While ABG results are indispensable to modern respiratory care, the process of both obtaining them and interpreting them is fraught with opportunity for error. Accurate results depend on what happens before, during, and after sampling.

By Larry H. Conway, BS, RRT, LRCP, FAARC

Arterial blood gas tests (ABGs) remain a critical component in the arsenal of diagnostic tools for critical patients. While fewer ABG tests tend to be required today, owing to improved methods of noninvasive assessment of oxygenation and ventilation, ABGs remain the gold standard for assessing these critical components, the bicarbonate level, and the cumulative impact of all these parameters on the body’s true holy grail, the pH.

If the body’s pH moves outside even a relatively small range from normal, critical chemical reactions cannot occur, and the body dies. If the pH falls much below 6.90 or much above 7.80, life literally is in the balance. Consider that the PaCO2 can fall by more than 75% of its normal value or increase to more than double normal and still not crash the fine balance of the body. By contrast, the pH range that can sustain life is only about ±7%. That is why the body intrinsically and primarily seeks to protect and normalize the pH.

We think of arterial sampling as relatively simple and error-free. Our instruments are incredibly precise; they auto-calibrate and await our needs. However, there are multiple potential pitfalls in obtaining accurate arterial blood gases and applying the results properly. While we generally focus on confusion over the interpretation of ABG results, there are issues that can come into play before, during, and after sample analysis. And of course, even if accurate, the results could be misinterpreted or misapplied.

Before we even get to the issue of interpretation, we must first recognize, consider, and prevent what are referred to as preanalytical errors, analytical errors, and postanalytical errors. If these are ignored, you might be puzzling for hours over ABG results that are inaccurate, useless, and, worse, perhaps even deadly if used as the basis of treatment.

And make no mistake, such errors can be deadly. Patients have crashed, coded, and died because of reductions in FiO2 based on reported—but false—improvements in PaO2, which were subsequently determined to be due to air-contamination of samples, not patient improvement. Other patients have been taken back to surgery to locate missed “bleeders” because of dramatically decreased hemoglobin and hematocrit values, which were later determined to be the result of massive dilution of the sample by liquid heparin in an exceedingly small sample. Similar sequelae have been averted in more fortunate cases by critical review of invalid ABG results.

Preanalytical Errors

We will begin, as they say, at the beginning. What are the most common preanalytical errors? They largely relate to the handling of the sample but also can include poor collection technique and lack of attention to detail.

The first potential error is in drawing the sample from the incorrect patient. This could play havoc with the course of treatment of a critical patient, if not detected before treatment is altered. Two other preanalytical errors are closely related to this: a) results posted to the incorrect patient record, or b) mislabeling a proper sample. Any of these three errors could lead to serious consequences for both patients. These consequences also could be devastating for the hospital, physician, nurses, and the respiratory therapist.

Perhaps the most likely potential preanalytical error is obtaining a nonarterial sample and not recognizing it. A variety of situations, from systemic blood pressure to close artery-vein alignment, might cause a syringe to fill with venous blood mimicking arterial filling. Further, in some cases a sample cannot be obtained without aspirating. The need for aspiration or very slow and/or nonpulsatile filling of the syringe should always bring into question the true arterial nature of the sample. If the results are borderline, it is best to raise the issue with the physician or obtain another, clearly arterial sample to assure valid results.

Clotting is a preanalytical error that makes the sample useless, and that, if introduced to the instrument, also could render the machine nonfunctional. If this is the only available instrument, this could delay results for all patients.

Another potential error is obtaining a sample on incorrect settings or support. This can mislead the entire medical team as to the patient’s needs. If an ABG ordered on room air is obtained while the patient is still on supplemental oxygen, or if the patient is on greater ventilator support than intended, the ABG results can alter the assessment of the patient’s true condition and needed level of support. At the very least, this might expose the patient to another arterial puncture; at worst, it could lead to bad support decisions based on incorrect information.

Air contamination of a sample can dramatically alter the reported PaCO2 and PaO2 of the ABG. Both will move toward the value of the PaCO2 and PaO2 in room air. For example, if the patient’s real PaO2 is 250 mm Hg, the measured PaO2 will fall toward room air PaO2, which is about 150 mm Hg. Alternatively, if the patient’s PaO2 is really 60 mm Hg, the measured PaO2 will rise toward 150 mm Hg. How far the PaO2 will move in either case depends on the surface area of the air/blood interface and the duration of the exposure of the blood to the bubble. In either case, this event would render the reported PaO2 unreliable and would require another sample, even if it did not result in an ill-informed change in treatment or supplemental oxygen. Since the PaCO2 of room air is essentially zero, the measured PaCO2 will always fall below the actual value.

Contamination by excessive liquid heparin is another potential preanalytical error. Most current blood gas kits now use dry heparin. However, liquid heparin may still be used for a variety of reasons. If the collection syringe has a large dead-space, if all extra liquid heparin is not expelled, or if a very small blood sample is obtained, the residual heparin can affect the results in a couple of ways. First, as in the case described above, an elevated ratio of liquid heparin to blood can decrease the hemoglobin/hematocrit measurement available on modern instruments. Or, second, the heparin could act as a “liquid bubble,” co-mingling its PaO2 and PaCO2 values with the blood sample just as an air bubble does.

Inappropriate mixing of the sample can alter results. The sample should be thoroughly mixed immediately upon collection to ensure that heparin is spread evenly through the sample to avoid clotting. The sample also should be remixed just before introduction to the instrument to assure a homogeneous sample and adequate protection against clotting. Inadequate or overly aggressive mixing can alter results through a nonhomogeneous sample or hemolysis. Samples should be rolled between the palms and/or inverted gently end-to-end, not shaken or vigorously agitated. Iced samples should be mixed somewhat longer to assure mobilization and mixing of all sample components.

The growing trend of adding electrolytes to ABGs analyte panels introduces at least six more potential preanalytical errors:

  1. alteration of sodium values if sodium heparin is used;
  2. alteration of lithium values if lithium heparin is used;
  3. alteration of pH if pH balanced heparin is not used;
  4. alteration of N+, K+, Cl-, Ca++, glucose, and lactate by dilution with liquid heparin;
  5. alteration of K+ values by hemolysis if the sample is too aggressively shaken or agitated (it should be mixed as described in the previous paragraph);
  6. alteration of electrolyte values if the sample is held while iced for a prolonged period (prolonged icing of the sample should be avoided if electrolytes are to be reported from that sample).

Finally, prolonged delays in analyzing a sample allow for changes in the PaO2 (generally lower) and PaCO2 (generally higher) due to the continuous metabolism of the red blood cells in the sample. Icing the sample slows this change, but as mentioned above, this could adversely impact reported electrolyte values. The degree of impact upon electrolyte values is subject to some debate. Un-iced, promptly analyzed samples are always the best solution if electrolytes are to be obtained from the sample and, in fact, for all ABGs.

Analytical Errors

Some of the more common analytical errors have already been touched upon in this article, specifically, allowing air contamination of the sample, and inappropriate or inadequate mixing of the sample.

A bubble at an electrode can wreak havoc on results. Many instruments protect against large air bubbles, but small bubbles within the sample can sometimes lodge at the sample/electrode interface. One must ensure that even the smallest air bubbles are dislodged and expelled prior to introducing the sample to the instrument.

The use of “out of control” instruments can be a major issue. All laboratory instruments, including blood gas instruments, are required to be evaluated for being “in control” at least every 8 hours. The phrase “in control” means that the response and results from the electrodes are linear and predictable within the functional range of the instrument. An instrument’s electrode response can “drift” from the intended response while still appearing to perform properly when being calibrated.

There are telltale patterns of response that can reveal that the instrument is “out of control.” Levy-Jennings charts help detect such patterns. In addition, Westgard has established a number of “Westgard Rules,” really patterns of response to control samples, that can indicate “out of control” performance of laboratory instruments.1 These help determine when a control sample result should be accepted as random variation or rejected as suspect and requiring further troubleshooting of the instrument. Some such patterns include the following:

  1. Any control value more than 3 standard deviations (SD) from the mean. This should always result in an investigation of the instrument’s performance.
  2. Two consecutive control values for the same analyte more than 2 SD from the mean.
  3. Four consecutive control values for the same analyte more than 1 SD from the mean on the same side of the mean.
  4. Ten consecutive control values for the same analyte on the same side of the mean.

There are other Westgard Rules and variants; however, a discussion and explanation of these are beyond the scope of this article. (Much detailed information on these rules is available through various search engines on the Internet.) Suffice it to say that an instrument displaying any of the four patterns above should be considered “out of control” and not used for producing patient results until their accurate performance can be ensured.

Failing to assure proper calibration is a possible source of analytical error. Blood gas instruments must be periodically calibrated using known value samples. A one-point calibration is required at least every 30 minutes and a two-point calibration is required at least every 8 hours. A test of linearity also is required at least twice a year. Failure to assure proper calibration within the required time of running a sample can result in inaccurate results being produced and reported for patient care. If an instrument cannot be successfully calibrated, it should not be used to produce any patient results until its function has been evaluated and accuracy assured.

Failure to recognize and question results that are strongly contradictory to the patient’s clinical presentation is a potential analytical and interpretive error. Results inconsistent with the patient’s condition should always be considered before blindly posting them to the patient chart. Extremely low PaO2 in a patient who is alert, awake, coherent, and perhaps ambulatory should be questioned. Likewise, a normal PaO2 in a patient showing clear signs of hypoxemia should be suspect. Reporting blood gas values without consideration of their agreement with the patient’s clinical appearance can lead to errors, regardless of the underlying cause of the error.

Finally, there must always be consideration of possible interfering substances within the patient and the sample. The addition of co-oximetry to standard blood gas instruments has increased the importance of this issue. There are few substances that directly interfere with PaO2, PaCO2, or pH. Carbon monoxide might be considered an interfering substance for PaO2, since it alters oxygen delivery without changing the PaO2, but true interfering substances change the results of the associated analyte. The number of potential interfering substances for co-oximetry or electrolytes is large and growing, and the precise impact on various results can vary. Such interfering substances may be introduced to a patient for specific diagnostic tests or as part of an operative procedure. Abnormal species of hemoglobin also can interfere with producing accurate determinations of CO-oximetry analytes or render these results useless for determining oxygen delivery. Failure to consider possible interfering substances can provide results one cannot explain, a false sense of comfort, or both. They can mask clinically dangerous realities.

Postanalytical Errors

The simplest postanalytical errors are miswritten, illegible, or mistyped results. A transposition of numbers can have significant impact on the care delivered to a patient if it is not recognized. Many facilities have moved to some kind of electronic reporting system, but often the results are still transferred from the blood gas instrument to the electronic record through manual entry, and that is fraught with error potential.

Another common postanalytical error is entering the results on the wrong patient. Electronic reporting systems that double-check accession numbers and order numbers decrease this risk, but it can still happen, and the impact on either patient can be huge.

Reporting incorrect information about what support the patient was on when the sample was obtained also can mislead those caring for the patient. A good PaCO2 for a patient reported to be on CPAP could lead to devastating errors in care if in fact the patient was still on ventilatory support at the time of the sampling. The impact grows even greater a few hours later when the care team has rotated out and the new team is trying to determine the patient’s progression and reaction to different types of therapy and support.

One must always review results critically before reporting them for patient care. Reporting results obviously inconsistent with the patient’s condition or that are physically impossible could cause serious harm if those acting upon them are unaware of the error. This could include either reporting PaCO2 levels that are higher than those possible based on the alveolar gas equation or reporting pH, HCO3-, and PaCO2 values that are mutually exclusive (inconsistent with the Henderson-Hasselbalch equation). The impact on the patient could be significant.

Finally, one could totally misinterpret the blood gas values. You might mess up on respiratory acidosis or metabolic alkalosis. Or perhaps the patient is in a prolonged asthmatic attack and the PaCO2 finally has begun to normalize; everyone breathes a sigh of relief, not realizing that this does not signal an end of the attack but impending respiratory failure and the need for immediate intubation and ventilator support. The patient is exhausted. This drives home the fact that blood gases must be interpreted not as discrete snapshots of the patient, but in the context of the patient’s course of illness, treatment, and total clinical presentation.

It is great to be able to determine the acid-base balance and oxygenation status. However, being able to correlate these to the patient—to determine if normalization is good or bad, or to know if the time for concern has passed or just gotten more critical—is the real key to making effective interpretation and use of the arterial blood gas test.


Larry H. Conway, BS, RRT, LRCP, FAARC, is the chief of respiratory service for the VA Medical Center in Washington, DC. He will be contributing a subsequent article on the proper interpretation of blood gas results. For further information, contact [email protected]


  1. Westgard JO. Westgard Rules. Westgard QC website, http://www.westgard.com/westgard-rules/