Measurement

14


Measurement




Introduction


The term measurement refers to the procedure of attributing qualities or quantities to specific characteristics of objects, persons or events. For example, how tall you are is an attribute which is measured by a score on the variable ‘height’. Accurate and standardized procedures are available such as a ruler to produce the score (e.g. 175 cm) which is your height.


Measurement is a key process in health research as well as in clinical practice. The same issues are important in both settings. If the measurement procedure used in a study or diagnostic procedure is inadequate, its usefulness will be limited. Similarly, in clinical practice, the validity of diagnostic and treatment decisions can be compromised by inadequate measurement processes and tools. The development of accurate instruments or tools is one of the foundations of scientific and clinical advances.


The aims of this chapter are to:




Operational definitions and measurement


Sometimes researchers start their projects with rather vague views of how to measure theoretical constructs/key factors in their study. For instance, if researchers are interested in collecting data on ‘levels of pain’ experienced by patients, then they must convert this general idea about pain to a tightly defined statement of how exactly this is to be measured. Depending on their theoretical interpretation of the concept of ‘pain’, and the practical requirements of the investigation, one of the many possible approaches to measurement of pain will then be selected.


The process of converting theoretical ideas to a precise statement of how variables are to be measured is called operationalization. It is important that researchers give exact details of how the measurements were taken in order that others may judge their adequacy and appropriateness and be in a position to repeat the procedures in a new study. Data collection is a very important stage of the research process. A quantitative study that is adequate in terms of design, sampling methods and sample size may nevertheless have limited value due to the use of inadequate measurement techniques. Let us now discuss operationalization.


The operational definition of a construct or associated variables is a statement of how the researcher conducting a particular study chooses to measure the variables being investigated. It should be unambiguous and reproducible by other researchers.


At the outset, let us note that in most circumstances there is no single best way of taking measurements. If a researcher claimed that her therapeutic techniques significantly increased ‘motor control’ in her sample of patients, the obvious question that arises is ‘What was meant by ‘motor control’ and how is it measured?’ If our researcher replied that she was interested in motor control as measured by the Plunkett Motor Dexterity Task scores, she has, in fact, supplied her operational definition. Another researcher may challenge the adequacy of this definition and substitute their own, stating that patients’ self-ratings of control on a ten-point scale is a more appropriate definition.


A good operational definition will contain enough information to enable another researcher or clinician to replicate the measurement techniques used in the original study. Similarly, a good operational definition of a clinically relevant variable will enable a fellow professional to replicate the original diagnostic or assessment procedures. An operational definition can be an unambiguous description, a photograph or diagram, or the specification of a brand name of a standard measuring tool. In describing the procedure for quantitative research, we must include operational definitions of the measurement instrument and how they are used, so that readers are quite clear as to what has been done to collect the data.



Objective and subjective measures


A distinction is commonly drawn between objective and subjective measures, often with overtones of suspicion of poor quality directed towards so-called ‘subjective’ measures. Let us make a much less value-laden distinction and define them as follows: objective measurements involve the measurement of physical quantities and qualities using measurement equipment; subjective measures involve ratings or judgements by humans of quantities and qualities. In general, subjective measures are observations (see Ch. 13) of values measured on nominal or ordinal scales, while objective measures are used to produce scores on interval or ratio scales. We will discuss levels of measurement at the end of this chapter.


One should not confuse the distinction between objective and subjective measures as corresponding to good quality or bad quality measurement techniques. Equipment might be improperly calibrated, complicated to use, or become damaged during an investigation. For instance, a researcher might have an absolutely terrible set of scales that gives results far at variance with the true weights of people. With the sophistication and complexity of much current measurement equipment, it is often difficult to calibrate equipment accurately without a complex calibration procedure. Just because a tool is involved in measurement does not mean that the results will be accurate. Furthermore, many quantities and qualities associated with persons and clinical phenomena are difficult to measure objectively, such as the personal attractiveness of individuals, or aspects of patient–therapist relationships, or the intelligence of a person. Ultimately, the issue is whether the best reliable and accurate data has been produced to answer the research and/or clinical question.



Desirable properties of measurement tools and procedures


Measurement tools and procedures ought to yield scores or values that are reproducible, accurate, applicable to the measurement task in hand and which are practical or easy to use. These properties are often given the technical terms of reliability, validity, applicability and practicability. These properties will be reviewed in detail in the following sections. Measurement theory and method are applied to the development of measurement tools that maximize these properties.


Before these specific test properties are reviewed, it is useful to review some basic concepts in test theory. In any measurement, we have three related concepts: the observed value or test score, the true value or test score and measurement error. Thus if I could be weighed on a completely accurate set of weighing scales, my true score might be 110 kg. However, the scales that I use in my bathroom might give me a reading of 100 kg. The difference between the observed score and my true score is the measurement error. This relationship can be expressed in the form of an equation such that:


image


Measurement tools are designed with a view to minimizing measurement error so that the observed value we obtain from our assessment process is close to the true or real value.



Reliability


Reliability is the property of reproducibility and consistency of the results of a measurement procedure. There are several different ways in which reliability can be assessed. These include test–retest reliability, inter-observer reliability and internal consistency. Let us examine each.




Inter-observer (inter-rater) reliability


A common issue in clinical assessment is the extent to which clinicians agree with each other in their assessments of patients. The extent of agreement is generally determined by having two or more clinicians independently assess the same patients and then comparing the results using correlations. If the agreement (correlation) is high then we have high inter-observer or inter-rater reliability.


Table 14.1 illustrates examples of both high and low inter-observer reliability on ratings of patients on a five-point scale. Let’s imagine that this scale measures the level of patient dependency and need for nursing support. As we mentioned earlier, the degree of reliability is quantitatively expressed by correlation coefficients. However, by inspection you can see that in Table 14.1 there is a high degree of disagreement in the two observers’ ratings in the ‘Low reliability’ column. In this instance the clinical ratings would be unreliable, and inappropriate to use in the research project. However, the outcome shown in the ‘High reliability’ column in Table 14.1 shows a high level of agreement.


Related posts:

  1. Surveys and quasi-experimental designs
  2. 11:Questionnaires
  3. Standard scores and normal distributions
  4. Measures of central tendency and variability

Stay updated, free articles. Join our Telegram channel
Tags:
Apr 12, 2017 | Posted by in MEDICAL ASSISSTANT | Comments Off on Measurement

Full access? Get Clinical Tree
Get Clinical Tree app for offline access