The aim of this study was to investigate a) rates and patterns of potentially inappropriate and avoidable ED visits by nursing home patients, b) ED resource use arising from such visits, and c) variables that affect their occurrence. Overall, we found that, while only 6% of ED visits were potentially inappropriate, 29% were potentially avoidable—assessed with ACSCs. The three highest ranked of these were COPD/pneumonia (26.3%), urinary tract infections (22.5%), and congestive heart failure (13.4%).
As might be expected, potentially inappropriate ED visits received significantly lower triage classifications; and visits initiated by ambulance calls were rarely potentially inappropriate. When compared to admissions by ambulance, patients who walked in were more likely to be there inappropriately (OR 3.42, 95%CI 1.79-6.55) as were those referred by their GPs (OR 2.13, 95%CI 1.16-3.90). The fact that referrals by GPs had higher odds of being inappropriate than ambulance calls seems counterintuitive: GPs are in a position to adequately assess and triage a situation in a nursing home. We assume that inappropriate GP referrals were more likely in situations where GPs had to rely on the assessments of nursing home care staff without being able to examine the patients personally. On the other hand, potentially avoidable visits were more often admitted as ambulance arrivals than as walk-in patients (OR for walk-ins 0.46, 95%CI 0.27-0.77), but more often still by GPs (OR 1.40, 95%CI 1.00-1.94).
As might also be expected, patients delivered by ambulance for potentially avoidable issues also had higher odds of being hospitalized than their walk-in counterparts. Compared to appropriate ED visits, potentially inappropriate ones led to significantly lower resource use; however, we found no overall difference between potentially avoidable and non-avoidable ED visits regarding ED resource use.
Compared to international findings, a 6% incidence of potentially inappropriate visits is rather low. E.g., a similar French study indicated 18.1% [15], an Australian one 13% [17], and a US one 36% [32]. Part of this comparative lowness might result from our use of an explicit evaluation protocol. In our sample, e.g., procedures that cannot be performed in nursing homes, e.g., radiological examinations, adjustment of indwelling urethral catheters, etc., accounted for more than 10% of all nursing home residents’ ED visits. Although visits for these procedures were not rated as potentially inappropriate, some might have been manageable in ambulatory care settings. As we did not have the information available to confirm such points, we treated them as appropriate. Similarly, some suspected fractures are manageable by GPs (during office hours) without reducing the resident’s quality of care. However, since in Switzerland not all GPs have the capacity for radiological examinations, only suspected fractures that did not require radiological examination were considered potentially inappropriate. Thus, while the low incidence of potentially inappropriate visits might accurately reflect the situation in nursing homes, it is also possible that we simply underestimated the number due to a lack of information.
One important point is that an appropriate visit could also be avoidable: whereas a visit’s appropriacy depends on the patient’s condition at the time of the visit, avoidability depends on possibilities for early detection and treatment of deteriorating symptoms. For example, following a diagnosis for congestive heart failure, appropriate medication and monitoring will normally prevent acute exacerbations that require ED visits. Compared to rates of avoidable visits in literature, which vary from 4% to 55% [8], our 29% incidence of avoidable visits was moderate.
Our distribution of ACSC frequencies was similar to those reported elsewhere. Our most common ACSC, pneumonia/COPD, accounted for 26.2% of our sample’s potentially avoidable visits. In comparison, this classification accounted for 30.1% of Gruneir and colleagues’ potentially avoidable visits [5] and 30.2% of Hsieh and colleagues’ [10]; where urinary tract infections accounted for 22.5%, Gruneir et al. found 20.3% [5] and Hsieh et al. 25.2% [10]. And where congestive heart failure accounted for 13.4% of our potentially avoidable visits, it represented 15.9% of Gruneir et al’s [5]. Similarly, our finding that ACSCs were related to higher hospitalization rates confirms the findings of several other studies [5, 10, 33]. In such cases, Gruneir et al. [5] suspected that the ACSCs’ deterioration had made hospitalization necessary.
Regardless of visits’ appropriateness or the presence of ACSCs, we found high levels of multimorbidity: our sample’s median CCI score was 6. This finding corresponds with other data from Switzerland, where nursing home residents have higher mean levels of multimorbidity than their community-dwelling peers [4]. Roughly 85.5% have at least two separate diagnoses; 22.8% have five or more chronic diagnoses [4]. Such high levels of comorbidity contribute to polypharmacy, which leads to increases in healthcare utilization, hospitalization rates [34] and ED overcrowding. Therefore, our group’s high comorbidity prevalence underscores the importance of interventions that address nursing home residents in holistic and coordinated ways, with due attention to overall chronic care management [35].
While the ED resource consumption of our sample’s potentially inappropriately admitted patients was significantly lower than for those who were appropriately admitted, they still used a total of 49’295 tax points (corresponding to approx. the same amount of US$) over the 3-year period assessed. Given the extra costs and limitations of emergency services entail regarding individualized care, every inappropriately-admitted patient could almost certainly have been treated both more cost-effectively and more safely elsewhere. Further, compared to the appropriately admitted group, potentially inappropriately admitted patients consumed smaller mean amounts of nurse work, laboratory and radiology resources; however, mean physician-patient time and medical reporting time did not differ between the two groups. I.e., inappropriate ED visits use resources that would clearly be better invested elsewhere.
Reviewing the reports for potentially inappropriate visits, we found that patients were often transferred for check-ups or examinations, while no acute event was listed in the patient’s history. Possible causes for such omissions include the lack of timely access to the patient’s treating physician [36] or a geriatric opinion [15], a failure to accurately assess the situation’s acuity, or simply the residents’ or their relatives’ preference to visit the ED. In a French study, Rolland et al. corroborated this possibility, observing that residents with access to both specialist medical advice and a mobile emergency medical unit were less likely to experience potentially inappropriate ED transfers [12].
Similarly, outreach teams or local interdisciplinary teams have been shown to reduce ED visits [37]. Given that 9% of our population’s visits were walk-in arrivals, another way of decreasing such ED visits might be to provide emergency walk-in doctors’ offices. Other patients in our sample whose visits were rated as potentially inappropriate were recommended admission for critical conditions, but opted to be transferred back to their nursing homes, as they wanted to abstain from invasive interventions. This indicates that some of the admissions might have been avoided by Advanced Care Planning (ACP), i.e., the assessment of and continuous conversation about residents’ and their families’ wishes for treatment in acute situations. ACP has been shown to be effective in reducing hospital admissions [38].
Unlike nursing home residents whose visits are potentially inappropriate, those transferred to the ED for potentially avoidable causes, especially poorly-managed ACSCs, incur the same mean resource costs as other appropriate visits (e.g., physician work, radiology resources), and more of others (physician-patient time, admin and medical report time, nurse work, laboratory resources). Therefore, over the 3-year study period, ACSC patients’ overall resource use was much higher than for inappropriate admissions. This corroborates the need to improve chronic care management in nursing homes as a strategy to reduce avoidable ED referrals. The current gap in management could be bridged via improved access to primary care providers—either GPs or nurse practitioners; or, within each nursing home, a structured medical system could be implemented to provide both on-site expertise and quick access to it.
Nursing home residents with regular opportunities to engage with their primary care providers are less likely to have ED visits [39]. To test and, if possible, optimize this connection, an ongoing US initiative is placing geriatric-specialized nurse practitioners in nursing homes. By supporting early recognition and initiating interventions to stabilize worsening conditions, these practitioners have reduced all-cause hospitalizations by up to 30% [40]. Especially in nursing homes with a majority of low-skilled care workers, this initiative shows the value of geriatric expertise (or other forms of geriatric-specialized services), strong advance care planning, and palliative care [41].
The present investigation is limited by several factors. Firstly, no classification system for potentially inappropriate hospitalizations has yet been used broadly to allow for international comparison. For the sample described, we distinguished appropriate from potentially inppropriate visits based on the development of a protocol by a clinical review panel in Australia [17]. Other appropriateness evaluation protocols have been used elsewhere [15, 16], which might increase the heterogeneity between our and earlier researchers’ results; and our use of a single instrument may bias the current analysis. However, when we were starting our data analysis, Finn and colleagues’ instrument was designed specifically for ED visits (as opposed to hospitalizations) [17], so we deemed it the most suitable yet published. Within the protocol's recommended procedure, one critical step was to determine whether it would have been possible to perform an observation or procedure in the nursing home (Supplement 1). Although we predefined scenarios for such observations or procedures, their validity remains unsure: we could not obtain data about the relevant resources’ availability at the nursing homes in question, e.g., their access to primary care and possibilities for risk assessments. To reduce the risk of incorrect classification, we had two trained raters classify the visits independently.
As for relying on the presence of ACSCs as avoidability indicators, the raters were evaluating only diagnoses. I.e., while other factors—e.g., nursing homes’ in-house treatment options, the availability of appropriately-qualified staff, the presence of a GP or how closely their policies adhered to their residents’ wishes—may have been influential to other decisions, the raters did not consider them. Furthermore, we had no access to ICD-10 diagnoses, which are not used in EDs. Instead, we had text versions of the involved physicians’ diagnoses. For future researchers, this limits the comparability of this study’s diagnoses to those obtained elsewhere.
An additional limitation was that we had access only to ED data: we had no information about nursing home staffs’ reasons for initiating transfers, the challenges or needs affecting them or the resources accessible to them at the time of the transfer. Therefore, we can give no definitive suggestions to improve the link between nursing homes and the medical care centers in the ED’s catchment area. Moreover, in the ED, data on our sample patients’ end-of-life planning were not systematically obtained. This hindered our assessment of problems regarding ACP or do-not-hospitalize orders. The study also has a risk of selection bias due to exclusion of patients who objected to the use of their data in health research; and as this is a retrospective chart review, it is prone to documentation error. Finally, some categories of the predictor variables contain relatively few cases, leading to large confidence intervals for certain variables (e.g., ‘No connective tissue disease’).