Presented at the American College of Emergency Physicians meeting, San Francisco, CA, 2011; and at the 5ème congress of Société Française de Médecine d'Urgence (SFMU), Paris, 2011.
Original Research Contribution
Unscheduled Return Visits to the Emergency Department: Consequences for Triage
Version of Record online: 13 JAN 2013
© 2013 by the Society for Academic Emergency Medicine
Academic Emergency Medicine
Volume 20, Issue 1, pages 33–39, January 2013
How to Cite
Sauvin, G., Freund, Y., Saïdi, K., Riou, B., Hausfater, P. (2013), Unscheduled Return Visits to the Emergency Department: Consequences for Triage. Academic Emergency Medicine, 20: 33–39. doi: 10.1111/acem.12052
The authors have no relevant financial information or potential conflicts of interest to disclose.
- Issue online: 13 JAN 2013
- Version of Record online: 13 JAN 2013
- Manuscript Accepted: 31 JUL 2012
- Manuscript Revised: 30 JUL 2012
- Manuscript Revised: 24 MAY 2012
- Manuscript Received: 3 FEB 2012
Vol. 20, Issue 3, 327, Version of Record online: 20 MAR 2013
Original Research Contribution: Correction: Unscheduled Return Visits to the Emergency Department: Consequences for Triage
Vol. 20, Issue 3, E3–E9, Version of Record online: 20 MAR 2013
The objective was to conduct a survey of unscheduled revisits (URs) to the emergency department (ED) within 8 days of a prior visit, to test the hypothesis that patients making these URs are disproportionately likely to suffer short-term mortality or manifest a need for any admission to the hospital (adverse events [AEs]) at the time of the UR, compared to patients triaged at the same level who did not have an unscheduled ED revisit within 8 days.
This was a 1-year retrospective study of patients with an UR to the ED of an urban, 1,600-bed tertiary care center and teaching hospital. The criteria for inclusion as an UR were: 1) making an emergency visit to our adult ED during 2008, without being admitted to our hospital nor being transferred to another hospital; and 2) subsequently making an UR to the same ED within 8 days following the first one. Patients who were contacted by members of our staff and specifically asked to make return visits to our ED (such as those who returned for wound care follow-up visits), and those who made more than five visits to our ED during 2008, were excluded. AEs were defined as death or hospitalization within 8 days of the second visit.
During 2008, there were 946 patients with URs (2% of patients treated and released after the first ED visit), and 931 were analyzed (n = 15 missing values). Associated with the second visit, an AE was noted for 276 (30%) patients. Eight variables were significantly associated with AE: age ≥ 65 years, previously diagnosed cancer, previously diagnosed cardiac disease, previously diagnosed psychiatric disease, presence of a relative at the time of the UR, arrival with a letter from a general practitioner at the time of the UR, a higher level of severity assigned at triage for the UR than for the first ED visit, and having had blood sample analysis performed during the first visit. The median triage score for the UR was not significantly different from that group's median triage score for the first ED visit, whereas the proportion of admissions to the hospital (29%) or to the intensive care unit (ICU; 2%) was greater overall in the UR group than in the patients making their first ED visit.
The authors observed that 2% of patients had an UR. This UR population was at greater risk of AE at the time of their URs compared to their initial visits, but the median triage nurse score was not significantly different between the first visit and the UR. This suggests that the triage score should be systematically upgraded for UR patients.
Las Reconsultas no Esperadas al Servicio de Urgencias: Consecuencias para el Triaje
Realizar un estudio de reconsultas no esperadas (RN) al servicio de urgencias (SU) en los primeros ocho días tras una primera consulta, para comprobar la hipótesis que los pacientes que hacen estas RN tienen mayor probabilidad de sufrir mortalidad a corto plazo o de necesitar ingreso en el hospital (evento adverso, EA) en el momento de la RN, en comparación con los pacientes clasificados en el mismo nivel de triaje que no han tenido una RN en el SU los ocho días siguientes.
Estudio retrospectivo de pacientes con una RN a un único SU urbano de un hospital terciario y universitario de 1600 camas durante un año. Los criterios de inclusión como RN en el estudio fueron: 1) realizar una consulta al SU para adultos durante 2008, sin ser ingresado en el propio hospital o trasladado a otro hospital y 2) realizar posteriormente una RN al mismo SU en los primeros ocho días siguientes a la primera visita. Fueron excluidos aquellos pacientes con los que los urgenciólogos se pusieron en contacto y a los que se les pidió específicamente que realizaran una reconsulta al SU (por ejemplo aquéllos que volvieron para el de seguimiento de una herida) y aquellos otros que hicieron más de cinco consultas al SU durante el 2008. EA se definió como muerte u hospitalización en los primeros 8 días tras la segunda consulta.
Durante 2008 hubo 946 pacientes con RN (un 2% de los pacientes tratados y dados de alta tras la primera consulta al SU) y se analizaron 931 (15 pacientes con datos perdidos). En relación con la segunda consulta, se documentó un EA en 276 (30%) pacientes. Ocho variables se asociaron significativamente con el EA: la edad ≥ 65 años, el diagnóstico previo de cáncer, la cardiopatía, la enfermedad psiquiátrica, la presencia de un familiar en el momento de la RN, el acudir con un informe del médico de atención primaria en el momento de la RN, un nivel de gravedad asignado por el triaje para la RN mayor que para la primera consulta al SU y el tener realizado un análisis de sangre durante la primera consulta. La mediana de puntuación del triaje para la RN no fue significativamente diferente a la mediana de puntuación del triaje de los grupos para la primera consulta al SU, mientras que el porcentaje de ingresos en el hospital (29%) o a la unidad de cuidados intensivos (2%) fue globalmente mayor para en el grupo de la RN que en los pacientes que hicieron su primera consulta al SU.
Se documentó que el 2% de los pacientes tenían RN. Esta población de RN tuvo un riesgo mayor de EA en el momento de su RN en comparación con su consulta inicial, pero la mediana de la puntuación del triaje enfermero no fue significativamente distinta entre la primera visita y la RN. Esto sugiere que la puntuación del triaje debería ser sistemáticamente incrementada para los pacientes con RN.
Short-term unscheduled revisits (URs) to the emergency department (ED) occur in 2% to 5% of patients.[1, 2] A UR differentiates a patient as being different from frequent users of the ED and from patients making EDs visit after recently being discharged from the hospital.[4, 5] The principal causes of URs to the ED are patient-related factors, illness-related factors, and medical errors.[1, 6, 7] Patients with UR suffer increased mortality. Therefore, URs remain an important concern for ED management, and URs are a key element of routine ED audits. However, no prior study has indicated how the knowledge that a patient is making an UR to the ED might influence the management of the patient during the UR visit. It is possible that UR patients are at increased risk not only of mortality, but also of other adverse events (AEs). Patients making URs may be distinguishably different from patients making their first ED visits.
Due to various factors, such as unpredictable arrivals of patients with high medical acuity, overcrowding, and other causes, many ED patients cannot be immediately evaluated by emergency physicians (EPs). Therefore, most EDs implement a triage process. Triage is usually performed by a triage nurse, who makes a brief assessment of the patient's apparent needs and priority of care, resulting in the patient being placed in a queue. Most countries have adopted a five-level triage and acuity scale,[9-12] and these triage tools have been shown to estimate a patient's degree of urgency with satisfactory interobserver reliability.[13-15] However, to date the presence of a UR is not a variable considered when nurses assign these triage scores.
We conducted a 1-year survey of URs to the ED at one hospital and tested the hypothesis that, stratified by triage score, UR patients are at greater risk of AEs than patients making a first ED visit. If the data support this hypothesis, it would be logical to suggest that UR patients should be triaged at a greater level of acuity than is usual by current triage scales and/or that a special alert should be sent to the EP to specifically notify him or her that the patient whom they are treating is making a UR, to increase the probability of appropriate medical decision-making by that physician.
We performed a single-center retrospective cohort study of ED patients. This study was authorized by the Conseil National Informatique et Libertés (Paris, France) and our institutional review board (Comité de Protection des Personnes se Prêtant à la Recherche Biomédicale, Paris, France). Because the protocol was observational, authorization to waive written informed consent was granted.
Study Setting and Population
The setting was an ED of a 1,600-bed tertiary care center and teaching hospital and covered 1 year (2008). The criteria for inclusion in the study were an initial emergency visit to our adult ED during 2008 without admission to our hospital nor transfer to another hospital, followed by a second (UR) ED visit within 8 days after the first. Patients who were asked to revisit our ED (e.g., scheduled posttraumatic medical consultation and nurse consultation for dressing changes) and those who were frequent users of the ED (more than five visits in 2008) were excluded from the UR cohort for statistical analysis.
Data were entered in a computerized database and inconsistency between data was systematically checked and solved. An AE associated with URs to the ED was defined as any patient death or hospital inpatient admission within 0 to 8 days of that UR. When an AE occurred, the complete chart was examined by two physician experts (GS, PH) who decided if the AE appeared to be due to a missed diagnosis at the first visit, a complication of a previously diagnosed disease, or an iatrogenic complication. The expert panel also adjudicated whether the same diagnosis was appropriate for the UR visit as that for the first visit, whether the UR was due to a different etiology than the reason for the first ED visit, or whether the patient left the ED during the first visit before the medical evaluation was complete, thus leaving without an initial diagnosis at the time of the first ED visit. When a disagreement between the two physician experts was observed, a third physician (BR) adjudicated the incident to resolve the disagreement.
Each patient visiting our ED had immediate triage scoring by a nurse trained for that assessment. Two-day triage scoring training was performed by the Paris University hospital network (Assistance Publique-Hôpitaux de Paris). The five-level scoring system employed (red, orange, green, dark blue, light blue) suggests the maximum acceptable delay between a patient's arrival and examination by a physician (from immediate to 4 hours; Table 1). Elements of the triage score include physiologic variables (arterial blood pressure, heart rate, pulse oximetry, respiratory rate, Glasgow Coma Scale score, and pain measured with either a visual analog scale or a numeric scale), the chief complaint of the patient, and information (comorbidities, worsening of outpatient condition, request for admission) communicated by the general practitioner's referral letter (when available). By analyzing the 2008 database of all patients visiting our ED, the probability of hospital admission, admission into intensive care unit (ICU), or death in the ED was determined for each triage scoring level (Table 1). In this manner, it became possible to calculate the expected percentage of patients having deaths in the ED, admissions to the hospital, or admissions into an ICU at the time of the UR, if UR patients were similar to initial ED visit patients for each level of triage scoring. We then compared the observed versus expected rates of death in ED, hospital admission, and ICU admission rates for the UR cohort, for each triage level. In patients with missing nurse triage scores, the lowest probabilities (0.025 for admission and 0.001 for death) were used. An increase of nurse triage severity assessment for a patient was defined by the presence of a higher acuity level of triage for the UR than occurred for the initial visit.
|Acuity Level||Coded Color||Maximum Expected Delay to Physician Examination||Prevalence,%||Proportion of Admission to the Hospital,%||Proportion of Admission Into ICU or Death in the ED,%|
|4||Dark Blue||<2 hr||25||8||0.1|
|5||Light Blue||<4 hr||8||2||0.1|
Parametric data are expressed as mean ± standard deviation (SD). Ordinal data are expressed as median and interquartile range (IQR). Nominal data are described by proportions. Comparison between first ED visit and UR ED visit was performed using the Student's t-test, the Mann-Whitney U-test, and Fisher's exact test, as appropriate. To avoid overfitting the predictive model developed, a conservative approach, using only variables with p-values of ≤0.10 from univariate analysis were employed for the predictive model. If the Pearson correlation coefficient between variables was 0.60 or greater, only the variables judged by the expert panel to be clinically relevant were entered into the multivariate model. Inclusion in the final model was determined by a backward stepwise process. The discrimination of the final model was assessed using the c-statistic, and its calibration was performed using the Le Cessie statistic. In addition, to validate this model internally, we performed a 10-fold cross-validation, which is recognized as one of the most efficient methods available. The internal validation of the model was expressed as the difference (optimism) between the c-statistic observed between the complete population and in the cross-validation samples. Agreement between experts was assessed using the kappa score. All statistical tests were two-tailed, and a p-value of less than 0.05 was required to reject the null hypothesis. Statistical analysis was performed using NCSS 2001 software (Statistical Solutions Ltd, Cork, Ireland).
During 2008, a total of 52,704 patient visits were recorded to our ED. A UR ED visit occurred for 946 patients (2.2% of nonadmitted patients). Of these, 15 were excluded due to missing data (vital parameters, final diagnosis, or empty medical file). Thus, 931 patients constituted the study sample (Figure 1). The mean (±SD) age was 46 (±19) years, 16% were older than 65 years, 64% were men, and 36% were women. At the first visit, the mean (±SD) systolic arterial pressure was 131 (± 19) mmHg, the mean (±SD) diastolic arterial pressure was 80 (±14) mm Hg, the mean (±SD) heart rate was 88 (±17) beats/min, the mean (±SD) temperature was 36.7°C (±0.6°C), and the median (IQR) pulse oximetry was 98% (97% to 99%). Self-presentation at the ED was noted for 616 (66%) patients and transportation by ambulance for 315 (34%) patients. A relative accompanied the UR patients in 267 (29%) cases. A letter from a general practitioner was present in 149 (16%) patients at the first visit and 80 (9%) at the UR. The most frequent chronic diseases in the UR population were psychiatric diseases (17%), alcohol abuse (16%), hypertension (14%), neurologic diseases (10%), cardiac diseases (9%), cancer (8%), and pulmonary diseases (7%); 109 (12%) patients were homeless. During the first visit, blood sample analyses were performed in 328 (35%) patients and radiologic imaging in 322 (35%) patients. The distribution of the delay between the first and the second visit is shown in Figure 2.
Associated with the UR visit, an AE was noted for 276 (30%). Admission to a conventional ward took place in 258 (28%) patients, admission into an ICU occurred for 18 (2%) patients, and no deaths were recorded within 8 days after a UR. Table 2 shows that several variables were associated with AEs. To identify independent predictor variables, we performed multivariate analysis using the stepwise backward logistic regression. Eight variables remained significantly associated with AEs in the multivariate model (Table 3): age ≥ 65 years, prior diagnosis of cancer, prior diagnosis of cardiac disease, prior diagnosis of psychiatric disease, arrival with a relative present at the time of the UR, presence of a letter regarding the initial visit with the patient from a general practitioner at the UR, a more acute triage severity assessment by the nurse during the UR than during the first visit, and blood sample analysis having been performed at the first visit. Calibration (Le Cessie statistics Z = 1.60, p = 0.11) and discrimination (c-statistics 0.727 [IQR = 0.708 to 0746], p < 0.001) of the model were appropriate. The observed difference between c-statistic in the entire derivation population (0.727) and that obtained after 10-fold cross-validation (0.705), i.e., optimism, was 0.022, indicating that our model was robust.
|Variables||No AE (n = 655)||AE (n = 276)||p-value|
|Age (years), mean ± SD||43 ± 17||52 ± 20||<0.001|
|Age ≥ 65 years, n (%)||76 (12)||76 (27)||<0.001|
|Men, n (%)||439 (67)||158 (57)||0.03|
|Women, n (%)||216 (33)||118 (43)|
|Relatives present, n (%)||156 (24)||111 (40)||<0.001|
|Homeless, n (%)||87 (13)||22 (8)||0.03|
|Letter from a general practitioner at the first visit, n (%)||79 (12)||70 (25)||<0.001|
|Transportation to the hospital, n (%)|
|Ambulance||211 (32)||104 (38)||0.11|
|Self-presentation||443 (68)||173 (62)|
|Pre-existing diseases, n (%)|
|Neurologic disease||59 (9)||33 (12)||0.19|
|Psychiatric disorder||95 (15)||55 (20)||0.04|
|Hypertension||74 (11)||53 (19)||0.002|
|Cardiac diseasea||35 (5)||45 (16)||<0.001|
|Alcohol abuse||118 (18)||31 (11)||0.01|
|Diabetes||30 (5)||30 (11)||<0.001|
|Cancer||33 (5)||38 (14)||<0.001|
|Respiratory disease||33 (5)||30 (11)||0.002|
|Physiologic variables at the first visit|
|Core temperature (°C), mean ±SD||36.7 ± 0.5||36.7 ± 0.6||0.63|
|Systolic arterial pressure (mm Hg), mean ± sd||130 ± 19||132 ± 20||0.17|
|Glasgow coma scale, median (IQR)||15 (15–15)||15 (15–15)||0.64|
|Oxygen saturation (%),b median (IQR)||98 (97–99)||98 (97–99)||0.15|
|Pain score (0–10),c median (IQR)||5 (3–6)||5 (4–6)||0.18|
|Delay between first and second visit (days), median (IQR)||3 (1–5)||2 (1–4)||0.97|
|Blood sample analysis at the first visit, n (%)||197 (30)||131 (47)||<0.001|
|Radiological exam at% the first visit, n (%)||203 (31)||119 (43)||<0.001|
|Rise in nurse triage assessment, n (%)||188 (29)||121 (44)||<0.001|
|Variables||Prevalence, n (%)||Odds Ratio (95% CI)||p-value|
|Cancer||71 (8)||2.06 (1.19–3.56)||<0.01|
|Letter from a general practitioner at the first visit||149 (16)||2.02 (1.38–2.97)||<0.001|
|Age ≥ 65 years||152 (16)||2.02 (1.33–3.06)||0.001|
|Previous cardiac diseasea||80 (9)||2.02 (1.19–3.44)||0.01|
|Rise in severity of nurse assessment||309 (34)||1.99 (1.46–2.73)||<0.001|
|Biological examination at first visit||328 (35)||1.85 (1.35–2.54)||<0.001|
|Psychiatric disease||150 (16)||1.81 (1.35–2.54)||0.003|
|Relative present||267 (29)||1.69 (1.22–2.35)||0.002|
Despite the increased percentage of AEs in the UR group, the median triage score was not significantly different between the two visits for triage levels 2 (“orange”) through 5 (“light blue”). However, the proportion of patients triaged at the highest level of acuity (“red”) was significantly greater for the UR group (Table 4). The numerical triage score at the time of the UR visit, compared to the initial visit, suggested an increase in acuity in 309 (34%) patients, a decreased acuity in 291 (32%) patients, and unchanged acuity in 320 (35%) patients. According to that score, taken as a whole, the probability of admission to the hospital was not significantly modified between first and UR visits, nor was the probability of admission into ICU or death in the ED. However, across all five levels of triage acuity for the UR population, the observed proportion of patients admitted to the hospital (29%) or to ICU (2%) was much higher than that expected from their triage scores, because the AE percentages for UR patients were similar to the admission percentage (29%) plus the percentage of patients having ICU admission or death in the ED (1.3%) for patients triaged to level 2 (“Orange”) during the first ED visit (Tables 1 and 4).
|Variable||First Visit||Second Visit|
|Nurse assessment score||19 (2)||45 (5)a|
|1 (red)||147 (16)||156 (17)|
|2 (orange)||428 (46)||370 (40)|
|3 (green)||254 (27)||253 (27)a|
|4 (dark blue)||78 (8)||96 (10)|
|5 (light blue)||5 (0.5)||11 (1.2)a|
|Not performed||5 (0.5)||11 (1.2)|
|Expected admission to hospital||153 (16)||156 (17)|
|Observed admission to hospital||0 (0)||271 (29)a|
|Expected admission to ICU/death||6 (1)||9 (1)|
|Observed admission to ICU/death||0 (0)||18 (2)a|
Table 5 shows the qualitative analysis of patients whose condition worsened between the two visits. Agreement between the two experts was observed in 266 (96%) patients (kappa score = 0.92, p < 0.001). The plurality of patients (41%) had no change of diagnosis, but 25% of patients were judged to have had missed diagnoses during their first visit. Complications of a previously diagnosed disease were present in 13% of patients, and an unrelated diagnosis was made at the UR in 8%. Another 8% of UR patients had left the ED before their evaluations were complete at the time of the first visits. Only 4% suffered iatrogenic complications, and all of these were due to post–dural puncture headache requiring blood patches. Among the 25% of UR patients with previously missed diagnoses, the most likely causes were related to abdominal pathologies (n = 24), psychiatric diseases (n = 9), extraabdominal infectious diseases (n = 6), traumatic lesions (n = 2), stroke (n = 2), or acute coronary syndromes (n = 2). The leading causes of abdominal missed diagnosis were intraabdominal septic processes (n = 8), gastrointestinal bleeding (n = 6), and occlusion (n = 5). The leading cause of psychiatric missed diagnosis was depression with suicidal attempt (n = 4).
|Root Cause of AEs||Number (%)|
|Missed diagnosis||70 (25)|
|Complication of the disease||36 (13)|
|Iatrogenic complication||12 (4)|
|Unchanged diagnosis||114 (41)|
|Unrelated diagnosis||21 (8)|
|Left the ED before medical examination at the first visit||23 (8)|
In this large cohort study we observed that in our ED, 2.2% of initially nonadmitted patients had URs to our ED within 8 days. The overall population of UR patients was at an increased risk of AEs, compared to initial visit patients. However, the median nurse-assigned triage scores did not significantly differ between the first and the UR visits. We believe that this may suggest that the nurse-assigned triage score should be systematically upgraded for all UR patients.
The inclusion criteria used in our study merit clarification. The definition of UR has markedly varied among other studies of UR patients. Most studies have used a UR delay of up to 72 hours, but others have employed a 7- or 14-day delay between the two visits.[4, 18] An 8-day delay was used in our study, but 63% of our UR patients revisited within 72 hours (Figure 2). Some studies have only considered patients who made both ED visits for the same reason or those who failed to improve from their initial conditions (including missed diagnosis), whereas most studies have not limited the definition of an UR in this matter. In our study we excluded frequent users of the ED because we considered these to be a different population, requiring dedicated management. The studies that included patients with very long delays between the two visits (up to 28 days) probably included frequent users of the ED and are a very different population and phenomena than was considered by our study. Last, in our study, we only considered ambulatory patients. Our results may not apply to patients who return after short-duration observation periods in a dedicated ED observation unit. The proportion of UR visits in our study (2.2%) was similar to those previously reported (from 1.9% to 3.4%).[1, 6, 21-23]
The choice of our criteria of judgment (AE) also merits clarification: the decision to admit a patient does not only depend on disease severity at the time of UR, but also on the putative severity and/or potential to have worsening of the initial condition. We have thus considered that the decision to admit a patient was a summary of what the EP analyzed as an aggravation of the initial condition or a failure of appropriate medical decision-making at the time of the first visit.
Our study provides a clear indication that patients making URs to the ED have a higher probability of harboring a more severe condition at the time of the UR that was not evident during the first visit. This was confirmed by the fact that the proportion of all UR patients admitted to the hospital (29%) or admitted into the ICU (2%) was nearly equivalent to the percentage of patients admitted to the hospital or into the ICU for patients triaged to level 2 (“orange”) at the time of the first ED visit. The proportions of patients with each of the five possible nurse-assigned triage assessment levels was not significantly different between first visit patients and UR patients, except for a significant increase in the proportion of patients ranked to the highest score (“red”) at the time of the UR. This disparity between AE rates between first visit and UR patients supports the idea that the nurse-assigned triage score should be systematically upgraded for UR patients. Because the overall probability of AE in the UR population was similar to the AE proportion for patients who received an “orange” (level 2) triage at the time of the first ED visit, we suggest that each UR patient should be triaged to “orange” (level 2), except those previously coded as “orange” at the first visit. These patients, as well as those coded as “red” (Level 1) at the time of first ED visit should be systematically triaged to “red” (level 1) at the time of the UR. Although the proportion of UR visits may be an important quality measure of ED management, it has recently been shown that multifaceted interventions to decrease crowding in the ED failed to significantly decrease the return visit rate. Thus, URs may be considered to be unavoidable to some degree. However, this study suggests that upgrading the nurse triage score for URs may be used as a tool to better predict the probability of an AE apart from efforts to decrease incidence of return visits and as a tool to decrease the waiting time to be seen by a physician for patients having a UR.
We have identified several variables that were significantly associated with an AE at the time of the UR (Table 3). There is probably no need to include these variables in the triage acuity scoring system for UR patients, if the suggestion of assigning these patients an “orange” or “red” score, as described above, is performed. However, these variables may be useful to alert the physician that there is an increased risk of an AE. These variables identified either an inherent frailty characteristic of the patient (cardiac disease, psychiatric disease, elderly patients, or cancer) or a previous concern raised by a physician (letter from a general practitioner or additional blood sample analysis requested by the EP during the first visit) or a clear modification of the nurse-assigned triage score between the two visits.
When considering the causes of AEs, comparison with several previous studies is difficult because symptoms (for example, dehydration) and diseases were both considered in the analysis or because the analysis was mainly orientated to identify if the revisit was related to the patient, the physician, the disease, or the health system. Keith et al. reported a 15% incidence of medical management deficiencies, and Liaw et al., an 8% incidence. Most of the missed diagnoses were related to abdominal, infectious, and psychiatric diseases as previously reported.[26, 27] However, the pathologies associated with the missed diagnosis may have different significance, suggesting different options to decrease their incidence and/or consequences. While an infectious process might be detected earlier using appropriate biomarkers, the identification of patients at risk for suicidal attempts may be far more problematic. In the case of intraabdominal pathologies, using appropriate imaging may improve the accuracy and speed of arriving at an appropriate diagnosis, but it is also appropriate to inform the patient that any deterioration of their condition should prompt a return visit to the ED. In practice, many patients visit the ED for abdominal pain, and few of them require admission to the hospital. Moreover, many patients now present to an ED very early after the beginning of an abnormal abdominal process, at a time where precise diagnosis may be more difficult because there has not been sufficient time elapsed for important diagnostic clues to become evident.
Several studies have tried to assess the proportion of avoidable UR visits. This proportion was estimated at 32% by Keith et al. Although we did not assess this characteristic, while considering as avoidable missed diagnoses, iatrogenic complications, and patients who left the ED before medical examination, one can estimate that at least 38% of our revisits should be considered as potentially avoidable (Table 5). However, the prevention of these avoidable return visits requires very different action plans to be deployed for the different causes of UR. Improvement in the accuracy of diagnosis could improve the UR rate and should focus especially upon abdominal and infectious processes, while prevention of iatrogenic complications should mainly focus on the quality of lumbar puncture techniques employed in the ED. The problem of patients who leave the ED before their medical examination has been completed has a high likelihood of being related to ED overcrowding. However, the potential solutions are not easy to apply, and previous attempts to decrease crowding have failed to significantly decrease revisits to the ED.
The nurse triage score was derived to determine an assessment of a patients' disease severity, to identify patients requiring rapid medical examination. However, these triage scores have not been validated to predict the requirement for admission. Nonetheless, our data show a positive correlation between scores derived from our triage scoring system and the observed rate of hospitalization and ICU admission. Therefore, we think that data derived from a triage scoring system may be one of the numerous variables taken into account that should be considered by EPs when deciding whether or not to admit a patient.
Because our ED received only adults, our results may not apply to pediatric patients. The pediatric population in the ED differs markedly from the adult population in terms of the chief complaint that prompts a visit to the ED and the proportion of patients admitted to the hospital. Moreover, the parents may play a major role in the decision to make an unscheduled revisit to the ED. Second, this study conveys the experience of a single institution. These findings should be validated in other EDs from different countries and different health care systems. Third, there was no accounting for socioeconomic factors that are known to influence the probability of a visit to an ED.[20, 31]
We observed that 2.2% of our patients made an unscheduled return to the ED within 8 days. This population was at risk of deterioration of their condition. Despite this, their median triage nurse score was not significantly greater at the time of the unscheduled return visit, suggesting that this triage score should be systematically upgraded for unscheduled return patients to “orange” (level 2) or “red” (level 1), as has been described above in the Discussion. Eight variables have been identified that may serve as indicators to help the EP to identify more rapidly the patients at higher risk of an adverse event.
The authors thank Dr David J. Baker, DM, FRCA (Department of Anesthesiology, CHU Necker-Enfants Malades, Paris, France), for reviewing the manuscript and Yannick Le Manach (Department of Anesthesiology and Critical Care, CHU Pitié-Salpêtrière, Paris, France), for statistical advice.
- 4Emergency department visits by persons recently discharged from U.S. hospitals. Natl Health Stat Rep. 2008; 24:1–9., , .
- 10Australasian College of Emergency Medicine. The Australasian triage scale. Emerg Med. 1994; 6:145–6.
- 14Assessment of inter-observer reliability of two five-level triage and acuity scales: a randomized controlled trial. Can J Emerg Med. 2004; 6:240–5., , , et al.
- 19Emergency unscheduled returns: can we do better? Singapore Med J. 2009; 50:1068–71..
- 31Factors associated with ED use by “repeater” and “non-repeater” patients. J Emerg Nurse. 1982; 8:243–7., .