Assessment tools and incidence of hospital-associated disability in older adults: a rapid systematic review

Registration

The protocol of this rapid systematic review registered at OSF registries with the following identifier: DOI 10.17605/OSF.IO/9JEZ4. Portions of this text were previously published as part of a preprint (https://www.medrxiv.org/content/10.1101/2022.09.22.22279726v1.full).

Study design

A rapid systematic review was performed according to the recommendations of the Cochrane Rapid Reviews Methods Group (Garritty et al., 2021). Reporting was conducted in accordance with the PRISMA statement (Page et al., 2021).

Search strategy and selection criteria

A literature search was performed in Medline (via Ovid), EMBASE, and Cochrane Central Register of Controlled Trials (CENTRAL) databases on 26th August 2021. All authors were involved in the literature search. The search strategy comprised five search terms related to: (i) study setting (i.e., hospital), (ii) observed disability in ADL, (iii) incidence and prevalence, sensitivity to change and responsiveness, (iv) population identification (i.e., older adults), and (v) articles that cover the aspect of disability acquired in hospitals. The search terms, combined with the Boolean operator “AND”, were applied to titles and abstracts, and MeSH terms were added when available and relevant. The full search strategy is shown in the additional material (see Table S1). The review included prospective and retrospective cohort studies. The control group of randomized controlled trials (RCTs) was eligible when performing a usual or a sham intervention. Inclusion criteria were: studies investigating a general older population (≥65 years) who were admitted to hospital for an acute disorder; studies had to assess the individual items of the ADL measurement tool before hospitalization (retrospectively or prospectively) and at the end of hospitalization or after hospital discharge. Exclusion criteria were: studies investigating a specific condition that could have an effect on functional decline (e.g., stroke, brain injury, heart failure, COVID-19, and acute respiratory failure); and studies that primarily examined a population with cognitive impairment.

One reviewer (GK) independently screened all the records based on titles and abstracts (phase 1) and the full-text of the eligible studies (phase 2). A second reviewer (SKM) screened 20% of the same records. If Cohen’s kappa coefficient of agreement was >0.80 for both phases, only 20% of the studies were planned to be screened independently by two reviewers. Disagreements were resolved through discussion.

Data extraction

Study characteristics (authors, country, study sample size, population age, type of ward, proportion of women, proportion living alone, proportion living in a nursing home, Acute Physiology and Chronic Health Evaluation (APACHE II) score, Charlson Comorbidity Index, number of days of hospitalization, any scale of mental status) were extracted by one reviewer (GK), while incidence data were extracted by three reviewers (HR, SKM, GK) at the same time. Disagreements were resolved through discussion between all three reviewers. The following information was extracted: item type, response options, criteria for the response options, baseline, and discharge assessment (i.e., who performed it and how), baseline and discharge prevalence of ADL dependency, operationalization, and definition of HAD and incidence per item and for the overall score.

Methodological quality assessment

Methodological quality assessment was performed using the critical appraisal checklist for Studies Reporting Prevalence Data (Munn et al., 2015) from the Joanna Briggs Institute (JBI). The JBI checklist was completed independently by two reviewers (SKM, GK). Differences in rating were discussed and resolved. We considered the methodological quality and the risk of bias of a study as low if all questions were rated ‘yes’, some concern if one of the questions was rated ‘unclear’ and high if one of the questions was rated ‘no’.

Incidence calculation of hospital-associated disability

The incidence of HAD (total score) was pooled using the statistical software R (R Core Team, 2022) and its package meta (Schwarzer, 2007). A random effects model was applied based on an inverse variance model with a logit transformation.

Heterogeneity was assessed by the I², which is the proportion of total variability due to between-study heterogeneity (Deeks, Higgins & Altman, updated February 2022), to estimate inter-study variability. Tau was estimated with the DerSimonian-Laird estimator (Schwarzer, 2007). Migliavaca et al. (2022) recommended avoiding the use of an arbitrary cut-off for the I² statistics, as it may not be discriminative in incidence studies. Therefore, heterogeneity was explained by discussing the level of dependency at baseline and the method of assessment.

The data required to calculate the incidence of HAD are the number of patients who are dependent and independent at baseline and the evolution of these groups at discharge. If these values were given, the following formula was used to calculate the incidence of the individual ADL task or set of tasks (total score): ${\displaystyle HADincidence=\frac{a}{b}}$

where a and b represents the number of newly dependent patients (requiring the help of someone else), and the number of patients at risk of developing or increasing dependency respectively. For example, in the study by Inouye et al. (1993), 51 patients were newly dependent at discharge according to the overall score and 188 patients had the potential to decline in one of the ADL tasks. ${\displaystyle \text{HAD incidence}=\frac{51}{188}=0.27.}$

If these values were not reported, an unbiased estimate of the incidence of HAD in that study could not be calculated. In this case, an estimate of the incidence was made by subtracting the percentage of dependency at discharge minus the percentage of dependency at baseline.

Tools used for assessment of activities of daily living in hospitalized older patients

ADL assessment tools are made up of several items each evaluating individual tasks (e.g., bathing, dressing, transferring, etc.) which together form a tool. The Barthel or Katz Index is an example of a tool composed of several tasks. The literature also mentions task sets that are not defined by the original authors as an index or tool. Other authors propose a set of individual tasks without creating an index.

The ADL tools assess whether patients are dependent or independent. For example, the Barthel Index scores the individual tasks from 0 points (dependent) to independent with a maximum of 15 points for transferring, 10 points for climbing stairs and 10 points for using the toilet.

All ADL tools or sets of tasks used in the assessment of hospitalized older patients were reported narratively.

Outcome results

The pooled incidence of HAD for the overall score included two studies reporting the Katz Index of ADLs 1963 (Covinsky et al., 2003; Inouye et al., 1993) and two others used the Katz Index of ADLs 1970 (Mudge, O’Rourke & Denaro, 2010; Sager et al., 1996) for a total of 4,020 patients. Figure 2 presents the pooled incidence of HAD (total score) of 37% (95% CI [0.30–0.43]). Heterogeneity was substantial at 90%.

The calculated and estimated incidences of HAD are categorized per ADL task and set of tasks (see Table S2). Calculated and estimated incidences of HAD categorized per ADL task and set of tasks). Adequate data for the incidence calculation at the item level was reported only by one study (Covinsky et al., 2000). These authors separated the patient reported and the proxy reported dependency in ADL. The incidence of HAD for the individual items was respectively 32% and 70% for bathing, 27% and 66% for dressing, 27% and 70% for toileting, 30% and 61% for eating, 27% and 59% for transferring. For the other studies, it was not possible to calculate the incidence of HAD (either total or single-item level) due to insufficient data. One RCT could not be integrated into Table S2 because the authors reported only the mean change in ADL score, and the results of this study were described narratively (Martinez-Velilla et al., 2021). The control group worsened in all items of ADL, negative values indicating a decline in the Barthel mean change score. The three items with the highest mean change score from baseline to discharge were: transferring (−2.06 points; 95% CI −1.44 to −2.71), climbing stairs (−1.36 points; 95% CI −0.84 to −1.91), and toileting (−1.23 points; 95% CI −0.76 to −1.69).

Ten studies were included in the identification of the tools or functional tasks used to assess ADL in hospitalized older patients (Covinsky et al., 2000; Covinsky et al., 2003; Dharmarajan et al., 2020; Hirsch et al., 1990; Inouye et al., 1993; Martinez-Velilla et al., 2021; Mudge, O’Rourke & Denaro, 2010; Park et al., 2021; Sager et al., 1996; Zelada, Salinas & Baztan, 2009). The review identified four assessment tools, two sets of tasks, and individual items added to original tools. The identified tools were: two references of the Katz Index (1970, 1963), the Barthel Index (Mahoney & Barthel, 1965), the nine-item Care Needs Assessment tool (Hirsch et al., 1990) and 7 ADL tasks form the Frailty Index (Rockwood & Mitnitski, 2011). The first set of tasks is composed of 2 activities coming from the Nagi (1976) and the Rosow & Breslau scale (1966) which were integrated in the Frailty Index (Rockwood & Mitnitski, 2011). The second sets of tasks was proposed by Gill (2014) (bathing, dressing, transferring and walking across the room). Walking was assessed in one study (Mudge, O’Rourke & Denaro, 2010), while one other study (Inouye et al., 1993) assessed the task of grooming by referring to the Katz Index of ADLs. However, the item walking and grooming were not in the original version of the Katz Index, and were considered separately.

Methodological quality

Figure 3 presents a summary plot of the assessment of methodological quality using the JBI checklist. All the included studies and study sub-groups reported an adequate sample frame and recruitment procedure. Sixty percent of the included study sub-groups reported an inadequate study sample size. All studies and sub-groups described in detail the study subjects and setting. Adequate analysis with sufficient coverage of the identified sample was conducted in 20% of studies sub-groups. All study sub-groups were unclear regarding the methods used for identification of the condition. Sixty percent of study sub-groups were rated unclear for the reliability of the measurement of participant’s condition. The statistical analysis was inadequate in approximately 86% of sub-groups. The response rate of the subgroups was unclear or inadequate in 86%.

Overall, the Covinsky et al. (2000) study showed some concern for methodological quality. All the other studies showed a high risk of bias. The rating of the individual items for each study is available in the additional material (see File S3).

Implications and further research

Further studies should investigate the reasons for the overestimation of HAD. As reported above, there is a need to develop a more sensitive tool that reflects the true functional status of older patients (over 65 years) before hospitalization for acute care. The systematic integration of proxies in the evaluation of functional status before hospitalization, in addition to the patient self-reported assessment, needs to be deepened.

This study highlighted the fact that there is a lack of systematic reporting of data in order to assess the incidence of HAD in older patients aged over 65 years. This means that future cohort or intervention studies will have to report in detail the trajectory of such patients between before hospitalization, on admission and on discharge. Indeed, when retrospectively assessing patients’ ADL at home, it is necessary to detail the number of people being dependent for each item of ADL. On admission, it is important to describe which of the independent and dependent patients have improved, worsened, or maintained their status prior to hospitalization. At discharge, it is essential to describe which of the independent and dependent patients on admission have improved, worsened, or maintained their status compared with admission. Only in this way will it be possible to determine the number of people with a new decline and to determine its incidence. This review also raises public awareness of the need for careful interpretation of study data.

HAD is a relevant problem, and a systematic appraisal of existing intervention studies addressing this problem is missing. Future research should consider interview methods to help patients better remember their abilities at home in order to reflect their true ability in ADL function.