The basic principles of potentiometric measurement of ion concentration using the ion-selective electrode chain are discussed in greater detail elsewhere. Similarly, the marvellous properties of ion-selective electrode membranes are interesting enough to merit their own chapter. Additionally, as a main reference for this topic, I refer the readers to Nallanna Lakshminarayanaiah's Membrane Electrodes (2012), as well as Martin Frant's two articles.
Structure of the sodium-sensitive electrode
One cannot speak too broadly, having experience of only one blood gas analyser. The locally available unit apparently features a ceramic pin in lieu of a PVC-coated membrane (though the latter is a valid alternative, and is used in numerous industrial applications).
Unfortunately, Radiometer aren't very specific in their description of their electrode. Given the high degree of selectivity of the electrode, and the absence of a solvent reservoir, one might surmise that this is not one of those "saturated wick" liquid membrane electrodes, where the selectivity is conferred by a hydrophobic oil-based solvent impreganted with ionophores, soaked into a porous ceramic frit. Rather, it may be a ceramic loosely based on one of the old-school glass sodium electrodes, a staple of the electrochemistry lab since the 1950s.
In view of this uncertainty, one cannot help but take the following diagram with a grain of salt. It is not the result of careful scrutiny of the modern ABG sodium electrode, but rather the composite of many ancient articles, and therefore possibly quite incorrect.
Glass seems to be an effective sodium-selective membrane material. All glasses containing more than 0.01mol Al2O2 seem to have a high cation selectivity. What's more, the selectivity for a given cation species changes predictably with changing glass composition, and the glass seems indifferent to anion species. An excellent article by George Eisenman (1962) exists to inform and delight the ion-selective electrode enthusiast. Building on this work, In 1963, Moore and Wilson were the first to evaluate such glass membranes in electrodes designed for measuring the cation concenration of huma body fluids. "Samples of urine, serum, whole blood, and plasma were obtained from normal hospital personnel" who (it seems) were glad to provide everything but CSF (which was obtained over many months "from hospitalized patients with a variety of diseases"). The specific glass used in this early work references Eisenman, and uses his NAS11-18 glass, containing 11mol of Na2O, 18mol of A12O3 and 71mol of SiO2 per every 100ml. The electrode was essentially a slightly modified glass electrode with a membrane thickness of around 0.25mm.
A diagram of a similar setup is available in Richard A. Durst's "Ion Selective Electrodes (1969)", pp. 304.
As you can see, the tip forms a pin of sorts. On can use this fragile relationship to imagine that the modern radimerter E722 sodium electrode has a similar sort of "pin", even though the above-depicted device is only pin-like to suit the purposes of Khuri et al (1963), who shoved these things up into the proximal tubules of pet salamanders.
Anyway. The properties of the cation-selctive glass membrane are discussed in a classic article by Lee and Fozzard (1974). In brief, it would appear that the amorphous crystal lattice of glass contains numerous negatively charged regions into which cations gladly fit. The lattice has enough gaps in it (perhaps about 4 angstrom wide) that cations are able to migrate from gap to gap. Perhaps no single ion would ever migrate across the comparatively astronomical width of the entire membrane, but enough of them change position that some end up in the reference solution inside the electrode, and - more importantly - a potential difference develops across the membrane, which is proportional to the cation concentration in the sample solution.
Generally speaking, glass sodium electrodes tend to have rather extreme measurement ranges. For instance, the linear response range for the Metrohm 6.0501.100 ISE is from 1 × 10-5 mol/L to 1mol/L.
The ABG machine's sodium electrode follows this trend; its linear response range is well outside the normally expected physiological limuts (from 10mmol/L to 250mmol/L).
Other cations can interfere with measurement. Unlike the highly selective ionophore membranes, the glass electrodes only discriminate according to size of the ion, and to a lesser extent charge. This means the sodium-selective glass electrode can be confused by high concentrations of other cations, particularly monovalent ones like lithium and potassium. The selectivity is fortunately high enough that at survivable lithium or potassium concentrations this should never be a clinically relevant source of error.
But is it any good?
The other methods of sodium measurement are flame spectrophotometry and the indirect ion-selective electrode method. There is a well-known difference in results between these methods and the direct ISE, which is probably entirely due to the protein and lipid content of the plasma. The indirect methods measure a sample diluted 1/10th, and then relate the result to the original volume. This routinely returns higher results, and the difference seems to be greatest in patients with the lowest protein levels (conversely, the presense of hyperproteinaemia leads to pseudohyponatremia). The difference is not trivial - Siggaard-Andersen reports a theoretical discrepancy of 20mmol/L in patients who are severely hyperproteinaemic (eg. where the mass concentration of water has decreased by 15% to 0.8kg/L). Additionally, the γ-value (activity coefficient) of sodium changes between theoretical equations and direct measurement, which introduces inaccuracy into both direct and indirect ISE results.
In this case, the more accurate measurement is actually the point-of-care ABG result, because it measures undiluted whole blood. Some authors have even gone on to suggest that central laboratories consider changing to direct ISE measurement techniques.