Concordance of the analysis of basic biochemical and hematological parameters using a blood gas analyzer compared to laboratory analyses: a retrospective method-comparison study from a tertiary center in Morocco

¹Intensive Care Unit, Fifth Medical-Surgical Center, Errachidia, Hassan II University, Fez, Morocco, ²Critical Care Department, Moulay Ismail Military Hospital, Hassan II University Hospital, Fez, Morocco, ³Anesthesiology Department, Moulay Ismail Military Hospital, Hassan II University Hospital, Fez, Morocco, ⁴Central Laboratory, Moulay Ali Cherif Hospital, Errachidia, Moulay Ismail University, Meknes, Morocco, ⁵Medical Director of Fifth Medical-Surgical Center, Moulay Ismail University, Errachidia, Morocco, ⁶Anesthesiology and Critical Care Division, Moulay Ismail Military Hospital, Hassan II University Hospital in Fez, Fez, Morocco

^&Corresponding author
Maaroufi Ayoub, Intensive Care Unit, Fifth Medical-Surgical Center, Errachidia, Hassan II University, Fez, Morocco

Introduction    

The blood gas analysis (BGA) is a valid biological test that involves taking an arterial blood sample to measure the partial pressure of oxygen and carbon dioxide in the blood, the oxygen content, oxygen saturation, bicarbonate concentration, and blood pH [1,2].

It allows clinicians to assess a patient's acid-base disorders and hypoxia. So the BGA plays an essential role in the management of acute respiratory failure, intraoperative monitoring, and the critical care management of patients [3]. Some manufacturers offer additional tests, notably evaluations of electrolyte levels such as sodium (Na+), potassium (K+), hemoglobin (Hb), and hematocrit (Hct), and all kinds of biological or hematological assays, thereby providing clinicians with more detailed information on the patient's condition and facilitating their management in terms of accuracy and time savings [4].

In intensive care units (ICU), emergency departments, and even in the operating room, the speed of a physician's therapeutic decision-making is paramount. By use of "point-of-care" (POC) monitoring, like BGA, important variables can be measured in real-time, allowing for immediate results to be provided to make rapid decision [5,6]. It is easy to use and provides rapid results that can modify or influence patient management, particularly for critical parameters such as hemoglobin, hematocrit, electrolyte- and acid-base-balance, and the oxygenation status of the blood [7,8]. Discrepancies, particularly concerning electrolytes and hematocrit, can mislead clinical reasoning as well as therapeutic management if inaccurate POC results are used alone to guide treatment decisions.

Our study aimed to evaluate the reliability of the results obtained from the BGA by comparing them to those provided by the central laboratory (CL) and to identify any potential significant discrepancies between the two analysis methods. The objective of our study was to determine whether the results obtained by the BGA were sufficiently valid to be used as a basis for clinical decision-making in our ICU, especially for critical parameters like sodium, potassium, hemoglobin, and hematocrit.

Methods     

Study design: this retrospective, analytical, quantitative, and cross-sectional study was conducted to evaluate the concordance between results obtained from a point-of-care blood gas analyzer (BGA) using arterial whole blood and those provided by the central laboratory using venous blood.

Setting: the research was carried out in the intensive care unit of the 5^th Medical-Surgical Center in Errachidia, Morocco, over a period of one year and four months, from January 2024 to April 2025.

Participants: patient selection was conducted consecutively and non-probabilistically throughout the study period. A total of 200 patients admitted to the intensive care setting who underwent simultaneous sodium, potassium, hemoglobin, and hematocrit measurements via both BGA and the central laboratory were initially screened for eligibility. Of these, 147 patients were excluded due to failure to meet the strict thirty-minute time window between arterial and venous sampling (n = 92), incomplete central laboratory data or missing parameters (n = 35), or the occurrence of major resuscitation interventions and invasive procedures between draws that could confound results (n = 20).

Consequently, a final cohort was included in the analysis. The inclusion criteria encompassed all hospitalized patients who benefited from simultaneous blood sampling allowing the measurement of sodium, potassium, hemoglobin, and hematocrit through both arterial blood gas and central laboratory analyses. Conversely, exclusion criteria removed patients with incomplete biological datasets lacking at least one of the core parameters, cases where arterial and venous samples were not collected simultaneously, instances where the time interval between draws exceeded thirty minutes, and situations involving major resuscitation measures or invasive interventions-such as chest compressions-that could alter physiological parameters and bias comparative findings (Figure 1).

Variables: the evaluated variables comprised paired arterial and venous measurements of sodium (Na+), potassium (K+), hemoglobin (Hb) and hematocrit (Hct).

Data sources/measurement: the analytical equipment utilized for the study comprised an i-STAT Blood Gas Analyzer (Abbott Laboratories, United States) situated in the intensive care unit, complemented by a Sysmex XN-550 automated hematology system (Sysmex Corporation, Japan) and an ARCHITECT ci4100 chemical analyzer (Abbott Laboratories, United States) housed within the central laboratory. The blood gas analysis protocol was strictly standardized across pre-analytical, analytical, and post-analytical phases, with all personnel thoroughly trained in its execution. For the pre-analytical phase, blood collection for the central laboratory and the BGA was performed via separate sampling routes from different sites, with venous blood collected without a tourniquet and arterial blood collected independently in strict adherence to transportation protocols. Heparinized syringes were utilized to preserve the arterial samples before immediate analysis on pre-calibrated devices, whereas central laboratory samples were delivered within fifteen minutes and processed immediately upon receipt. When analysis was deferred, samples were systematically remixed by rolling the syringes horizontally between the palms for at least five seconds and subsequently inverting them consecutively for an additional five seconds. Prior to testing, the initial two drops of blood were discarded to eliminate air-exposed blood at the tip that lacked homogeneity with the rest of the sample, and a strict ratio not exceeding ten units of heparin per milliliter of blood was maintained during manual heparinization. During the analytical phase, cartridges and handheld readers were verified to be at room temperature, and cartridge barcodes were scanned prior to opening pouches to ensure immediate utilization and prevent quality control failures. The handheld analyzer was activated, and the i-STAT cartridge option was selected in accordance with standard operating protocols, followed by barcode lot scanning via automated on-screen prompts. Samples were prepared following established institutional guidelines involving collection, cartridge filling, and secure closure, after which the cartridge was inserted entirely into the analyzer port until fully seated to complete the automated analysis cycle. Post-analytically, numerical concentration values were retrieved from the display profile in customized standard units. Data were extracted from the hospital's patient medical records using a standardized data collection form.

Bias: to minimize potential selection and confounding biases, strict inclusion and exclusion criteria were applied, including a mandatory thirty-minute time window between arterial and venous draws and the exclusion of patients undergoing major resuscitation interventions between samples. Ethical considerations were fully addressed through authorization obtained from the Health Service Inspectorate of the Royal Armed Forces under approval number 8101/EMG/ISS in Rabat on April 18, 2025, with a waiver of informed consent granted for this retrospective review to ensure patient anonymity and data confidentiality.

Study size: a formal a priori sample size calculation was not conducted; instead, the final cohort represented a convenience sample of eligible individuals meeting all stringent inclusion criteria and time-window constraints during the study period.

Quantitative variables: to ascertain whether measurement differences exceeded acceptable performance thresholds, predetermined total allowable error and performance criteria were established based on recognized international guidelines, specifically referencing Clinical Laboratory Improvement Amendments (CLIA) proficiency testing criteria and relevant international standards including ISO and IFCC specifications for sodium (+/- 4 %), potassium (+/- 0.5 mmol/l or +/- 10 %), hemoglobin (+/- 7 %), and hematocrit (+/- 6 %).

Statistical methods: data analysis was performed using GraphPad Prism 10.4.2 and Microsoft Excel Office Professional Plus 2016 software. The normality of distribution for all paired measurement differences was tested using the Shapiro-Wilk test. Normally distributed data were compared using the paired $t$-test, whereas non-normally distributed variables were analyzed via the Wilcoxon signed-rank test, with a p-value of < 0.05 considered statistically significant.

Results     

Participants: the study cohort comprised the evaluated patients, yielding a total of 100 paired blood sample measurements (comprising 50 arterial and 50 venous samples). Among the study population, there were 43 men and 10 women, representing 81% and 19% of the cohort, respectively. The mean age of the patients was 64.61 years with a standard deviation of 15.92.

Descriptive data: reasons for hospital admission are detailed in Figure 2, and were primarily dominated by respiratory diseases (68%) and heart conditions (9%).

Outcome data: sodium (Na+) measured by BGA was, on average 7.3 mmol/L, lower than that obtained by the CL (129.75 ± 7.77 vs. 137.05 ± 4.39), with a median difference of 6.5 mmol/L, which was statistically significant (p < 0.001). Similarly, potassium (K+) was, on average, higher with BGA (3.85 ± 0.89) compared to the CL (3.5 ± 0.70), with a median difference of 0.55 mmol/L, which was also significant (p < 0.0001). Regarding hematocrit (Hct), BGA showed a mean value lower than that of the CL (35.90 ± 8.49 vs. 38.77 ± 8.71), with a median difference of 3.5%, which was also a significant difference (p = 0.0079). Although that hemoglobin (Hb) values were slightly higher with BGA (16.33 ± 2.30 vs. 13.8 ± 2.77 for the CL), there were no statistically significant difference (p = 0.562), with a modest median difference of 0.45 g/dL (Table 1).

Main results: in the Bland-Altman analysis, the mean bias along with its 95% confidence intervals, the 95% limits of agreement, and the exact percentage of points falling outside these limits were calculated for each parameter. The analysis demonstrated a mean negative bias between values obtained by the BGA and those from the central laboratory across parameters. Specifically, the bias was -7.30 ± 7.13 mmol/L for sodium, with 95% limits of agreement ranging from -21.28 to 6.67 mmol/L. For potassium, the bias was 0.55 ± 0.73 mmol/L, with limits ranging from -1.99 to 0.88 mmol/L. Hemoglobin showed a bias a bias of 0.30 ± 1.95 g/dL (ranging from -4.14 to 3.52 g/dL), while hematocrit exhibited a bias of -2.55 ± 6%, with limits ranging from -14.33 to 9.22% (Figure 3). The Bland-Altman plots revealed a minimal mean difference for potassium, whereas sodium demonstrated a notable negative mean bias (-7.30 mmol/L), representing a clinically meaningful difference that warrants careful consideration when interpreting point-of-care sodium values in the intensive care setting. For hemoglobin and hematocrit, the observed mean biases were -0.30 g/dL and -2.55%, respectively, with several measurements falling outside the limits of agreement, indicating potential divergences in specific clinical situations.

Other analyses: correlation analyses between the two measurement methods revealed a moderate correlation for electrolytes (Figure 4), with an r-value of 0.3853 for sodium Na+ (Na+; 95% CI: 0.1203 to 0.5990; p < 0.0001) and 0.5059 for potassium (K+; 95% CI: 0.2653 to 0.6872; p < 0.0001). In contrast, a strong and statistically significant correlation was observed for hematocrit (Hct; r = 0.8294; 95% CI: 0.7167 to 0.8999; p = 0.0079) and hemoglobin (Hb; r = 0.7923; 95% CI: 0.6597 to 0.8771; p = 0.562), indicating good overall reliability between the two methods for these hematological parameters (Table 2).

Discussion     

Point-of-care testing devices are classified based on their analytical modality and size [9]. Among the most compact models are portable devices, such as test strips and blood glucose meters [10]. The latest versions of these devices integrate cartridges capable of performing multiple analyses simultaneously, such as measuring cardiac markers, blood gases, and various hematological and endocrine parameters from a blood sample [11].

These benchtop units offer increased versatility by integrating different testing modalities, thus allowing for various diagnostics on the same device. Among the most common analyses are general biochemical tests, serological tests, C-reactive protein (CRP) and Various hematological and biological assays [12]. The growing demand for more compact and accurate benchtop POCT devices stimulates technological innovation [13], leading to the development of equipment that is both smaller and more efficient [12,14,15]. Point-of-care testing encompasses several distinct technological modalities designed to meet diverse diagnostic needs, they can be classified according to their practical use as well as based on the testing modality and the test size [10,16,17].

The simplest forms of POCT are the testing strips which rely on an interaction between an analyte and a reactive chemical, typically present or impregnated within a support medium in a manner that permits the managed addition or mixing of the sample [18]. A characteristic example includes test strips, particularly those utilized for urinalysis [10]. Further advancing diagnostic capabilities, lateral-flow testing utilizes a support material containing capillary beds that transport liquid samples to specific zones. Within these zones, the sample contacts reactive substances that interact with present analytes, with home pregnancy tests serving as an emblematic example through their reliance on immunoassays detecting human chorionic gonadotropin in urine [10].

Additional POCT methodologies are based on immunoassays, which utilize antibodies binding to specific targets when their concentrations exceed established threshold [18]. These targets can be proteins, drugs, or pathogens, offering a wide range of substances to detect [10]. Antigen-based testing specifically involves the detection of known antigens or antibodies tied to particular diseases or disease states, such as Helicobacter pylori-specific IgG antibodies, among others [19].

Finally, the demand for molecular POCT characterized by high sensitivity, high specificity, and relatively short turnaround times has spurred significant development in this domain. Although these instruments are generally more complex than unit-use devices, they incorporate diverse analytical principles including spectrophotometric substrate and enzyme-activity measurement, hematological particle counting, immunoassays, and sensor-based blood-gas analysis (BGAs) [16,19].

Based blood-gas analysis use potentiometric/amperometric or optical sensors for pH, pO2 and pCO2. Additional ion-sensitive electrodes for the measurement of electrolytes and other substrates are available[16]. Most BGA are equipped with a CO-oximetry unit to determine the O2-Hb saturation. BGA remains the gold standard method for oxygenation and ventilation monitoring [2].

In general, Blood gas devices are an integral part of POCT and allow clinicians to respond to acute clinical situations, particularly in the critical care unit, thanks to additional analysis modules. These modules enable the measurement of other parameters such as sodium, potassium, hemoglobin levels, hematocrit, troponin, and even hemostasis profiles. A single blood sample is used for both gas analysis and complementary assays [4].

In this research, population's statistical study shows a clear male predominance, with 81% men compared to 19% women. This difference can be partially explained by the nature of the pathologies managed in this type of unit, notably respiratory diseases, which account for 68% of admissions to our ICU. The predominance of respiratory diseases among admissions in our cohort can be attributed to several key epidemiological, environmental, and clinical factors specific to our intensive care setting: respiratory emergencies predominate due to our unit's role in managing severe military and civilian critical care cases requiring immediate ventilation, regional seasonal shifts and environmental triggers drive high volumes of winter admissions for severe respiratory decompensations and triage protocols preferentially route patients with profound hypoxia and respiratory crises to our specialized resuscitation unit for advanced cardiorespiratory support.

This is a pathology more frequent in men, particularly in advanced age groups [20,21]. The average age of the patients was 64.61 ± 15.92 years, indicating a predominantly elderly population. This is consistent with studies that show an increased frequency of intensive care unit admissions among older adults [22-24]. The high frequency of male patients is explained too by the military nature of this medical structure which is located in a military zone and is primarily dedicated to offer treatments to active and retired military personnel in the region [25], who are predominantly male with an unstable medical condition [26,27].

Comparative results between the BGA and the CL highlight significant discrepancies for several biological parameters. While Spearman correlation coefficients were initially calculated to evaluate the strength of association between the point-of-care blood gas analyzer and central laboratory measurements, our primary interpretation of concordance relies on Bland-Altman analysis. Bland-Altman plots provide a more robust assessment of agreement by quantifying the mean bias, the precision of differences, and the limits of agreement relative to predefined clinical tolerance limits, thereby avoiding the limitations inherent to correlation-based evaluations.

Regarding sodium, our findings demonstrate a significant divergence and wide limits of agreement between blood gas analysis (BGA) and central laboratory measurements, reflecting notable variability. This discrepancy carries important clinical implications, particularly in the management of fluid and electrolyte disorders. While previous literature reports varying degrees of correlation-ranging from moderate to strong across different studies-our results reinforce the consensus that BGA sodium values should be interpreted with caution [28-30].

For potassium, BGA measurements showed a tendency toward overestimation with moderate correlation compared to central laboratory results. Although these findings align with several previous studies, others report higher correlation coefficients [29-31]. This variability is frequently compounded by preanalytical factors, such as the use of combined blood gas and electrolyte devices and sample storage conditions, which can elevate potassium concentrations [32]. Furthermore, while findings regarding bias direction differ slightly across some published cohorts, studies consistently highlight high variability and limited clinical reliability [33]. Consequently, while BGA potassium values can assist in guiding clinical reasoning-especially in patients predisposed to electrolyte imbalances-they should not serve as the sole basis for clinical intervention without accounting for potential underlying bias [34].

Regarding hematocrit, significant differences were observed between the two analytical methods, accompanied by a strong correlation that aligns with previous findings [28,29,30,]. The conductivity-based methodology utilized by blood gas analyzers is inherently susceptible to interference from unmeasured variables such as hemolysis, abnormal protein levels, or hyperlipidemia, which can falsely elevate hematocrit values, whereas factors like cold agglutinins may lead to underestimation [35]. Despite strong correlation coefficients reported in the literature, broader Bland-Altman analyses frequently reveal insufficient agreement, underscoring the limitations of these methods in clinical practice [29].

Conversely, hemoglobin measurements showed no statistically significant difference or major discordance between the two approaches, demonstrating a strong and clinically acceptable correlation consistent with prior studies [28,36]. Nevertheless, as noted by Altunok et al. high correlation coefficients can coexist with insufficient limits of agreement, indicating that analytical discrepancies can still compromise clinical reliability if measurements are used interchangeably without critical evaluation [29].

Another point to consider when interpreting the results obtained by the central counterparty, in our study, is that it relies on an automated hematology analyzer (Sysmex XN-550), which typically employs the sodium lauryl sulfate (SLS)-hemoglobin method or automated cyanmethemoglobin-equivalent principles. These distinct underlying technologies can introduce analytical variations, particularly in critically ill patients or in the presence of interfering substances (such as hyperlipidemia, hyperbilirubinemia, or abnormal leukocyte counts), which should be taken into account when evaluating clinical concordance. Serum measurement of Hct and Hb via ABG can be used reliably for monitoring patients in cases of hemorrhage, transfusion, or rapid hemodynamic changes [29]. These data suggest that Na+, K+, and Hct measurements differ significantly between the two methods, while those for Hb are relatively concordant, a result that is consistent with other studies suggesting that these results cannot be entirely relied upon [31,37].

To overcome these result discrepancies and make them representative, integrating a local correction coefficient, based on internal information, could strengthen the credibility of clinical analyses and facilitate a more precise interpretation of POC results in daily practice. In fact, misuse or overuse of these devices may generate deceptive results and escalate healthcare expenditures [38,39].

Point-of-care testings (POCTs) hold a major place in technical and medical progress; the results obtained guide clinicians in decision-making and therapeutic management, particularly in emergency and intensive care departments. Examples include blood gas analysis, capillary blood glucose measurement via glucometer [16], and the detection of ketones in urine [16]. The latter two constitute essential tools in all medical facilities and even in outpatient settings (pharmacies, diabetic patients for home monitoring), which reflects a 'democratization' of technology for the benefit of our patients.

Blood gas analyzer (BGA) devices can be used alone without concern and without the need for confirmation by the CL for the interpretation and analysis of blood gases, determining hypoxia and classifying its type, and specifying the acid-base balance based on PCO2, PO2, pH; BE (base excess). These parameters are of major importance and guide management measures based on the observed disorder and its cause (non-invasive ventilation (NIV), high-flow oxygen therapy, invasive ventilation, alkalinization, dialysis, etc.). Indeed, these devices are validated for this purpose, and their results are valid and interpretable, provided that usage recommendations, calibration, and storage conditions are respected [1,40].

Limitations: our study has several limitations that should be taken into account when interpreting the results: it is a single-center study, conducted in one intensive care unit, which restricts the applicability of the conclusions to other clinical situations. Furthermore, even if the sample size (53 patients) is acceptable, it remains too small to perform detailed analyses according to clinical or pathological subgroups. Furthermore, the lack of an a priori sample size or power calculation introduces the potential for a type II error. With 53 included patients, the study may have been underpowered to detect minor, yet clinically meaningful, systematic biases or proportional differences between the point-of-care analyzer and the central laboratory. Future prospective studies with larger, pre-determined sample sizes are warranted to confirm these findings.

The use of a single BGA model also prevents us from extrapolating or inferring the results to other POC devices that might have different performance. Pre-analytical factors (sample collection, handling, and transport conditions) were not always standardized or verified, which could have affected the accuracy and validity of the analyses, especially for sensitive electrolytes like potassium. Neither a quantitative hemolysis index nor a visual hemolysis check was systematically recorded at the time of sampling. Because in vitro hemolysis significantly affects extracellular potassium concentrations and other parameters, this unmeasured variable represents a potential confounding factor that should be addressed in future prospective studies.

Finally, the lack of a personalized assessment for each patient makes it impossible to verify whether certain specific clinical characteristics are linked to a greater discrepancy between the results obtained by the BGA and those provided by the central laboratory. In addition, our research did not examine the direct therapeutic impacts of the observed divergences between the two techniques, which would have made it possible to appreciate their real impact on patient management. Finally, it would also be advisable for each medical institution to conduct an agreement evaluation tailored to its context, by considering the type of analyzer used, the collection conditions, as well as the sample processing procedures.

Conclusion     

The BGA allows us a significant reduction in processing time, especially in emergency situations, but the observed variation in certain fundamental parameters questions the reliability that can be placed in its values to guide a clinician's reasoning and therapeutic choices, particularly in emergency situations with unstable and fragile patients. The comparison between the results provided by the BGA and those of the CL for sodium, potassium, hemoglobin, and hematocrit revealed marked discrepancies, from both a statistical and clinical standpoint. The findings suggest that caution should be exercised when interpreting electrolyte and hematocrit measurements obtained by this specific BGA device under the study conditions.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

Conceptualization: Maaroufi Ayoub and Ouazzani Ibtissame; investigation: Bouramdane Fatima Ezzahra, Ouazzani Ibtissame, and Bouaboula Fath-Allah; methodology: Diai Abdellatif, and Bouaboula Fath-Allah; formal analysis: El Akef Taoufik; writing original draft: Maaroufi Ayoub and Ouazzani Ibtissame; writing review and editing: Maaroufi Ayoub, and Kechna Hicham. All the authors read and approved the final version of this manuscript.

Tables and figures     

Table 1: sodium, potassium, hemoglobin, and hematocrit values obtained by the gas blood analysis analyzer compared to those of the central laboratory of study participants, recruited from the intensive care unit of the fifth Medical-Surgical Center Errachidia (Morocco), from January 2024 to April 2025 (N=53)

Table 2: Bland-Altman analysis of bias and 95% limits of agreement between blood gas analysis and central laboratory values for sodium, potassium, hemoglobin and hematocrit of study participants, recruited from the intensive care unit of the fifth Medical-Surgical Center Errachidia (Morocco), from January 2024 to April 2025 (N=53)

Figure 1: flow chart of patient selection and inclusion/exclusion process

Figure 2: distribution of patients admitted to the intensive care unit according to the reason for admission of study participants, recruited from the intensive care unit of the fifth Medical-Surgical Center Errachidia (Morocco), from January 2024 to April 2025 (N=53)

Figure 3: bias plots for: A) sodium (Na⁺), B) potassium (K⁺), C) hematocrit (Hct); D) and hemoglobin (Hb), the Bland-Altman plot shows the bias of the parameters for each pair of samples of study participants, recruited from the intensive care unit of the fifth Medical-Surgical Center Errachidia (Morocco), from January 2024 to April 2025 (N=53)

Figure 4: Spearman's rho correlation between values from the gas blood analysis and the central laboratory, shown separately for Hct, Hb, Na⁺, and K⁺ of study participants, recruited from the intensive care unit of the fifth Medical-Surgical Center Errachidia (Morocco), from January 2024 to April 2025 (N=53)

References