Take blood pressure measurement as an example of a measurement task considered pretty well solved by the existence of good instruments. I choose this example because it is not in the realm of social science, perhaps making the issues to be dealt with somewhat starker but less effectively arousing than would an example from psychology or another social science. In fact, however, even more than 100 years after the invention of the sphygmomanometer, problems in measuring blood pressure persist, along with uncertainties about the meaning of the values obtained (Parati 2004
). The medical literature on blood pressure measurement and meaning grows by dozens of articles each year.
The central problem of validity for psychosocial measures is, to reiterate, almost always construct validity, the extent to which we can legitimately claim that a measure reflects variability in the construct it purports to measure. Constructs are hypothetical factors that underlie (cause) behaviors, including those reflected in measuring devices or processes, and they cannot be measured directly, but only estimated. A major problem arises from the possibility that constructs of interest may be conceived at several different levels of meaning, and construct validity at one level may not apply to validity at another level. Another major problem, however, arises from the fact that what constitutes a “measure” may be conceived at different levels, and construct validity may obtain across those levels. To take one simple example, self-reported physical functioning may be taken as a “measure” of what the subject wishes a clinician or researcher to know about his physical functioning, and the construct validity of the measure at that level would probably be quite good. Or, the same self-report could be taken as a “measure” of self-perceived physical functioning, and construct validity at that level would usually be good as long as one could assume that the subject did not have any strong reason to mislead the clinician or researcher. Or, the same self-report could be taken as a “measure” of “actual” physical functioning, as might be done in a study of treatment outcomes. Construct validity at that level might be quite variable, depending on the conditions of measurement, e.g., whether the subject wanted to flatter the clinician or whether the subject wanted to preserve secondary gain from disability.
To anticipate just a bit, the construct “blood pressure” has different meanings; our interest in it is, probably most of the time, in relation to its meaning as a measure of fitness/illness rather than blood pressure per se, e.g., we make allowance for anything that might have had a temporary effect of raising blood pressure. And “the validity” of an instrument is not the same as the validity of the measure (data) it produces.
What Blood Pressure Are We Interested in?
Blood pressure actually involves two elements (concepts), not just one: diastolic and systolic blood pressures. They are measured in the same general way and during the same process, but they are quite different concepts and phenomena and may not be measured equally well (Kay 1998
). Measuring diastolic blood pressure requires a judgment about just when the artery is no longer constricted, and errors may be more frequent or larger in measurement of diastolic pressure. Potentially complicating matters is that blood pressure may be measured, with different results, when subjects are in a standing, sitting, or supine position. Some recommendations for accurate measurement of blood pressure call for measures to be taken in all three positions (e.g., Goldman 2002
). So, that provides for the possibility of six different blood pressures. If the issue were simply one of the dependability (stability) of the measurement, it would do as well to measure blood pressure three times in, say, the sitting position. But the values from standing and supine position measures are thought to carry additional information, i.e., information about an at least somewhat different construct.
Blood pressure is conventionally measured in the upper arm, but measurement at the ankle may also be desirable. Clinicians sometimes measure the blood pressure gradient
between arm and ankle sites (brachial-ankle) in order to assess arterial occlusion in the lower extremities (Baccellis et al. 1997
). And simply to extend the picture, there is some reason to believe that pulse pressure
, the difference between diastolic and systolic pressures, may be critical information (Engvall et al. 1995
; DeStefano et al. 2004
). So, what is a blood pressure device, e.g., the sphygmomanometer, a “valid” measure of? Apparently it is to be taken as a measure of several different constructs, or, perhaps more accurately, as part of a measurement process for different constructs.
Diastolic and systolic blood pressures, although different concepts, are to some extent related to each other (but try to find data on the correlation between them!). The relationship, however, may be conceived as conceptual, or as empirical, or both. That is, diastolic and systolic blood pressures both represent pressure on the walls of arteries but at opposite points in the pumping cycle of the heart. They are probably correlated to some degree, but is that simply an empirical fact or is it inherent in the definitions of the two concepts? It could be that the cardiovascular system is constructed in such a way that, in “normal” people, the stronger the contraction forcing the blood through the arteries, the stronger the residual pressure when the ventricular muscle relaxes. What should the diastolic/systolic relationship be, then? High? Moderate? Low? But maybe the system is not constructed in that way at all, and there is no particular reason that there should be any relationship at all, and it is simply an empirical fact (if it is so) that the two pressures should correlate, and we use the data we gather to determine what the correlation is.
The Instrument: Sphygmomanometer
Sphygmomanometer is a strain gauge rigged to turn mechanical pressure into a rise or fall of mercury in a tube. That was an arbitrary but convenient choice. The tube could have contained any other visible liquid. The point is that the pressure of blood on the arterial wall (actually on the tissue within which the artery is embedded) has to be transduced in some way so it can be converted into a useful metric. Blood pressure per se has nothing to do with millimeters of mercury. In fact, because of risks of toxic exposure, mercury-tube sphygmomanometers are disappearing from use in this country today. Yet, the metric for registering blood pressure is still related to the height of a mercury column. Blood pressure devices being sold today depend on one or another of several different mechanisms for detecting physical changes associated with arterial pressure and for transducing the signal to produce number calibrated in terms of millimeters of mercury, e.g., even if the output is in the form of a pointer on a dial or a digital readout. These various devices yield values that are highly related to one another in correlational terms, but they are not all equally good “measures” of blood pressure. That is because, although the values from two devices may be highly correlated, the absolute differences between estimates of blood pressure may be sufficiently large as to imperil correct interpretation of the findings.
The foregoing are details of the measuring device (instrument), somewhat like variations in the ways in which personality scales might be presented or responded to. The results from scales with the same name presumably are intended to be equivalent—actually equal in the case of blood pressure since the metric is fixed. The equivalent for most social science measures is to fix the metric by converting scores into percentiles or some such. That metric depends for its interpretability, although, on the equivalence of the populations on which norms are established or on whether it makes sense to compare scale values of a person with those of other persons …
Blood Pressure Is Estimated
It is important to remember that under most circumstances blood pressure is not measured directly. As directions for one set of devices note, “When you take a patient's blood pressure, you're measuring the pressure in the cuff—only indirectly are you measuring the pressure in the blood vessel.” That same principle applies even more cogently to most measures in psychology and other social sciences. Characteristics of persons in which social scientists are interested are rarely measured directly but must be inferred from indicators, most of which are much more tenuously linked to underlying psychosocial constructs than cuff pressure is to blood pressure.
The relationship between the way the measure of blood pressure is structured and the underlying variable is reasonably transparent, but it is to some extent arbitrary, and a standard measure of blood pressure could have taken some other form. The average of blood pressures for standing, sitting, and supine positions could have become standard. Alternatively, an average of morning and evening blood pressures or from the two arms, or blood pressure taken after moderate exercise could have become standard. With modern equipment and computers it is possible to have a measure of average blood pressure over time, a measure of maximum systolic blood pressure, of resting blood pressure, fasting blood pressure, stress-induced blood pressure. In fact, devices and arrangements are available to make such measurements. Convenience and expense, however, dictated the directions taken over many decades in developing the standard way of measuring blood pressure. It is worth keeping in mind that blood pressure is not merely a screener, the results of which are necessarily to be checked by a series of other, more precise measures. People are put on antihypertensive medications solely on the basis of blood pressure readings taken in a doctor's office, maybe during only a single visit.
In fact, the sphygmomanometer, used by a properly trained person is often considered the gold standard
for blood pressure, e.g., in evaluating home blood pressure devices.2
The provision that it should be used by a properly trained person is an essential codicil in the definition of the proper measurement of blood pressure. Validity is not invariably and simply a property of an “instrument.” Rather, validity must be considered to inhere in a system or process of which the instrument itself is only a feature. An enlightening view of reliability argues persuasively that reliability is a characteristic of data
not measures (Thompson and Vacha-Haase 2000
). The same is necessarily true of validity, if for no other reason than that validity of data is limited by its reliability. Is a tape measure a valid measure of length of pieces of lumber? Only if the person using the tape measure understands its use and follows the usual conventions of its application. Is the Hamilton Rating Scale for Depression (Hamilton 1967
) a valid measure of depression? It can be considered so if it is properly used, but not otherwise.
One of my friends, a diabetologist, will measure blood pressures of his patients only after they have sat quietly in the waiting room for 20 minutes. He has noted that blood pressure readings are affected by the ambient temperature outside, by whether and how far patients have walked across parking lots and campus to get to his office, and by other activities. He questions them about such matters as how much coffee or other stimulant they may have taken and when they took it before coming to his office. His use of the instrument is quite unlike that of many other clinicians. Maybe (or maybe not) he produces more reliable, and hence more valid, blood pressure readings than those of other clinicians. So what is the validity of the measuring instrument?
Blood Pressure as a Latent Variable
Blood pressure as measured is widely recognized by physicians and other medical personnel—but by no means by all of either group—as a latent variable, even though they have not much idea at all of what a latent variable is. That is, they recognize that measured blood pressure is one fallible indicator of a construct that is not easily, if at all, accessible. Physicians know, for example, that blood pressure measured in the doctor's office may not be the same as blood pressure measured elsewhere—the “white coat” phenomenon (e.g., Godwin et al. 2004
)—let alone being the same as “walking around” blood pressure. Probably many physicians seeing a patient will measure blood pressure more than once, but very few will adhere to the recommended standard that blood pressure should be measured in standing, sitting, and supine positions—too much trouble, one supposes. These observations do suggest, however, that medical personnel realize that the blood pressure readings they get are merely indicators of some hypothetical, unknowable blood pressures characterizing any given patient. That is, we assume that each person is characterized by some “real,” underlying blood pressure, e.g., perhaps conceived as the 24 hours average of all possible systolic pressures (and the same for diastolic pressures or pulse pressures). What we get from measurements is one or more fallible indicators of that underlying reality.
The constructs of diastolic and systolic blood pressures are not as well delineated as they might be, at least in terms of their implications. That is in part because their implications are not completely understood. After more than 100 years of measuring blood pressures, their construct validity is still open to doubt, in large part because the focal construct is often not explicit. In recent decades it has gradually come to be accepted that systolic blood pressure may have implications for health problems different from those of diastolic blood pressure (e.g., Pocock et al. 2001
). It may be that the level of the maximum pressure on the artery walls is more important than the level of the minimum pressure. Apparently, it still is not known whether sharp spikes of blood pressure may be more important than levels averaged over time. Recent articles have altered interpretations of blood pressures once considered in the “high normal” range so that they are no longer regarded as healthy.