Increasingly sophisticated instruments are being used for monitoring and guiding patient care but the onus is on the observer to be aware of the limitations of measurements and the causes of error.
Confounding errors in clinical measurements are readings that are not true reflections of the underlying signal. They may be caused by instrumental, sampling and patient factors.
The measuring system must be accurate and precise to produce reliable clinical measurements. Accuracy is the difference between the measurement and a ‘gold standard’ measurement of the underlying signal. Precision is the reproducibility of repeated measurements of the same biological signal. A repeated consistent measurement may be precise but inaccurate, for example an erroneous arterial pressure in an un-zeroed system. A continuous measuring system needs:
• high signal-to-noise ratio
• good zero stability and constant gain to prevent slow drift
• linearity of amplification over a range of signal
• no amplitude or phase distortion of fundamental frequencies
up to the 10th harmonic
• correct calibration by the user.
The ECG measures potentials of 0.5–2 mV at the skin surface. Noise and interference that obscure these tiny signals are the main cause of a poor ECG trace. The most obvious error is incorrect electrode placement relative to the heart because QRS complexes vary with position of the electrodes (Figure 1). The ECG signal must be isolated from noise and interference before amplification and display (Figure 2).
Noise originating from the patient – high frequency electromagnetic activity caused by shivering or movement is filtered by low-pass filters.
Noise originating from the patient–electrode interface – high skin impedance is the most common reason for a poor ECG signal and can be reduced by de-greasing the skin with alcohol. Modern disposable electrodes are made of silver coated with silver chloride with stable impedances reducing slow drift. Gel-impregnated foam pads decrease movement artefact and lower skin impedance.
Noise originating from the environment – the operating theatre is cluttered with sources of AC current that may distort ECG recordings (Figure 3).
The very small currents caused by capacitative and inductive coupling are exaggerated if electrode impedance is high, causing severe 50 Hz interference. Radiofrequency signals may saturate the input stages of the amplifier and block any output. Newer diathermy machines with higher frequency output may produce less interference.
Modern ECG monitors often allow selection of a diagnostic filtering band up to 100 Hz while the monitoring mode is a narrower band up to 50 Hz, which reduces motion artefact and mains interference. These high pass filters attenuate slow drift and maintain a stable baseline, but low-frequency elements of the ECG, such as the T wave, may appear biphasic or distorted.
Invasive blood pressure measurement
Intra-arterial cannulation with electromechanical transduction provides continuous monitoring of arterial blood pressure with greater accuracy in patients with unstable blood pressures or hypotension. The arterial pressure wave narrows and increases in amplitude in peripheral vessels, so the systolic pressure is higher in the dorsalis pedis than in the radial artery. Improper damping and calibration account for a large percentage of the errors in direct arterial pressure monitoring.
The fluid and diaphragm of the pressure transducer constitute a mechanical system that oscillates at the natural resonant frequency (Fn). This determines the frequency response of the measuring system. Factors that decrease Fn to within the range of frequencies of the arterial waveform result in sine wave oscillations being superimposed on the blood pressure trace. Depending on the shape of the arterial pressure wave, this distortion can introduce a 20–40% overshoot error in systolic blood pressure readings; it is worse with tachycardia. The most common reason for an under-damped spiked arterial trace is soft tubing inserted to extend the arterial line.
Damping results from friction of the fluid moving within the tubing which tends to extinguish any oscillations and decrease the frequency response of the transducer system. Excessive damping causes loss of detail in the waveform and underestimation of pressures. Air bubbles in the system, clotting or kinking in the cannula and arterial spasm increase damping. The use of relatively wide-bore cannulae and minimizing stopcocks improves the frequency response of the system.
The amount of distortion in a system is assessed by snapping the flush valve and observing the response (Figure 4). Commercial damping devices (Accudynamic) are available to avoid the risks of air embolism. It is essential to calibrate and zero the transducer system correctly to the right atrium. Baseline drift of the transducer’s electrical circuits may occur requiring periodic checks and re-zeroing.
Non-invasive blood pressure measurement
Indirect methods of blood pressure measurement provide intermittent readings that tend to overestimate at low pressures and underestimate at high pressures. Readings vary according to the site of measurement owing to hydrostatic effects and increased arterial pulse wave velocities in the peripheries.
Manual methods – errors in the auscultatory method occur because of deficiencies in the transmission of sound energy owing to overlong or loose stethoscope tubing. Korotkoff sounds may be difficult to hear in hypotension and arteriosclerosis, resulting in underestimation of blood pressure. Readings in shivering patients require an excessively high occluding pressure and produce pseudohypertension. Further measurement errors may occur due to the auscultatory gap and confusion over whether phase IV or V is to be taken as the diastolic point. Phase V is closer to directly measured diastolic pressure with better inter-observer agreement, but may be absent in high output conditions (e.g. aortic regurgitation, pregnancy or after exercise).
Common sources of error are wrong cuff size, zero and calibration errors in aneroid manometers, and leaks from the pneumatic system that prevent a slow and controlled deflation of the cuff. The width of the cuff should be 20–30% of the circumference of the limb, of which the pneumatic bladder should span at least half and should be centred over the artery. Narrow cuffs yield erroneously high readings while wide cuffs underestimate the pressure.
Automatic methods based on the oscillometric principle rely on a regular cardiac cycle, and inconsistent readings may occur in patients with atrial fibrillation. Over-frequent readings may cause impeded blood flow and can produce inaccurate results. The appropriate cuff size must be used. The apparatus fails to record at pressures below 50 mmHg.
Systolic pressure correlates well with a bias towards overestimation at low pressures and underestimation at high ressures. Mean arterial pressure shows a less satisfactory correlation with a similar systemic bias as does diastolic pressure. Newer devices deflate the cuff continuously rather than in steps, thus achieving a shorter measurement cycle.
Cardiac output measurement
Thermal indicator dilution with saline at room temperature is used to measure cardiac output. Sources of error relate to incorrect pulmonary artery catheter placement and incomplete mixing of the bolus with venous blood. The thermistor probe must be matched to the cardiac output processor and errors related to data input are reduced due to automated computer control that rejects any artefacts or non-exponential curves. Errors may arise when there is tricuspid incompetence or cardiac arrhythmias, and measurements should be averaged over several beats. Injections during the inspiratory or expiratory phases of mechanical ventilation may vary by up to 50% in calculated cardiac output. To minimize this error, injections are made at end-expiration.
Central venous pressure (CVP)
Low-pressure CVP measurements require that the saline manometer or tip of the stopcock in a transducer system are accurately referenced to the right atrium. If the manubriosternal junction is used the measured pressure is 0.5–1.0 kPa (5–10 cm H2O) lower. Errors are also caused by the catheter being inserted too far or being blocked.
Oximeters determine the oxygen saturation of haemoglobin by measuring the absorbance of light by arterial blood. The pulse oximeter is accurate to within plus/minus 2% but is less accurate below saturations of 85% (Figure 5). Algorithms in commercial devices are based on studies in volunteers breathing low concentrations of oxygen, and calibration points are in the range 80–100%. Values below 70% are extrapolated from higher readings and may be grossly inaccurate. The absolute measurement of oxygen saturation in arterial blood (SaO2) may vary between probes due to variability in the centre wavelength of light-emitting diodes (LEDs). This produces probe-to-probe variability in the absolute measurement of SaO2 but trends are accurate. Sophisticated devices have a mechanism for identifying the sensor LED wavelengths allowing internal correction for different wavelengths.
The most common causes of errors are signal artefacts, which produce a poor signal-to-noise ratio. Artefacts have three major sources: ambient light, low perfusion and especially movement. The light detectors cannot differentiate the red light LED wavelength from other sources of light such as infrared heaters and room lights. Manufacturers have incorporated minimum values for signal-to-noise ratios below which the device displays no value for oxygen saturation (SpO2) and modern oximeters display a low signal strength error message. Patient motion may be the most difficult artefact to eliminate. The device can average its measurements over a longer period to reduce the effect of an intermittent artefact but this slows the response time to an acute change in SaO2. Increasingly sophisticated algorithms are being incorporated to identify and reject spurious signals.
Pulse oximetry is a global measure of functional oxygen saturation; it reveals nothing about adequacy of ventilation (e.g. patients with supplementary oxygen with a rising PaCO2), or oxygen content (e.g. a grossly anaemic patient with severely deficient oxygen content and delivery to the tissues with normal saturations). Another source of error is a delay in response caused by instrumental and circulatory delay. The delay in response to step reduction in alveolar gas concentration is longer with a finger probe than with a ear probe. Cold-induced vasoconstriction, hypovolaemia and venous engorgement increase the response time. Atrial fibrillation may also cause measurement errors. Signals may be inadequate in the elderly with arterial disease.
Modern CO2 analysers are based on the absorption of infrared light by CO2 in a gas sample based on the Beer–Lambert law. There are many sources of error. There must be no leaks in the breathing system, the non-rebreathing valve should be functioning perfectly and the inspired gas well mixed and of constant composition.
Errors related to the machine include the overlap of absorption wavebands of different gases so that the nitrous oxide (N2O) absorbs some of the infrared energy within the CO2 bandwidth and the CO2 measurement is falsely high. A second error arises from ‘collision broadening’ in which the absorption spectrum for CO2 is widened by the presence of other gases such as N2O and N2.. The simplest method of eliminating these errors is to calibrate the instrument with gas mixtures containing the same background gas concentration as that to be analysed. Modern infrared analysers provide a simultaneous breath-by-breath analysis of CO2, O2, N2O and the volatile agents. Microprocessor technology can automatically identify and correct for the presence of other gases. Exhaled alcohol can confound vapour concentration in these machines.
Response time – a capnograph needs a rapid response time because a slow response may result in failure to reach the true maximum values at normal breathing frequencies. This alters the capnogram profile with prolongation of phase II with a steeper phase III (similar to asthma), decreases in peak end-tidal CO2 measurements and elevation of the baseline during rebreathing.
Water vapour – a slow response time is caused by long tubing and blockage in the sampling line by secretions or condensation of water vapour, hence the water trap. Liquids and particulate matter entering the measuring chamber cause erroneous readings of CO2 owing to their high infrared absorbance.
Calibration – it is essential to calibrate the machine with the background gas and intended sampling system in place. Modern CO2 analysers should have a three-point calibration with known concentrations of CO2 at regular intervals. Frequent calibration checks are necessary to minimize errors from changes in atmospheric pressure.
Ram-gas effect – because infrared analysers act as partial pressure detectors of the number of molecules of absorbent gas in the sample chamber, changes in atmospheric pressure affect output. A pressure drop across the sampling line (e.g. inadequate suction along an extended sampling tube) causes underestimation of end-tidal CO2.
Errors related to sampling: patient size, respiratory rate and site of sampling may contribute to sampling errors. The optimal sampling site is at the top of the tracheal tube. Errors caused by dilution of the end-tidal sample by fresh gas flow can occur with the Bain (Mapleson D) breathing system; a right-angled connector should be used to prevent this. Sampling of fresh gas flow during expiration may occur inadvertently if the gas sampling flow rate is too high, particularly at lower tidal volumes and faster respiratory rates (e.g. in neonates and infants). The response time of the analyser should be less than the respiratory cycle time to achieve predictable CO2 values and waveform. In smaller children and neonates a sine wave type of capnogram can occur during spontaneous ventilation where there is no clear alveolar plateau or with controlled ventilation in neonates (Figure 6).
Figure 6: cardiogenic oscillations in a patient requiring assisted ventilation after opioid administration.
Errors related to the patient: during a prolonged expiration or end-expiratory pause, when the gas flow exiting the trachea approaches zero, the sampling of the monitor may aspirate gas alternately from the trachea and the inspiratory limb causing ripples on the expired CO2 trace. These are called cardiogenic oscillations and appear during the alveolar plateau in synchrony with the heartbeat and are thought to be due to mechanical agitation of deep lung regions that expel CO2-rich gas. Such fluctuations are smoothed over when the lung volume is increased by application of positive end-expiratory pressure. In patients with asthma or chronic obstructive airway disease, the end-tidal CO2 may under-read, especially if the response time is slow.
Dry spirometers require patient cooperation and depend on voluntary effort. Gas meters such as the Wright respirometer over-read at high tidal volumes and under-read at low tidal volumes because of inertia of the moving parts.
© 2003 The Medicine Publishing Company Ltd