Authors: Adrian D. Haimovich (1Department of Emergency Medicine, Beth Israel Deaconess Medical Center & Harvard Medical School, Boston, MA), Sydney Mulqueen (1Department of Emergency Medicine, Beth Israel Deaconess Medical Center & Harvard Medical School, Boston, MA), Jossie Carreras-Tartak (1Department of Emergency Medicine, Beth Israel Deaconess Medical Center & Harvard Medical School, Boston, MA), Cameron Gettel (2Department of Emergency Medicine, Yale School of Medicine, New Haven, CT), Mara A. Schonberg (3Department of Internal Medicine, Beth Israel Deaconess Medical Center & Harvard Medical School, Boston, MA), Susan N. Hastings (4ADAPT Center of Innovation, Durham VA Health Care System, Durham, North Carolina; 5Departments of Medicine and Population Health Sciences, Duke University School of Medicine, Durham, North Carolina; 6Center for the Study of Aging and Human Development, Duke University, Durham, North Carolina; 7Geriatrics Research Education and Clinical Center, Durham VA Health Care System, Durham, North Carolina), Christopher Carpenter (8Department of Emergency Medicine, Mayo Clinic, Rochester, MN), Shan W. Liu (9Department of Emergency Medicine, Massachusetts General Hospital and Harvard Medical School, Boston, MA), Stephen H. Thomas (1Department of Emergency Medicine, Beth Israel Deaconess Medical Center & Harvard Medical School, Boston, MA; 10Blizard Institute, Barts & The London School of Medicine, London, UK)
Categories: Article
Source: Academic emergency medicine : official journal of the Society for Academic Emergency Medicine
Doi: 10.1111/acem.15013
Authors: Adrian D. Haimovich, Sydney Mulqueen, Jossie Carreras-Tartak, Cameron Gettel, Mara A. Schonberg, Susan N. Hastings, Christopher Carpenter, Shan W. Liu, Stephen H. Thomas
Older adults are at high risk of adverse health outcomes in the post-emergency department (ED) discharge period. Prior work has shown that discharged older adults have variable understanding of their discharge instructions which may contribute to these outcomes. To identify discharge comprehension gaps amenable to future interventions, we utilize meta-analysis to determine patient comprehension across five domains of discharge diagnosis, medications, self-care, routine follow-up, and return precautions.
Using Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines, two-reviewers sourced evidence from databases including Medline (PubMed), EMBASE, Web of Science, CINAHL, and Google Scholar (for gray literature). Publications or pre-prints appearing before April 2024 were included if they focused on geriatric ED discharge instructions and reported a proportion of patients with comprehension of at least one of five predefined discharge components. Meta-analysis of eligible studies for each component was executed using random-effects modeling to describe the proportion of geriatric ED cases understanding the discharge instructions; where appropriate we calculated pooled estimates, reported as percentages with 95% confidence interval (CI).
Of initial records returned (N = 2,898), exclusions based on title or abstract assessment left 51 studies for full-text review; of these, seven constituted the study set. Acceptable heterogeneity and absence of indication of publication bias supported pooled estimates for proportions comprehending instructions on medications (41%, 95% CI 31–50%, I^2^ 43%), self-care (81%, 95% CI 76–85%, I^2^ 43%), and routine follow-up (76%, 95% CI 72–79%, I^2^ 25%). Key findings included marked heterogeneity with respect to comprehending two discharge diagnosis (I^2^ 73%) and return precautions (I^2^ 95%).
Older patients discharged from the ED had greater comprehension of self-care and follow-up instructions than about their medications. These findings suggest that medication instructions may be a priority domain for future interventions.
Older adults discharged from the emergency department (ED) are at high risk for near-term adverse outcomes. In 2021, there were over 27 million ED visits by adults aged 65 and older,^1^ representing over 19% of total ED volume and, historically, over 60% of these visits resulted in discharge.^2^ In the 30 days post-discharge, 17.4% of older adults will have another ED visit, 9.4% be admitted to the hospital, and 1.4% die.^3^ The importance of ED discharge transitions of care has been highlighted by multiple professional societies including the Society for Academic Emergency Medicine and the American College of Emergency Physicians and is included as an area of emphasis in the Geriatric Emergency Department Guidelines.^4^
A previous systematic review of the ED general population identified a broad range – from 8% to 94% – of patients with correct recall of discharge instructions, but few studies have focused on older adults.^5^ Prior work has revealed multi-dimensional challenges older patients face with respect to their discharge instructions. First, older adults report feeling that the ED discharge process is abrupt with inadequate explanation of the cause of presenting symptoms or the results of performed testing.^6^ Second, many older adults do not follow their discharge instructions, with particularly limited adherence to medication instructions and poor recollection of return precautions, “red flag” symptoms suggesting a need to return to the ED.^7^ Third, cognitive impairment and sensory deficits become more common in older adults with increasing dependence on concurrent communication with care partners throughout an episode of emergency care which is often neglected.^8^
The overarching goal of our work is to improve ED discharge transitions of care for older adults by addressing challenges with discharge instruction comprehension. Discharge instructions, which are most often a combination of verbal and written information, are often grouped into five domains including diagnosis (which may include details of the ED workup and treatment), medications, self-care, routine follow-up, and return precautions.^5^ Prior studies of discharge instruction comprehension, however are from single sites and focus on varied areas of discharge comprehension, limiting comparison and generalizability. To address this gap and to identify the discharge instruction domain at greatest need for intervention, we performed a meta-analysis of prior literature reporting older ED patient comprehension. We then used these results to perform a power calculation in preparation for a future intervention in this space.
We conducted this systematic review following the framework of the Preferred Reporting Items for Systematic reviews and Meta-analyses (PRISMA).^9,10^ As this study is not human subjects research, it does not require IRB review.
The patient, intervention, control, and outcome (PICO, see Table S1–2) characterization for this review drove determination of study eligibility. Since the purpose of our project was to provide point estimates to inform power calculations for future geriatric ED studies, this analysis did not include a comparator. Although there was known potential for clinical heterogeneity in definitions of patients, interventions, and outcomes, these terms were defined broadly to optimize both study inclusion and generalizability.
We included studies of ED patients discharged home who were “geriatric” or “older” without specific age cutoff for the terms. We included studies of “routine discharge instructions” which may include verbal, written, or both forms of discharge instructions as is common clinical practice. To assess discharge instruction comprehension in routine clinical practice, we included studies of novel interventions (e.g., of dedicated and specially trained discharge coordinators) if the study outlined discharge comprehension in a control group of routine discharge processes. In such a case, only the control population receiving routine care were used in this analysis. Further details on individual study methodologies are provided in the Supplement’s Risk Of Bias In Non-randomised Studies (ROBINS) tables (Stage 2). As patient transitions between EDs and long-term care facilities are subject to specific process guidelines, they were excluded from our study.^11^
Given that our aim was to determine baseline discharge instruction across multiple domains, we defined our outcome as patient or surrogate care partner comprehension of their ED discharge instructions in each of the five discharge parameters of diagnosis, medications, self-care, routine follow-up, and return precautions. The metric was the proportion who exhibited comprehension (Supplementary Information, ROBINS Tables). Studies were not excluded based on their approach to defining comprehension, but these differences were tabulated as a source of potential confounding. Recognizing that cognitive impairment might introduce significant variability to study outcomes, our primary analysis is of cognitively intact older adults or cognitively impaired older adults with caregivers present with a planned sensitivity analysis of all older adults regardless of cognitive status.
The basic search strategy followed the Cochrane Group^12^ recommendations for broad, multi-database literature identification. We used the Harvard On-Line Library Information System (HOLLIS) which leverages multiple databases including PubMed, Excerpta Medica (EMBASE), Web of Science, Cochrane Trials, Cumulative Index to Nursing and Allied Health Literature (CINAHL), and Google Scholar, among others. The broad search strategy included the keyword terms (anywhere in records) linked by AND: geriatric, emergency department, discharge, instructions, and (comprehension OR understanding). We searched records published from January 1900 through March 2024 (search 28 March 2024)
We did not use any automated search filters (e.g. language, study design). The HOLLIS system largely eliminates duplicate record returns.
Articles were downloaded into reference manager software (EndNote version 21, Clarivate, www.clarivate.com). No reference-manager automated processes were used for records review.
Initial title and abstract screening was performed by one author (ST), with 10% of studies screened for eligibility (with full agreement) by a second author (AH). The remaining records’ full-text review, as well as subsequent data extraction where applicable, was executed independently by all review authors.
Examples of studies of potential relevance which were screened out included qualitative studies (e.g. perceptions of barriers to transitions of care^6^) and thematic analyses.^13^
Study characteristics (e.g. geography, inclusion/exclusion criteria) were tabulated, and bias was assessed using the framework of the Risk of Bias in Non-randomized Studies (ROBINS), which is the standard tool recommended by the Cochrane group for tabulation and evaluation of bias in non-randomized studies.^14^ The ROBINS-I (intervention) table was used. All study evaluations were independent. After the initial independent reviews, discussion among authors resolved initial disagreements about study interpretation or classification (e.g. ROBINS-I tabulation).
Study data were entered into statistical software (Stata version 18MP, Stata Corp, College Station, TX) used for all analysis and plotting. For all analyses, significance was defined at the p of <.05 level and confidence intervals (CIs) were calculated at the 95% level. Failure of 95% CIs to overlap was predefined as an indicator of statistical significance at the p <.05 level. We did not establish an a priori cutoff for a “clinically important” absolute (or relative) percentage difference in estimated proportions of patients with comprehension.
We anticipated a low study N, i.e., <10. The Sidik-Jonkman approach was thus selected as a random-effects model well-suited to low-N analyses.^15^ Pre-planned sensitivity analysis included use of omitted-study meta-analysis. In case of a larger sample size of studies (N = 10) we also planned to perform advanced calculations such as formal funnel-plot testing (Egger) to evaluate potential publication bias, meta-regression, or generation of prediction intervals.^15^
With regard to heterogeneity, we report Q but emphasize I^2^ as an overall measure that could be compared across different discharge parameters, which were assessed in differing numbers of studies. We used Cochrane I^2^ 30–60% was defined as a level that “may represent moderate heterogeneity.”^15^
As a principle reason for the meta-analysis was to provide an estimated baseline proportion of comprehension for each discharge parameter, we generated pooled proportion estimates for all discharge parameters regardless of heterogeneity findings. By a priori plan, we highlighted as having concerning heterogeneity and thus suspect reliability, any estimates in which the parameter’s I^2^ exceeded 60%. As a secondary analysis, we planned a pooled analysis of the percent of participants who understand the entirety of their discharge instructions.
To explore sources of bias (e.g. publication bias, outlier studies), we employed non-parametric trim-fill analysis with imputation. The trim-fill approach was based on the Sidik-Jonkman random-effects modeling used for the main meta-analysis. Small-study and outlier biases were additionally assessed using Galbraith plotting.
We perform a power analysis in STATA using the following parameters using a two-sample alpha = 0.05, beta = 0.20, attrition = 10%.
Our search identified 2,898 records. After removing duplicates not removed by HOLLIS (e.g. an abstract and a report of the same study), title and abstract screening yielded a set of 51 records requiring full-text review. As depicted in the PRISMA diagram (Figure 1), full-text review yielded N = 7 studies^7,16–21^ providing eligible data for at least one of the five discharge parameters of interest.
Of the seven publications, one^20^ was a PhD dissertation and the others appeared in journals in the fields of EM^7,16,17,21^ or healthcare quality and safety.^18,19^ Five studies^7,17–20^ were from the USA, with one each from Canada^16^ and the Netherlands.^21^ Studies’ age minima defining older patients ranged from 60 to 75 years. Table 1 provides general summary information regarding studies providing data for this analysis.
Figure 2 shows which studies contributed data to each of the five discharge instruction domains. Marked differences were identified in the estimates for proportions of patients’ comprehension of discharge instructions across domains. Patients had significantly better comprehension of self-care (pooled estimate (95% confidence interval); 81% (76–85%)) and follow-up instructions (76% (72–79%)), than of medication instructions (41% (31–50%)). Forest plotting focused on portrayal of findings for each of the five discharge parameters. These plots, shown in Figure 2, indicate heterogeneity (I^2^) as well as pooled estimated proportion.
The overall estimated proportion for diagnosis (Figure 2A) and return precautions (Figure 2E) discharge parameters should be interpreted as preliminary (high I^2^), but the forest plots suggest different etiologies of heterogeneity. It is noted that the Figure 2 forest plot depicts relatively low overall heterogeneity – 13% absolute variation in various studies’ point estimates. This can inflate I^2^, which reports the proportion of overall heterogeneity attributable to inter-study variation. In the forest plot for return precautions comprehension, the studies’ estimates were markedly different and the associated high I^2^ finding is more clearly indicative of substantial difference in studies’ estimates (Figure 2E). We note that the range of return precaution comprehension was between 35–71%.
The Supplementary Information provides detailed information for bias evaluation and outlier assessment. For each of the discharge parameters, the omitted-study, funnel, and Galbraith plots are presented. Omitted-study plots did not suggest high leverage, and funnel plots suggested potential small-study bias primarily for the return precaution comprehension. Galbraith plotting showed studies within the 95% CI of the plot, except for the return precaution comprehension. The overall result was that the return precaution parameter was characterized by very high heterogeneity and potential bias, whereas the other discharge parameters’ pooled estimates were likely relatively reliable.
Our planned sensitivity analysis including older adults regardless of cognitive status was relevant only to Han et al., and did not meaningfully change relevant findings to understanding of diagnosis, follow-up instructions, and return precautions (Supplementary Information 3).^17^
Power calculations using the findings of our meta-analysis show that detection of a 15% absolute improvement in medication, follow-up, and self-care instruction comprehension requires 192, 105, and 77 subjects per arm (Supplementary Information 4). We were unable to calculate a pooled estimate of overall discharge instruction comprehension as the reported studies did not report this figure across their study cohorts.
This meta-analysis found that older adult discharge instruction comprehension varied by domain. Medication and return precaution instructions were less likely understood than diagnosis, self-care, or follow-up instructions. We found that even where heterogeneity decreased reliability of pooled effect estimates, the confidence intervals of individual studies showed clear parameter-associated differences in discharge instruction comprehension.
Through an emphasis on older adults, this work adds to prior ED literature on discharge comprehension. Studies in the general ED adult population have suggested that home care and return-to-ED were the areas with greatest comprehension problems; follow-up, medications, and diagnosis were less frequently characterized by knowledge gaps.^22,23^ The results of our analysis suggest knowledge of and adherence to medication instructions may be a significant challenge for older adults, a challenge likely to be exacerbated by high rates of polypharmacy.^24^ This problem has been the focus of recent efforts to improve geriatric ED patient discharge prescriptions.^25^
The etiologies of limited discharge comprehension likely vary.^18^ Healthcare providers overestimate health literacy and the related concept of numeracy, and ED studies have shown these patient characteristics are not always easily measured with desired precision.^26,27^ For older adults, it may be that cognitive decline is a risk factor for low health literacy.^28^ Older adults may face other barriers to discharge comprehension including hearing loss.^29^ Concerningly, prior work suggests that, in the general ED population, lack of comprehension is often accompanied by patient unawareness that they do not understand their instructions.^30^
Efforts to improve discharge comprehension and, subsequently, discharge outcomes, have had varied success. A 2020 systematic review of ED discharge comprehension by Hoek and colleagues^5^ differentiated the method of discharge instruction (verbal, written, video, telephone follow-up), noting that the addition of written to verbal discharge instruction improved recall from 47% to 57%. A more recent randomized trial of a care transition intervention for older adults comprised of a post-discharge home visit and phone call did not reduce ED revisits, but did improve identification of return precaution symptoms,^23^ while a randomized trial of post-ED telephonic follow-up improved neither ED revisits nor medication compliance.^31^ Outside of emergency medicine, a 2021 systematic review and meta-analysis of discharges from inpatient wards found that communication interventions at discharge significantly reduced the risk of readmission.^32^ This literature is supported by a 2021 consensus statement from the Geriatric Emergency care Applied Research (GEAR) Network concluding that “no care transition intervention addressing social needs in older adult populations consistently reduced subsequent health care utilization or other patient-centered outcomes.”^33^ The GEAR Network also noted barriers to effective communication included discordant messaging between ED staff and outpatient healthcare teams, as well as inadequate support for non-English speaking patients.^8,34^ Taken together, these studies highlight the challenges of improving post-ED discharge outcomes for older adults.
New measures are needed to both evaluate the quality of discharge instructions provided to older adults (or care partners when patient cognitive impairment is present) in a high-throughput, systematic manner, but also to enable more robust comparison of interventions. For example, a multivariate scoring system that incorporates aging-centric factors like reading level, completeness of clinician-provided instructions, usability,^35^ and preferred language concordance would enable targeted quality improvement efforts in this space. Such a measure would benefit from clinician- and patient- stakeholder input and leverage advances in natural language processing to enable broad implementation.
Our analysis includes multiple limitations. We sought to be broadly inclusive in terms of our study inclusion criteria, which may have led to a more heterogeneous study selection. Moreover, we defined five discharge comprehension domains, while other studies have approached these differently.
For instance, the four domains of discharge instructions were defined in one study^19^: diagnosis, self-care instructions, expected duration of symptoms or illness, and return precautions. Particularly since we found that comprehension of discharge instructions varied across domains, the results we calculated should be considered applicable only to the parameters we studied. While noting evidence for correlations between comprehension and patient outcomes,^19^ we did not consider patient outcomes directly. We purposefully restricted assessment to “routine discharge processes” to elucidate a baseline from which to design interventions. Finally, we did not specifically consider differential impacts among at-risk populations including minoritized patients, individuals with low English proficiency, or those with cognitive impairment.
The limited efficacy of prior interventions combined with the persistent vulnerability of this patient population highlight the need for further efforts to improve discharge comprehension and post-discharge outcomes of older adults. The results of this meta-analysis may be useful for power calculations of large-scale studies aimed at this intersection. While acknowledging the non-definitive link between better geriatric instructions comprehension and patient-centered outcomes, comprehension is a necessary – if not sufficient – step in the process of effective transition of care to home.