1. IntroductionError in the medical field is usually referred to as “any preventable event that may cause or lead to inappropriate medication use or patient harm”, and prevention of such events must be the primary goal of the healthcare system [1]. An error event can lead to misdiagnosis and delayed or failed treatment, having great impact on healthcare performance and patient safety; it can occur due to malpractice and depend on human behavior, but it could also result from an error occurring in the processing of laboratory data. Laboratory errors can be difficult to identify without a specific test, and for this reason, the issue of laboratory errors is receiving increasing attention [1]. The process of laboratory data analysis consists of a pre-analytical phase, an analytical phase, and a post-analytical phase, and each of these steps are not free from the possibility of error [2]. Majority of errors occur in the extra-analytical phases of sample collection and storage. The frequency of error in this step accounts for 60–70% of the total errors that occur throughout the entire process [3].Hemolysis represents the most evident error in the pre-analytical phase, as 40–70% of sample rejections are primarily due to the presence of hemolysis, followed by insufficient sample volume (10–20%), the use of an inappropriate container (5–15%), and the evidence of visible clots within the blood sample (5–10%) [3]. The term hemolysis refers to the breakage of red blood cells, resulting in the release of free hemoglobin into the plasma. After centrifugation of the blood sample, a process that helps to identify the phenomenon, a distinct pinkish discoloration of the serum may be noticed. This coloration is visible if the sample contains at least 0.5% of lysed erythrocytes and a free hemoglobin concentration greater than 0.3–0.5 g/L [1].The emergency department (ED) is one setting of care where hemolysis phenomena are more frequent [1]. A survey promoted by the Working Group on Laboratory Errors and Patient Safety (WG-LEPS) of the International Federation of Clinical Chemistry and Laboratory Medicine (IFCC) [4] and the Global Preanalytical Scientific Committee (GPSC) [5], involving 391 clinical laboratories around the world, showed that approximately 53% of hemolyzed samples are ED samples [6]. This issue is determined by the working environment in the ED, which is often overcrowded and is a place where speed and efficiency are essential. While central laboratories systematically checked blood samples for hemolysis to limit this problem, the same does not happen for a point-of-care system (POCT) such as a blood gas analyzer [7]. Arterial blood gas analysis (BGA) is one of the most widely used technologies in the ED [8] based on the large amount of crucial information that it can provide in the critical area and, when used as a POCT system directly in the emergency room, reduces the waiting time for results and can improve decision-making and diagnostic processes [9]. Recently, this new technology was also found to be applicable in venous sampling analyses for metabolites and electrolytes detection (bilirubin, glucose, creatinine, urea, and Troponin I) [10]. However, the blood samples in POCT cases are not usually checked for hemolysis. This is critical if we consider that the parameters most affected by hemolysis are potassium, pH, PO2, and pCO2 [6].However, hemolysis in arterial samples have not been studied in-depth, and even it’s frequency is not clear, having a range between 1.2% [11] and 7.9% [9]. Thus, given the limited data on this topic, we believe that our study may be of interest and bring consistent benefits to clinical practice.

The aims of this study were to: (1) estimate if the percentage of hemolysis in arterial samples collected using a POCT system accessible in ED in real life might be much higher than 8%; and (2) evaluate the hemolysis-induced error in the dosage of some values (i.e., potassium) and the chance of its relevant clinical impact.

3. ResultsA total of 525 arterial samples were collected, of which an initial statistical analysis was carried out, starting from the percentage of hemolysis of each individual operator in the study (Figure 1).The mean hemolysis was 18% (±19% SD), but three operators (n° 5, n° 8, and n° 13) reported a percentage of hemolysis that was +2SD out of the means, and they were thus eliminated. The final group included 472 arterial samples and hemolysis, defined as those with more than 0.5 g/L of free hemoglobin in plasma, was present in 12% (±13% SD) of all samples tested (Table 2). No differences between the non-hemolyzed and hemolyzed samples were found on the distribution of the main symptoms that led patients to the emergency visit (Table 3).Table 3 classifies the sample according to the symptom that led the physician to perform the BGA and shows the result of the analysis of BGA, with the electrolytes and metabolites divided into two groups (hemolyzed and non-hemolyzed samples). Of the six values analyzed, in the hemolyzed samples compared with the non-hemolyzed samples, we found a statistically significant increase in potassium by 15.29% (+0.61 mEq/L; p-value +) of the same patient in a contextual sampling. The pCO2 was also increased in the hemolyzed samples (+ 4.20%) as well as the lactate (+26.3%), but these were not statistically significant. The pO2 value was non-statistically significantly decreased in the hemolyzed samples (−8,6%). The pH and Ca2+ values remained unchanged in the two groups under analysis. The concordance between the two measures (BGA and venous) was also evaluated (Figure 2), and we found a strong correlation in potassium dosage between arterial and venous sample (R2 0.89 in non-hemolyzed, p2 0.74 in hemolyzed, pFigure 3, plot A and plot B) shows that in the case of the non-hemolyzed group (plot A), the arterial values are almost identical to the values measured in the venous (Δ(a-v) K+: −0.07, 95% CI (−0.58,0.38)), with an average bias of about 0.5 mEq/L. Meanwhile, in the hemolyzed group (plot B), the mean of the differences is frankly above zero (Δ(a-v) K+: 0.29, 95%CI (−0.70,128)), with an average bias of about 1 mEq/L.In Figure 4, the potassium values were categorized in five different intervals according to the values of the venous sample: hypokalemia (5.3 mEq/L); the error bars represent 95% confidence intervals of the hemolyzed (full black circles) and non-hemolyzed (open white circles) groups. As shown in the figure, in the hemolyzed group, the mean error in the potassium level measured on BGA of the

More than 77% of hemolyzed samples have a higher value of K+ in arterial blood than in venous blood. The 15.2% of the K+ values change from venous hypokalemia to artetial normokalemia, while there was only a 5.08% shift from venous normokalemia to arterial hyperkalemia. The same analysis in the non-Hemolyzed samples does not show any change.

Finally, Figure 5 shows the distribution of ranges of hypokalemia, normokalemia and hyperkalemia between the hemolyzed and non-hemolyzed samples. 4. DiscussionThe main finding of our study is that the average rate of POCT BGA hemolysis in the ED is higher than expected and that it could lead to the misdiagnosis of electrolyte imbalance [9,14].The normal concentration of free hemoglobin is usually around 0.22–0.25 g/L in the serum and between 0.10 and 0.13 g/L in the plasma. However, consensus has established that the threshold for effectively defining a hemolyzed sample is the presence of 0.5 g/L of free hemoglobin in the specimen [3]. In fact, some authors define the cut-off of 0.5g/L free hemoglobin in the sample as the limit above which it can cause interference in sample analysis. Many studies also suggests that it is better to avoid processing hemolyzed samples [15].Based on the mechanism that determines the phenomenon, a distinction can be made between in vivoand in vitro hemolysis. In the first case (in vivo), hemolysis is attributable to the patient’s clinical conditions (autoimmune diseases, mechanical heart valves, adverse reaction to transfusion, etc.) and typically accounts for less than 2% of the hemolysis found. Meanwhile, the in vitro hemolysis is mainly caused by inadequate collection or improper transport and storage of specimens before the analysis. Both kinds of hemolysis produce the same type of interference in the sample analysis [1]. In vitro hemolysis could be a clinical problem, as it interferes with the reliability of the results provided to physicians and is unrelated to the patient’s clinical condition. This result is an error that risks endangering the patient’s health as well as their discomfort and an increase in costs due to the need to repeat the blood sample [16]. The analytes most affected by hemolysis due to the loss of red blood cell content in plasma are potassium, LDH, and AST. Some authors also describe the interference of hemolysis in troponin analysis [17].Potassium is the most relevant indicator that an incorrect dosage of this electrolyte could have for clinical implications [16]. Our results show a hemolysis rate of 12% among the arterial samples analyzed. This percentage is higher than the percentage described in the literature, as we can see from the studies conducted by Duhalde et al. in 2019, where a hemolysis rate of 7.9% was found in a total sample of 1270 specimens analyzed in laboratory [9], and by Salvagno et al., which found that in a previous study, only 4% of specimens were hemolyzed in a sample of 487 arterial BGA [18]. Similarly, Wilson et al. compared the occurrence of hemolysis in arterial samples from the ED and ICU; they found that out of 100 samples, the percentages of hemolysis in the ED and ICU samples were 6% and 3%, respectively [14].The same authors discussed that the reason for these differences could be on account of factors that can cause in vitro hemolysis due to mechanical factors, such as an inappropriate needle size, excess blood flow rate, presence of arterial lines, or the need to perform an arterial puncture. They also considered the difference between the expertise of staff in the ICU and ED [14].This study therefore confirms the need to accurately identify hemolyzed arterial samples, as even the Clinical and Laboratory Standards Institute C46-A2 guidelines suggest that arterial samples with hemolysis should not be analyzed [19].The hemolysis of arterial samples can affect patients in a major way, especially in the ED. It can result in prolonging hospitalization due to the need to repeat the sampling, which confuses the diagnostic process with unreliable data. Furthermore, it creates a false sense of security in the clinician who does not know whether the sample is hemolyzed, as POCTs do not assess the presence of hemolysis [20]. In addition, hemolysis can significantly interfere with several analytes, in particular altering potassium values with the risk of increasing potassium concentrations, leading to the diagnosis of pseudo-hyperkalemia or mimicking normal potassium values and thus masking hypokalemia [21].

We found a statistically significant increase in potassium levels by +16% (+0.61 mEq/L; p-value < 0.001) in the hemolyzed samples compared withthe non-hemolyzed samples. Thus, the hemolyzed samples tended to have higher potassium values, both as an absolute value and as a difference (Δ(a-v) K) of the measured value of venous and arterial potassium of the same patient, in a contextual sampling. Our data confirm the possibility of incorrect therapeutic intervention (or non-intervention) if the clinician is unaware of the presence of hemolysis in the arterial sample analyzed.

Potassium disorders are common in patients attending EDs. Hypokalemia and hyperkalemia are very common electrolyte disorders, and if untreated, they can lead to potentially fatal cardiac conduction disturbances and neuromuscular disorders [22]. Pseudo-hyperkalemia can result in iatrogenic hypokalemia. In fact, the incorrect administration of therapy can lead to death through ventricular fibrillation/tachycardia, paralytic ileus, and respiratory depression [23]. Similar factors can cause masked hypokalemia, i.e., potassium being within normal ranges despite the patient being hypokalemic [24]. Unrecognized hypokalemia can also become potentially fatal for the patient, especially in the population group taking drugs that interfere with potassium homeostasis [25], who are administered sodium bicarbonate as a therapy for metabolic acidosis [26]; however, it is also a factor frequently associated with a worsening outcome, even in patients that are initially at low risk of deterioration [27]. Furthermore, our results seem to show that hemolysis can lead in some cases to a missed or incorrect diagnosis. As shown in Figure 4, when the potassium values were below normal, the average of the hemolyzed samples fell into the range of normokalaemia. In contrast, for samples at the high limits, the upper limit of the systematic error fell into the range of hyperkalemia.

From our sample, we eliminated the samples collected from three operators. This was because the personal hemolysis rate of each of them exceeded the mean of the total sample by +2SD. This finding may be associated with the fact that the correct use of POCT for the detection of hemolysis is associated with a large proportion of operator-related error, a major limit of this instrument.