Predictive value of venous bicarbonate levels for survival to hospital discharge in out-of-hospital cardiac arrest patients

Abstract

Background

Acid-base disturbances significantly impact cardiac function and prognostic outcomes in cardiac arrest. Previous studies have highlighted the correlation between pH levels from arterial blood gas (ABG) analyses during cardiopulmonary resuscitation (CPR) in out-of-hospital cardiac arrest (OHCA) and survival outcomes. However, ABG measurements are often impractical in resource-limited settings. This study explores the relationship between serum bicarbonate levels and survival outcomes in patients with OHCA.

Methods

This retrospective cohort study examined patients with OHCA who presented at Srinagarind Hospital (Thailand) between 2015 and 2021. We analyzed venous bicarbonate levels and other laboratory markers (Na+, K+, BUN, Creatinine). Demographic and clinical data were extracted from electronic medical records. The primary objective was to assess the association between venous bicarbonate levels and survival and to determine the optimal cutoff values for predicting survival in these patients.

Results

Of the 461 identified patients, 19% survived hospital discharge. Survivors exhibited higher bicarbonate and BUN levels but lower potassium levels. Bicarbonate levels ≥ 12.6 demonstrated a sensitivity of 74% and specificity of 47%, with an 88.44% negative predictive value (NPV) for survival. A sensitivity analysis, which reclassified patients who left against medical advice as non-survivors, revealed that a bicarbonate cutoff of 13.9 mmol/L yielded the best predictive value, with a sensitivity of 93.8% and a specificity of 52.1%. Factors associated with increased survival included BUN ≥ 19.5, bicarbonate ≥ 12.6, private transport, and initial PEA or VT/VF rhythms, while potassium ≥ 5.1 decreased survival likelihood.

Conclusion

Bicarbonate levels, particularly with a threshold greater than 12.6 mmol/L, may be effective prognostic indicators. Other factors influencing survival include BUN, potassium levels, private transport, and initial cardiac rhythm. These insights can help clinicians improve resuscitation strategies and prognosis assessment, especially in resource-limited settings.

Introduction

Acid-base disturbances are common and can be complex during cardiac arrest [1,2,3]. Following a cardiac arrest, lactate and other unmeasured anions build up within the cells when circulation stops, exacerbating metabolic acidosis [4]. This build-up increases hydrogen ion concentrations, depleting bicarbonate (HCO3-), a critical buffering agent, and lowering the pH [5]. These changes notably impair cardiac function by reducing contractility and cardiac output, causing arterial vasodilation, leading to hypotension, and reducing the effectiveness of catecholamine infusions [4]. These effects significantly impact prognostic outcomes, affecting both recovery and survival in patients with cardiac arrest [5].

Multiple studies have explored the role of acidosis in predicting treatment outcomes in out-of-hospital cardiac arrest (OHCA), focusing mainly on pH levels obtained from arterial blood gas (ABG) analyses [6,7,8]. These studies have shown that pH levels, as determined by ABG during cardiac arrest, are independent factors associated with survival to hospital discharge and neurological recovery [6, 7]. In contrast, one study identified pCO₂ as an independent predictor of sustained return of spontaneous circulation (ROSC) [8]. However, the effectiveness of ABG analyses depends on the availability of point-of-care ABG machines, which may not be accessible in resource-limited settings, especially in low- to middle-income countries. In such settings, healthcare providers might only have access to basic laboratory tests, such as serum bicarbonate, sodium (Na+), potassium (K+), blood urea nitrogen (BUN), and creatinine, limiting their ability to use comprehensive blood gas data as a basis for determining treatment.

In case of uncertain onset of cardiac arrest, deciding to withhold or terminate resuscitation can be challenging. The pathophysiology of metabolic acidosis indicates a depletion of the critical buffering agent, bicarbonate, particularly in cases of prolonged or severe cardiac arrest. We hypothesize that bicarbonate levels would decrease more significantly in these situations. However, to our knowledge, no study has specifically investigated this aspect. This study aims to fill this gap by examining the relationship between blood bicarbonate levels and treatment outcomes in OHCA.

Methods

Study design and ethical approval

This retrospective cohort study was conducted at Srinagarind Hospital, a tertiary care facility affiliated with Khon Kaen University, which receives about 60,000 emergency department (ED) visits annually. The study focused on OHCA patients who presented in the ED from January 1, 2015, to November 30, 2021.

Patient population and eligibility criteria

Upon arrival at the Emergency Department (ED), each Out-of-Hospital Cardiac Arrest (OHCA) patient undergoes resuscitation in accordance with Advanced Cardiac Life Support (ACLS) guidelines, administered by the Emergency Physician and their team. During resuscitation, venous blood is drawn as quickly as possible to evaluate critical laboratory parameters. This evaluation, called the “Cardiac Arrest Lab Protocol,” includes a complete blood count, basic blood chemistry, and a coagulogram. The protocol ensures that essential laboratory results are available within 15 min. Additionally, Point-of-Care Testing (POCT) for arterial blood gas (ABG) analysis is performed to assess acid-base status, with results typically available within 5 min.

This study included all OHCA patients aged 18 years or older, encompassing both trauma and non-trauma cases. Patients were excluded if they had do-not-resuscitate (DNR) orders, were not candidates for CPR, or had missing bicarbonate level results or clinical data regarding survival to hospital discharge.

Ethical approval

The study received approval from the Center for Ethics in Human Research, Khon Kaen University (HE651200). Notably, informed consent was not obtained from patients due to the retrospective nature of the data collection, which utilized pre-existing data.

Data collection

Data Collection: To ensure comprehensive and accurate data extraction, we established a predetermined set of variables, which included demographic information such as age, sex, presumed cause of cardiac arrest, patient’s mode of arrival, initial cardiac arrest rhythm, blood chemistry results (including bicarbonate level), and resuscitation outcomes. The data collection process consisted of several key steps: (1) Patients who met the predefined inclusion criteria were identified using the electronic medical record (EMR) database at Srinagarind Hospital. (2) Trained data collectors used standardized forms to systematically collect additional information not available from the EMR from the identified medical records. (3) Rigorous validation procedures were implemented to ensure the completeness and accuracy of the data. Instances of missing or inconsistent data underwent manual review, with supplementary information garnered through further examination of medical records, direct communication with relevant healthcare providers, and a review of the national database to confirm resuscitation outcomes, particularly for patients who were discharged against medical advice (AMA).

Discharged AMA patients

Patients whose outcomes were classified as discharged AMA were those who, after achieving ROSC in the ED and had their initial prognosis discussed with their families. In some cases, families, influenced by cultural beliefs, chose to withdraw life-sustaining treatment at home, preferring that their loved ones spend their final moments in familiar surroundings rather than in a hospital. To ensure accurate outcome assessment, we collected survival data from the national database.

Outcomes

This study aimed to determine the correlation between bicarbonate levels and the rate of survival to hospital discharge among OHCA patients presenting to the emergency department (ED). Secondary aims included identifying other factors associated with the survival rate of OHCA patients (e.g., the presumed cause of cardiac arrest, mode of arrival, initial rhythm, initial labs) and determining the optimal cutoff point for bicarbonate levels in OHCA cases to indicate survival to hospital discharge.

The power analysis for this study was based on previous research by Shin et al., which analyzed the initial blood laboratory test results during CPR of OHCA patients and guided our sample size determination [6]. Specifically, at least 21 survivors and 126 non-survivors were required to detect significant differences in bicarbonate levels between the two groups. This approach ensured the study’s robustness in exploring the associations and outcomes of interest.

Statistical analysis

The patient’s baseline characteristics and clinical data were presented as the median and interquartile range (IQR) for continuous variables and as the count and percentage for categorical variables. Continuous variables were compared using the Mann–Whitney U test, while categorical variables were assessed with Pearson’s chi-square test or Fisher’s exact test.

Blood chemistry variables differed significantly between survivors and non-survivors and were further analyzed to identify optimal cutoff points. The “cutpoint” package in R Statistical Software facilitated this process by determining the most effective threshold for various variables in predicting survival to hospital discharge. Univariate and multivariate analyses were conducted to identify factors associated with survival to hospital discharge.

We conducted a sensitivity analysis to account for the high rate of patients who discharged AMA and subsequently died. In this analysis, all discharged AMA patients were reclassified as non-survivors, yielding 16 patients who survived to hospital discharged. We recalculated the optimal cutoff point for bicarbonate levels using this adjusted survival outcome. The receiver operating characteristic (ROC) curve was then replotted, and the updated sensitivity, specificity, and area under the curve (AUC) were determined.

All statistical analyses were conducted using R Statistical Software version 4.2.3 (www.R-project.org, R Foundation for Statistical Computing), ensuring a comprehensive and robust examination of the data.

Results

Patient characteristics

During the study period, 648 cardiac arrest patients presented to the emergency department. Of these, 77 patients were excluded due to withholding CPR (not initiated in the ED), 72 were excluded due to missing bicarbonate level results (caused by failure to collect specimens, inadequate specimens, or result errors), and 38 patients had Do Not Resuscitate (DNR) orders. Clinical outcomes were obtained for all eligible patients, resulting in a final analysis cohort of 461 patients, as illustrated in Fig. 1. Of these, 88 patients (19%) survived until hospital discharge, while 373 patients (81%) did not. Most of the patients were male (67%), with a median age of 59. Notably, survivors had a higher proportion of myocardial infarction, respiratory problems, other cardiac diseases, private transport, pulseless electrical activity, and ventricular tachycardia/ventricular fibrillation (VT/VF) compared to non-survivors. Conversely, non-survivors had a higher proportion of major trauma, arrival by emergency medical service (EMS), and asystole.

Survivors tended to have higher blood urea nitrogen (BUN) and bicarbonate levels but lower potassium levels. Interestingly, although all survivors experienced a sustained return of spontaneous circulation (ROSC) and were admitted to the hospital, 81.8% chose to discontinue medical care and returned home against medical advice, where they ultimately succumbed.

(Table 1 provides a detailed breakdown of patient characteristics and outcomes.)

Fig. 1

Study flow

Table 1 Baseline characteristics of Out Of Hospital Cardiac Arrest (OHCA) patients

Optimal cutoff point and factors associated with survival to hospital discharge

Notably, a bicarbonate level cutoff above 12.6 mmol/L demonstrated a sensitivity of 74% and a specificity of 47% for predicting survival. Despite an area under the curve (AUC) of 0.607, the negative predictive value was 88%, as shown in Fig. 2. The cutoff points for BUN and potassium levels were 19.5 mg/dL and 5.1 mmol/L, respectively. However, both BUN and potassium levels showed low sensitivity and specificity for detecting survival to hospital discharge (Table S1).

Fig. 2

Receiver operating characteristic (ROC) curve of bicarbonate level and survival to hospital discharge

Factors associated with increased chances of survival to hospital discharge included BUN levels ≥ 19.5 mg/dL, bicarbonate levels ≥ 12.6 mmol/L, private transport, and an initial rhythm of PEA or VT/VF, as detailed in Table 2. Conversely, patients with potassium levels ≥ 5.1 mmol/L had a decreased likelihood of survival to hospital discharge. After multivariate analysis, showing only patient with private transport was associated with survival to hospital discharge. These findings highlight critical thresholds and factors that influence survival outcomes among OHCA patients.

Table 2 Factors associated with survival to hospital discharge in OHCA patients who presented at the ED

Patients with bicarbonate levels below the cutoff point

Table 3 shows that among patients with a bicarbonate level below 12.6 mmol/L, electrolyte abnormalities were the most prevalent presumed cause of cardiac arrest, with a significantly higher proportion compared to patients with bicarbonate levels above 12.6 mmol/L. Conversely, significantly lower proportions of patients with bicarbonate levels below 12.6 mmol/L had myocardial infarction, respiratory problems, and stroke as presumed causes.

Table 3 Comparing presumed causes of cardiac arrest and resuscitation outcomes above and below an optimal bicarbonate cutoff point

Sensitivity analysis

After reviewing the national database, all of the patients who were discharged AMA died within one month. In response to the high mortality rate among discharged AMA patients, we conducted a sensitivity analysis reclassifying all discharged AMA patients as non-survivors. This analysis identified an optimal bicarbonate cutoff of 13.9 mmol/L, with a sensitivity of 93.8% and a specificity of 52.1%. The area under the curve (AUC) was 0.720, indicating moderate predictive accuracy.

Using this refined survival definition, only 16 patients were identified as survivors to hospital discharged AMA. The revised receiver operating characteristic (ROC) curve, shown in Fig. 3, confirms that a bicarbonate cutoff of 13.9 mmol/L effectively distinguishes survivors from non-survivors in this cohort.

Fig. 3

Receiver operating characteristic (ROC) curve of bicarbonate level and survival to hospital discharge, after reclassifying patients who discharged AMA

Discussion

This study examined the relationship between venous bicarbonate levels and survival outcomes among patients with OHCA presenting in the ED. We found that a venous bicarbonate threshold of ≥ 12.6 mmol/L is associated with survival in these patients. The sensitivity of this marker was 74%, with a specificity of 47%, and it demonstrated an 88% negative predictive value for survival. These findings underscore the potential utility of bicarbonate levels as a prognostic tool, although the low specificity limits its efficacy in definitively ruling out non-survivors.

Bicarbonate (HCO3) serves as the principal buffer in the human body’s acid-base regulation system. Sodium bicarbonate is commonly administered to treat severe metabolic acidosis in critically ill patients, including those experiencing cardiac arrest, although its use remains a subject of ongoing debate [9, 10]. Our study examines the utility of bicarbonate in a different context, assessing its prognostic value rather than its therapeutic role in emergency medical situations. We found that higher initial bicarbonate levels (≥ 12.6 mmol/L) are significantly correlated with improved survival to hospital discharge in patients with OHCA. This finding aligns with previous research. Torres et al. [11] analyzed venous blood at the scene of OHCA and found minimal difference in bicarbonate levels between patients with and without ROSC (22.55 vs. 22.31 mmol/L), while Lopez-Sobrino et al. [12] observed lower bicarbonate levels in their study, with a wider differentiation between survivors and non-survivors (18.48 vs. 13.93 mmol/L). These discrepancies may be partially explained by the time interval between cardiac arrest and venous blood analysis, suggesting that ongoing cardiac arrest may intensify metabolic acidosis [2, 4, 5].

Furthermore, our study identifies a cutoff point of 12.6 mmol/L for bicarbonate, which demonstrates a notable negative predictive value of 88%. This finding suggests its potential utility in identifying patients at higher risk of mortality when bicarbonate levels fall below this threshold. However, this value is lower than the range found in previous studies (16.5–21.0 mmol/L) [12, 13]. This variation may be due to the fact that optimal thresholds can differ depending on healthcare setting and patient demographics. For instance, a previous study conducted in Srinagarind Hospital noted that the time from cardiac arrest to the initiation of CPR was relatively long (median 18 to 20 min) [13], which could influence bicarbonate levels and affect the optimal threshold for predicting outcomes.

Upon closer examination of patients with bicarbonate levels below 12.6 mmol/L, a notable distinction emerged in the presumed causes of cardiac arrest. Conditions characterized by a gradual onset, such as electrolyte abnormalities and gastrointestinal bleeding, were significantly more prevalent compared to patients with bicarbonate levels above 12.6 mmol/L. Conversely, presumed causes of cardiac arrest associated with a sudden onset, such as myocardial infarction and stroke, exhibited a significantly lower incidence in patients with bicarbonate levels below 12.6 mmol/L. This divergence may be attributed to an extended no-flow time in conditions with more gradual onset, potentially delaying the detection and resuscitation of the arrest, thereby exacerbating anion accumulation. This accumulation, in turn, contributes to the reduction of bicarbonate levels, aligning with existing literature that highlights the impact of prolonged no-flow time on acid-base balance [8].

We also found that patients with respiratory problems were more likely to exhibit higher bicarbonate levels, a phenomenon indicative of compensatory mechanisms for respiratory acidosis. This physiological compensation underscores the intricate interplay between respiratory and metabolic factors in influencing bicarbonate levels [14].

Our study also reaffirms the importance of a multifactorial approach to OHCA prognostication and management by identifying additional predictive factors for survival. These include elevated BUN levels (≥ 19.5 mg/dL), lower potassium levels (< 5.1 mmol/L), private transport, and specific initial cardiac rhythms. This aligns with findings from previous studies and emphasizes the need for holistic assessment to optimize outcomes [15, 16].

Interestingly, we observed a unique phenomenon where a significant proportion of survivors opted out of additional medical care and returned home against medical advice, where they subsequently succumbed to their condition. This finding likely reflects cultural beliefs prevalent in our region, particularly northeastern Thailand, where there is substantial value placed on dying within familiar surroundings and a belief that it is ethical to withdraw life support at home but not in the hospital [17]. Similar observations have been reported in studies from geographically diverse regions with distinct cultural contexts [18, 19]. This underscores the importance of considering cultural factors when interpreting patient decisions and suggests the potential need to tailor resuscitative strategies to align with individual preferences and values.

The sensitivity analysis conducted in our study, which reclassified patients who discharges AMA as non-survivors, offered critical insights into the robustness and reliability of our initial findings. Following this reclassification, the optimal bicarbonate cutoff increased from 12.6 mmol/L to 13.9 mmol/L, accompanied by a substantial improvement in sensitivity (from 74 to 93.8%) and a modest rise in specificity (from 47 to 52.1%). The AUC also improved from 0.607 to 0.72, reflecting moderate predictive capability.

This shift in the bicarbonate threshold highlights the significant impact of including discharged AMA patients in survival analyses, as their exclusion or reclassification can substantially influence study conclusions. The increased sensitivity observed in the refined analysis suggests that a bicarbonate level of 13.9 mmol/L may serve as a more reliable prognostic marker in clinical settings, especially for identifying patients with a higher likelihood of survival when discharged AMA cases are classified as non-survivors.

The improved AUC and high sensitivity associated with this new cutoff point underscore the importance of refining predictive models to address real-world complexities, including patients opting out of care. This is particularly pertinent in settings where cultural factors influence medical decision-making, as observed in our cohort. Moreover, the higher cutoff aligns more closely with thresholds reported in previous studies, bridging the gap between our findings and existing literature [12, 20]. Therefore, a cutoff point between 12.6 and 13.9 mmol/L may be used to guide the termination of resuscitation once the bicarbonate result is obtained.

Our study has several limitations. First, its retrospective design, single-center setting, and relatively small sample size limit its generalizability and make it challenging to account for confounding factors. Second, the lack of standardized criteria for terminating resuscitation and the inherent complexity of out-of-hospital cardiac arrest (OHCA) may have influenced patient outcomes. Third, the timing of cardiac arrest could affect bicarbonate levels; however, the exact time of onset was unknown for most cases, preventing its inclusion in our analysis. Future research incorporating multi-center collaborations and prospective study designs could help validate and expand upon our findings.

Our study contributes significantly to the understanding of the association between venous bicarbonate levels and critical outcomes in OHCA patients. The potential utility of readily available bicarbonate levels as a prognostic marker, particularly with the identified optimal cutoff of greater than 12.6 mmol/L, may improve outcome prediction in OHCA patients. Additional factors influencing survival outcomes include BUN, potassium, private transport, and initial cardiac rhythm. These findings offer valuable insights into resuscitation strategies, equipping clinicians with practical tools for assessing severity and estimating prognosis, especially in resource-limited settings.