COPD Assessment Test –Rationale, Development, Validation and Performance

Rationale for CAT development

Health status measurement has become a standard component of assessment in clinical studies using instruments such as the Chronic Respiratory Questionnaire (CRQ) (1) and SGRQ (2). However these are impractical for use in routine clinical practice, and there is a need for shorter simpler measurements. The first of these was the Clinical COPD Questionnaire (CCQ) (3), which began life as the COPD Control Questionnaire. Relevant items were identified from patient interviews, and selected for inclusion in the final instrument based upon the judgement of physicians. In some respects it can be seen as an instrument built from the clinician's perspective. An attempt was made to develop another instrument (the COPD-AQ) that could assess stability in COPD, but the concept of stability in this disease proved to be a methodological challenge, so its development was discontinued (4). The objective of the CAT development program was to develop an instrument that could be used in clinical practice to aid communication between patient and clinician, cover all the important symptomatic areas of COPD in a simple to use, reliable way and have very good measurement properties (5).

Development

The CAT's development has been described elsewhere (5, 6). Some principles underlying its development will be described, since they are key to understanding what the CAT is and what it measures. Questionnaires such as the CAT should capture the whole patient experience, by carrying out interviews with patients until no new items are identified. These are then consolidated into generalizable items that capture the essence of the patients’ descriptions. In the case of the CAT, this amounted to 21 items. The process of item reduction is key since it will determine the measurement properties of the instrument. Questionnaires are used as if they were a ruler or thermometer, but such properties cannot be assumed, they must be built in during its development.

The core methodology used for the CAT is known as item-response theory or Rasch analysis. There are three keys steps. The first is to test whether the various response categories (in the case of the CAT: 0, 1, 2…6) are correctly ordered. Which means that, when the patients indicate a response of 3, it they are clearly worse than 2, but not as severe as 4. The next step is to test how representative each item is of its measurement of the overall concept ‘overall symptomatic impact of COPD’. Rasch provides a number of ways of doing this, but this can be summarised by the ‘item response characteristic curve’, which tests how well the responses to one item fit with the responses of all the other items (6, 7). The item process is iterative, removing the less well-performing item and re-testing until the measurement properties start to deteriorate. Rasch models provide a number of statistics to identify this point, which with the CAT occurred at 8 items (6). The third key step uses a process similar to the item response curve to identify items that men and women answered differently or were influenced by language or country. The end result is the smallest set of the best items that describe the overall symptomatic impact of COPD, uninfluenced by gender or language that, when used together, provide a ruler-like measure. The 8 CAT items together provide a scoring range of 0–40.

Validity

The CAT correlated well (typically r≈0.8) with the SGRQ in every country in which it has been tested (8–11) (12) and also with the SGRQ component scores (12). When specifically tested, the slope of the CAT-SGRQ relationship was not significantly different between four Asian countries (12). A Bland-Altman plot showed that the relationship between CAT and SGRQ was consistent across the scaling range of the two instruments (13). The CAT has also been shown to correlate with each of the components of the CRQ, both between patients and within patients following pulmonary rehabilitation (14). Finally, the CAT and the CCQ have also been shown to be correlated (15). The CAT is minimally influenced by the presence of comorbidities (8, 16), only being slightly higher (2.4 units) in patients with three vs. two comorbidities (8).

Two studies have shown that the CAT is responsive to pulmonary rehabilitation, with significant changes of 2.2-2.9 units reported (14, 17). Differences in score between stable patients and those with an exacerbation have been reported in the range 4.7-6.6 units (6) (8, 11, 16). Stable ‘frequent exacerbators’ had scores that were 2.7 units worse than ‘infrequent exacerbators’ (16). In another study, the score was 20.2 in patients with 0–1 exacerbations; 23.5 in those with 2-4 per year and 28.5 with >4 per year (18). With the onset of an exacerbation the score worsened by 4.7 units (16), and during recovery in patients judged by clinicians to have responded after two weeks the score improved by 2.8 units, whereas there was no change in non-responders (14). Four weeks after discharge following hospitalization for a COPD exacerbation the improvement in CAT was 8.9 in patients reporting their health state as “much better”, and 4.6 in those “slightly better”. A recent study reported only in abstract form that the CAT score predicted future exacerbations (19)

CAT scores in different patient populations

In primary care patients in Europe, the mean score was 17.7 units, and the range across the seven countries in which they were recruited was 14.6 –18.9 (8). In patients recruited from 1y and 2y care in four Asian countries, the mean score was also 17.7 units (12), and in a UK university hospital outpatient population the mean score was 23.1 (18). In a recent epidemiological study in nine Middle Eastern countries and Turkey, the CAT score in Arabic speaking participants who met the study criteria for a diagnosis of COPD was 16.2 and in Turkey it was 20.9 (20).

Scores in subjects without COPD are now being reported. In 481 Canadian subjects without COPD the mean CAT score was 6.9 (21). In an epidemiological study of 5639 subjects in Turkey and the Middle East, the mean scores were 7.0 in the Arabic speakers and 9.9 in Turkey (20). In female non-smokers the mean score was 7.6 and 9.0 in smokers; in males scores the corresponding scores were 5.0 and 7.4. In the Arabic speaking participants 81% had a score ≤ 10; in Turkey this proportion was 65%.

Application in routine practice

The CAT requires only 2-3 minutes to complete and score, and a recent study showed that it can be used in routine hospital clinic practice (18), however there were no data to show that its use altered clinical outcome. To partially address this gap an innovative randomised controlled trial was designed to test whether it improved primary care physicians’ assessment of COPD (22). This found a small but statistically difference in the physicians’ evaluation of COPD symptoms when given the CAT.

The CAT has been incorporated in the GOLD 2011 combined assessment, to establish a threshold at which patients become sufficiently symptomatic to justify regular treatment (23). The selection of a cut point ≥ 10 units for this purpose was based on analysis of the broader impact of COPD associated with different CAT scores (13).

Pharmacological studies

No pharmacological studies of the response of the CAT to pharmacological intervention have been reported.

Minimum clinically important difference (MCID)

The MCID of the CAT has not yet been established. There is no agreed standard method of assessing the MCID, but mapping the CAT to the SGRQ gives an MCID at a population level of 1.6 (24). No MCID at an individual patient level has yet been identified, but it is likely to be 2, rather than 1 (individual CAT scores do not have decimal places). When used in responder analyses to compare two treatments (or active treatment with placebo), both estimates are likely to use similar results (25) (26).

Summary

In the three years since it was published the CAT has triggered a surprisingly large number of papers. In cross-sectional and longitudinal studies it correlated well with scores obtained with the more complex CRQ and SGRQ. Scores in COPD patients appear to be similar across the world, and are higher in people with more frequent exacerbations. It is responsive to pulmonary rehabilitation and changes with the onset and recovery from an exacerbation. Progress is also being made in establishing values to be expected in people without COPD both, smokers and non-smokers. It has been incorporated into assessment guidelines, although its contribution to the routine monitoring of patients has yet to be established. Perhaps most importantly, it enables rapid and reliable assessment of a patient's health status in routine practice.