Prognosis tools for short-term adverse events in older emergency department users: result of a Québec observational prospective cohort

Overcrowding and delays in providing care have become increasingly problematic in emergency departments (EDs) in high-income countries [1, 2]. The limited access to primary care in Canada, in part related to the lack of primary care physicians, force people in the community to seek care in EDs ([1, 2]. A continuously enlarging proportion of ED users are aged over 75 and account for up to a quarter of all ED users. These ED users, compared to their younger counterparts, are at higher risk for short-term ED adverse events including prolonged length of stay in ED and in hospital, as well as hospital admission during an index ED visit ([2], http://​www.​csbe.​gouv.​qc.​ca/​fileadmin/​www/​2016/​Urgences/​CSBE_​Rapport_​Urgences_​2016.​pdf). A major concern of ED healthcare workers is the identification of older patients who are the most vulnerable and are at highest risk of adverse outcomes (e.g. long ED and hospital stay) and treating them in a timely manner [1‐4]. The occurrence of adverse events may be reduced if the older ED users’ risk is determined early in the process of ED care (i.e., upon their ED arrival) [1, 3, 4]. One way for preventing short-term ED adverse events is to screen older users with the highest risk and provide timely and individualized geriatric interventions [5‐9]. Therefore, the prognostic tools designed for this purpose may provide relevant information regarding potential adverse events early in the ED care plan, with the goal of improving the decision-making process and to reduce adverse events [10‐15]. The identification of older patients at risk for short-term ED adverse events is, therefore, an important objective with regards to optimizing the care provided in the EDs. The “Program of Research on the Integration of Services for the Maintenance of Autonomy” (PRISMA-7) is a tool used to assess older ED users in the province of Québec (Canada) [16, 17]. PRISMA-7 was designed to screen disabilities in older community-dwellers and it stratifies them into two risk levels: low versus high-risk for disabilities. Disabilities are the result of chronic diseases burden, among which musculoskeletal and cardiovascular diseases are the main contributors [18]. Regardless the reasons of the ED visit, disabilities and related burden of chronic diseases characterize older ED users and this poorer overall health status may explain in part the high prevalence of the adverse events, suggesting that PRISMA-7 may be used as a prognosis tool to predict short-term ED adverse events [1‐4, 16, 17]. To date, the association of the PRISMA-7 high-risk level with the occurrence of short-term ED adverse events has never been examined and therefore validated as a prognosis tool for this purpose. “Emergency room evaluation and recommendations” (ER²) is another clinical tool currently used in Québec EDs [4, 19]. Compared to PRISMA-7, ER² has been particularly designed for assessing the risk for short-term ED adverse events in older patients [12‐14]. The objective of the present study is to determine which clinical tool, between PRISMA-7 and ER² is the most appropriate tool to use in EDs for detecting the older ED users most at risk for short-term ED adverse events. No previous study has compared PRISMA-7 and ER² high-risk levels with regards to predicting the occurrence of the short-term ED adverse events. Firstly, we hypothesised that both PRISMA-7 and ER² high-risk levels would be associated with the occurrence of short-term ED adverse events. Disabilities of older ED users and related chronic diseases burden may explain in part the occurrence of the short-term ED adverse events [1‐4]. PRISMA-7 screens disabilities, suggesting that it may be used as a prognostic tool of these ED adverse events [16, 17]. Secondly, we hypothesized that ER² would be a significantly better prognostic tool compared to PRISMA-7 for predicting prolonged length of stay in ED and in hospital, as well as predicting hospital admission, because it was designed and validated for this specific goal unlike PRISMA-7 [4, 12‐14, 19].

The study aimed to: 1) examine whether the PRISMA-7 and ER² high-risk levels were associated with length of stays in ED and in hospital, as well as hospital admission, an 2) compare the performance criteria (i.e., sensitivity, specificity, positive predictive value [PPV], negative predictive value [NPV], likelihood ratios [LR] and area under receiver operating characteristic curve) of the PRISMA-7 and ER2 high-risk levels for the three previously stated ED adverse events in Québec older patients visiting ED on a stretcher.

Table 1 shows the comparison of baseline patient’s characteristics separated into different subgroups based on their PRISMA-7 (i.e., low versus high) and ER² risk levels (i.e., low, moderate and high). Patients with a PRISMA-7 high-risk level compared to those with low-risk level were older (P ≤ 0.001) and more frequently male (P ≤ 0.001); lived more often at home (P = 0.001) with support (P = 0.009), and had more frequently polypharmacy (P ≤ 0.001), temporal disorientation (P = 0.001) and walking aid (P ≤ 0.001). They had also a longer length of stay in ED (P = 0.002) and in hospital (P = 0.004), and were more frequently admitted to hospital (P = 0.003) compared to those with a PRISMA-7 low-risk level. Similar results for age, sex, home support, polypharmacy, temporal disorientation and use of walking aid were observed for comparisons between the three ER² risk levels. More serious presenting conditions and higher incidence of adverse events were found in high-risk levels compared to moderate-risk and low-risk levels, as well as between moderate-risk and low-risk levels (P ≤ 0.003). There was no significant different for place of living (P = 0.784) between mild and moderate-risk levels, as well as no significant difference for sex (P = 0.453) between mild and high-risk levels, and no significant difference for home support (P = 0.651) and polypharmacy (P = 0.430) between moderate and high-risk levels. Furthermore, neuropsychiatric disorders as a reason for ED visit was more prevalent in patients with low-risk levels compared to those in moderate-risk group (P = 0.022), but less prevalent compared to high-risk group (P = 0.036). Prevalence of this reason of ED visit was also higher in high-risk group of patients compared to those in moderate-risk group (P = 0.005). Organ failure was also a more prevalent reason of ED visit in patients with a moderate compared to a high-risk level (P = 0.025). In addition, longer stay in ED and hospital as well as hospital admission were observed in patients considered to have high-risk level compared to those with low-risk level (P ≤ 0.001). Hospital admission was also more prevalent in patients with the moderate-risk level compared to those with the low-risk level (P = 0.040).

Regardless of the regression model used, the PRISMA-7 high-risk level was associated with a prolonged length of ED (P ≤ 0.001) and hospital stay (P ≤ 0.007) (Table 2). The PRISMA-7 low-risk level was associated with a short length of stay in ED and in hospital for linear regression model (P ≤ 0.007). The ER² high-risk level was associated with a prolonged length of ED and hospital stay, as well as elevated hospital admission rates (P ≤ 0.001) for all regression models. The ER² low-risk level was associated with short length of stay in ED and in hospital, as well as a lower hospital admission rate (P ≤ 0.001).

Kaplan-Meier distributions showed significant differences in the time elapsed before hospital discharge between patients with a PRISMA-7 low and high-risk level; those with a high-risk level were discharged from ED (P = 0.010) and from hospital (P = 0.002) later compared to those with low-risk (Fig. 1). The three ER² participants’s risk levels differed significantly for length of stay in ED and in hospital (P ≤ 0.001), those with a high-risk level being discharged later than those in the low-risk (P ≤ 0.001) (Fig. 2). Furthermore, patients with a high-risk level were discharged later when hospitalized than those in the moderate-risk level (P = 0.026).

All performance criteria were poor (i.e., sensitivity, specificity, PPV and NPV ≤0.74 LR ≤ 1.40 and area under receiver operating characteristic curve ≤0.62) with worse performances for PRISMA-7 compared to ER² (Table 3).

The findings showed that PRISMA-7 and ER² risk stratifications were successfully associated with length of stay in ED and in hospital, as well as hospital admission rates. High-risk levels were associated with longer stays and higher rates of hospital admission, whereas the contrary association was found in low-risk levels for both tools. The strength of association was stronger for ER² compared to PRISMA-7. However, performance criteria for the high-risk level of these ED adverse events were low, with the lowest performance being reported with PRISMA-7.

PRISMA-7 and ER² high-risk levels were both associated with prolonged ED and hospital stays and higher rates of hospital admission. This association was expected for ER² because this clinical tool has especially been designed to determine prognosis for these short-term ED adverse events [4, 12‐14, 19]. In particular, prolonged ED and hospital stays have previously been reported with the ER² high-risk levels, similar to the findings in the present study [4, 12, 14, 19]. In contrast, to the best of our knowledge, it is the first time that an association with prolonged ED and hospital stays is found with the PRISMA-7 high-risk level. PRISMA-7’s initial objective was to screen disability in community dwelling older adults [16, 17]. This tool has also been used to identify frailty in older ED users, but not related adverse events [21]. Frailty may be defined as a state of vulnerability related to homeostasis following a stressor and is independently associated with adverse events [22]. Frailty results from an accumulation of severe comorbid conditions and physiological aging [22, 23]. A tool which assesses frailty and its severity indirectly evaluates prognosis for frailty-related adverse events. The main characteristic of older ED users compared to younger users is an accumulation of severe and chronic morbidities and related-disabilities  [3]. These particular health and functional conditions greatly influence the ED care plan [2, 3, 8]. Although older adults undergo more diagnostic tests and procedures than younger ED patients, their ED diagnoses tend to be less accurate [2, 7]. This has been attributed to atypical disease presentation due to multiple comorbidities, which can complicate the clinical presentation, diagnosis and care plan [1, 3, 5, 7]. Additionally, chronic morbidities may decompensate in a cascade and lead to disabilities and complications, regardless of the nature and the severity of the acute condition leading to the ED visit  [3, 22]. This statement explains why the PRISMA-7 could be used as a prognostic tool to predict prolonged ED and hospital stays. Our findings also underscore that PRISMA-7 and ER² low-risk levels were associated with shorter ED and hospital stays. This finding is consistent with the former because it suggests that both tools also detect older ER users with a low level of frailty. The early identification of this category of older ED users may be useful for the ED care continuum, because these ED users with low levels of frailty have a higher propensity to be discharged to their place of living compared to those with higher stages of frailty [5‐9, 24].

Our study revealed that, even if significant associations of PRISMA-7 and ER² risk levels with short-term ED adverse events were found, the predictive performance was poor for both tools. No study has assessed the ability of PRISMA-7 or ER² of predicting short-term ED adverse events, therefore, the comparison of this result with previous studies is difficult. However, our findings demonstrated similar results to other screening tools, such as The Identification of Seniors at Risk (ISAR), the Triage Risk Screening Tool (TRST), the eight-item questionnaire of Runciman, and the seven-item questionnaire of Rowland, if we take into consideration the ability of predicting return visits of older patients after they have been discharged from the ED. [25] Poor predictive performances were also reported in this study. The imbalance between a significant association of the PRISMA-7 and ER² high-risk levels with short-term ED adverse events and the poor performance criteria highlight that a significant statistical association between a high-risk level and an adverse event does not necessarily guarantee that the prognostic tool is accurate nor reliable. Similar results have been reported with prolonged hospital stays in older patients admitted to geriatric wards [4]. Despite the limited predictive performance, improvement in cost-effectiveness of targeted geriatric interventions using a screening tool has been previously demonstrated with low positive predictive value (i.e., between 20 and 30%) [24]. In addition, screening tools are known to be particularly useful in decision-making procedures to prevent events that depend on multiple factors [25, 26]. This result confirms that screening older patients with the highest risk level and providing them a timely and appropriate intervention is an efficient strategy.

The important number of older ED users recruited and the appropriate design to achieve the research objectives are the strengths of our study. However, some limitations need to be underscored. Firstly, the recruitment has been performed in a single centre, which limits the external validation of our results. Secondly, although adjustment for clinical characteristics influencing the studied short-term adverse events was performed, there are still potential residual confounders that may influence the short-term ED adverse events. Thirdly, both PRISMA-7 and ER² do not take into consideration the reasons of ED visits which is usually an acute disease directly influencing the occurrence of short-term ED adverse events. For instance, an acute disease may decompensate in a cascade related to chronic morbidities cumulated in older ED users, leading to multiple acute organ decompensations and disabilities and exposing this group to prolonged length of stays and hospitalisations [5‐9]. However, in the regression analysis used to examine the association of between PRISMA-7 and ER² risk levels with the three short-terms ED adverse events, all models were adjusted by the reasons of ED visits and a significant association was found.

In conclusion, even if PRISMA-7 and ER² high-risk levels were associated with incidental short-term ED adverse events, their performance criteria for these risks were relatively low, suggesting that they cannot be used as prognostic tools for this purpose. However, they may be the starting point of an assessment that could provide pertinent information pertaining to ED tailor-made geriatric interventions.