The Impact of Hemolysis-Index Thresholds on Plasma and Serum Potassium Measurements

Abstract

Background:

Modern clinical laboratory analyzers measure a hemolysis index (H-index) because test results can be inaccurate when intracellular contents from erythrocytes leak into serum or plasma. In 2020, Roche Diagnostics decreased the H-index from 90/100 to 20 for potassium, recommending that laboratories avoid using specimens with an H-index >20; however, there are a limited number of studies investigating the impact of this recommendation on patient testing.

Methods:

Out of 113 916 serum or plasma potassium tests performed within a 6-month interval, 72 patients with potentially hemolyzed potassium specimens (H-index >20) and a second non-hemolyzed specimen (H-index ≤20) within 2 h were identified. The clinical impact of decreasing the H-index and the utility of applying a corrective formula for adjusting potassium results were evaluated.

Results:

The majority of initial test results either had small differences between original and corrected results that would not have affected clinical management or H-indices above the threshold previously recommended by Roche. We estimated the second sample was reported an average of 3 h 23 min after the initial sample was collected, with 95% CI [2 h 37 min to 4 h 8 min], and the median time delay was 2 h 44 min.

Conclusions:

Our analysis does not show a clear benefit from avoiding the use of potassium specimens above an H-index threshold of 20. Our findings suggest these practices may be detrimental in terms of patient safety due to increased turnaround time for a critical analyte.

INTRODUCTION

Red blood cell hemolysis complicates the interpretation of potassium testing because the intracellular concentration of this analyte is roughly 25 times higher than the concentration in serum or plasma. In 2020, Roche Diagnostics issued a bulletin decreasing the recommended H-index threshold for potassium due to concerns that low levels of hemolysis could adversely impact test results. The bulletin advised clinical laboratories to avoid using hemolyzed specimens and defined the acceptable limit as an H-index of 20 (1). Their recommendations were influenced by a study from Martínez-Morillo and Álvarez who analyzed tests from an emergency department laboratory and derived a correction equation to estimate potassium in hemolyzed samples (2). Their study recommended using a correction formula to report qualitative comments describing potassium levels for samples with H-indices between 50 and 500 rather than providing quantitative potassium results (2).

Several subsequent studies have noted that avoiding the use of specimens with an H-index threshold above 20 would impact the reporting of 3% to 12% of potassium results, which could have significant financial implications for hospitals, costing an additional $120 000 annually for phlebotomy and supplies (3, 4). By comparing changes in potassium before and after spiking samples with hemolysate, 2 studies observed that a modest H-index cutoff between 57 and 75 might be more feasible and clinically relevant for laboratories (3, 4).

The purpose of our analysis was to investigate how lowering the H-index threshold to 20 would impact potassium reporting and turnaround time at a large academic medical center using specimens from inpatients, outpatients, and the emergency department.

METHODS

This study was conducted at Beth Israel Deaconess Medical Center (BIDMC), a 673-bed academic medical center in Boston, MA. Serum and plasma potassium and H-index measurements were performed on a Roche Cobas c501 instrument using the ion-selective electrode and serum index generation 2 reagents. At the time of this study, specimens with an H-index greater than 1000 were rejected. Potassium results from samples with moderate (H-index 101 to 300) or gross (H-index 301to 999) hemolysis were reported with a comment stating that the specimen was hemolyzed, and that hemolysis can falsely elevate the test result. Retrospective analysis of potassium testing was performed by analyzing test results obtained between July 1, 2020, and December 31, 2020, and procedures were approved under Institutional Review Board protocol 2021P000531.

Potassium v alues for patient samples with an H-index >20 were compared with those from a second sample collected within 2 h having an H-index ≤20. Parameter queries to extract this in formation were performed using a Microsoft Access (version 2016) database that warehouses test results generated by our laboratory analyzers. Similar to the approach taken in the literature (2), a simplified linear regression equation K_corrected = K_measured – (0.0051 × H-index) was used to categorize potassium results. Corrected potassium values are not reported in the laboratory information system at our institution.

We analyzed the degree of concordance between the corrected potassium concentration based on the H-index from the first sample relative to the potassium concentration in the subsequent non-hemolyzed redraw using 5 categories relevant to our reference intervals: low-critical (K < 3.0); low (K: 3.0 to 3.4); normal (K: 3.5 to 5.4); high (K: 5.5 to 6.0); and high critical (K > 6.0). Turnaround times were evaluated to estimate the impact of rejecting initial specimens on clinical workflow.

RESULTS

Beginning with 113 916 potassium results, we identified 72 patients with potassium results within 2 h for whom the first sample was potentially hemolyzed (H-index >20) and the second sample was non-hemolyzed (H-index ≤20) as indicated in Fig. 1. The majority of these specimens (74% or 53/72 initial samples) had H-index values in the same qualitative category before and after using the correction formula, however, the remaining 19 samples would have been categorized incorrectly and are listed in Table 1. The 95% confidence intervals (CI) for the potassium and H-index in the initial sample were (5.4–6.8) and (206–423), respectively, and the 95% Cis for potassium and the H-index in the second sample were (4.0–4.8) and (4–7), respectively.

Fig. 1.

Overview of workflow analysis to examine the impact of low levels of hemolysis (H-index 20 to 100) on serum/plasma potassium reporting using the Roche Cobas c501 module.

Table 1.

Serum/plasma potassium samples with discordant qualitative interpretations along with H-indices and repeat measurements within 2 h.^a

First sample					Second sample
Potassium (Kmeasured)	H-index	Qualitative interpretation	Corrected potassium (Kcorrected)	Qualitative interpretation (corrected)	Potassium	H-index
3.0	28	Low (3.0–3.4)	2.9	Low critical (<3.0)	3.9	5
3.0	31	Low (3.0–3.4)	2.8	Low critical (<3.0)	4.4	15
3.5	37	Normal (3.5–5.4)	3.3	Low (3.0–3.4)	3.4	6
5.6	64	High (5.5–6.0)	5.3	Normal (3.5–5.4)	6.1	1
6.2	108	High critical (>6.0)	5.6	High (5.5–6.0)	5.3	4
6.6	126	High critical (>6.0)	6.0	High (5.5–6.0)	6.2	4
5.7	147	High (5.5–6.0)	5.0	Normal (3.5–5.4)	5.2	1
6.8	174	High critical (>6.0)	5.9	High (5.5–6.0)	3.8	0
5.6	252	High (5.5–6.0)	4.3	Normal (3.5–5.4)	4.6	9
6.0	306	High (5.5–6.0)	4.4	Normal (3.5–5.4)	4.5	9
6.2	344	High critical (>6.0)	4.4	Normal (3.5–5.4)	3.8	6
6.2	374	High critical (>6.0)	4.3	Normal (3.5–5.4)	3.9	10
6.5	399	High critical (>6.0)	4.5	Normal (3.5–5.4)	3.9	7
6.1	401	High critical (>6.0)	4.1	Normal (3.5–5.4)	4.0	5
8.1	429	High critical (>6.0)	5.9	High (5.5–6.0)	5.5	8
6.6	557	High critical (>6.0)	3.8	Normal (3.5–5.4)	3.4	9
8.4	689	High critical (>6.0)	4.9	Normal (3.5–5.4)	4.1	5
6.5	722	High critical (>6.0)	2.8	Low critical (<3.0)	3.5	2
9.1	789	High critical (>6.0)	5.1	Normal (3.5–5.4)	4.4	9

^aThe correction formula K_corrected = K_measured – (0.0051 × H-index) derived in reference (2) was used to adjust potassium values. Specimens with H-indices greater than the threshold of 90 previously recommended for the Roche Cobas c501 analyzer are shaded in light grey.

Of these 19 samples, 15 had H-indices greater than the threshold of 90 previously recommended for the Roche Cobas c501 analyzer, so the new threshold would not have affected reporting. Of the remaining 4 samples, using a correction formula would alter the potassium result by 0.1 to 0.3 mmol/L and would not have affected patient management. However, one sample had an initial H-index of 64 and reporting qualitatively from the corrected potassium value for this case might have been misleading. The sample taken 2 h later had a potassium value of 6.1, a high-critical concentration that is closer to the original level of 5.6 relative to the corrected value of 5.3, and the clinical note indicated that the patient required urgent hemodialysis after admission due to several missed outpatient dialysis appointments.

We also analyzed the turnaround time and found that the average time to result reporting for the initial sample was 1 h 38 min (95% CI, 1 h 14 min to 2 h 2 min), a figure that incorporates specimen draw, transport, login, and analytic time. To estimate the impact of not reporting results from specimens with an H-index >20 we calculated the average delay between specimens. We found the second sample was reported an average of 3 h 23 min (95% CI, 2 h 37 min to 4 h 8 min) after the initial sample was collected. The median time difference between the initial and repeat draws was 2 h 44 min. This delay is significantly above the goal of <60 min for acceptable turnaround time from registration to result reporting for many common laboratory tests (6).

DISCUSSION

Recent literature suggests several alternative approaches to reporting results for specimens that may have elevated potassium due to ex vivo hemolysis. While a modest decrease of the H-index threshold to between 50 and 75 may be reasonable based on empirical testing and total allowable error estimates (3, 4, 7), our institution does not reject specimens with an H-index between 20 and 100 or replace test values with qualitative comments. These approaches would not be feasible in our practice environment due to the increased turnaround time for this potentially critical test result, and because text appended to test results appears in footnotes in our laboratory information system where lengthy qualitative comments might be overlooked by clinical providers.

Unfortunately, it was not possible to confirm whether recollected samples were performed on the same tube type due to limitations in the specimen data collected in our laboratory database. Therefore, it is possible that there may have been differences in measured potassium values between the first and second draw if the ordering provider switched between serum and plasma specimens. However, we do not think this would have a major impact on our study because the expected differences between serum and plasma samples are relatively small between 0.1 to 0.4 mmol/L (5), there is no rationale for providers to switch tube types between draws, and hemolysis has a similar impact on both sample types.

While hemolysis is a major preanalytical source of error and many studies have proposed correction formulae to distinguish in vivo from ex vivo red blood cell lysis, there is significant variability in these equations due to the semiquantitative nature of serum indices performed across different analyzers and institutions (7). Using the H- index to calculate a corrected potassium level is not optimal because this parameter is not subject to the same calibration and control analyses that are performed on reportable analytes such as potassium (8).

While there may be a benefit to using a correction formula so that potassium results for mildly hypokalemic samples do not fall within the reference interval, laboratories should be cautious with applying these formulae broadly so that patients with hyperkalemia and mild hemolysis are not corrected to a lower potassium level. For a large hospital with a complex laboratory testing menu, adding comments rather than numerical results could lead to missed and delayed results and our analysis does not show a clear benefit from lowering the acceptable H-index to 20 in terms of patient safety.

IMPACT STATEMENT.

Accurate reporting of serum/plasma potassium results is of great importance because severe hypokalemia or hyperkalemia requires urgent medical treatment. Unfortunately, this analyte can be falsely increased due to ex vivo hemolysis. This study analyzes the impact of slightly elevated hemolysis (H)-indices between 20 and 100 on potassium levels in clinical samples at a large academic teaching hospital. Our manuscript contributes to general knowledge because it integrates data from a broad patient population, weighing the need for redrawing potassium samples that have H-indices between 20 and 100 with the necessity of maintaining a rapid turnaround time.