (4) self-reported pre-admission ability to walk with or without personal or assistive device. Exclusion criteria included (1) admission diagnosis of acute or chronic neurologic disorder, (2) acute orthopedic condition (e.g., fractures), (3) diagnosis of dementia (4) delirium, the presence of delirium was evaluated at admission and daily during hospitalization using the 4'A's Test (4AT) [38], (5) need for mechanical ventilation, (6) prescribed isolation, (7) admitted for end-of-life care. A total of 145 patients were approached for participation in the current study, of them 45 declined, leaving 100 eligible participants. Of the100 participants recruited, eight withdrew their consent during hospitalization or one month after discharge, three were transferred to another ward or discharged after less than 24 h, three were discharged to a rehabilitation facility, and two participants were unable to contact at one month, leading to 84 participants that completed one month follow up. In the three months’ follow-up, one participant withdrew consent and one died. In addition, the three months’ follow-up of five participants was during the first outbreak of the COVID-19 pandemic therefore their data were excluded from analysis, leaving 77 participants at three months follow-up.
The sample size was calculated (using G*Power) for linear multiple regression R2 increase and the following parameters: significance level of 0.05, power of 0.80, and effect size of f2 = 0.15 for four stages (blocks) and a total of five predictors. Based on this calculation, the total number of participants needed for this study was 85. Nevertheless, a rule-of-thumb suggests that 10 subjects per variable is the minimum required sample size for linear regression models to ensure accurate prediction [39], indicating an acceptable sample size for the present study.
Data collection
Trained research assistants administered the assessment measures and questionnaires during the first 48 h of hospitalization in the internal medicine ward, and after obtaining an informed consent. At one-month post-discharge the follow up was conducted at the participant’s homes and at three months, the follow up was conducted by a telephone interview. The same research assistant (an experienced occupational therapist) administered the measure of participation during hospitalization and after discharge, as well as the functional cognition screening measure during hospitalization.
Measurements
Main outcome-participation
Participation was measured by the Activity Card Sort (ACS) version 1 that is valid in Israel [40]. An interview version was conducted to determine the impact of the acute illness on the individual's level of activity and participation compared with before hospitalization. The ACS consists of 88 daily activities divided into four categories: IADL (e.g., grocery shopping, meal preparation, household care, financial management, driving and using public transportation), social and leisure activities (e.g., meeting with family and friends, attending the cinema, theater and restaurants, participating in religious activities), leisure activities with low physical demands (e.g., watching television, reading, listening to music, painting) and leisure activities with high physical demands (e.g. walking, exercising, swimming, travelling, gardening). During hospitalization, participants were asked whether they participated in each activity before the current hospitalization (no = 0; yes = 1) i.e. baseline participation level. In the follow-up visits after one and three months participants rated the activities as currently doing (same = 1 or less = 0.5) or not doing anymore (= 0). A total “retained activity level” score (0–100) is calculated as a percentage, that is the sum of weighted scores of activities in which a person is currently engaged divided by those they were engaged with before hospitalization. Lower percentages indicate more significant withdrawal from previous participation. The ACS is used worldwide and has been translated, adapted and validated in Israel [41, 42]. The original ACS has a good test–retest reliability (r = 0.90) in community dwelling adults [40]. The ACS has a good construct validity, as demonstrated in studies comparing activity levels in different age groups and populations with neurological disabilities [41, 43].
Predictors
In-hospital characteristics and risk factors were chosen according to the well-documented predictors of acute hospitalization outcomes described in previous studies [4, 44]. In the present study these factors include cognitive status at admission, emotional status, functional status in BADL during hospitalization, Length of Hospital Stay (LOS), severity of the acute illness, and symptoms severity. Baseline characteristics i.e. before current hospitalization include comorbidities and premorbid functional status. Sociodemographic characteristics include age, gender, years of education and living situation. All measurements were administered at admission to the internal medicine ward. Functional status in BADL was assessed also at discharge.
Basic cognitive status was assessed by the Mini Mental State Examination Telephone version (MMSET) [45]. The original Mini-Mental State Examination (MMSE) [46] is a widely used screening tool for cognition in older, community-dwelling, hospitalized, and institutionalized adults. In the current study we used the 22-point telephone version of the MMSE (MMSET). Total scores for the original face-to face MMSE and telephone versions of the MMSE correlate strongly (Pearson’s r = .85, P < .001).
Executive functions were assessed using the Color Trails Test (CTT) [47], a neuropsychological paper and pencil test which measures executive skills, sustained and divided attention. The CTT is a variant of the validated Trail Making Test and was developed to minimize cultural and language bias. It consists of two subtests, for the CTT part 1, the respondent uses a pencil to rapidly connect, in sequence, circles numbered 1 through 25. For the CTT part 2, the respondent rapidly connects numbered circles in sequence, but alternates between pink and yellow colors. Scoring is calculated by measuring completion time (up to 240 s). One of the advantages of the CTT is that it is quick to administer. Standardized T scores are derived from normative data correcting for age and years of education.
Functional cognition was measured by the medication management subtest of the alternative version of the Executive Function Performance Test (aEFPT) [48]. The aEFPT is an addition of the valid EFPT [31], a performance-based assessment which measures executive functions while carrying out daily tasks. The subtest used in the current study assess medication management abilities necessary for independent daily function. The task involves sorting medications into a 7-day pill sorter, using 3 prescription bottles. The patient has to ignore distractors (other bottles) and to follow the specific sorting instructions. The examiner evaluates the individual’s executive functioning on five components: task initiation, organization, sequencing of task components, safety and judgment, and task completion. The EFPT uses a standardized hierarchical cueing system that is scored from 0 (no cue required) to 5 (the participant cannot do the task). The score for each executive function component ranges from 0 to 5, and the total score of the task ranges from 0 to 25, such that a higher score indicates the need for more assistance. The medication management task from the original EFPT has good interrater reliability (.87) [31]. The alternate medication management task from the aEFPT is more complex and discriminates between patients with stroke and healthy controls [48].
Emotional status was assessed using the Hospital Anxiety and Depression Scale (HADS), a brief questionnaire containing 14 items [49]. The HADS was originally designed to detect emotional disturbances in non-psychiatric patients, it measures anxiety and depression on two separate scales (seven items for each scale). The scores range from 0 to 3 points for each question, with a maximum score of 21 for each scale. Depressive symptoms were coded as normal (0-7 points), borderline (8-10 points) and positive symptoms of depression (11 points and above) categories.
BADL were measured by the Modified Barthel Index (mBI) [50]. The mBI consists of individual's subjective assessment of their independence level in 10 items of BADL: feeding, bathing, grooming, dressing, bowel and bladder care, toilet use, ambulation, transfers, and stair climbing. The total score ranges from 0 to 100, with lower scores indicating increased disability and more assistance required. Baseline (premorbid) mBI was assessed on admission retrospectively for 2 weeks before admission and at discharge. A functional decline score was calculated and defined as a decrease of 5 points or more on the mBI total score from baseline pre-morbid status to discharge, which express a loss of independence in one of the BADL domains [51].
Severity of symptoms was assessed at admission using 11 self-rated common symptoms (e.g., dyspnea, cough, pain) [52]. Patients rated the presence and intensity of symptoms experienced during the past 24 h on a 0–10 scale with higher scores indicating greater severity.
The following information was obtained from patient's medical records: LOS, comorbidities and severity of the acute illness. Comorbidities were assessed using the Charlson Comorbidity Index (CCI) [53]. The CCI weights the number and severity of 19 health conditions representing patient chronic health status. Weighted scores for each condition range from 1 to 6, resulting in a maximum score of 33, with higher scores indicating greater comorbidity. Severity of illness was measured using the National Early Warning Score (NEWS) [54], assessing the objective severity of the patient's condition by the degree of deviation from the norm (0 to 3) of seven physiological parameters: respiratory rate, oxygen saturation level, supplemental oxygen use, temperature, heart rate, systolic blood pressure, and conscious state resulting in a maximum score of 21, with higher scores indicating greater severity.
Statistical analysis
Descriptive data were presented as frequencies for categorical variables and mean and standard deviation for continuous variables. A log10 transformation was performed to normalize the left skewed distribution of the participation outcome ACS total retained score. After transformation, distribution became normal and the direction of the scale has changed indicating that higher numbers represent lower participation. Transformation of the mBI data did not correct the non-normal distribution. All other study variables were normally distributed. Independent sample t-test and Mann–Whitney tests for continuous variables and chi-squared tests for categorical variables were used as sensitivity analysis to test for significant differences in study variables between participants that completed the study versus those who withdrew. A bivariate Pearson correlation analysis (Spearman for the mBI) was performed to examine relationships between study variables with participation and between the independent variables. A hierarchical linear regression model was used to examine whether the functional-cognitive screening explains additional variance in participation beyond other cognitive measures while controlling for confounders. The ACS total retained score was used as the dependent variable and independent variables were entered in blocks according to the bivariate correlations. For all analyses significance was accepted at level p ≤ 0.05. All analyses were performed using SPSS version 25 statistical program (SPSS Inc., Chicago, IL).