Exploring the association between happy hypoxia and Coronavirus disease 2019 in the triage phase

Abstract

Background

Patients with severe coronavirus disease 2019 (COVID-19) have been reported to show hypoxia without displaying typical clinical signs or symptoms, called “happy hypoxia.” To explore the potential of happy hypoxia as a distinctive symptom of COVID-19, we compared vital signs in the triage phase between patients with and without COVID-19.

Methods

We retrospectively identified emergency patients with and without COVID-19 admitted to Rakuwakai Marutamachi Hospital, Kyoto, Japan, between January 2021 and December 2021.

Results and Conclusions

317 patients were analyzed. Multivariate logistic regression analysis, including all vital signs, demonstrated that the respiratory rate was not statistically associated with COVID-19 (odds ratio, 0.94, p = 0.058), suggesting that happy hypoxia may not be a distinct hallmark of COVID-19.

PLAIN LANGUAGE SUMMARY

This study examined a phenomenon referred to as “happy hypoxia,” in which patients with severe coronavirus disease 2019 (COVID-19) exhibit low oxygen levels without the typical signs of distress, such as rapid breathing. Our objective was to determine whether happy hypoxia is a distinct symptom of COVID-19 compared to patients with other medical conditions.We analyzed data from 317 patients treated at a hospital in Kyoto, Japan, during 2021. Our findings revealed that the breathing rates of patients with COVID-19 were comparable to those of patients who did not have COVID-19, suggesting that happy hypoxia may not serve as a reliable indicator of COVID-19.In conclusion, this research highlights that relying solely on breathing rates during initial assessments in emergency situations may not effectively identify COVID-19. Understanding these findings can enhance how healthcare providers triage and manage patients.

SUMMARY POINTS

Patients with severe coronavirus disease 2019 (COVID-19) have occasionally exhibited hypoxia without displaying typical clinical signs or symptoms, called “happy hypoxia.” However, there is no evidence that has investigated the relationship between “happy hypoxia” and COVID-19 by analyzing vital signs in triage settings. This study is the first clinical research that demonstrated the respiratory rates during the triage phase in patients with COVID-19 were not markedly different from those in patients with other diseases. Happy hypoxia may not be a clear characteristic of COVID-19.

ARTICLE HIGHLIGHTS

• Happy hypoxia, characterized by severe hypoxia without an increase in respiratory rate, was investigated as a potential distinctive symptom of severe coronavirus disease 2019 (COVID-19).

• This study retrospectively analyzed 317 patients with respiratory failure, comparing those infected with COVID-19 to those with other conditions at Rakuwakai Marutamachi Hospital in 2021.

• Multivariate logistic regression analyses indicated that respiratory rates in patients with COVID-19 did not significantly differ from those of other patients.

• The inclusion of respiratory rates in the diagnostic models did not notably enhance COVID-19 diagnosis during the triage phase, as assessed through receiver operating characteristic analysis.

• No significant difference was observed in the correlation coefficient between the percutaneous oxygen/inspired oxygen (SpO2/FiO2) ratio and respiratory rate between COVID-19 and non-COVID-19 patients.

• The results indicate that happy hypoxia may not be a distinct hallmark of COVID-19, and that the respiratory rate has limited utility in triage decisions.

1. Introduction

Coronavirus disease 2019 (COVID-19) was initially documented in Wuhan, China, in 2019 [1]. Following the inception of the disease, the incidence of such cases has increased worldwide. Clinicians face the challenge of managing large cohorts of infected individuals, necessitating triage protocols primarily predicated on overall health status and vital sign assessment, particularly during pandemics [2]. Shortness of breath or an elevated respiratory rate with reduced arterial oxygen saturation, reflecting an increase in the work of breathing or symptoms of dyspnea, is a well-established physical finding in patients with acute respiratory distress syndrome, as observed in patients with COVID-19 [3].

However, “happy hypoxia,” characterized by a lack of an elevated respiratory rate despite severe hypoxia, was reported as a clinical feature of COVID-19 during the early phase of the pandemic [4,5]. Although no consensus exists regarding a precise definition, happy hypoxia is considered a clinical feature of COVID-19. The phenomenon of happy hypoxia poses significant challenges to triage and treatment strategies globally, as it can lead to rapid deterioration of the patient, although it is less severe in general appearance and may be an afterthought in diagnosis and treatment priorities at the triage stage [6,7]. Moreover, the Japanese Ministry of Health, Labor and Welfare highlights happy hypoxia and emphasizes caution in the recognition and management of the condition in their guidelines for COVID-19 in 2021. Potential explanations for happy hypoxia include reports suggesting that SARS-Cov-2 target receptors involved in oxygen chemosensitivity [8] and intrapulmonary shunting resulting from vascular injury [9]. This results in hyperventilation failing to sufficiently elevate arterial oxygen levels (PaO₂), despite leading to a marked reduction in arterial carbon dioxide (PaCO₂) due to the more efficient diffusion of carbon dioxide than that of oxygen.

In triage settings such as emergency rooms, arterial blood gas analysis is often unavailable, and physical examination findings are important for diagnosis. The absence of an elevated respiratory rate with severe hypoxia could be a diagnostic clue for the distinction between COVID-19 from other diseases, facilitating the prompt implementation of infection control measures, isolation, and treatment. However, limited evidence exists to definitively establish a correlation between COVID-19 and respiratory rate thus, further research is warranted to support this hypothesis. To elucidate the characteristics of happy hypoxia, a validation study analyzing vital signs collected at the initial hospital visit preceding admission is imperative. The primary objective of this study was to ascertain the association between COVID-19 and respiratory rate and investigate whether happy hypoxia is a characteristic symptom of COVID-19, providing valuable insights for triage decision-making.

2. Methods

2.1. Study design and patients

This was a single-center, cross-sectional study. We retrospectively identified outpatients with < 92% oxygen saturation in room air, which is indicative of respiratory failure [10,11], in the triage phase with and without COVID-19 admitted to Rakuwakai Marutamachi Hospital, Kyoto, Japan, between January 2021 and December 2021. Ethics approval for this study was provided by the Ethics Committee of Rakuwakai Marutamachi Hospital (ethical approval number: 02-23-00059), and the study was conducted in accordance with the guidelines of the Declaration of Helsinki. Informed consent was obtained in the form of opt-out on the web-site.

2.2. Measures and protocols

This research was conducted in a real-world setting, grounded in clinical practice, with data collected in the following environment. Upon arrival, the patients underwent an immediate assessment of vital signs, including body temperature, blood pressure, heart rate, pulse rate, respiratory rate, and oxygen saturation for patient triage, by a healthcare professional. If oxygen saturation was below 92% in ambient air, oxygen therapy was promptly initiated to maintain levels above 92%. The choice of oxygen delivery method (nasal cannula, oxygen mask, or non-rebreather mask) was determined by the attending physician. When not explicitly noted, the fraction of inspired oxygen (FiO₂) was deduced from the documented oxygen flow rate and the type of delivery device employed.

Following the initial assessment, the saturation of percutaneous oxygen (SpO₂)/FiO₂ ratio was calculated. Subsequently, the physician obtained a comprehensive medical history to ascertain the details of the patient’s medical background, current medications, alcohol consumption, and smoking habits. Concurrently, a physical examination was conducted, and appropriate diagnostic tests were performed to establish a primary diagnosis upon admission.

All admitted patients were tested for COVID-19 using nucleic acid amplification tests (Abbott ID-Now, nicking enzyme amplification reaction). The diagnosis of COVID-19 was confirmed if the nucleic acid amplification test yielded positive results.

2.3. Data collection and outcomes

For each patient, we retrospectively collected the data on vital signs including body temperature, blood pressure, heart rate or pulse rate, respiratory rate, and oxygen saturation at the time of the patient’s arrival at the hospital, along with the following patient background information: age, sex, past medical history (cardiovascular disease, cerebral vascular disorder, respiratory disease, hepatobiliary disease, diabetes mellitus, renal disease, collagen disease, mental disorder, and tumor), and primary diagnosis (bacterial pneumonia, viral pneumonia, exacerbation of chronic lung disease, other respiratory diseases, biliary disease, and pancreatitis, ischemic heart disease, left-sided heart disease, other heart diseases, malignancy, trauma, and others) based on medical records.

2.4. Statistical analysis

Continuous variables were summarized using mean and standard deviation, and categorical variables were summarized using frequencies and proportions. Using univariate and multivariate logistic regression analyses, we examined the association between the respiratory rate and the presence/absence of COVID-19 at the time of the patient’s arrival at the hospital as the primary outcome. The multivariate logistic model included the continuous variables on vital signs: body temperature (°C), systolic blood pressure (mmHg), heart rate (or pulse rate) (/min), SpO₂/FiO₂, and respiratory rate (/min). In these analyses, odds ratios (OR), 95% confidence intervals (CIs), and p-values were estimated for each variable. We also assessed the utility of respiratory rate in the triage phase as a criterion for determining whether to immediately perform a diagnostic test for COVID-19 using receiver operating characteristic (ROC) analysis. The area under the curve (AUC) of the ROC analyses was estimated for three logistic models, Model 1 included only the respiratory rate, Model 2 included all vital signs, and Model 3 included all vital signs except for the respiratory rate. Delong’s test was used to compare the AUCs of the two models [12]. To compare the performance of discrimination and calibration among the three models, bootstrap-based adjusted AUC, calibration-in-the-large, and calibration slope were estimated based on bootstrap sample sizes of 10,000, along with their 95% CIs [13]. Furthermore, subgroup analyses were conducted using univariate and multivariate logistic regression analyses to compare the respiratory rate between patients with COVID-19 and patients with five primary diagnoses other than COVID-19 (e.g., pneumonia, other lung diseases, left-sided heart disease, other heart disease, or others). Specifically, bacterial and viral pneumonia were combined as “pneumonia”; exacerbation of chronic lung disease and other respiratory diseases were combined as “other lung diseases”; ischemic heart disease and other heart diseases were combined as “other heart diseases”; and biliary disease and pancreatitis, malignancy, trauma, and other were combined as “others” (Table 4). Furthermore, to analyze the relationship between the SpO2/FiO2 ratio and respiratory rate, we performed a correlation analysis between the COVID-19 and non-COVID-19 groups, as well as between the COVID-19 group and each primary diagnosis group listed in Table 4. Statistical significance was set as p < 0.05. All statistical analyses were performed using the statistical software R version 4.2.2 (R Core Team, R Foundation for Statistical Computing, Vienna, Austria) and the logistf R package version 1.26.0 [14].

Table 4.

Comparison of respiratory rate between patients with COVID-19 and patients with primary diagnosis other than COVID-19 based on univariate and multivariate logistic regression analyses.

Subgroup	Primary diagnosis	Univariate			Multivariate under model 2
		N	OR (95%CI)	p-value	OR (95%CI)	p-value
Pneumonia	Bacterial pneumonia, viral pneumonia	122	0.92 (0.85, 1.00)	0.041	0.92 (0.85, 1.00)	0.064
Other lung diseases	Exacerbation of chronic lung disease, other respiratory diseases	31	0.87 (0.78, 0.96)	0.008	0.90 (0.80, 1.00)	0.060
Left-sided heart disease	Left-sided heart disease	86	0.88 (0.80, 0.96)	0.004	0.91 (0.81, 1.02)	0.107
Other heart diseases	Ischemic heart disease, other heart diseases	29	0.88 (0.79, 0.98)	0.021	0.96 (0.83, 1.09)	0.506
Others	Biliary disease and pancreatitis, malignancy, trauma, others	83	0.94 (0.88, 1.01)	0.090	0.94 (0.87, 1.02)	0.136

OR, odds ratio; CI, confidence interval; SpO₂, saturation of percutaneous oxygen; FiO₂, fraction of inspired air.

3. Results

A total of 330 patients were identified during the study period. After excluding 13 patients owing to missing vital sign data, our analysis focused on the remaining 317 individuals (32 with COVID-19 and 285 without COVID-19). Except for one patient, all participants were of Japanese ethnicity. Table 1 presents vital signs, along with background information regarding medical history and primary diagnosis upon admission, stratified by COVID-19 status. The average age of the patients in the COVID-19 group was 59 years, which was markedly less than that in non-COVID-19 group, with an average age of 88 years. Additionally, the proportion of male patients in the COVID-19 and non-COVID-19 groups was 63% and 43%, respectively. Histograms of the patients’ proportions for vital sign data in the COVID-19 and non-COVID-19 groups are displayed in Figure 1. The histograms of raw vital sign data in the two groups are shown in Figure S1.

Table 1.

Vital signs and background information of patients in COVID-19 and non-COVID-19 groups.

Variable	COVID-19 (n = 32)	Non-COVID-19 (n = 285)
Age, mean (SD)	59.1 (18.2)	87.9 (44.9)
Age, minimum (max)	23 (97)	18 (105)
Male, n (%)	20 (62.5)	122 (42.8)
Body temperature (°C), mean (SD)	37.6 (0.8)	37.2 (1.1)
Systolic blood pressure (mmHg), mean (SD)	128.7 (23.1)	144.3 (35.7)
Heart rate or pulse rate (/min), mean (SD)	93.8 (14.7)	98.0 (22.8)
Respiratory rate (/min), mean (SD)	21.8 (4.4)	24.7 (6.7)
SpO2 (%), mean (SD)	92.8 (3.8)	91.5 (6.8)
FiO2, mean (SD)	29.3 (10.2)	31.5 (12.6)
SpO2/FiO2, mean (SD)	3.4 (0.8)	3.3 (1.0)
Past medical history
Cardiovascular disease, n (%)	9 (28.1)	199 (69.8)
Cerebral vascular disorder, n (%)	2 (6.3)	61 (21.4)
Respiratory disease, n (%)	6 (18.8)	42 (14.7)
Hepatobiliary disorder, n (%)	5 (15.6)	23 (8.1)
Diabetes mellitus, n (%)	7 (21.9)	78 (27.6)
Renal disease, n (%)	3 (9.7)	45 (15.8)
Collagen disease, n(%)	2 (6.3)	31 (10.9)
Mental disorder, n (%)	3 (9.4)	109 (38.2)
Tumor, n (%)	3 (9.4)	42 (14.7)
Primary diagnosis
Bacterial pneumonia, n (%)	N/A	121 (42.5)
Viral pneumonia, n (%)	N/A	1 (0.4)
Exacerbation of chronic lung disease, n (%)	N/A	6 (2.1)
Other respiratory diseases, n (%)	N/A	25 (8.8)
Biliary disease and pancreatitis, n (%)	N/A	1 (0.4)
Ischemic heart disease, n (%)	N/A	18 (6.3)
Left-sided heart disease, n (%)	N/A	86 (30.2)
Other heart diseases, n (%)	N/A	11 (3.9)
Malignancy, n (%)	N/A	6 (2.1)
Trauma, n (%)	N/A	11 (3.9)
Other, n (%)	N/A	66 (23.2)

SD, standard deviation; SPO_2, percutaneous oxygen saturation; FiO₂, fraction of inspired oxygen.

Figure 1.

Histograms of patient proportion for vital sign data in the COVID-19 and non-COVID-19 groups.

According to Table 2, the statistically significant association between respiratory rate and the presence of COVID-19 was observed in the univariate logistic regression analysis (OR, 0.92, 95% CI, 0.86, 0.98, p = 0.016 in Model 1), but there was no significant association in the multivariate logistic regression analysis (OR, 0.94, 95% CI, 0.87, 1.00, p = 0.058 in Model 2). We additionally performed multivariate logistic regression analysis for Model 2, including age, and did not identify a significant impact of the respiratory rate on COVID-19 (OR, 0.975; 95% CI, 0.899–1.057; p = 0.054). However, the respiratory rate in non-COVID-19 groups was higher than that in the COVID-19 group. The discrimination performances between the COVID-19 and non-COVID-19 groups based on the AUC of the ROC curve in Models 1–3 were 0.64, 0.73, and 0.70, respectively (Figure 2) (Table 3). No statistically significant difference was observed in the AUCs between Models 2 and 3 (p = 0.248 by DeLong’s test). Moreover, the increase in discrimination performance by including respiratory rate was as small as -0.03 (95% CI, -0.07, 0.02). Similar results were also obtained when comparing the bootstrap-based adjusted AUC between the two models. The corrected calibration-in-the-large and calibration slopes were almost identical between the two models (Table 3). Table 4 shows the odds ratio of the respiratory rate in patients with COVID-19 relative to the patients with other primary diagnosis using univariate and multivariate logistic regression analyses. According to the results of multivariate logistic regressiona analysis shown in Table 4, there was was no statistically significant difference in the respiratory rate between the two groups.

Table 2.

Results of univariate and multivariate logistic regression analyses of the relationship between the respiratory rate and presence/absence of COVID-19 at the time of the patient’s arrival at the hospital, including all vital signs (Model 2), and the vital signs except for the respiratory rate (Model 3).

	Univariate		Multivariate
			Model 2		Model 3
	Or (95%CI)	p-value	Or (95%CI)	p-value	Or (95%CI)	p-value
Body temperature (°C)	1.46 (1.04, 2.05)	0.030	1.57 (1.10, 2.24)	0.014	1.58 (1.10, 2.27)	0.014
Systolic blood pressure (mmHg)	0.99 (0.97, 1.00)	0.017	0.99 (0.97, 1.00)	0.029	0.99 (0.97, 1.00)	0.018
Heart rate or pulse rate (/min)	0.99 (0.97, 1.01)	0.311	0.99 (0.97, 1.01)	0.357	0.99 (0.97, 1.01)	0.172
SpO2/FiO2	1.19 (0.79, 1.80)	0.393	1.06 (0.69, 1.62)	0.799	1.12 (0.73, 1.70)	0.605
Respiratory rate (/min)	0.92 (0.86, 0.98)	0.016	0.93 (0.87, 1.00)	0.058	N/A	N/A

OR, odds ratio; CI, confidence interval, SpO₂, saturation of percutaneous oxygen; FiO₂, fraction of inspired air.

Figure 2.

Receiver operating characteristic curve for each model.

Table 3.

Performances on discrimination and calibration among the three logistic models including only the respiratory rate (Model 1), all vital signs (Model 2), and the vital signs except for the respiratory rate (Model 3).

	Model 1	Model 2	Model 3
AUC	0.64 (0.55, 0.73)	0.723 (0.65, 0.81)	0.70 (0.62, 0.79)
corrected AUC*	0.64 (0.61, 0.73)	0.70 (0.69, 0.79)	0.67 (0.66, 0.77)
difference AUC ref. to Model 1		0.09 (0.00, 0.17)	0.06 (-0.06, 0.18)
difference AUC ref. to Model 2			−0.03 (-0.07, 0.02)
calibration in-the-large	0.00 (-0.37, 0.37)	−0.08 (-0.45, 0.30)	−0.06 (-0.44, 0.31)
corrected calibration in-the-large*	0.01 (0.01, 0.01)	−0.08 (-0.09, −0.06)	−0.06 (-0.07, −0.05)
calibration slope	1.00 (0.83, 1.17)	1.04 (0.85, 1.23)	1.04 (0.85, 1.23)
corrected calibration slope*	1.00 (1.00, 1.00)	1.01 (1.01, 1.02)	1.01 (1.01, 1.02)

OR, odds ratio; CI, confidence interval; AUC, the area under the curve.

^*Optimism was corrected by the bootstrap method (10,000 bootstrap samples), and the confidence interval was calculated by the location-shifted bootstrap confidence interval.

In the correlation analysis between SpO2/FiO2 ratio and respiratory rate, the correlation coefficients for the COVID-19 group and non-COVID-19 group were -0.11 (95% CI, -0.44, 0.25) and -0.20 (95% CI, -0.31, -0.09), respectively, with no significant difference was observed (p = 0.636) (Figure 3). Similarly, the correlation coefficients between COVID-19 group and each primary diagnosis groups listed in Table 4 showed no significant differences (Pneumonia: p = 0.932, Other lung diseases: p = 0.450, Left-sided heart disease: p = 0.467, Other heart diseases: p = 0.468, Others: p = 0.703) (Figure 4).

Figure 3.

Results of correlation analysis between SpO2/FiO2 ratio and respiratory rate in COVID-19 and non-COVID-19 groups.

Figure 4.

Results of correlation analysis between SpO2/FiO2 ratio and respiratory rate in patients with COVID-19 and those with primary diagnosis other than COVID-19.

4. Discussion

To the best of our knowledge, this study is the first clinical research that investigated the relationship between “happy hypoxia” and COVID-19, analyzing the vital signs in triage settings. Despite the limitations posed by the heterogeneity of the population and the small sample size, this research is a single-center, cross-sectional study encompassing all emergency patients requiring hospitalization due to respiratory failure during the pandemic. We contend that conducting and interpreting the results of a study grounded in real-world clinical practice can be more beneficial for the field of emergency medicine since it complements the evidence obtained from studies conducted in idealized environments with homogeneous populations. In this study, we focused on the utility of respiratory rate in the triage phase as a criterion for determining whether to immediately perform a diagnostic test for COVID-19. However, multivariate logistic regression analysis showed no statistically significant difference in terms of the respiratory rate between patients in the COVID-19 and non-COVID-19 groups. A similar tendency was also observed in the subgroup analyses comparing the respiratory rate between the patients with COVID-19 and patients with five other primary diagnoses. These results indicate that happy hypoxia is not considered a diagnostic marker for COVID-19.

As displayed in Table 1, differences in the distributions of age and sex between the groups were observed; therefore, we additionally performed multivariate logistic regression analysis for Model 2, including age, and identified no significant impact of respiratory rate on COVID-19. In addition, a significant difference was observed in body temperature between the two groups (Table 1); however, this discrepancy may stem from the uncertainty surrounding the use of antipyretics before the patients arrived at the hospital. We also discovered that the contribution of respiratory rate to the discrimination and calibration performances for COVID-19 was negligibly small, based on the results of Table 3 and Figure 2, suggesting that the utility of respiratory rate in the triage phase as a criterion in determining whether to immediately perform a diagnostic test for COVID-19 would be limited.

In addition, to analyze the relationship between the degree of hypoxia and the work of breathing, we conducted a correlation analysis between SpO2/FiO2 ratio and respiratory rate. No significant differences were found in the correlation coefficients between the COVID-19 and non-COVID-19 groups, nor between the COVID-19 group and each primary diagnosis group listed in Table 4 (Figures 3 and 4). In COVID-19, the respiratory rate and work of breathing have emerged as useful indicators for predicting the risk of intubation [15]. Validation scores for assessing the severity of COVID-19, such as the 4 C score [16], quick COVID Severity Index [17], and NEWS2 [18], incorporate the respiratory rate as a component. A high respiratory rate typically indicates a worse condition in these scoring systems. In harmony with these assessment tools, our study implies that the importance of assessing the work of breathing with respiratory rate seems to be equally important in the triage phase.

Our findings are consistent with those of a recent report from the United Kingdom suggesting that happy hypoxia may not be a distinctive feature of COVID-19 [19]. This report also considers post-hospitalization vital signs, which could potentially obscure the original COVID-19 condition due to the subsequent treatment received after hospitalization. Our findings are purely from the triage phase, and imply that happy hypoxia is not a distinctive hallmark of COVID-19.

5. Limitations

This study had some limitations. As a single-center, cross-sectional study conducted during the pandemic, the analysis was inevitably limited to a small sample size with variations in patient backgrounds between the two groups. In addition, the variant of COVID-19 in this study was not identified microbiologically using patient specimens, although the patients included in this study were infected with COVID-19 during the delta variant epidemic. The impact of respiratory rate on the presence of COVID-19 may vary depending on the variant. Information regarding vaccination history for COVID-19 was not utilized in the analyses.

6. Conclusion

The respiratory rates during the triage phase in patients with COVID-19 were not markedly different from those in patients with other diseases, suggesting that happy hypoxia may not be a distinct hallmark of COVID-19.

Funding Statement

This paper was not funded.

Authors’ contributions

Mihiro Kaga (Literature search/Data collection/Study design/Manuscript preparation). Masanao Sasaki (Analysis of data). Takahiro Masuda (Literature search/Manuscript preparation/Review of manuscript). Hiroyuki Sato (Review of manuscript). Takeshi Ueda (Review of manuscript). Akihiro Hirakawa (Study design/Analysis of data/Manuscript preparation/Review of manuscript).

Disclosure statement

Ethical disclosure

Ethics approval for this study was provided by the Ethics Committee of Rakuwakai Marutamachi Hospital (ethical approval number: 02-23-00059), and the study was conducted in accordance with the guidelines of the Declaration of Helsinki. Informed consent was obtained in the form of opt-out on the web-site.

Financial disclosure

The authors have no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, con­sultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.