Statistics from Altmetric.com
Editor,—Brown et alcompared the quality of life of patients after myocardial infarction with age and sex adjusted population norms from Oxford (age < 65 years) and Sheffield (age > 65 years).1 This takes no account of social class or place of residence, which are known to influence health profile results.2 Why not use controls and patients from the same community? Also, a comparison of the change in physical functioning score between the two age ranges shows a much greater fall in the controls (24.65 v12.06). This suggests that the Oxford and Sheffield norms are not comparable and therefore confounds any attempt to make inferences by age group. The eight (short form) SF-36 scales can be summarised into physical and mental components, which are standardised to a mean score of 50, the population norm.3 This allows interpretation of the quality of life of patients in relation to a general population and has been validated for the UK version of the SF-36.4Surely this is preferable, and more clinically meaningful, to using something as obscure as principal components analysis, which few readers are likely to understand.
Patients who have had a myocardial infarction commonly have a cluster of coronary risk factors such as diabetes mellitus, hypertension, and obesity. Furthermore, atherosclerosis is a systemic disease with many manifestations, and these patients may also suffer from other smoking related conditions. This total burden of illness is likely to have a profound effect on their health profile, swamping the contribution of any single condition.5 Brown et al provide no detailed information on comorbidities, instead focusing on their patients' treatment, which may be an inadequate surrogate. Additionally, employment status is often considered a measure of health related quality of life for men of working age, but failure to return to work can be a cause as well as a consequence of declining health. How many of their patients who were initially employed were still working at follow up?
What was Cronbach's α for each scale? This coefficient assesses reliability by looking at the internal consistency of item responses and is an important measure of data quality.6 The UK SF-36 has a few ambiguously worded items and suffers from floor and ceiling effects in the two role performance scales. The improved UK SF-36 version 2 has eliminated these problems but was not used by Brownet al. Healthy survivor and volunteer effects clearly made the study patients unrepresentative of the initial group. In addition, there is evidence that, as patients come to terms with chronic illness, psychological adaptation occurs such that patients can consider their quality of life as good even when severely limited.6 An earlier administration of the instrument would have mitigated these problems. Finally, the SF-36 health profile adopts a fixed format “medical model” of health related quality of life. Newer questionnaires, such as the schedule for the evaluation of individual quality of life (SEIQoL), take account of individual patient preferences and priorities, and may therefore improve the appropriateness and responsiveness of these outcome measures. Concurrent administration of the SEIQoL, rather than the Nottingham health profile, would have been more interesting and informative.
This letter was shown to the authors, who reply as follows:
Quality of life issues and their measurement are rightly assuming an increasingly important role in health outcomes; however, they are relatively new, not without limitation, and are subject to continuing modification. As Mazeika points out, social class and place of residence can influence health profile results. Even so, assigning social class to housewives and the retired, for example, can be difficult and recent evidence suggests that ecological data are prone to error.1-1
We do not share Mazeika's concerns over the use of normative data from two cities in the UK. We had intended to use controls from the same community but we rejected this as the logistics of generating a potential list of age and sex matched “historical” controls four years later for a cohort that was initially assembled in 1992 were enormous. The regional differences between younger patients from these cities are small and not of the order of magnitude suggested by the designers of the tool as significant. Mazeika's interpretation of the change in physical functioning scores between controls compared to patients is feasible, but we believe that four year survivors of myocardial infarction over age 65 have a quality of life similar to their peers. This may be due either to increasing comorbidity with age or to reduced expectations in the elderly “norms” as we originally discussed.
Patients with atherosclerotic disease may indeed have significant comorbidity and we did attempt to measure this, albeit using surrogates. Approximately 16% of our cohort described their main physical limitation as non-cardiac. We discussed return to work in the original text and accept that it is influenced by economic, social, and personal factors. However, establishing causality in the relation between quality of life and ability to return to work is contentious.
It is ironic that Mazeika describes principal components analysis as “obscure” when this technique was used to standardise scores into the summary scales of physical and mental components of health which he recommends. The analysis of our data took place before validation data on the summary scale scores he cites were published. Because of space constraints, we omitted Cronbach's α, a measure of internal consistency, from the final draft of our paper. For items in the same domain, α exceeded the recommended value1-2 of 0.8 for patients younger and older than 65 years, with the exception of the domain mental health in patients over 65 years where α was 0.74. All domains were significantly correlated with each other, with Spearman's correlation coefficients exceeding 0.3 for all domains as recommended in the SF-36 manual.
Mazeika expresses surprise that we did not use the improved UK SF-36 version 2.1-3 Research projects take time to design, implement, analyse results, and finally undergo peer review and modification before publication. Our questionnaires were distributed in 1996, before the UK SF-36 version 2 was developed. At that time, the original SF-36 was recommended and considered the most appropriate tool for this type of study.
Mazeika states that “Healthy survivor and volunteer effects clearly made the study patients unrepresentative of the initial group”. There is no suggestion in our paper that these four year survivors are representative of all patients with myocardial infarction. The purpose of our study was to describe medium to long term survivors, whether healthy or not. Survivorship may form part of the explanation for some of our findings in the elderly, nevertheless younger patients' demonstrably poor quality of life is hardly likely to be described as “healthy survival”.
Following our experience with quality of life tools, we believe that the combination of a disease specific tool, such as the quality of life after myocardial infarction instrument,1-5 or perhaps the schedule for the evaluation of individual quality of life (a new, patient weighted measure, not without limitation) and a generic tool such as the SF-36 may well offer a more complete assessment of the impact of illness and comorbidity on health related quality of life. Even so, the SF-36 did provide us with evidence that a myocardial infarction makes a young man feel old and an old man feel a bit older. Perhaps most important, Mazeika seems to be missing the essential point of our paper: the quality of life of infarctsurvivors younger than 65 is significantly impaired four years after their acute illness.