Abstract
Background/Aim: Several risk scores can stratify patients with acute heart failure (AHF) at the Emergency Department (ED). Registration of vital signs, such as blood pressure (BP), heart rate (HR) and respiratory rate (RR) upon admission is mandatory. Nevertheless, measurement of RR remains neglected worldwide. Patients and Methods: The predictive value of RR in classifying patients with AHF was investigated by processing several vital signs recorded in the ED. Results: HR and RR individually did not discriminate patients according to hospitalization length, Intensive Care Unit (ICU) admittance, mechanical respiratory support or death. The derivative indices, HR:RR and Respiratory Efficacy Index (REFI) (=RR×100/SatO2), differentiated study patients regarding hospitalization length. Receiver operating characteristic curves predicting mortality and ICU admission for REFI and HR:RR revealed high accuracy, sensitivity and specificity for cut-off values of REFI >27 and HR:RR ≥4. Conclusion: The RR and its derivative indices are easily accessible vital signs monitored at the ED which merit ‘revitalization’.
The vast majority of patients (80-90%) presenting at the Emergency Department (ED) with acute heart failure (AHF) are hospitalized (1, 2). Half of them are in mild condition, rendering their hospitalization unessential and financially unwarranted (3-5). On the other hand, patients misclassified as being at low-risk and discharged directly from the ED are prone to early complications, with a 7-day mortality reaching 4% (6, 7). Several risk scores have been suggested in order to stratify patients with AHF at the ED (8, 9). Registration of vital signs upon admission, such as blood pressure (BP), heart rate (HR), body temperature, and respiratory rate (RR) is mandatory (10, 11). Nevertheless, measurement of RR remains neglected worldwide (12, 13). In this study, the predictive value of RR in classifying patients with AHF hospitalized through the ED was investigated in order to emphasize on the need for monitoring RR at the ED.
Patients and Methods
The registration charts and ED-monitored parameters of all patients presenting with AHF at the ED of Sotiria Chest Diseases General Hospital, Athens, Greece during 2014 were reviewed. The study protocol conformed to the ethical guidelines of the 1975 Declaration of Helsinki and to Title 45, U.S. Code of Federal Regulations, Part 46, Protection of Human Subjects (effective December 13, 2001) as reflected in a priori approval by the Institution's Human Research Committee (approval number AM 3/2013).
Among 985 consecutive patients who presented with AHF at the ED, RR was recorded at the ED in only 262 (26.5%) and 100 were hospitalized, constituting the study population. Therefore, selection bias was minimized.
For each patient, demographics, comorbidities, triggering factors, vital signs, arterial oxygen saturation (SatO2) at the ED, inotropic support and length of hospitalization were collected. The novel derivative indices, heart rate to respiratory rate (HR:RR) and respiratory efficacy index (REFI=RR×100/SatO2), were also introduced using the values measured at the ED. Need for intubation, admittance to the intensive care unit (ICU) and death constituted the study endpoints. Length of hospitalization was also one of the clinical outcome measures and a nominal variable was created for statistical purposes: ≤6 days, 7-10 days, ≥10 days.
Statistical analysis. The SPSS 12.0 package (SPSS Inc, Chicago, IL, USA) was used. Normally distributed continuous variables were expressed as the mean±standard deviation and categorical variables as percentages. The independent t-test and the Mann-Whitney U-test were applied to compare means and medians. Pearson and Spearman correlations were also investigated. Differences between categorical variables were assessed by the chi-square test. Multiple logistic regression analysis with backward selection investigated associations between various variables. Several measurable parameters acting as possible confounders (age, sex, HR, BP, SatO2, RR and length of hospitalization) were included in the initial variable set. Percentile analysis and contingency tables were prepared and Kaplan-Meier survival curves were developed to study the relationship of HR:RR and REFI with the number of deaths recorded and ICU admission. The prognostic ability of HR:RR and REFI on ICU admission was also evaluated by receiver operating characteristic (ROC) curves. p-Values less than 0.05 were considered statistically significant.
Demographical characteristics and medical condition of the study population.
Results
Table I summarizes the demographical characteristics and the medical evaluation of the study population in total and by subgroup for the study endpoints (need for intubation, admittance in the ICU and death). All vital signs presented were recorded at the ED upon hospital admission.
RR was increased by 18% (p=0.03) in patients admitted to ICU compared to the total population of the study and was 33% (p=0.02) higher in patients treated with mechanical respiratory support due to severe AHF that caused respiratory insufficiency compared to the general study population, while the increase of RR in patients who finally died was even higher, reaching 44% (p=0.04).
Percentile analysis and multiple logistic regression analysis with backward selection showed that HR and RR individually were not able to discriminate patients according to the AHF severity indices (ICU admittance, mechanical respiratory support, death).
A high discriminatory ability was detected for the introduced HR:RR and REFI indices. Using the most relevant median=values revealed by percentile analysis, HR:RR=4 and REFI=27 were selected as cut-off points and were used to dichotomize the data in multiple logistic regression analyses. Both HR:RR and REFI differentiated study patients regarding the duration of their hospitalization (p=0.007 and p=0.001 respectively, Figure 1A and B). However, multiple logistic regression analysis revealed that only REFI as a dichotomous variable was associated with the length of hospitalization (beta=0.361, 95% CI=0.285-0.438, p=0.01) among the various parameters tested (age, sex, HR, BP, SatO2, RR).
Furthermore, Kaplan–Meier curves showed longer hospitalization in patients who died and presented a HR:RR <4 (median=10 days, range=8.3-11.7 days) compared to patients with HR:RR≥4 (median=6 days, range=5.6-6.4 days, p<0.001). Similarly, Kaplan–Meier curves showed higher hospitalization length for patients who died and presented a REFI>27 (median=14 days, range=13.1-14.9 days) compared to those with REFI≤27 (median=9 days, range=7.5-10.5, p=0.007). Kaplan–Meier curves with endpoint admission to ICU revealed longer hospitalization for patients who presented to the ED with REFI>27 (median=11.1 days, range=10.1-11.90 days), compared to those with REFI≤27 (median=7.2 days, range=5.5-8.5 days, p=0.007) (Figure 2). Use of the HR:RR index did not lead to statistically significant results for patients admitted to ICU regarding the duration of hospitalization (p=0.095).
There was a strong bivariate correlation between ICU hospitalization and HR:RR <4 (83.3% of patients, p<0.001). The same applied to REFI>27 (73.3% of patients, p=0.002).
The prognostic value of REFI measured at the ED in identifying patients needing ICU admission was evaluated by ROC curves (Figure 3). At REFI=27, accuracy of 79.4%, sensitivity of 73.3% and specificity 60% (p<0.01 respectively) were found. Similarly ROC curve analysis proved the prognostic value of HR:RR measured at the ED in identifying patients needing ICU admission and showed that for HR:RR=3.98, accuracy, sensitivity and specificity were 81.1%, 83.3% and 70.0%, respectively (p<0.001). The cut-off values for HR:RR and REFI were similar to those obtained from percentile analysis regarding the prediction of ICU admission. In multiple logistic regression analysis, however, only REFI was associated with ICU admission (beta =0.512, 95% CI=0.487-0.567, p<0.015).
In addition, a strong bivariate correlation was noted between recorded deaths and HR:RR<4 (100% of patients, p=0.009). The same applied to REFI>27 (100% of patients, p=0.012).
ROC curve analysis for the predictive ability of REFI with a cut-off of 38 regarding mortality gave accuracy, sensitivity and specificity of 96.7%, 89.4% and 100% (p<0.001), respectively. ROC curve analysis for the predictive ability of HR:RR index with a cut-off value of 3.85 for mortality gave accuracy, sensitivity and specificity of 77.7%, 87.8% and 70.0%, respectively (p=0.001). It is worth noting that the HR:RR cut-off value of 3.85 was statistically equal to the median of 4 from the percentile analysis.
Length of hospitalization based on respiratory efficacy index (REFI) (A) and heart rate to respiratory rate (HR:RR) index (B).
In the subgroup of patients with atrial fibrillation (AF) in the present study, a statistically significantly higher mean HR (13) was recorded (110±29.1 beats/min, p=0.01). The AF patients of our study also had lower systolic BP (118±40.6 mmHg, p=0.05), but all other vital signs, such as RR, diastolic BP, SatO2, along with the length of hospitalization, were statistically similar to those of the rest of the study population. In addition, patients with AF did not differ from the rest of the study population regarding the study endpoints, namely the need for intubation, admittance to ICU and death. Removing patients with AF from the analysis did not qualitatively alter the results.
Discussion
In the present study, medical practitioners who first treated dyspneic patients with AHF were reluctant to register RR at the ED, similar to the worldwide recorded percentage of 30%, mainly due to the lack of awareness among the ED staff (14). The average RR in our study population is similar to those reported for large AHF cohorts (15-17). Similarly the incidence of death in the present study was comparable to previous reports (18). Increased RR values in this study population were also related to intubation incidence, admission to ICU and in-hospital death. Similarly, the presence of tachypnea in the ED was correlated in a recent study with the need for mechanical ventilation, cardiac arrest incidence and early failure of outpatient treatment following discharge from the ED (19). Even in children, impaired lung function can lead to heart failure (16). Risk scores evaluating the severity of AHF in the ED, usually co-examine tachypnea with tachycardia and hypoxemia, in a triad reflecting the grade of cardio-respiratory distress (6, 20). In the presence of all three, the risk for 1- and 7-day mortality rises to 10% and 17%, respectively (10). HR, RR and SatO2 were combined in this study to develop the HR:RR and REFI indices, which were shown to be efficient in the prognostic discrimination regarding patient predisposition to ICU admission, length of hospitalization and final outcome, as expected from clinical practice experience, even in the AF subgroup (21). The HR of the patients with AF of our study was also higher than the mean HR of a recently published analysis on a large cohort of patients with AF presenting HF, probably due to the acute stress of our patients (13). Recently, new approaches regarding the breathlessness of patients with AHF at rest needing hospitalization have been developed; interestingly the belief that shortness of breath at admission is closely linked to mortality is challenged, while breathlessness at slight exertion might be more significant in risk both for in-hospital mortality and even up to 180 days post-hospital presentation (18). As breathlessness in patients with AHF is again in the spotlight, viewing it from a different angle with the new indices suggested in the current study, which combine different aspects of sympathetic activation, such as HR and RR, might further contribute to the early characterization of patients with AHF.
Kaplan–Meier curve for length of hospitalization in patients who died according to the heart rate to respiratory rate (HR:RR) index (A), and respiratory efficacy index (REFI) (B), and for length of hospitalization in patients admitted to Intensive Care Unit according to REFI (C).
Receiver operating characteristic curves for the prognostic ability of the respiratory rate to heart rate (RR:HR) index (I), the respiratory efficacy index (REFI) (II) and the heart rate to respiratory rate (HR:RR) (III) index in predicting Intensive Care Unit admission of patients with acute heart failure.
In conclusion, despite the relatively small number of patients and the fact that the study was conducted only at one hospital, our findings might assist the risk stratification of AHF in the ED, contributing to a more cost-effective treatment of patients with AHF without the need for support from analytical hospital departments. At the same time, further medical complications as a result of down-grading the risk of patients with decompensated AHF risk in the ED could also be avoided. The RR and its derivative indices are easily accessible vital signs, which merit ‘revitalization’.
Footnotes
This article is freely accessible online.
Conflicts of Interest
The Authors report no conflicts of interest in regard to this study.
- Received March 24, 2018.
- Revision received April 21, 2018.
- Accepted April 26, 2018.
- Copyright © 2018 The Author(s). Published by the International Institute of Anticancer Research.
References
In this issue
Jump to section
Related Articles
Cited By...
- The use of the raw pulse/breathing rate ratio (PBR) as a predictor of mortality and criticality in the emergency department: a retrospective study
- Nomograms for Predicting the Risk of Acute Heart Failure in ICU Patients with Atrial Fibrillation: A Retrospective Cohort Study
- Measurement of Respiratory Rate using Wearable Devices and Applications to COVID-19 Detection