Article Text

Download PDFPDF
Near-patient testing of potassium levels using arterial blood gas analysers: can we trust these results?
  1. R J P José,
  2. J Preller
  1. Dr R J P José, Department of Intensive Care Medicine, Broomfield Hospital, Court Road, Broomfield CM1 7ET, UK; rjpj{at}


Background: Near-patient testing allows rapid availability of results to enable prompt decision-making. Potassium abnormalities are common in acutely ill patients and can be associated with life-threatening complications. At times there is uncertainty whether clinical decisions can be based on the potassium result obtained from arterial blood gas (ABG) analysers or if laboratory values should be awaited.

Objectives: To determine the opinion of clinicians regarding the use of blood gas analysers to measure potassium and to determine the level of agreement between blood gas analyser and laboratory measurements of potassium in arterial blood samples.

Method: Survey of 64 doctors using a questionnaire and a retrospective comparative study of 529 paired results of ABG and arterial laboratory measurements of potassium in 121 critically ill patients.

Results: 51.6% of the doctors would wait for laboratory confirmation and 48.4% would base clinical decisions on results obtained from the blood gas analyser. The difference between the means of potassium values from the two methods is 0.03 mmol/l (95% CI 0.011 to 0.056; p = 0.0041). The 95% limits of agreement were from −0.485 mmol/l (95% CI −0.524 to −0.447) to 0.551 mmol/l (95% CI 0.513 to 0.590). 95% of the results fell within the difference limits of 0.5 mmol/l.

Conclusions: Most clinicians still await laboratory confirmation of results obtained from blood gas analysers but in this setting there is sufficient agreement between the results obtained from the authors' blood gas analyser and a laboratory analyser to enable effective clinical desisions to be made.

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Potassium is a major intracellular ion and plays an important role in stabilising cell membranes. Changes in extracellular potassium levels can have profound effects on the cardiovascular and neuromuscular systems that may result in cardiac arrhythmias and neuromuscular symptoms, which may be life threatening. Abnormalities in potassium may need to be corrected rapidly, especially when the change in potassium levels occurs over a short period of time or in the presence of electrocardiographic changes. In the presence of abnormal potassium levels, which may require urgent clinical intervention, the clinician needs to know whether the results rapidly obtained from their near-patient testing are sufficiently close to those obtained in the laboratory.

The aims of the study were to determine the opinion and practice of clinicians towards near-patient analysis of potassium and to test the agreement between the potassium results from arterial blood, using a near-patient blood gas analyser and a laboratory biochemistry analyser.


To determine the opinion and practice of clinicians, an e-mail was sent to all doctors working in our hospital in acute specialties. It asked them to complete an anonymous, online, short questionnaire on near-patient potassium testing. The questions asked were: (1) Do you find this method useful to obtain rapid results? (2) How often do you use this method to measure potassium? (3) In your opinion, how often are the potassium results reliable? (4) Do you base important clinical decisions on the potassium results obtained or do you prefer to wait for laboratory confirmation before making clinical decisions? (5) Your grade is? (6) Your specialty is? A total of 64 doctors answered the survey.

We retrospectively compared the potassium levels obtained from arterial blood as measured by our intensive care unit (ICU) near-patient blood gas analyser and a laboratory-based biochemistry analyser. Measurements on arterial blood samples obtained from 1 January 2007 to 28 of May 2007 were extracted from our ICU (general ICU population) database (Metavision version 5.43.17). We searched for arterial blood gas (ABG) samples that were performed within one hour of a laboratory sample. The reason this time difference was used in the search is because of the way in which the results for the two methods are entered into the database and not that the samples were taken one hour apart. If more than one ABG sample was available then the one performed nearest to the time of the laboratory sample was chosen. The usual practice in our ICU is to take the (paired) samples at the same time. Samples were taken on admission and at least every morning.

In our unit the ABG samples are directly taken into an electrolyte balanced pre-heparinised syringe via an arterial line after the first 5 ml of blood has been discarded. The sample is immediately analysed in our ICU blood gas machine (Bayer Rapidlab 865). The results are then automatically uploaded into our database via a computer connection between the database server and the blood gas machine. The entire process takes approximately 2–3 minutes before the results are available. The arterial blood sample that is sent to the laboratory for analysis is taken into a non-heparinised syringe, also after the first 5 ml of blood has been discarded. It is rapidly transported to the laboratory and analysed in an Olympus AU 640 or Olympus AU2700 analyser. The analysis of this specimen is done on plasma and the results are usually available after 40 minutes. The daily morning arterial blood samples for the laboratory are taken between 06:30 and 08:00 hours together with the arterial sample analysed by the blood gas analyser and inserted into our database in the 07:00 hours time slot. Therefore, even though the samples are both obtained at the same time and therefore are directly paired samples, when searching for the pairs in the database a time difference of 60 minutes was used in the search criteria to ensure maximum sample retrieval. There would, however, be a minority of samples that were not taken at the same time. This could happen in particular when multiple blood gas samples were taken around the ICU admission. These would then not differ by more than one hour.

A population of 121 patients with a total of 533 pairs of potassium results from the ABG and arterial laboratory biochemistry analyser were obtained. Four of the paired readings were excluded. In one the values could not be found on manual search through the results, in two there were clear input errors of the values into the database and in one there was a clear sampling error, with all the values obtained from that sample being incorrect. A total of 529 paired results were compared.

The statistical analysis was performed using the statistical program, Analyze-it version 1.73. The two methods were compared using the paired sample t-test and the Bland–Altman statistical method1 of determining the agreement between two clinical measurements.


Sixty-four doctors, comprising 24 consultants, nine specialist registrars, 16 senior house officers, six foundation year 2 doctors and nine foundation year 1 doctors answered the survey. A total of 93.4% (60/64) of the respondents said that they found ABG analysers a useful method to obtain rapid results; 3.1% of the doctors said that they always use this method to obtain rapid measurements of potassium, 31.3% use it often, 34.4% use it sometimes, 28.8% use it rarely and 3.1% never use this method. Some 6.3% of the doctors had the opinion that the potassium results obtained from this method are always reliable, 60.9% said often reliable, 26.6% said sometimes reliable, 4.7% said rarely reliable and 1.6% said that the results are never reliable.

Only 48.4% would base important clinical decisions on the potassium results obtained from an ABG analyser when managing acutely ill patients and 51.6% would wait for laboratory confirmation of the results.

The distribution of the data obtained from our population of potassium measurements is represented in figs 1 and 2. For the laboratory measurements the mean was 3.96 mmol/l (SD 0.605 mmol/l, 95% CI 3.906 to 4.010) and median 3.90 mmol/l (95% CI 3.80 to 3.90) within a range of 2.6–7.1 mmol/l (p<0.001). The mean for the ABG measurements was 3.99 mmol/l (SD 0.584 mmol/l, 95% CI 3.941 to 4.041) and the median was 3.88 mmol/l (95% CI 3.82 to 3.94) within a range of measurements from 2.66 mmol/l to 6.98 mmol/l; p<0.001 (table 1). The difference between the means obtained using the paired sample t-test (table 2) was 0.033 mmol/l (95% CI 0.011 to 0.056; p = 0.004).

Figure 1 Distribution of the arterial blood gas analyser potassium measurements and normality plot. ABG, arterial blood gas.
Figure 2 Distribution of the laboratory potassium measurements and normality plot.
Table 1 Descriptive summary of the mean and median values for potassium using the laboratory versus the near-patient testing method
Table 2 Paired t-test

The measurements were plotted (fig 3 and 4) using the Bland–Altman statistical method and the 95% limits of agreement were from −0.485 mmol/l (95% CI −0.524 to −0.447) to 0.551 mmol/l (95% CI 0.513 to 0.590).

Figure 3 Arterial blood gas and laboratory measurements of potassium with the line of equality. ABG, arterial blood gas.
Figure 4 Difference between the arterial blood gas and laboratory measurements against the mean of both methods.


A previous survey looking at clinician attitudes to near-patient testing has reported that only 34% of clinicians relied on near-patient tests to guide their clinical decisions.2 Our survey found that even though the majority of clinicians find the method of measuring potassium levels on a blood gas analyser a useful method to obtain rapid results, only 48.4% would rely on these results to make important clinical decisions. The rest would prefer to wait for laboratory confirmation when managing acutely ill patients. These clinicians use the blood gas analyser from the ICU. We therefore looked at the level of agreement between our blood gas analyser and the laboratory biochemistry analyser in “real life” clinical practice.

In our study only 27/529 (5%) of the paired values disagreed by more than 0.5 mmol/l and of these outliers only five (0.95% of the total population) did not agree by more than 0.99 mmol/l. Of the 27 outlying paired values 11 (40.7%) were on samples at admission to the ICU. Three out of these five (60%) were on the admission blood samples in the ICU. It may be argued that in this group the samples were not time matched as well as in the morning samples, and that some of the most significant changes in serum potassium levels may have occurred during the admission period as a result of fluid therapy, potassium shifts in and out of cells due to serum pH changes during resuscitation and the initiation of mechanical ventilation and active correction of abnormal potassium levels. A limitation may be that blood gas analysers do not recognise haemolysis and report the potassium regardless of this occurrence. For this study, however, we feel that it is unlikely that when both samples are taken at the same time from the arterial line haemolysis would occur in the sample that goes to the blood gas analyser and not in the sample that goes to the laboratory. If the laboratory reported a sample as haemolysed and did not report a result then that paired sample was not used for the study, as it would not be picked up during our search. The samples were taken from arterial lines without the use of tourniquets and small-bore needles. The larger bore of the cannula and the low velocity of the blood being drawn into the syringe would make haemolysis during the taking of arterial blood less likely. In a real working scenario we would be more likely to get haemolysis in a venous sample taken with a syringe and small-bore needle, using the negative pressure of the syringe to draw the blood and then inserting the blood via the needle into a vacuum container.

Previous studies, looking at smaller numbers of samples, concluded that the results obtained from ABG analysers for potassium should be used with caution due to the limits of agreement being too wide.3 4 In our study we have shown that there is no significant difference between the means. Clinical management decisions, however, usually entail responding to abnormally high or low levels. The Bland–Altman statistical method showed that even in states of hypokalaemia and hyperkalaemia 95% of the results differed by less than 0.5 mmol/l. A prospective study with all samples verified as taken at the same time might prove an even closer agreement at the extremes of the measured values.


Only 48.4% of the clinicians that we surveyed would base important clinical decisions on the potassium results obtained from a blood gas analyser. Our study supports the assumption that there is sufficient agreement between the results obtained from our blood gas analyser and a laboratory analyser to enable effective clinical decisions to be made. Differences in a small percentage of paired samples are probably due to weaknesses in the pairing of the samples rather than in the analytical processes. Clinicians should consider the possibility of haemolysis when interpreting results and repeat any result that does not fit with the clinical impression. Potassium levels obtained from a blood gas analyser are useful, valid and more rapidly available than results from formal laboratory analysis.


View Abstract


  • Competing interests: None declared.

Linked Articles

  • Primary survey
    Steve Goodacre