Predictors of invasive mechanical ventilation requirement and mortality in hypercapnic respiratory failure: A retrospective analysis

Abstract

INTRODUCTION: While hypercapnic respiratory failure (HRF) has been widely studied in emergency department (ED) populations, the prognostic significance of metabolic compensation and lactate levels remains unclear. This study aimed to evaluate the association of creatinine, bicarbonate (HCO₃), chloride, and lactate levels with the need for invasive mechanical ventilation (IMV) and inhospital mortality in patients with HRF.

METHODS: This single center, cross sectional study included adult patients who presented to the ED with respiratory distress and arterial partial pressure of carbon dioxide >50 mmHg between 2020 and 2023. The relationships between initial laboratory parameters and clinical outcomes were analyzed.

RESULTS: A total of 420 patients were included (median age: 77 years [interquartile range: 68–85]; 51.4% female). The mortality rate among bedridden patients was 55.9% and 34.3% among patients diagnosed with pneumonia. Creatinine ≥1.6 mg/dL was associated with inhospital mortality (odds ratio [OR]: 3.7; 95% confidence interval [CI]: 2.044–6.696) and IMV requirement (OR: 2.323; 95% CI: 1.216–4.439). Lactate >1.5 mmol/L was also associated with higher mortality (OR: 10.441; 95% CI: 5.739–18.996). Delta HCO₃ (ΔHCO₃) <−7.5 mEq/L predicted mortality (OR: 2.965; 95% CI: 1.756–5.008) and IMV need (OR: 10.181; 95% CI: 5.709–18.156). Low ΔHCO₃ , hemoglobin, pH levels, elevated lactate, creatinine levels, and immobility are the independent risk factors for mortality. The AUC values of lactate levels for predicting mortality were higher than those of pH (0.662 vs. 0.655).

CONCLUSIONS: In patients with HRF, ΔHCO₃ , hemoglobin, pH, lactate and creatinine levels, and immobility are strong predictors of poor outcomes. Lactate is a robust and independent predictor of poor outcomes, with prognostic accuracy comparable to that of pH, and may be valuable for clinical risk stratification.

Introduction

Hypercapnic respiratory failure (HRF) is one of the common causes of emergency department (ED) visits and is characterized by an arterial partial pressure of carbon dioxide (PaCO₂) exceeding 45–50 mmHg, accompanied by respiratory distress. Patients presenting with HRF often have a high risk of requiring invasive mechanical ventilation (IMV) and are associated with significant mortality.

Although numerous studies have investigated prognostic indicators in patients with HRF, the relationship between the effectiveness of metabolic compensation and clinical outcomes remains insufficiently elucidated. In patients with HRF, metabolic compensation is expected to be activated in response to respiratory acidosis resulting from elevated CO₂ levels. In the absence of renal dysfunction, these compensatory responses typically include increased bicarbonate (HCO₃) levels and decreased chloride (Cl) levels.[1] The prognostic significance of the delta HCO₃ (ΔHCO₃, difference between the actual HCO₃ level and expected HCO₃ level) and Cl levels has not yet been clearly established.[1-5] Although elevated lactate levels have been shown to be a poor prognostic factor in metabolic disturbances such as sepsis, acute renal failure, and hemorrhage, there is a lack of sufficient research in patients with HRF.[6,7]

The impact of the effectiveness of metabolic compensation on adverse outcomes in HRF patients represents a gap in the current literature. We aimed to identify the factors associated with inhospital mortality and the need for IMV in patients with HRF presenting to the ED, with particular focus on metabolic compensation parameters, including delta HCO₃ , creatinine, Cl, hemoglobin, and initial lactate levels.

Material and Methods

This retrospective, cross sectional study was conducted at the ED of Dokuz Eylül University Hospital. Adult patients presenting to the ED with HRF between March 2020 and March 2023 were included. Eligibility criteria required patients to be aged 18 years or older and to have an arterial blood gas (ABG) analysis showing a PaCO₂ ≥50 mmHg within the first hour of ED admission.

Exclusion criteria were as follows: (1) hypercapnia due to central nervous system (CNS) depressant drug intoxication; (2) hypercapnia secondary to CNS or neuromuscular diseases; (3) trauma related hypercapnia; (4) presentation following out of hospital cardiac arrest; (5) inaccessible outcome data; and(6) patients whose blood gas analyses were performed using venous samples.

The primary outcome of the study was inhospital mortality, and the secondary outcome was the need for IMV during the ED stay. Eligible patients were identified using the hospital’s electronic health record system based on PaCO₂ levels. The study team reviewed patients’ electronic medical records to collect data on demographics, comorbidities, vital signs at presentation, ABG parameters, other laboratory results, final diagnoses, and outcome data. All data were recorded in a standardized data collection form.

The expected HCO₃ levels were calculated based on the classification of hypercapnia as either acute or chronic. For acute hypercapnia, the expected HCO₃ was calculated as: 24 mEq/L + (1 mEq/L × [PaCO₂ −40]/10). For chronic hypercapnia, the expected HCO₃ was calculated as: 24 mEq/L + (4 mEq/L × [PaCO₂ −40]/10). The difference between the expected and measured HCO₃ levels was defined as ΔHCO₃. The expected HCO₃ value was calculated based on the 2021 Guideline for the Management of Chronic Obstructive Pulmonary Disease Exacerbations developed by the Emergency Medicine Association of Turkey/ Turkish Thoracic Society Clinical Practice Guideline Task Force, as well as a previous study conducted by Marcy et al. [8,9] The final diagnosis and classification of hypercapnia as acute or chronic were determined by two members of the study team(SÖ and AT) based on patients’ medical histories, previous hospital admission data, and blood gas analysis reports. In case of disagreement, a third physician (EA) adjudicated to reach a consensus.

Ethical approval for our study was obtained from the Ethics Committee of Dokuz Eylul University (Number: 7982, Date: December 4, 2023).

Analysis

SPSS 29.0 (IBM® Corporation, Armonk, New York, United States) was used for data analysis. Descriptive statistics were presented as numbers and percentages for categorical variables. Differences in proportions of categorical variables between independent groups were analyzed using the Chi square test or Fisher’s exact test.

The normality of numerical variables was assessed using the Kolmogorov–Smirnov test. Numerical variables were expressed as mean and standard deviation if they followed a normal distribution, whereas those not following a normal distribution were presented as median and interquartile range (IQR). For comparisons of numerical variables between independent groups, the Student’s t test was used if the data were normally distributed, and the Mann–Whitney U test was used if they were not.

Odds ratios (ORs) were calculated to evaluate the associations of pH, lactate, blood urea nitrogen (BUN), Cl, hemoglobin, HCO₃, ΔHCO₃, and platelet levels with the requirement for IMV and inhospital mortality. Receiver operating characteristic (ROC) analyses were conducted for pH, lactate, creatinine, HCO₃, and ΔHCO₃ to evaluate their predictive value for the requirement of IMV and inhospital mortality, with area under the curve (AUC) values reported. Univariate analyses were performed on the variables sex, platelet count, ΔHCO₃, HCO₃, hemoglobin, pH, lactate, creatinine, and immobility. Variables with P < 0.2 in the univariate analyses were subsequently included in a logistic regression model. The model’s goodness of fit was assessed using the Hosmer– Lemeshow test, which yielded P = 0.722, indicating an adequate fit.

A confidence level of 95% was applied, and P < 0.05 was considered statistically significant.

Results

During the study period, a total of 1086 patients with a PaCO₂ level ≥50 mmHg were identified. Among them, 666 patients were excluded; as a result, 420 patients were included in the final analysis [Figure 1].

Figure 1. Patient flow diagram. PaCO₂: Partial pressure of carbon dioxide, IMV: Invasive mechanical ventilation

The median age of the patients was 77 years (IQR: 67–84), and 216 (51.4%) patients were female. The most common final diagnoses were decompensated heart failure (n = 127, 30.2%), pneumonia (n = 105, 25%), and acute exacerbation of COPD (AECOPD) (n = 91, 21.7%). Inhospital mortality was observed in 82 patients (19.5%), and IMV was required in 67 patients (16%) during their ED stay. A total of 288 patients (68.6%) were admitted to the intensive care unit (ICU). Non IMV (NIMV) was applied to 212 of our patients (50.5%).

Table 1 summarizes the relationships between patients’ demographic characteristics, comorbidities, vital signs at ED admission, final diagnoses, and clinical outcomes. Mortality and IMV rates were significantly lower among patients with COPD, chronic hypercapnia, or those using home oxygen therapy or noninvasive ventilation. In contrast, bedridden patients and those with active malignancy had significantly poorer outcomes. The prognosis was more favorable in patients with AECOPD or heart failure compared with those with pneumonia. Table 2 compares ABG parameters and laboratory markers of metabolic compensation with patient outcomes. Elevated lactate, creatinine, and BUN, as well as lower HCO₃ and hemoglobin levels, were significantly associated with both increased mortality and higher IMV requirement.

Table 1. The relationship between patients’ characteristics, comorbidities, and vital signs on admission with inhospital mortality and invasive mechanical ventilation requirement in the emergency department

Table 2. The relationship between patients’ blood gas analysis, laboratory findings, and clinical outcomes

Table 3 presents the optimal cutoff values of pH, lactate, creatinine, platelet count, HCO₃, ΔHCO₃, hemoglobin, and Cl levels for predicting inhospital mortality and the need for IMV. AUC values for pH, lactate, creatinine, ΔHCO₃, and HCO₃ in predicting mortality and the need for IMV are shown in Table 4. ROC curves for the need for IMV and inhospital mortality are shown in Figures 2 and 3.

Table 3. Optimal cutoff values of the predictors for mortality and invasive mechanical ventilation

Table 4. Area under the curve values for pH, lactate, creatinine, ΔHCO₃ , and HCO₃ in predicting mortality and the need for invasive mechanical ventilation

Figure 2. Receiver operating characteristic curve for invasive mechanical ventilation. ROC: Receiver operating characteristic, HCO₃: Bicarbonate

Figure 3. Receiver operating characteristic curve for inhospital mortality. ROC: Receiver operating characteristic, HCO₃: Bicarbonate

Results of univariate and multivariate analyses evaluating major clinical variables and laboratory markers of metabolic compensation as predictors of inhospital mortality are shown in Table 5. Ten variables were included in the univariate analysis, of which eight were entered into the logistic regression model. Multivariate analysis demonstrated that ΔHCO₃, hemoglobin, pH, lactate, creatinine, and immobility were independent predictors of inhospital mortality.

Table 5. Univariate and multivariate analyses of predictors of inhospital mortality

Discussion

This study aimed to investigate the risk factors associated with the need for IMV and inhospital mortality in patients presenting to the ED with HRF. Particular attention was given to the role of metabolic compensation, which has not been extensively explored in previous literature. In addition, the prognostic value of lactate levels was assessed, given their potential association with disease severity and clinical outcomes in HRF.

Respiratory acidosis, whether compensated or uncompensated, is a hallmark of HRF. It is reasonable to expect a worse prognosis in patients with impaired renal function, where metabolic (renal) compensation is insufficient. In the largest study to date involving 1768 patients with AECOPD, the presence of acute kidney injury (AKI) was associated with significantly higher rates of mechanical ventilation (both noninvasive and invasive), ICU admission, and inhospital mortality. Multivariable analysis showed that Stage 1, 2, and 3 AKI were associated with 1.9 , 2.1 , and 6.0 fold increased risks of inhospital mortality, respectively.[10] We also observed that patients with creatinine levels ≥2 mg/dL had a 4.7 fold higher risk of inhospital mortality and a 3.3 fold higher likelihood of requiring IMV. Similarly, Ucgun et al. found that elevated creatinine levels were an independent risk factor for mortality in ICU admitted AECOPD patients.[4] In that study, patients with low HCO₃ levels (<20 mEq/L) had a mortality rate of 59%, whereas those with higher levels (>28 mEq/L) had a rate of 19%. Similarly, HCO₃ levels < 22 mEq/L were associated with a 2.4 fold increased risk of inhospital mortality and a 6.6 fold increased risk of requiring IMV. These results suggest that kidney function should be taken into account when predicting the prognosis of patients with HRF. It should be noted that patients with low HCO₃ levels and elevated creatinine values are more likely to experience a complicated clinical course. To evaluate the adequacy of metabolic compensation in respiratory acidosis, we calculated the expected HCO₃⁻ levels based on established formulas for acute and chronic hypercapnia. We then determined the ΔHCO₃ (measured minus expected). Negative ΔHCO₃ values indicated insufficient HCO₃ compensation. According to our literature review, only one study has investigated ΔHCO₃ levels in patients with HRF. This study examined 498 patients with AECOPD who required respiratory support. In patients with Stage I and II AKI (AKIN), the actual HCO₃ levels exceeded the expected values, whereas in patients with Stage III AKIN or those requiring hemodialysis, the actual HCO₃ levels remained below the expected levels. However, this study did not investigate the relationship between ΔHCO₃ levels and poor clinical outcomes.[9]

We identified the optimal discriminative cutoff value for ΔHCO₃ as −7.5 mEq/L. Patients with values below this threshold had approximately a 2.9 fold increased risk of requiring IMV and a 10 fold higher mortality rate. To the best of our knowledge, this is the first study to demonstrate an association between ΔHCO₃ levels and both the need for IMV and mortality.

Cl⁻ may also play a compensatory role in respiratory acidosis. An increase in Cl levels alone can cause a normal anion gap metabolic acidosis. Theoretically, in response to hypercapnia, the kidneys increase HCO₃⁻ reabsorption and production, leading to a decrease in serum Cl levels.[1] Therefore, in patients with respiratory acidosis who exhibit effective renal compensation, lower or normal Cl levels are expected. Terzano et al. investigated the factors influencing the need for NIMV in 68 hospitalized patients with HRF.[11] Contrary to expectations, the Cl levels in patients who required NIMV were lower than those in patients who did not (95 mmol/L vs. 100.2 mmol/L, P < 0.001). Among our patients, those with Cl levels within the range of 98–107 mEq/L had lower mortality and a reduced need for IMV compared to patients with hyperchloremia or hypochloremia. Specifically, in hyperchloremic patients, the OR for mortality was 3.5, and for IMV, it was 4.2.

In patients with chronic HRF, the presence of coexisting hypoxemia is common. Therefore, an increase in hemoglobin levels is expected to ensure adequate oxygen delivery to the tissues. However, since the underlying cause of chronic HRF is often a chronic disease, a significant proportion of patients may not achieve the expected polycythemic response due to anemia of chronic disease.[12] A study conducted on 300 hospitalized AECOPD patients found that 37% were anemic.[13] They reported that the mean survival time of anemic patients (defined as hemoglobin levels <13 g/dL in men and <12 g/dL in women) was 31 months (95% CI: 27.7–34.3), whereas the mean survival time of nonanemic patients was 41 months. Our study population included patients with pneumonia and congestive heart failure in addition to those with AECOPD. We demonstrated that anemia (hemoglobin levels below 9 g/dL) is an independent risk factor for mortality and the need for IMV.

Several studies have investigated the use of lactate levels as a prognostic marker in patients with HRF. Terzano et al. reported significantly higher admission lactate levels in patients requiring NIMV (3.1 vs. 0.7 mmol/L).[11] Durmuş et al. examined whether lactate clearance could help determine the need for hospitalization in patients presenting to the ED with AECOPD. Admission lactate levels were similar between groups, but follow up showed 11.8% clearance in hospitalized versus 14.7% in discharged patients.[14] Kasapoğlu et al. investigated the association between lactate levels and NIMV failure in patients with acute HRF. In this study, a lactate level >2.1 mmol/L was found to have an AUC of 0.680 (95% confidence interval [CI]: 0.578–0.791), with a sensitivity of 53.49% (95% CI: 38.6–68.4) and a specificity of 87.67% (95% CI: 80.1–95.2) for predicting NIMV failure.[15] In our study, elevated lactate levels at ED admission were identified as an independent risk factor for both the need for IMV and mortality. In patients with a lactate level >1.5 mmol/L, the ORs for mortality and IMV were 2.291 and 10.441, respectively.

pH is one of the most used laboratory parameters in clinical practice for predicting mortality and determining the need for NIMV in patients with HRF. According to our data, the lactate level measured at the time of ED admission is also a prognostically valuable parameter, at least as significant as pH. Furthermore, lactate levels demonstrated a higher AUC for predicting mortality compared with pH. This underscores the potential utility of lactate as a rapid and readily available marker for early risk stratification in emergency settings. While arterial pH is a well established prognostic indicator in respiratory failure, the prognostic value of lactate in this specific context has been less thoroughly explored. Given the comparable OR observed in our study, we propose that lactate levels should be incorporated into clinical assessment algorithms alongside traditional blood gas parameters when evaluating patients with HRF.

Most previous studies on HRF have focused on ICU patients with AECOPD. However, Chung et al. showed that lower respiratory tract infections and congestive heart failure were also common causes of HRF in ED settings. In their study, mortality was lower in patients with COPD (OR: 0.59) and higher in those with lower respiratory tract infections (OR: 1.68).[5] In our cohort, AECOPD accounted for only 21.7% of HRF cases, and mortality in this group was just 3.3%, compared to 34.3% in patients with pneumonia and 7.9% in those with decompensated heart failure. The overall mortality rate among all patients was 19.5%. Among those using home oxygen therapy, mortality was 12.1%; in patients with chronic hypercapnia, it was 13.9%, and among those using home NIMV, it was 9.1%. These findings suggest that the prognosis of patients with isolated AECOPD without concomitant pneumonia or decompensated heart failure is significantly better than that of other HRF patients. It should be considered that patients with chronic hypercapnia due to COPD, who are already on home oxygen and NIMV therapy, tolerate hypercapnia relatively well. In contrast, hypercapnia that develops in the context of pneumonia or congestive heart failure appears to be a significantly poorer prognostic indicator.

Although this was not among our primary or secondary objectives, we found significantly higher mortality in bedridden patients. The need for NIMV (47.1%) and mortality rate (55.9%) in bedridden individuals were significantly higher compared to patients with comorbidities such as diabetes, active malignancy, or coronary artery disease. These findings indicate that the prognosis of HRF in bedridden patients is particularly poor.

Limitations

The primary limitation of our study is its single center design and the retrospective nature of data collection from patient records, which may affect the generalizability of the findings and introduce potential information bias. Interrater reliability analysis was not conducted to assess the consistency between physicians’ decisions regarding the final diagnosis and classification of hypercapnia as acute or chronic. The classification of hypercapnia as acute versus chronic is somewhat subjective and may also introduce bias. Furthermore, in cases where patients developed acute HRF, it is possible that metabolic compensation had not yet been fully established, potentially influencing some of the biochemical parameters analyzed.

Another important limitation is that due to high ICU bed occupancy, many patients received IMV or NIMV in the ED. This may have led to an underestimation of the total IMV requirement, as patients intubated after transfer might not have been captured. The lack of an external validation cohort and potential changes in clinical management over the 3 year study period should be considered. Since a power analysis was not conducted in our study, it is not known whether the sample size was adequate.

Due to the retrospective nature of the study, we were unable to standardize the indications for initiating IMV. Physicians’ clinical experience and the individual patient’s condition may have influenced the decision to start IMV, potentially introducing bias to the primary outcome.

Conclusion

In patients presenting to the ED with HRF, metabolic compensation status and initial lactate levels appear to be important prognostic indicators for both the need for IMV and inhospital mortality. Worse outcomes were associated with lower HCO₃ levels, elevated creatinine and lactate levels, and anemia. In addition, bedridden status and a final diagnosis of pneumonia were associated with significantly higher mortality rates and IMV requirements. Low ΔHCO₃ , hemoglobin, pH levels, elevated lactate and creatinine levels, and immobility are the independent risk factors for mortality. These findings highlight the importance of incorporating easily obtainable clinical and laboratory parameters, such as lactate, HCO₃ , ΔHCO₃, and hemoglobin levels, into early risk stratification models for HRF patients.

How to cite this article: Özarslan S, Aksay E, Saray S, Tarhan A. Predictors of invasive mechanical ventilation requirement and mortality in hypercapnic respiratory failure: A retrospective analysis. Turk J Emerg Med 2026;26:234-41.

The study was approved by the ethics committee of Dokuz Eylül University (Number: 7982, Date: December 4, 2023).

ŞÖ (Conceptualization, Data curation, Methodology, Project administration, Supervision, Visualization, Writing – original draft, Writing – review and editing). EA (Conceptualization, Methodology, Project administration, Supervision, Visualization, Formal analysis, Writing – original draft, Writing – review and editing). SS (Conceptualization, Data curation, Supervision, Writing – original draft). AT (Conceptualization, Data curation, Supervision, Writing – original draft).

None Declared.

None.

References

1. Palmer BF, Clegg DJ. Respiratory acidosis and respiratory alkalosis: Core curriculum 2023. Am J Kidney Dis 2023;82:347 59.
2. Chen D, Jiang L, Li J, Tan Y, Ma M, Cao C, et al. Interaction of acute respiratory failure and acute kidney injury on in hospital mortality of patients with acute exacerbation COPD. Int J Chron Obstruct Pulmon Dis 2021;16:3309 16.
3. Chen L, Chen L, Zheng H, Wu S, Wang S. The association of blood urea nitrogen levels upon emergency admission with mortality in acute exacerbation of chronic obstructive pulmonary disease. Chron Respir Dis 2021;18:14799731211060051.
4. Ucgun I, Oztuna F, Dagli CE, Yildirim H, Bal C. Relationship of metabolic alkalosis, azotemia and morbidity in patients with chronic obstructive pulmonary disease and hypercapnia. Respiration 2008;76:270 4.
5. ChungY, Garden FL, Marks GB, Vedam H. Causes of hypercapnic respiratory failure and associated in hospital mortality. Respirology 2023;28:176 82.
6. Aliustaoglu Bayar AE, Aksay E, Oray NC. Lactate measurements accurately predicts 1 week mortality in emergency department patients with acute kidney injury. Turk J Emerg Med 2019;19:136 40.
7. Kaçar AA, Aksay E, Bayram B, Kıran E, Güldalı BE. Identifying high risk undifferentiated emergency department patients with hyperlactatemia: Predictors of 30 day in hospital mortality. Turk J Emerg Med 2024;24:158 64.
8. Doğan NÖ, Varol Y, Köktürk N, Aksay E, Alpaydın AÖ, Çorbacıoğlu ŞK, et al. 2021 guideline for the management of COPD exacerbations: Emergency medicine association of Turkey (EMAT)/Turkish Thoracic Society (TTS) clinical practice guideline task force. Turk J Emerg Med 2021;21:137 76.
9. Marcy F, Goettfried K, Enghard P, Piper SK, Kunz JV, Schroeder T. Impact of AKI on metabolic compensation for respiratory acidosis in ICU patients with AECOPD. J Crit Care 2024;83:154846.
10. Wan X, Chen D, Tan Y, Ma M, Zhang F, Liu Z, et al. Incidence, risk factors, and prognostic implications of acute kidney injury in patients with acute exacerbation of COPD. Int J Chron Obstruct Pulmon Dis 2020;15:1085 92.
11. Terzano C, Di Stefano F, Conti V, Di Nicola M, Paone G, Petroianni A, et al. Mixed acid base disorders, hydroelectrolyte imbalance and lactate production in hypercapnic respiratory failure: The role of noninvasive ventilation. PLoS One 2012;7:e35245.
12. Alisamir M, Ebrahimi M, Rahim F. Anemia in chronic obstructive pulmonary disease: A systematic review. Respir Investig 2022;60:510 21.
13. Cireli E, Mertoğlu A. The impact of anemia on the mortality of COPD patients hospitalized for acute exacerbation resulting in respiratory failure. Monaldi Arch Chest Dis 2022;93:2254. [doi: 10.4081/monaldi. 2022.2254].
14. Durmuş U, Doğan NÖ, Pekdemir M, Yılmaz S, Yaka E, Karadaş A, et al. The value of lactate clearance in admission decisions of patients with acute exacerbation of COPD. Am J Emerg Med 2018;36:972 6.
15. Kasapoglu U, Gunay B, Unlu C, Cinar C, Arikan H, Yalcinkaya E, et al. Lactate albumin ratio: A novel predictor of noninvasive mechanical ventilation failure in acute hypercapnic respiratory failure. Cureus 2025;17:e83314.