EMS non-conveyance: A safe practice to decrease ED crowding or a threat to patient safety?

Background The safety of the Emergency Medical Service’s (EMS’s) non-conveyance decision was evaluated by EMS re-contacts, primary health care or emergency department (ED) visits, and hospitalization within 48 h. The secondary outcome was 28-day mortality. Methods This cohort study used prospectively collected data on non-conveyed EMS patients from three different regions in Finland between June 1 and November 30, 2018. The Adjusted International Classification of Primary Care (ICPC2) as the reason for care was compared to hospital discharge diagnoses (ICD10). Multivariable logistic regressions were used to determine factors that were independently associated with adverse outcomes. Results are presented with adjusted odds ratios (aORs) together with 95% confidence intervals (CIs). Data regarding deceased patients were reviewed by the study group. Results Of the non-conveyed EMS patients (n = 11,861), 6.3% re-contacted the EMS, 8.3% attended a primary health care facility, 4.2% went to the ED, 1.6% were hospitalized, and 0.1% died 0–24 h after the EMS mission. The 0–24 h adverse event rate was higher than 24–48 h. After non-conveyance, 32 (0.3%) patients were admitted to an intensive care unit within 24 h. Primary non-urgent EMS mission (aOR 1.49; 95% CI 1.25 to 1.77), EMS arrival at night (aOR 1.82; 95% CI 1.58 to 2.09), ALS unit type vs BLS (aOR 1.43; 95% CI 1.16 to 1.77), rural area (aOR 1.74; 95% CI 1.51 to 1.99), and older patient age (aOR 1.41; 95% CI 1.20 to 1.66) were associated with subsequent primary health care visits (0–24 h). Conclusions Four in five non-conveyed patients did not have any re-contact in follow-up period. EMS non-conveyance seems to be a relatively safe method of focusing ED resources and avoiding ED crowding.

a healthcare facility [7], rural area, younger patient age [3,10,11], low National Early Warning Score (NEWS2), and alcohol use increase the likelihood of nonconveyance [3]. A recent review showed that, after a non-conveyance decision, re-contact with the EMS or the ED, hospitalization and mortality rates varied a lot [5]. However, whether re-contact with EMS was for a similar reason as the initial EMS contact is unclear [5].
Assessment and triage are a central part of the EMS work process [4]. Under-triaging may put patients' safety at risk, whereas over-triaging leads to inappropriate use of limited resources [12]. NEWS2 developed by the Royal College of Physicians is a simple, widely adopted scoring system [13]. NEWS2 may help identify patients at risk of deterioration who need to be treated and conveyed by the EMS, but whether it can be used as an indicator regarding EMS non-conveyance decisions is controversial [14,15].
Patient safety is a priority of the EMSs. Decisions not to convey patients may represent a risk to patient safety. These safety factors are unclear in the prehospital setting [4,16] and relevant studies are lacking [5,17]. However, some studies indicate that EMSs are able to make accurate preliminary diagnoses [18]. In the EMS context, a great number of adverse outcomes are associated with difficulties in clinical judgement [16] and patient groups with non-specific reasons for care [19,20]. Older age and abnormal vital signs are common predictors of adverse outcomes after non-conveyance and, therefore, pose a threat to patient safety [21]. Finally, from the patient safety perspective, little is known about nonconveyance decisions and related adverse outcomes [5,17,22].
The aim of this study was to identify the rate and predictors of adverse outcomes after non-conveyance by the EMSs to determine whether the current practice of nonconveyance ensures patient safety.

Design
This is a prospective cohort study.

Finland's health care system and EMS
Finland is one of the five Nordic welfare states. The health care system is financed by public funds and mainly organized by public sector. Health services are divided into primary and specialized medical care.
Organized by 21 hospital districts, EMSs are part of specialized care. In a four-tiered system, ambulance units are normally at the ALS level manned by at least one paramedic-nurse with a 4-year bachelor-level education. The other person in an ALS unit or personnel in a BLS unit can be a firefighter, an emergency medical technician (EMT) or a practical / registered nurse. A non-conveyance decision can be made supported by regional or national guidelines or by consulting a 24/7 oncall EMS or primary care physician. If EMS conveyance is needed, the target ED or other health care facility is decided upon by the EMS care provider with consultation of the EMSs or primary care physician if necessary [23].
A national dispatch authority operates with the common 112 emergency number in six regional emergency medical communication centers (EMCCs). Medical emergency calls are classified into four categories (A, B, C, and D), with A and B being urgent calls with lights and sirens. All dispatchers have completed 18 months of education, but they are usually not health care providers.

The EMS data
The EMS data were collected between June 1 and November 30, 2018, from the data systems of the Finnish hospital districts of South-Savo, Kanta-Häme, and Päijät-Häme (Fig. 1). The study area, consisting of 32 municipalities and both urban and rural areas, has 482,805 inhabitants, which is 8.8% of the total Finnish population. The average population density is 26.1 people per square kilometer. The adjusted ICPC2 classification as the main reason for EMS care was taken in to use. The original ICPC2 coding is used in primary care and published by WHO [24,25]. The adjusted classification (around one hundred ICPC2 codes) was created by the Nordic Collaboration (Benchmarking) Group for the context of prehospital emergency care [26] and it is published in the code server of the Finnish Institute for Health and Welfare [27]. The EMS care providers were trained in the use of the codes before the study period. More detailed description of the adjusted ICPC2 classification, measurement and interpretation of the NEWS2 scores, urban-rural classification, distance to health care facilities, use of alcohol, and additional data collection were described previously [3].

Study protocol
Non-conveyed patients who were discharged at the scene after EMS assessment and treatment were included in this analysis. Exclusion criteria are presented in Fig. 2. Included patients were identified using unique 10-digit personal identity numbers and linked to the registries described below.
The first EMS re-contact was recorded between 0 and 24 h and 24-48 h from the initial non-conveyance EMS mission. If the re-contact did not lead to conveyance, a new follow-up period was started. The main reason for care (ICPC2; preliminary diagnosis) was compared between the initial contact and the EMS re-contact, i.e., whether the re-contact was related to the initial nonconveyance mission.
In Finland, hospital districts are required to submit care notifications to registries. Therefore, registry information on visits to primary health care facilities or EDs and hospitalization is available from the Finnish Institute for Health and Welfare [28,29]. Unscheduled visits to primary health care and EDs (0-24 h or 24-48 h) were collected. If the exact time of the visit was missing, the initial non-conveyance case was judged to have occurred first and the 0-24 h health care visit to have occurred the same day or the day after. However, in the Register of Primary Health Care Visits [28], the data also include chronic disease monitoring. The first visit was analyzed and combined with the latest non-conveyance case if there were many. Whether the patient went to hospital by ambulance or by other means was not recorded. The ICPC2 code chosen by the EMS was compared to the main discharge diagnosis (ICD10) based on ICPC2 and ICD10 mapping charts to determine whether the visit was for the same or a related complaint [25]. The Finnish Causes of Death registry data, including death certificates from Statistics Finland [30], were used to identify deceased patients. The deaths were considered for a longer time to gain deeper insight (28 days from initial non-conveyance mission). In the data, the time of death was registered only by date, not by hour. Thus, the 0-24 h mortality includes deaths that occurred the same and the following day as the initial nonconveyance case. Only unexpected deaths were analyzed; end of life patients were excluded from the analysis because they normally have formal arrangements for dying at home. The death was connected with the last nonconveyance mission. Two experienced emergency physicians (JK and TI) analyzed all cases independently; if the judgement differed, the case was discussed until consensus was reached (JP, JK, and TI). Deceased patients were evaluated as follows: 1) Was the death related to end-oflife care? 2) Did the patient refuse conveyance to the ED or primary health care facility? 3) Was the death credibly connected to the initial non-conveyance case? 4) Would the patient have benefited from conveyance to an ED or primary health care facility?

Outcome measures
The primary outcomes were EMS re-contact, unscheduled primary health care or ED visit, and hospitalization in 0-24 and 24-48 h. The secondary outcome was 28day mortality.

Data analysis
Categorical variables were characterized using frequencies and percentages and continuous variables using medians and interquartile range (IQR). The age groups were defined based on the Finnish national classification provided by Statistics Finland. Distance to nearest primary health care facility or ED was classified for the purposes of the analysis.
Univariate associations between outcome variables and categorical study variables were studied using logistic regression analysis. Multiple logistic regression analysis included variables that were clinically and statistically significant after univariate analysis. The NEWS2 score is suitable only for patients > 16 years of age, and distance to the nearest primary health care facility or ED measures the same thing as rural-urban classification; therefore, these factors were excluded from the model. Moreover, a non-specific reason for care (ICPC2) as the categorical variable and hospitalization in 24-48 h and 28-day mortality as dependent variables were infrequent in the data. Thus, these analysis were not performed. Results are presented with univariate and adjusted odds ratios (ORs and aORs) together with 95% confidence intervals (CIs) and p-values. All analyses included only patients who did not have missing values for variables included in the model. Statistical analyses were carried out using SAS for Windows version 9.4 (SAS Institute Inc., Cary, NC, USA), and p < 0.05 was considered significant.

Results
A total of 48,297 EMS missions were identified, with 35,250 EMS missions included for the first-phase analysis. Of these patients, 42% (n = 14,874) were treated and discharged at the scene. Of these non-conveyed patients, 11,861 met the final inclusion criteria for this study (Fig. 2, Table 1). The reasons for non-conveyance are given in Table 2. The median age of the included patients was 67 (IQR 44-80) years, and 52.6% were females. Regarding EMS contacts during the study period, 85% (n = 7887) had 1 contact, 14.4% (n = 1334) had 2-6 contacts, and 0.6% (n = 60) had at least 7 contacts. Overall, 16% of the patients were under the influence of alcohol. The NEWS2 score was low, with 54.5% of patients having zero points (Table 3).
In the case of EMS re-attendance in 0-24 h, the recontact was related (same ICPC-2 code) to the initial mission in 36.1% of cases, for a different reason in 50.5%, and the ICPC2 code was missing in 13.4% of missions. The corresponding figures for EMS re-attendance in 24-48 h were 29.0, 55.0, and 16.0%, respectively. Of these EMS re-contacts in 0-24 h, in 4 of 5 cases (80.4%) the mission priority was non-urgent, ending in a new non-conveyance decision in 39.7%; the corresponding figures for 24-48 h are 79.5 and 50.6%, respectively. The NEWS2 points were as follows: at 0-24 h, 50.8% had zero points and 83.3% had 0-4 points, whereas at 24-48 h 53.0% had zero points and 85.5% had 0-4 points. The median duration of visits to primary health care was 15 min and 65% of the patients were discharged from the ED within 24 h. The majority of primary health care or ED ICD10 diagnoses were not mapped to the initial EMS ICPC2 codes. Among the non-conveyed patients, 0.3% were admitted to an intensive care unit in 0-24 h (0.03% in 24-48 h) and 0.5% to high dependency units (0.1% in 24-48 h; Table 6).
Predictors of adverse events are shown in Table 7. EMS arrival time at night and older patient age were common predictors of many events. Furthermore, EMS re-contact (0-24 h) was more likely when the patient had refused conveyance than when the EMS had treated  In addition, non-specific reasons for care increased the probability of EMS re-attendance in 24-48 h (OR 1.304; 95% CI 1.00 to 1.70). These predictors based on univariate analyses did not show any significant associations with other outcomes. Gender (p-value ≥0.054) and if there was less than an hour to complete a shift (p = ≥0.094), did not predict any outcomes in this study.
After exclusion of patients for whom an end of life decision had been made either before or during EMS attendance via telephone consultation with EMS physician (n = 55, median age 85 years), there were 126 patients who died within 28 days (Fig. 3). In this group, the mission leading to the non-conveyance decision occurred most often at home (61%, n = 77). The remaining missions occurred at health care or social service units (37%, n = 47) and public places (2%, n = 2). Among the deceased patients, the place of death was at home in 17% (n = 22), at health care or social service units in 81% (n = 102), and in public places in 2% (n = 2). Overall, 10 patients refused conveyance to the ED. The median age of the deceased patients was 83 years. Emergency physicians or primary care physicians were consulted in 51% of the cases. Based on clinical reevaluation by JP, TI, and JK, 25% (n = 32) of the deaths were related to the initial non-conveyance mission. In four cases (0.03% of the initial non-conveyance missions), the non-conveyance decision was not appropriate according to the clinical re-evaluation. In the first case, the patient with symptoms of gastroesophageal reflux had ST-elevation myocardial infarction but a 12-lead ECG had not been recorded. In the three remaining cases, the patients had shortness of breath and swelling of the foot due to coronary disease leading to heart failure, non-specific symptoms due to pneumonia, and aortic dissection with typical back pain. In two of these four cases, a physician had been consulted. In addition, one patient who refused treatment and later died would have clearly benefited from conveyance to the ED.

Discussion
The main findings of this study were as follows. Firstly, 4 in 5 (83.9% in 0-24 h) of non-conveyed patients had no adverse events after the non-conveyance mission. Secondly, patients were mainly in good condition; the NEWS2 scores were low and duration of visits to primary health care or EDs short. The reasons for the adverse events seem to be different than the reason for the initial EMS mission. Thirdly, 0.03% of the nonconveyance decisions seem to be related to a patient's death, where re-evaluation showed poor clinical judgement and/or clinical treatment protocol violation.
From the perspective of patient safety, it is important that the majority of non-conveyed patients did not have any subsequent events during the follow-up period, which is in line with previous studies [21,31]. We found that EMS re-contacts, visits to primary health care or ED and hospitalization were relatively rare after nonconveyance missions. Two previous reviews indicate that there are many studies focusing on specific populations in which the sample size is small or the follow-up is by telephone, which may lead to bias because the follow-up of a large number of patients may fail [5,17]. Thus, comparisons with our findings are challenging. However, the subsequent event rates in this study were roughly the same as in other similar studies [21,31]. Our study indicates that the 24-h period after an EMS visit seems to be critical. The highest incidence of all outcomes occurred within 0-24 h after the initial nonconveyance mission, which was noted previously [21]. A longer follow-up period could have provided more information, but the likelihood of an adverse event being due to some other reason would have increased.
Our study shows that, in the case of EMS re-contact, primary health care or ED attendance, and hospitalization, the patients were mainly in good condition. The new mission (EMS re-contact) was commonly non-urgent, the patients' NEWS2 scores were low, and almost half of these cases ended in a new non-conveyance decision. The visits to primary health care facilities or EDs were mostly short. On the other hand, 32 patients (0.3%) needed intensive care and 62 patients (0.5%) were treated in highdependency units after the initial non-conveyance case. These findings are similar to a Swedish study [10]. However, it is difficult to assess whether these patients' critical condition was derived from the EMS's poor clinical judgement and incorrect non-conveyance decision. When comparing the preliminary diagnosis (adjusted ICPC2) to new ICPC2 codes in case of EMS re-contact or discharge diagnoses (ICD10) from primary health care, ED, or after hospitalization, these adverse events were usually not related to the initial non-conveyance mission. In addition, we found that there were many subsequent visits to primary health care, which may be an indication of impaired access to primary health care, especially in the evening and at night. Our logistic regression model (Table 7) and previous studies [21,32] indicate that older age is a risk factor for adverse outcomes. This may be due to elderly patients' complex symptoms and many comorbidities and medications. We also found that patients' refusal of conveyance and the use of alcohol predicted an EMS re-contact in 0-24 h. One explanation may be that these patients do not know how to handle their problems and an EMS re-call is the easiest choice. Other studies have reported that the refusal is associated with ED visits [33] and alcohol increases the likelihood of non-conveyance [3].
Decision-making at night is challenging [34]. Our results show that EMS arrival at night increases the likelihood of 3 in 4 primary outcomes of this study. However, the end of a work shift was not associated with recontacts, which may indicate that EMS providers consider the patient's needs even though the shift is close to its end. However, EMS arrival at night, non-urgent mission, ALS unit, rural area, and longer distance to a health care facility or ED were related to subsequent visits to primary health care. There is a possibility that this demonstrates appropriate use of health care resources to avoid unnecessary conveyance to the ED. Geographic variation in the EMS context is high [35], but the impact of geography on a patient's outcome is unclear. However, the on-scene time is reported to be high in rural areas [36] and in the case of nonconveyance [3]. Understandably, visits to primary health  care were related to non-urgent missions. In contrast, it seems that EMSs can safely discharge urgent missions such as hypoglycemia and other chronic diseases at the scene after appropriate assessment and treatment. Previous studies have shown that non-conveyance is challenging and requires competence [5,34]. Indeed, there are a number of factors that are related to nonconveyance decisions [3,8,[37][38][39]. Our study demonstrated that ALS units are associated with re-contact in primary health care. This may be due to the longer  Moreover, an Australian study indicated that, when the EMS discharged patients at the scene, there was an increased risk of adverse events compared to patients discharged from the ED [21]. Notably, subsequent contacts with health care do not automatically mean that patient safety is compromised [5,30].
Abnormal vital signs have been found as a common predictor of adverse events after EMS non-conveyance [21]. We found that, if the patient's NEWS2 score increases by 1 point or the score is high (> 7 points), the risk of EMS re-contact increases. Non-specific complaints lead to a number of adverse outcomes in both the EMS context and EDs [19,20,40]. Surprisingly, we did not find similar results. Based on univariate analyses, non-specific complaints were only related to EMS recontact in 24-48 h.

Limitations
This study has several limitations, three of which were described previously (excluded patients, the use of adjusted ICPC2 classification, and NEWS2 score calculation) [3].
The register of ED visits included the date, but the exact visit time was mainly missing. Thus, the initial non-conveyance case was judged to have occurred first and the 0-24 h ED visit to have occurred on the same day or the following day. Furthermore, the register of visits to primary health care includes chronic disease monitoring. Therefore, the rate of visits to primary health care and EDs for real adverse events after the non-conveyance mission may be lower. On the other hand, some patients may have sought further care at private clinics, but all patients with severe or even moderate deterioration would have been sent to an ED.
The reasons for care (adjusted ICPC2) were compared to the discharge diagnoses (ICD10) from the primary health care facilities or EDs or to diagnoses after hospitalization in order to determine whether these events were related to the initial non-conveyance mission. Notably, the ICPC2 chosen by EMSs is based on symptoms and signs present at the time of the EMS contact. The time between ICPC2 and discharge diagnosis, further examination, and the treatment given may affect the discharge diagnosis. Thus, the real rate of adverse events due to the same reason as the initial nonconveyance case may be higher.
Only unexpected deaths were analyzed, and therefore patients with end of life decisions were excluded. However, it is possible that this approach did not identify all of these patients due to missing information. When assessing the risk factors for adverse events, 28-day mortality and hospitalization in 24-48 h were excluded from our multivariable logistic regression model as dependent variables because these events were very rare in our data set.
In Finland, EMSs and EDs are encouraged to plan emergency patient pathways together. Therefore, triaging and assessing the need for conveyance and nonconveyance decisions are commonly made by the EMS. This practice and the level and scope of education differs between countries; therefore, the generalizability of the results may be limited [23].