Association between heart rate on admission and in-hospital mortality among general inpatients

Abstract

Association between heart rate (HR) and in-hospital mortality in general patients irrespective of underlying diseases were not well scrutinized. We assessed the relationship between HR on admission and in-hospital mortality among general inpatients.

We used data from Japan Adverse Drug Events (JADE) study, a prospective cohort study. One tertiary care hospital in Japan with 13 medical and 12 surgical wards, and an intensive care unit (ICU). Patients (n = 2360) were ≥12 years old and admitted to this hospital within 3 months; and pregnant women were excluded. We assessed the relationship between HR and mortality in five (<60, 60–79, 80–99, 100–119, ≥120 beats per minutes [bpm]) groups. We also compared the five HR groups according to the age (<70 years; ≥70 years) and wards (medical; surgical; ICU).

We enrolled 2360 patients (median age, 71 [interquartile range (IQR) 58–81] years) including 1147, 1068, and 145 patients in the medical and surgical wards, and the ICU, respectively. The median (IQR) HR on admission was 78 (68–91) bpm. Ninety-five patients died during hospitalization. Mortalities in the <60, 60–79, 80–99, 100–119, and ≥120 bpm groups were 2.9% (5/175), 2.7% (28/1047), 3.4% (26/762), 8.2% (24/291), and 14.3% (12/84), respectively (P < .001). The adjusted odds ratios of in-hospital mortality was 3.64 (95% CI 1.88–7.05, P < .001) when HR was ≥100 bpm in the medical ward; and 5.69 (95% CI 1.72–18.82, P = .004) when HR ≥120 bpm in the surgical ward. There was no statistically significant relationship with the ICU.

In conclusion, higher HR should be associated with in-hospital mortality among patients with general diseases. Even with less severe condition or outside ICU, HR should be directed attention to and patients with high HR on admission should be taken additional therapy to reduce the further risk of deterioration.

1. Introduction

High heart rate (HR) is associated with cardiovascular mortality and all-cause mortality in the general population.^[1–3] A recent meta-analysis reported the relative risk of having 10 beats per minutes (bpm) resting HR as 1.08 (95% CI 1.06–1.10) for cardiovascular mortality among the general population.^[1] The association between HR and in-hospital mortality has also been reported in patients with cardiovascular comorbidities.^[4] Patients with acute ischemic stroke and HR ≥83 bpm on admission had higher risk of in-hospital mortality with adjusted odds ratio (OR) of 4.42.^[4] However, the association between HR and in-hospital mortality in general patients irrespective of underlying diseases were not well scrutinized.

With any relationship between HR and in-hospital mortality, or any threshold to trigger it, the risk in such patients could be addressed to reduce mortality efficiently. In addition, any intervention aimed at reducing HR, or the sympathetic nervous system, might offer supplementary therapy for patients with abnormal HR on admission. Thus, we analyzed data from a prospective cohort study to clarify the association between HR and in-hospital mortality among general patients.

2. Methods

2.1. Study design and patient population

The Japan Adverse Drug Events (JADE) study involves series of cohort studies conducted to evaluate adverse drug events and medication errors in Japan.^[5–8] In this study, we used the data from a tertiary care hospital. There were 13 medical and12 surgical wards, and an intensive care unit (ICU). We included patients aged ≥12 years old, admitted to this hospital during a 3-month period from September through November 2013, while excluding pregnant women; because they were generally considered healthy people. Those aged <12 years were excluded because the median HR among them was higher than those ≥12 years.^[9] Patients were followed-up until transfer, discharge, or death. The study protocol complied with the Declaration of Helsinki and the Ethical Guidelines for Medical and Health Research Involving Human Subjects issued by the Ministry of Health, Labor, and Welfare in Japan.

The institutional review board of the hospital approved the study and the board waived the requirement of informed consent because all data were obtained as part of daily routine practice.

2.2. Data collection and definitions

In the JADE study, data on all clinical symptoms and signs as well as laboratory data were extracted from the electronic medical record from the admission to the discharge. Because patients were admitted due to a number of diseases or medical conditions, we categorized the primary diseases on admission into 15 groups by the International Classification Diseases 10th revision.^[10]

The study primary endpoint was in-hospital mortality, and HR at admission was compared with in-hospital mortality. HR was treated as a continuous variable or categorized into five (<60, 60–79, 80–99, 100–119, ≥120 bpm) groups; because we hypothesized that there was threshold to the risk of mortality. We analyzed the relationship between HR and mortality as a whole, stratified by the age (<70 years; ≥70 years) and wards (medical; surgical; ICU). The threshold of age was determined by the median value. The missing values were treated as missing without imputations and we analyzed the data without the missing variables.

2.3. Statistical analyses

The descriptive statistics were shown as median (interquartile range [IQR]) for continuous variables, and as numbers and percentages for categorical variables. We used Wilcoxon rank sum test or chi-square test to compare patients’ characteristics between “died” and “survived” patients. We compared HR groups and mortalities by chi-square test.

To assess the association between HR on admission and in-hospital mortality adjusted for possible confounders, we constructed multivariable logistic regression models including the following independent variables; age, gender, systolic blood pressure, hemoglobin, total protein, creatinine, and white blood cell count as well as HR. Because there were a number of diseases or medical conditions, we could not adjust for such comorbidities, rather, we stratified by wards and adjusted the surrogates, which were associated with the severity in patients. We constructed four models by treating HR as continuous or dichotomized variable with different thresholds. We conducted all analyses using JMP 13.1 (SAS Institute Inc., Cary, NC, USA) software. Two tailed P-values < .05 were considered statistically significant.

3. Results

We enrolled 2360 patients among the 3120 patients who were admitted during the study period. We excluded 365 patients with pregnancy or pregnancy complications, and another 395 patients who were <12 years old. The median age was 71 (IQR, 58–81) years; and men accounted for 54% (1266) (Table 1). The median HR was 78 (IQR 68–91) bpm. The missing values occurred in two patients with HR, and were generally observed in <100 patients for other variables, aside respiratory rate (1031), total bilirubin (108), γ-glutamyltranspeptidase (760), lactate dehydrogenase (272), alkaline phosphatase (538), and creatinine kinase (854). Common diseases or medical conditions included the circulatory system (26.7%), neoplasms (21.8%), and digestive system (19.4%) in the medical wards; neoplasms (33.2%), injury, poisoning, and certain other consequences of external causes (19.9%); and digestive system (13.2%), in the surgical wards; and injury, poisoning, and certain other consequences of external causes (38.6%), respiratory system (15.2%), and circulatory system (12.4%), in the ICU (Table 2).

Table 1.

Patient characteristics.

Table 2.

Conditions on admission.

During the hospital stay, the 95 patients who died were older than those who survived (median: 83 vs 70 years, P < .001). Median HR among dead patients was higher than those who survived (92 vs 78 bpm, P < .001) (Table 1). Patients who died during the hospital stay had significantly lower blood pressure, lower hemoglobin level, higher white blood cell count, higher urea nitrogen level, and higher lactate dehydrogenase level (Table 1).

Overall, in-hospital mortality was significantly elevated when the HR increased (Fig. 1A). Among patients with age less than 70 years, mortality in the ≥120 groups was 12.1% (4/33) (P < .001; Fig. 1B), whereas those ≥70 years, mortality in the 100–119 and ≥120 groups were 12.2% (20/164) and 15.7% (8/51) (P < .001; Fig. 1C), respectively.

Figure 1.

Heart rate on admission and in-hospital mortality in total cohort and according to age. (A) All patients; (B) patients younger than 70 years old; (C) patients equal to or older than 70 years old.

In terms of wards, significant associations were observed in the medical and surgical wards (Fig. 2A and B, respectively). Mortalities in the <60, 60–79, 80–99, 100–119, and ≥120 groups were 0% (0/90), 2.3% (11/484), 3% (11/367), 10.7% (18/169), and 10.8% (4/37) in the medical wards, respectively (P < .001; Fig. 2A). In the surgical wards, for the ≥120 group, this was 14.8% (4/27) (P = .009; Fig. 2B). Although mortalities in the <60 and ≥120 groups were relatively high in the ICU (30% [3/10] and 20% [4/20]), the difference was not statistically significant (P = .31; Fig. 2C).

Figure 2.

Heart rate on admission and in-hospital mortality according to ward. (A) Medical wards; (B) surgical wards; (C) intensive care unit.

Multivariable logistic regression model showed that the adjusted OR for one increment in HR was 1.03 (95% CI 1.01–1.04, P < .001) in the medical ward and 1.02 (95% CI 1.00–1.04, P = .03) in the surgical ward, but not statistically significant in the ICU (Table 3, Model 1). With HR ≥100 bpm, adjusted OR for this category compared to HR <100 bpm was 3.64 (95% CI 1.88–7.05, P < .001) in the medical ward; however, this was not statistically significant in the surgical wards (P = .10) and ICU (P = .30) (Table 3, Model 2). Similarly, at HR ≥120, adjusted OR was 5.69 (95% CI 1.72–18.82, P = .004) compared to HR <120 bpm in surgical ward, but not statistically significant in medical ward (P = .10) or ICU (P = .70) (Table 3, Model 3). There was no association between HR <60 bpm and in-hospital mortality either in the medical (P = .81) or surgical wards (P = .80), or ICU (P = .30) (Table 3, Model 4).

Table 3.

Multivariable model for the effect of heart rate on in-hospital mortality.

4. Discussion

Similar to previous reports in the general population or among patients with specific diseases, we found higher HR was associated with higher in-hospital mortality among general in patients. This association was statistically significant in the medical and surgical wards with adjusted ORs of one bpm increment of 1.03 and 1.02, respectively. The threshold in this cohort was 100 bpm in the medical and 120 bpm in the surgical wards. However, such thresholds, either higher or lower ones, were not apparent in the ICU, although some lower and higher threshold values were graphically implied.

Recent meta-analysis showed that the general people with resting HR of ≥80 bpm had higher risk of cardiovascular and all-cause mortality with relative risks of 1.33 and 1.45, respectively.^[2] Another meta-analysis showed that the resting HR was an independent predictor of coronary artery disease (hazard ratio: 1.12), stroke (HR, 1.05), all cancer types (hazard ratio, 1.09), and other diseases (hazard ratio, 1.25).^[2] Long-term follow-up cohort from the Framingham Heart Study also reported the association between higher HR and cardiovascular events with hazard ratio of 1.15 for 11 bpm increment in the baseline HR during a median follow-up of 19 years.^[11] The hazard ratio (1.32) for the same increment in HR was also reported for heart failure.^[11] These observations were derived from epidemiological studies among the general population, but other reports also suggested the associations among inpatients similar to our reports.

The association between baseline HR and in-hospital mortality has been reported in patients with cardiovascular diseases. HR on admission was associated with in-hospital mortality with hazard ratio of 4.42 for 10 bpm increment in HR, in patients with acute ischemic stroke. The recalculated OR for 10 bpm increment in HR in the medical wards in our study was 1.33, and the effect was smaller than that reported in patients with acute ischemic stroke. The reason for this discrepancy is probably because our study enrolled all patients including relatively healthier ones and thus the effect of HR was diluted by such patients.^[4] Not only HR on admission, but high HR 24–36 h after admission was also associated with in-hospital mortality in patients with heart failure.^[12] Although we did not find any relationship between HR on admission and in-hospital mortality in patients in ICU, mortality was reported to decrease in such patients when HR was kept less than 100 bpm within the first admission day.^[13] Thus, to the best of our knowledge, there has been no report to suggest the existence of a relationship between HR and in-hospital mortality among the general patients; and our study is the first to address this important clinical issue.

The association between HR and mortality was well documented, but the reasons for this association were not well clarified.^[14] Several explanations were proposed such as low physical fitness, higher blood pressure, or reduced variability in HR, and diminished baroreceptor sensitivity in those with high HR.^[1,2,14] The immune or endocrine systems are impaired by the dysregulation of the autonomic nervous system in patients with chronic diseases, and HR variability is the indicator for the autonomic nervous system.^[15–18] However, these explanations could not fully account for the pathophysiology of higher mortality, especially, within the context of short-term effect, observed among inpatients.

Association between baseline HR and in-hospital mortality should shed light on the effective risk stratification of inpatient care, among patients with cardiovascular diseases and the general inpatients. Patients with such elevated HR and with no apparent reasons for the high HR should be closely investigated to determine the reasons, and be monitored to prevent the deterioration of underlying diseases. This strategy could be more effective in the general medical or surgical wards, due to the apparent trend of in-hospital mortality in our study. On the other hand, this approach might not be effective in ICU, because ICU patients are being closely monitored due to their serious conditions already. Another approach is the use of medications, which weakens the sympathetic nervous tone and decreases the HR. Such medications could be used as supplementary treatment to avoid fatality, in addition to therapy, targeting the underlying diseases. The effectiveness of beta-blockers in non-cardiac surgery patients remains controversial,^[19] however, supportive treatments in this direction, should be investigated.

Several limitations must be addressed in this study. First, there were a number of diseases with significantly varied severity, in this cohort; because we enrolled all admitted adult patients. Mortality was primarily associated with underlying diseases and their severity. Because it was unrealistic to adjust for all disease categories in the multivariate models, we adjusted for the surrogate markers of mortalities such as blood pressure or critical laboratory parameters. Second, the number of patients were relatively small, especially in the ICU; therefore, the relationship, other than the categories used, might not have been well scrutinized. In addition, there were no statistically significant associations with the ICU. Third, we utilized the HR on admission only. HR changes over time and the first measurement of HR on the admission day might not be a precise indicator. However, there were no distinct rules to determine which timing of HR measurement is best for use as an indicator; therefore, it was inevitable to use the first measurement to stratify the patients’ risk. Finally, the JADE study only enrolled Japanese patients; and this analysis utilized the data from just one hospital. To generalize and validate our results, it is necessary to conduct similar study with enough sample size in several settings.

5. Conclusion

We confirmed that higher HR was associated with higher in-hospital mortality among patients with general diseases in the medical and surgical wards. Even with less severe condition or outside ICU, HR should be directed attention to and patients with high HR on admission should be taken additional therapy to reduce the further risk of deterioration. Our findings should be attested to by further studies in other settings, or studies using interventional designs.

Acknowledgments

This study used part of the data from JADE Study with Decision Support. We are indebted to Dr. Shinji Kosaka, Dr. Kiyoshi Kikuchi, Dr. Tsukasa Nakamura, Mr. Eisaku Hirano, Mr. Kazuo Takeshita, Mr. Tomohiro Sonoyama, Mr. Shuichi Kurahashi, Mr. Koichiro Ichiki, and Mr. Akito Uchio for their management of JADE Study with Decision Support.

Author contributions

Conceptualization: Takeshi Morimoto.

Data curation: Yoshinori Ohta, Mio Sakuma, Jiro Takeuchi, Chisa Matsumoto, Takeshi Morimoto.

Formal analysis: Marumi Yamamoto, Jiro Takeuchi, Takeshi Morimoto.

Funding acquisition: Takeshi Morimoto.

Investigation: Marumi Yamamoto, Yoshinori Ohta, Jiro Takeuchi, Chisa Matsumoto, Takeshi Morimoto.

Methodology: Mio Sakuma, Takeshi Morimoto.

Project administration: Takeshi Morimoto.

Resources: Takeshi Morimoto.

Supervision: Takeshi Morimoto.

Validation: Yoshinori Ohta, Jiro Takeuchi, Takeshi Morimoto.

Visualization: Yoshinori Ohta, Jiro Takeuchi, Takeshi Morimoto.

Writing – original draft: Marumi Yamamoto, Takeshi Morimoto.

Writing − review & editing: Yoshinori Ohta, Mio Sakuma, Jiro Takeuchi, Chisa Matsumoto.

Takeshi Morimoto orcid: 0000-0002-6844-739X.