Diagnostic anticipation to reduce emergency department length of stay: a cohort study in Ferrara University Hospital, Italy

Background : Emergency Department (ED) crowding reduces staff satisfaction and healthcare quality and safety, which in turn increase costs. Despite a number of proposed solutions, ED length of stay (LOS) - a main cause of overcrowding - remains a major issue worldwide. This cohort study was aimed at evaluating the effectiveness on ED LOS of a procedure called “diagnostic anticipation”, which consisted in anticipating the ordering of blood tests by nurses, at triage, following a diagnostic algorithm approved by physicians. Methods : In the second half of 2019, the ED of the University Hospital of Ferrara, Italy, adopted the diagnostic anticipation protocol on alternate weeks for all patients with chest pain, abdominal pain, and non-traumatic bleeding. Using ED electronic data, LOS independent predictors were evaluated through multiple regression. Results : During the weeks when diagnostic anticipation was adopted, as compared to control weeks, the mean LOS was shorter by 18.2 minutes for chest pain, but longer by 15.7 minutes for abdominal pain, and 33.3 for non-traumatic bleeding. At multivariate analysis, adjusting for age, gender, triage priority and ED crowding, the difference in visit time was significant for chest pain only (p<0.001). Conclusions : The effectiveness of the anticipation of blood testing by nurses varied by patients' condition, being significant for chest pain only. Further research is needed before the implementation, estimating the potential proportion of inappropriate blood tests and ED crowding status Background:

urgent [4] or frequent visits [5]. One of the main causes of prolonged ED LOS involves the patients flow within the ED and is defined as "throughput" [6]. This period starts from patient's arrival in ED (triage) to the decision of the physician (admission or discharge).
Many interventions have been tested to improve ED waiting times and LOS [7], including deployment of physicians at triage [8], use of trained scribes to assist ED physicians [9], nurse-initiated diagnostic ordering at triage, based on physician approved algorithms [10], and resident-initiated advanced triage [11]. A systematic review concluded that nurse-initiated diagnostic ordering were effective in reducing ED LOS, but the available evidence was limited, as studies were scarce and of poor methodological quality [12].
Given that the Italian and Regional healthcare government recommended a maximum threshold of six hours for ED LOS, the Ferrara University Hospital introduced nurse-initiated diagnostic ordering at triage at alternate weeks, thus allowing an evaluation of the effectiveness and feasibility.
This cohort study was aimed at evaluating the effectiveness on ED LOS of a procedure called "diagnostic anticipation", which consisted in anticipating the ordering of blood tests by nurses, at triage, following a diagnostic algorithm approved by physicians.

Methods Ethics
The study protocol was approved by the Independent Ethical Committee of Area Vasta Emilia Centrale (CE-AVEC, study code: 840/2019/Oss/AOUFe; date of approval CE: 11/12/2019).

Study design and setting
This cohort study was performed at the Emergency Department of the Ferrara University Hospital, a tertiary care hospital in Emilia-Romagna region, Northern Italy, from July 1st, 2019 to December 31, 2019. All participants were monitored during the ED stay, from triage registration to physician's decision (hospital admission or discharge).

Study population
Inclusion criteria were: -Hour of visit between 8:00 am and 8:00 pm; -chest pain, abdominal pain or non-traumatic bleeding; -Triage priority color code yellow or green. In Italy, triage involves assigning a priority color code to patients arriving at the hospital ED: White = The situation is not an emergency, the patient is safe or does not have a life-threatening condition. Green = The situation is not an emergency; the patient has an acute but stable pathology, and vital signs are normal; Yellow = The situation is a medical emergency. Intervention cannot be delayed; Red = the situation is an absolute emergency. The patient's vital signs have deteriorated or indicate an immediate threat to patient's life [13].
Exclusion criteria were: -Death or leave of ED before physician's decision.

Procedure
The diagnostic anticipation protocol was implemented on alternate weeks to evaluate its effectiveness before full implementation (and to avoid the history bias that typically afflicts before/after evaluations). During the weeks in which the diagnostic anticipation was adopted, following an algorithm made by the physicians, the nurses at triage ordered the blood tests listed in Table 1 for all eligible patients, before physician's visit. In control weeks, the routine diagnostic pathway -with physicians (eventually) ordering blood tests after the visit -was followed. Table 1 Nurse-initiated blood test ordering at triage, based on a physician-approved diagnostic algorithm. A multidisciplinary team including the hospital risk manager, ED physicians and nurses, laboratory physicians and IT technicians defined the standard operating procedure: whenever an eligible patient was accepted to ED triage, the nurse selected the above listed blood tests within 15 minutes [14].
When blood tests results become available, the nurse delivered them to the physician for interpretation.

Data Analysis
Data have been collected from administrative ED electronic database. For each visit, the following variables were recorded: age, gender, symptoms at triage, diagnosis, date and hour of triage registration, priority code, medical imaging, specialist consultations, blood tests, physician's decisions about the patient, date and hour of the decision. ED crowding was estimated for each visit through the National ED Overcrowding Study (NEDOCS) score [15].
The statistical significance of the differences between intervention and non-intervention weeks was evaluated using Fisher's exact test for categorical variables, and t-test for continuous variables.
Separately for triage conditions, the potential independent association between diagnostic anticipation and ED LOS was evaluated using multiple regression, adjusting for age, gender, priority access codes and NEDOCS score. All analyses were performed using Stata 15.1 (StataCorp, College Station, Texas, USA, 2017). A two-tailed p-value < 0.05 was defined as statistically significant for all analyses.

Results
From July 1st, 2019 to December 31, 2019, 3242 visits were included in the study (1695 during control weeks, 1547 during diagnostic anticipation weeks), out of a total of 30,532 ED visits (Fig. 1).
As shown in Table 2, some of the demographic and clinical characteristics of the patients significantly differed during DA weeks, as compared with control weeks. In specific, during DA weeks, the NEDOCS score was higher for all clinical conditions, as well as the number of prescribed blood tests (100%), the mean age of the patients with chest pain was slightly lower, whereas patients with non-traumatic bleeding were older by more than 5 years (which probably explains, for these subjects, the higher rate of hospitalization and yellow priority codes). During DA and control weeks, respectively, the following mean ED LOS were recorded (

Multivariate analyses
Multivariate analyses showed that, for the patients with chest pain, ED LOS was significantly reduced during DA weeks: regression coefficient: -29.2 minutes; 95% Confidence Interval -CI: -44.9; -13.5 (adjusted p < 0.001 - Table 3). In contrast, ED LOS did not significantly differ during DA and control weeks for the patients with abdominal pain (p = 0.50) and non-traumatic bleeding (p = 0.26). The other independent predictors of ED LOS were higher age and NEDOCS score (for all patients), yellow priority code (only for the patients with chest or abdominal pain), and male gender (for patients with abdominal pain only).

Discussion
In this field, observational study, the introduction of protocol of blood testing diagnostic anticipation in ED showed contrasting results: although the ED was reduced by approximately 30 minutes for the patients presenting with chest pain, no impact was observed for the patients with abdominal pain and non-traumatic bleeding. Also, with regard to chest pain, the observed reduction in LOS was shorter than the mean difference of 51 minutes reported in a systematic review on triage-nurse ordering [12].
The potential explanations for the observed smaller, or zero impact, are manifold. First, the average ED LOS in the study hospital was long for all patients, approaching 6 hours, which may dilute the impact of anticipating blood testing. Second, the studies included in the above mentioned review mostly regarded triage initiated x-rays, and only 2 studies out of 14 also considered blood tests [12].
Moreover, of the two studies including blood tests, one was an unpublished dissertation, and the other had a weak methodology [10]. Third, the DA protocol was implemented for the first time during the six months of the study, and the adoption of the algorithm by triage nurses was certainly suboptimal, especially in the first months. Finally, with regard to the different findings on chest pain and abdominal pain or non-traumatic bleeding, this may be due, at least in part, to the lower proportion of blood testing that were performed during control weeks for the subjects with abdominal pain or nontraumatic bleeding, as compared to those with chest pain. Performing a lower number of blood tests could clearly result into a shorter LOS, jeopardizing the potentially positive impact of anticipation.
Certainly, further research is needed to clarify these points, as well as to confirm of disprove the benefit of diagnostic anticipation for the patients with chest pain.
The other results of the multivariate analyses were straightforward: a longer ED LOS was observed for older patients, with upper priority code, during the periods of higher ED crowding (higher NEDOCS score). Noteworthy, female patients with abdominal pain showed a significantly longer LOS than males. This could be explained by the fact that abdominal pain has gender-specific diagnostic differences (for example gynecological conditions). Again, further, specific studies are warranted to investigate the potential gender difference on LOS and its potential organizational consequences.

Limitations
First, in this study the diagnostic anticipation protocol was limited to the daily hours of service from 8:00 am to 8:00 pm, due to a limited availability of resources (nurses in service) during night shifts.
However, during the nights, ED crowding is typically lower.
Second, triage-initiated blood testing requires a crowded ED in order to detect a positive impact on LOS: in uncrowded ED patients are immediately, or after a very short waiting time, addressed to physician's evaluation, and it may not be observed any LOS reduction from anticipated testing. In this study, the mean NEDOCS score ranged from 120 (overcrowded) to 160 (severely overcrowded). Thus, the findings of this study cannot be generalized to Emergency Departments with low crowding status and short waiting times before physician's evaluation.
Conclusions: 10 The introduction of a protocol of diagnostic anticipation of blood tests at triage, into a crowded Emergency Department, showed contrasting results: the LOS was significantly reduced, by   approximately 30 minutes, for the patients reporting chest pain, whereas no impact was observed for   the patients with abdominal pain or