Risk of unplanned emergency department readmission after an acute-care hospital discharge among geriatric inpatients: Results from the geriatric EDEN cohort study

Abstract

Objectives

The study aims 1) to examine whether items of the brief geriatric assessment (BGA) or their combinations predicted the risk of unplanned emergency department readmission after an acute care hospital discharge among geriatric inpatients, and 2) to determine whether BGA could be used as a prognostic tool for unplanned emergency department readmission.

Methods

A total of 312 older patients (mean age, 84.6 ± 5.4 years; 64.1% female) hospitalized in acute care wards after an emergency department visit were recruited in this observational prospective cohort study and separated into 2 groups based on the occurrence or not of an unplanned emergency department readmission during a 12-month follow-up period after their hospital discharge. A 6-item BGA was performed at emergency department admission before the discharge to acute care wards. Information on incident unplanned emergency department readmission was prospectively collected by phone call and by consulting the hospital registry. Several combinations of items of BGA identifying three levels of risk of unplanned emergency department readmission (i.e., low risk, intermediate risk and high risk) were examined.

Results

The unplanned emergency department readmission was more frequently associated with a temporal disorientation (P=0.004). Area under receiver operating characteristic curves of unplanned emergency department readmission based on BGA items and their combinations ranged from 0.53 to 0.61. The best predictor of unplanned emergency department readmission was the temporal disorientation (hazard ratio>1.65, P<0.035), which defined the high-risk group. Inpatients classified in high-risk group of unplanned emergency department readmission were more frequently readmitted to emergency department than those in intermediate- and low-risk groups (P log Rank <0.004). Prognostic values for unplanned emergency department readmission of items and their combinations were poor with sensitivity below 67%, specificity ranging from 36.4 to 53.7, and positive likelihood ratio below 1.4.

Conclusions

The items of BGA and their combinations were significant risk factors for unplanned emergency department readmission, but their prognostic value was poor.

Introduction

Geriatric patients (i.e., patients aged 75 years and older) comprise a growing proportion of emergency department users (1). For instance, in Europe they account for up to 20% of all emergency department-visits (1, 2). A visit to the emergency department is not a trivial event for a geriatric patient compared to younger patients because of a high prevalence of discharge to acute care wards, an increased risk of prolonged length of hospital stay and unplanned readmission to emergency department (3, 4, 5, 6, 7, 8, 9, 10). This last adverse event is usually considered as a marker of quality of care (e.g., lower rates of emergency department readmissions are associated with better quality of cares) and also as an economic marker for consumption of costly cares (11, 12, 13, 14, 15, 16). Screening of emergency department users at risk of unplanned emergency department readmission is helpful to target early the interventions able to reduce the rate of readmission and related-health expenditures (16).

Previous screening tools have been developed to predict unplanned emergency department readmission (3, 5, 16, 17, 18, 19, 20). The ‘identification of seniors at risk’ tool (ISAR) and the ‘triage risk stratification tool’ (TRST) are two well know and validated instruments in emergency department to screen users aged 65 years and over (3, 5). Unfortunately, both these tools predict with a poor accuracy unplanned emergency department readmissions after a first visit in emergency department (2, 3, 5, 21). Therefore, one main challenge for emergency departments is the development of an efficient screening tool of older emergency department-users at high risk of unplanned emergency department readmission, in order to initiate early specific interventions able to reduce the risk of readmission.

ISAR and TRST scores are based on a simple accumulation of clinical characteristics either self- or hetero-reported that did not take into consideration the potential complex interplay between items (3, 5). For instance, we recently showed that various combinations of the 6-item brief geriatric assessment (BGA), which uses items exploring similar domains as ISAR and TRST tools (i.e., age ≥ 85 years, gender male, polypharmacy, non-use of home services, history of falls, and cognitive decline), significantly predicted the risk of prolonged length of hospital stay (21). More specifically, the risk of prolonged length of hospital stay was stratified into 3 levels (i.e., high, intermediate and low) according to specific combinations of BGA items. Interestingly, geriatric inpatients with prolonged length of hospital stay had similar clinical characteristics comparable to patients at risk of unplanned emergency department readmission such as cognitive impairment, functional decline and poor health conditions (21, 22, 23, 24, 25).

Based on these previous results, we hypothesized that the 6-item BGA stratification of risk of prolonged length of hospital stay could also be used to stratify the risk of unplanned emergency department readmission in three levels (i.e., high, intermediate and low) after acute care hospital discharge among geriatric inpatients. The aims of this prospective longitudinal study were 1) to examine whether items of BGA or their combinations predicted the risk of unplanned emergency department readmission after an acute care hospital discharge among geriatric inpatients, and 2) to determine whether BGA could be used as a prognostic tool for unplanned emergency department readmission.

Methods

Study design

This study was a prospective observational cohort study with a 12-month follow-up period.

Study setting and population

A convenience sample of geriatric patients discharged from the emergency department of a university hospital was recruited between February and April 2011. Inclusion criteria were an unplanned emergency department-visit followed by a discharge to an acute-care ward and age 75years and over. Among the 1251 geriatric emergency department users who visited emergency department during the period of recruitment, 428 (34.2%) were discharged in acute-care wards and were asked to be included in the geriatric «Emergency Department Elderly populatioN» (EDEN) cohort study. A total of 6 patients refused to be part of the study, 18 died during hospitalization, 41 were lost during the 12-month follow-up period, and 51 had missing data. In final, 312 (72.9%) participants were included in the present analysis (Figure 1).

Figure 1.

Flow chart showing the selection of patients included in the analysis

Assessment

The 6-item BGA was performed upon the admission to emergency department (19). It was composed by the 6 following items: age coded as a binary variable (i.e., ≥ or < 85years), gender (i.e., male or female), number of drugs daily taken, history of falls during the past 6 months (i.e., yes or no), the use of formal (i.e., health professional and/or social worker) and/or informal (i.e., family or friend) home services coded as a binary variable (i.e., yes or no) and the presence of temporal disorientation defined as the inability to give the month and/or the year coded as a binary variable (i.e., yes or no). Polypharmacy was defined by a number of drugs taken per day above 4. A fall was defined as unintentionally coming to rest on the ground, floor, or other lower level. Information on falls and formal or informal home services were obtained from patient, or from a close person who lived with the patient. The non-use of home-help services was defined as living alone without using any formal or informal home services. The use of formal home and social services was sought in both participants who lived at home or in an institution (in France, residents of institutions may also use formal or informal care services). The usual place of living (i.e., home-living versus institution-dwelling defined as living in nursing home or in senior housing facilities) was also recorded. The reasons for emergency department admission were separated into acute organ failure (e.g., congestive heart failure, chronic lung disease, chronic kidney disease or cirrhosis), mobility (i.e., gait and/or balance disorders) and neuropsychiatric (i.e., delirium, and/or dementia and/or mental behavioural disorders) disorders, and social-related conditions defined as the absence of symptoms of acute organ failure combined with an acute increase of the use of formal and/or informal home and social services.

Unplanned emergency department readmission follow-up

Unplanned emergency department readmission in the University Hospital was followed-up during 12 months after the acute care hospital discharge of participants. This period of follow-up was chosen because it was the median follow-up period identified in a recent systematic review and meta-analysis published by Ellis et al., which examined the effectiveness of comprehensive geriatric assessment in hospital for older adults admitted as an emergency (26). Information on incident unplanned emergency department readmission was collected by a physician via a phone call using a standardized questionnaire and by consulting the hospital registry. In case of cognitive impairment of patients, information was obtained from a guardian, a nurse or a person who lived with the participant.

Stratification of risk of unplanned emergency department readmission

The risk of unplanned emergency department readmission was estimated using two consecutive strategies. First, a priori combinations of BGA items as previously reported was used to determine 3 levels of risk of unplanned emergency department readmission, without giving information to the physicians in charge of patients (22). The high risk of unplanned emergency department readmission was defined by the combination of cognitive decline + history of falls. The intermediate risk was defined by cognitive decline, or history of falls, or the combination of age≥85years + male gender + polypharmacy + no use of home services. The low risk was defined by the combination of 3 items or less among age≥85years, male gender male, polypharmacy, and no use of home services. Second, we explored all other possible combinations of BGA items that were associated with unplanned emergency department readmission. Combinations were selected using the number of participants and the value of the receiver operating characteristic (ROC) curve. Only combinations involving at least 10 participants were considered. A new classification into three levels of risk of unplanned emergency department readmission was therefore built using areas under ROC separated into three parts (i.e., highest, intermediate and lowest values). The three best combinations for each part were used to define the risk of unplanned emergency department readmission. The high risk of unplanned emergency department readmission was defined by the item temporal disorientation. The intermediate risk was defined by four combinations of items: history of falls + temporal disorientation; age ≥ 85 years + male gender + polypharmacy + no use of home services; age ≥ 85 years + polypharmacy + no use of home services. The low risk of emergency department readmission was defined by three combinations of items: age ≥ 85 years + male gender +polypharmacy; history of falls alone; male gender + polypharmacy + no use of home services.

Statistical analysis

The participants’ baseline characteristics were summarized using means and standard deviations or frequencies and percentages, as appropriate. Participants were separated into 2 groups based on the occurrence or not of unplanned emergency department readmission during the 12-month follow-up period after acute care hospital discharge. Between-group comparisons were performed using Chi-square test. In order to estimate the efficiency of prediction of unplanned emergency department readmission, for each BGA item and their combinations (a priori and newly combinations), sensitivity, specificity, positive (PPV) and negative (NPV) predictive values, likelihood ratios of positive (LR+) and negative (LR-) tests and accuracy (ACC) which is a global measure of the performance in term of prediction (i.e., ACC = (True Positive value + True Negative value) / (True Positive value + True Negative value + False Positive value + False Negative value); highest values being related to a better prediction) were calculated. Univariate and multiple Cox regression models were performed to identify the most significant items or their combinations predicting the one-year unplanned emergency department readmission. In final, the elapsed time to unplanned emergency department readmissions among participants classified at risk based on a priori and newly BGA stratification (i.e., low risk, intermediate risk and high risk) was studied by survival curves computed according to the Kaplan-Meier method. P-values less than 0.05 were considered statistically significant. All statistics were performed using SPSS (version 19.0; SPSS, Inc., Chicago, IL).

Standard Protocol Approvals, Registrations, and Participant Consents

The study was conducted in accordance with the ethical standards set forth in the Helsinki Declaration (1983). All participants recruited in this study provided a verbal informed consent because the study did not change the usual clinical practice. The verbal informed consent was obtained from the patients themselves in the presence of their trusted person, usually a family member, who helped them to make decision. The participant consent was recorded in the digital file of patients. The study protocol and the consent procedure were approved by the Ethical committee of Angers university hospital.

Results

The incidence of unplanned emergency department readmission was 41.7% (n=130) with a 95% confidence interval (95%CI) from 36.2% to 47.1%. Between-group comparisons showed that geriatric patients with unplanned emergency department readmission had more frequently a temporal disorientation (39.2% versus 24.2% with P=0.004) than those who were not readmitted to emergency department (Table 1). There was no significant difference for other baseline characteristics.

Table 1.

Baseline characteristics of geriatric inpatients separated into two groups based on the emergency department readmission during the 12-month follow-up period after acute care hospital discharge (n=312)

Characteristics	Overall (n=312)	Unplanned emergency department readmission		P-Value*
		Yes (n=130)	No (n=182)
Age >85 years, n (%)	161 (51.6)	73 (56.2)	88 (48.4)	0.174
Male gender, n (%)	112 (35.9)	50 (38.5)	62 (34.1)	0.425
Number of drugs daily taken ≥5, n (%)	250 (80.1)	109 (83.8)	141 (77.5)	0.164
History of falls during the past 6 months, n (%)	137(43.9)	65 (50.0)	72 (39.6)	0.067
Temporal disorientation†, n (%)	95 (30.4)	51 (39.2)	44 (24.2)	0.004
No use of home services‡, n (%)	129 (41.3)	47 (36.2)	82 (45.1)	0.115
Living at home, n (%)	224 (71.8)	87 (66.9)	137 (75.3)	0.106
Reasons for emergency department admission, n (%)
Acute organ failure§	207 (66.3)	85 (65.4)	122(67.0)	0.761
Neuropsychiatric disorders#	23 (7.4)	9 (6.9)	14 (7.7)	0.798
Mobility disorders¶	34 (10.9)	14 (10.8)	20 (11.0)	0.951
Social-related conditions**	48 (15.4)	22 (16.9)	26 (14.3)	0.524

P-value significant (i.e., P < 0.05) indicated in bold

^*Between-group comparison based on Chi-square test

^†inability to give the month and/or year coded as a binary variable (i.e., yes or no)

^‡Formal (i.e., health and/or social professional) or informal (i.e., family and/or friends)

^§congestive heart failure, chronic lung disease, chronic kidney disease or cirrhosis

^#delirium, dementia or mental behavioural disorders

^¶gait and/or balance disorders

^**defined as the absence of symptoms of acute organ failure combined with an acute increase of the use of formal and/or informal home and social services leading to an inability to stay in its place for life.

As shown in Table 2, among BGA items the best predictor of unplanned emergency department readmission was the temporal disorientation, this item having the highest value of ROC curve (0.58). The combination ‘history of falls’ plus ‘temporal disorientation’ had the highest value of ROC curve (0.61). Its sensitivity, specificity and NPV were low (<61.0) but the PPV was higher (87.9). The LR+ and LR- were low and ranged from 0.68 and 1.4. New BGA combinations compared to the a priori classification had highest values of ROC curve but the sensitivity and specificity were low under <64.0%. In addition, the highest LR+ calculated at 1.37 was shown for the temporal disorientation alone.

Table 2.

Diagnostic values of 6-item brief geriatric assessment and combinations for emergency department readmission during the 12-month follow-up period after acute care hospital discharge (n=312)

	Area under ROC	Sensitivity (%)	Specificity (%)	PPV (%)	NPV (%)	LR+	LR-	Accuracy (%)
Items of brief geriatric assessment
Age>85 years	0.538	62.25	45.34	51.65	56.15	1.14	0.83	53.53
Male	0.526	60.00	44.64	65.93	38.46	1.08	0.90	54.49
Polypharmacy*	0.530	66.13	43.60	22.53	83.85	1.17	0.78	48.08
No use of home help services†	0.536	55.56	36.43	54.95	37.01	0.87	1.22	47.57
History of falls‡	0.544	62.86	47.45	60.44	50.00	1.20	0.78	56.09
Temporal disorientation§	0.576	63.59	53.68	75.82	39.23	1.37	0.68	60.58
A priori combinations¶ of items of brief geriatric assessment combinations
Low risk
Age ≥85years+male gender+polypharmacy*	0.569	58.91	45.95	89.01	13.08	1.09	0.89	57.37
Age ≥85years+ polypharmacy* + no use of home services†	0.566	59.70	48.98	86.26	18.46	1.17	0.82	58.01
Male gender+ polypharmacy* + no use of home services†	0.578	57.93	39.02	86.26	12.31	0.95	1.08	55.45
Intermediate risk
History of falls‡	0.544	62.86	47.45	60.44	50.00	1.196	0.78	56.09
Temporal disorientation§	0.576	63.59	53.68	75.82	39.23	1.37	0.68	60.58
Age ≥85years + male gender + polypharmacy* + no use of home services†	0.582	58.59	46.67	95.60	5.38	1.01	0.89	58.01
High risk
History of falls‡ + temporal disorientation	0.605	60.38	53.19	87.91	19.23	1.29	0.75	59.29
Newly identified combinations of items of brief geriatric assess-
Low risk
Age > 85 years + male gender + polypharmacy†	0.569	58.91	45.95	89.01	13.08	1.09	0.89	57.37
History of falls§	0.544	62.86	47.45	60.44	50.00	1.196	0.78	56.09
Male gender + polypharmacy† + no use of home services‡	0.578	57.93	39.02	86.26	12.31	0.95	1.08	55.45
Intermediate risk
History of falls§ + temporal disorientation#	0.605	60.38	53.19	87.91	19.23	1.29	0.75	59.29
Age > 85 years + male gender + polypharmacy† + no use of home services	0.582	58.59	46.67	95.60	5.38	1.01	0.89	58.01
Age > 85 years + polypharmacy† + no use of home services‡	0.566	59.70	48.98	86.26	18.46	1.17	0.82	58.01
High risk
Temporal disorientation#	0.576	63.59	53.68	75.82	39.23	1.37	0.68	60.58

ROC: Receiver operating characteristic curve; PPV: Positive predictive value; NPV: Negative predictive value; LR+: Likelihood ratio of positive test; LR-: Likelihood ratio of negative test

^*Defined by a number of drugs taken per day above 4

^†Living alone without using any formal or informal home services and social help

^‡during the past 6 months

^§Inability to give the month and/or year

^¶Combinations described in previous published study (Beauchet et al. J Emerg Med. 2013;45:739-45) and combinations involving at least 10 participants were shown.

Table 3 shows the results of Cox model regression exploring the association between one-year unplanned emergency department readmission (dependent variable) and combinations of the BGA 6 items (independent variable) adjusted or not on the reason for emergency department admission and place of living. While considering a priori 6-item BGA combinations, low risk successfully predicted a low risk of unplanned emergency department readmission (hazard ratio [HR]=0.53 with P=0.001 without adjustment, and HR=0.55 with P=0.002 while adjusting), intermediate risk successfully predicted an increased risk of unplanned emergency department readmission whatever the adjustment used (HR ranged from 1.15 to 1.88 with P<0.04), and high risk successfully predicted the highest risk of unplanned emergency department readmission only while using low risk as the reference value (HR=2.05 with P=0.005 without adjustment, and HR=2.02 with P=0.010 while adjusting on reason for emergency department admission and place of living). Similarly while considering modified 6-item BGA combinations, low risk successfully predicted a low risk of unplanned emergency department readmission (HR=0.54 with P=0.001 and HR=0.56 with P=0.02 without and with adjustment), intermediate risk successfully predicted an increased risk of unplanned emergency department readmission whatever the adjustment (HR ranged from 1.51 to 1.76 with P<0.004), and high risk successfully predicted the highest risk of unplanned emergency department readmission without and with adjustment (HR ranged from 1.66 to 2.17 with P<0.04).

Table 3.

Risk estimates of one-year unplanned emergency department readmission after acute care hospital discharge using Cox regression models (n=312)

	HR [95% CI] (P-Value)
	Model 1	Model 2	Model 3	Model 4
A priori combinations* of items of brief geriatric assessment
Low risk	0.53 [0.36; 0.78] (0.001)	1.00 (Ref)	0.55 [0.37; 0.80] (0.002)	1.00 (Ref)
Intermediate risk	1.15 [1.03; 2.04] (0.035)	1.83 [1.22; 2.28] (0.003)	1.47 [1.04; 2.09] (0.031)	1.88 [1.25; 2.85] (0.003)
High risk	1.45 [0.94; 2.25] (0.094)	2.05 [1.24; 3.41] (0.005)	1.35[0.85; 2.12] (0.203)	2.02 [1.18; 3.46] (0.010)
Newly identified combinations¶ of items of brief geriatric assessment
Low risk	0.54 [0.38; 0.77] (0.001)	1.00(Ref)	0.56 [0.39; 0.80] (0.002)	1.00(Ref)
Intermediate risk	1.51 [1.06; 2.15] (0.021)	1.76 [1.21; 2.56] (0.003)	1.42 [0.99; 2.04] (0.054)	1.68 [1.14; 2.48] (0.008)
High risk	1.66 [1.04; 2.65] (0.034)	2.09 [1.27; 3.45] (0.004)	1.71 [1.05; 2.80] (0.032)	2.17 [1.28; 3.67] (0.004)

HR: Hazard ratio and p-value significant (i.e., <0.005) indicated in bold; CI: confidence interval; Model 1: Unadjusted model; Model 2: Model with low risk of prolonged length of hospital stay used as reference level without adjustment on baseline characteristic; Model 3: Model with low risk of prolonged length of hospital stay used as reference level with adjustment on reason for emergency department admission; Model 4: Model with low risk of prolonged length of hospital stay used as reference level with adjustment on reason for emergency department admission and place of living; *: Combinations described in previous published study (Beauchet et al. J Emerg Med. 2013;45:739-45) with low risk (three items among age >85 years old, gender mal, polypharmacy† and no use of home services‡), intermediate risk (history of falls§ or temporal disorientation# or age >85 years old + male gender + polypharmacy† + no use of home services‡) and high risk (history of falls§ + temporal disorientation#); ¶: Low risk (age > 85 years + male gender + polypharmacy†, or history of falls§, or male gender + polypharmacy† + no use of home services‡), (history of falls§ + temporal disorientation#, or age > 85 years + male gender + polypharmacy† + no use of home services, or age > 85 years + polypharmacy† + no use of home services‡) and high risk (temporal disorientation#); †: Defined by a number of drugs taken per day above 4; ‡: Living alone without using any formal or informal home services and social help; §: during the past 6 months; #: Inability to give the month and/or year

Kaplan-Meier distributions showed that participants separated in three groups based on 6-item BGA combinations, whatever the BGA stratification (i.e., a priori and modified) differed significantly (P<0.004), those classified at intermediate (P<0.003) and high risk (P<0.004) being more frequently readmitted to emergency department than those at low risk (Figure 2). There was no significant difference between those classified in intermediate and in high risk groups (P>0.350).

Figure 2.

Kaplan-Meier estimates of the probability of one-year unplanned emergency department readmission after acute care hospital discharge among geriatric inpatients separated in three groups based on a priori* (a) and newly identified stratification (b) using combinations of the 6 items of the brief geriatric assessment. *: Combinations described in previous published study (Beauchet et al. J Emerg Med. 2013;45:739-45)

Discussion

The results show that despite the significant predictive value for unplanned emergency department readmission of BGA items taken separately or combined, their prognostic value was poor.

The main finding of our study is that BGA items and their combinations were significant risk factors for unplanned emergency department readmission. However, the prognostic value of these combinations for unplanned emergency department readmission was poor. The ROC curve, sensitivity, specificity, NPV, PPV and LR values were low. This finding underscored that significant risk factors for unplanned emergency department readmission are not necessarily good parameters to build an accurate prognostic tool. First, risk factors should exhibit a much stronger association than those shown in our study (i.e., HR below 2) if the aim is to build an accurate prognostic tool for unplanned emergency department readmission (27, 28). Second, simple tools are unlikely to be highly accurate for the prediction of events related to several factors such as emergency department admissions. Better prediction would require far more complex data sets and modeling techniques. Indeed, unplanned emergency department admission may be considered as a manifestation of chaotic behaviour that has been thus far poorly understood by the classical linear methods. These results suggest that the 6-item BGA could be used to stratify elderly populations, albeit weakly, for trial or group-level purposes, but it should not be used alone for individual patient care planning.

The fact that BGA items are indirect or direct markers of frailty status may explain their predictive value for unplanned emergency department readmissions. For instance, cognitive impairment, functional decline and poor health conditions have been associated with emergency department readmission (16, 17, 18, 19, 20, 21, 22, 23, 24, 25). Besides, geriatric syndromes and functional decline constitute the main profiles of older emergency department users and are precisely the items that BGA intends to screen (29). Indeed, increasing age as well as polypharmacy are strongly associated with an accumulation of comorbidities exposing to unplanned hospitalisation (30, 31). Furthermore, history of falls is usually caused by gait and balance disorders, which lead to loss of independence and thus increase the risk of hospital admission among older adults (29). In addition, the temporal disorientation is a sensitive and specific symptom to assess the presence of severe cognitive disorders, either acute (i.e., delirium) and/or chronic (i.e., dementia), which provoke loss of autonomy and expose to hospitalisation (29, 32). In final, among the 6-item BGA, the absence of home-help services is a marker of social isolation (8, 22). A recent meta-analysis confirmed social marker as a major item to predict the risk of unplanned emergency department readmission, especially when there is an accumulation of comorbidities in older adults (9, 33, 34).

Our findings also underscore that the risk did not increase as a simple cumulative way according to the combination of items. Indeed, the higher risk of unplanned emergency department readmission was reported with the association of history of falls in the past 6 months combined with temporal disorientation for a priori combination, and also temporal disorientation alone for newly identified combinations. The fact that this single item was more predictive than associated to a history of falls remains unclear. We suggest that it could highlight an opposite effect of each item decreasing the risk of emergency department readmission. As mentioned above, temporal disorientation is a marker of severe cognitive decline, whatever its nature. Because of older age of studied population, cognitive vulnerability may be reported to Alzheimer disease and/or vascular dementia (35, 36). It has been shown that, in demented older patients, there is an accumulation of comorbidities, comorbidity being defined as “any distinct clinical entity existing or occurring during the evolution of the main disease” (36, 37, 38). The few French data available come mainly from the REAL.FR study (39) where a high prevalence of hypertension (45%), depression (39%) and sensory deficits (23%) has been shown. This accumulation of comorbidities may lead to a frail status defined as an unsteady medico-social state observed at a certain point of an individual’s lifetime (40, 41). Under the action of the pathologies and/or ageing that generate disabilities, this dynamic process occurring from reduced physiological resources results in impaired adaptation to stress, whether of medical, psychological or social origins. The frailty state results in a precarious medico-social balance likely to be disrupted by any stressful, even though benign event. The fluctuation in cognitive or mobility states (i.e., temporary or permanent disabilities) may expose to transitions in the frailty state that may lead to hospitalization (39, 40, 41). The complex interplay of BGA items, which are direct or indirect markers of geriatric syndromes or functional decline, is highlighted by the fact that the best combination of items is only composed by history of falls during the past 6 months and temporal disorientation. One explanation of this result could be that these two items are markers of both loss of independence and autonomy leading to an inability to stay in usual place of living due to not a single medical event but rather to an inappropriate environment (i.e., social event). This interaction between medical and social vulnerabilities therefore exposes patients to a major risk of hospitalization (9).

Limitations

Some limitations of this study need to be considered. First, participants were included from a single centre and, thus, they were probably not representative of all older adults admitted to emergency department. Second, although we were able to control for many characteristics likely to modify the risk of rehospitalization, residual potential confounders might still be present. We limited this confounding bias by adjusting our results on organ failure and place of living. Third, there was a potential recall bias about the history of falls, which is well known in the elderly. Falls are usually underreported because of a cognitive decline of fallers who forget to report the falls, and depends on the occurrence of fall-related adverse health outcomes (42). Fourth, only readmission to emergency department of one University hospital was considered. Because this hospital was the referent hospital of the area where all the participants lived, there is a low probability that participants were hospitalized in another hospital during the following year. An indirect point is the fact that the proportion of unplanned emergency department readmission was higher in our studied cohort compared to most previous data published on European populations (2).

Conclusions

Our results show that although the combinations of items of 6-item BGA are significantly associated with a risk of unplanned emergency department readmission among geriatric inpatients after their acute care hospital discharge, their prognostic value was poor and similar to ISAR and TRST tools. Furthermore, this risk depended on items combinations without necessarily an increased risk with the accumulation of items. We suggest that the BGA could be used to stratify populations, albeit weakly, for trial or group-level purposes, but it should not be used alone for individual patient care planning. Further work to predict outcomes in prospectively-followed, large populations should consider using far more complex data sets and non-linear modeling techniques.

Conflict of interest and disclosures: The authors have declared that no competing interests exist. The study was supported by Biomathics, Paris, France. The sponsors had no role in the design and conduct of the study, in the collection, management, analysis, and interpretation of the data, or in the preparation, review, or approval of the manuscript.

Funding: None.

Acknowledgments: The authors would like to thank the patients for their participation in the study.

Author contributions: Conceived and designed the experiments: CPL and OB. Performed the experiments: CPL, LDD and AK. Analyzed the data: CPL, LDD, CA, AK, and OB. Contributed reagents/materials/analysis tools: CPL, CA, AK, and OB. Wrote the paper: CPL, LDD, CA, AK, and OB.

Ethical Standards: The study protocol and the consent procedure were approved by the Ethical committee of Angers university hospital.