Use of relative survival to evaluate non-ST-elevation myocardial infarction quality of care and clinical outcomes
European Heart Journal - Quality of Care and Clinical Outcomes
Use of relative survival to evaluate non-ST-elevation myocardial infarction quality of care and clinical outcomes
Marlous Hall 2
Oras A. Alabas 2
Tatendashe B. Dondo 2
Tomas Jernberg 0 1
Chris P. Gale 0 2
0 York Teaching Hospital NHS Foundation Trust , York , UK
1 Department of Medicine, Section of Cardiology, Huddinge, Karolinska Institutet, Karolinska University Hospital , Stockholm , Sweden
2 Division of Epidemiology and Biostatistics, Leeds Institute of Cardiovascular and Metabolic Medicine, University of Leeds , Leeds , UK
Survival after non-ST-elevation myocardial infarction (NSTEMI) is high and non-cardiovascular death has become more frequent. Observational studies typically quantify quality of care and clinical outcomes using all-cause mortality, which nowadays may not reflect the impact of index NSTEMI. We review and investigate relative survival for quantifying longer term outcomes after NSTEMI. National cohort study of hospitalized NSTEMI (Myocardial Ischaemia National Audit Project; patients: n ¼ 346 546, hospitals: n ¼ 243, countries: England and Wales). Mortality rates derived from two relative survival techniques were compared with all-cause mortality, and the impact of relative survival adjusted patient characteristics compared with those from Cox proportional estimates. Cox proportional hazards models provide lower survival estimates because they include deaths from all causes, overestimate the impact of increasing age on survival, and underestimate temporal improvements in care. The Royston - Parmar model allows more accurate estimation of relative survival because it is flexible to the high early hazard of death after hospitalized NSTEMI. All-cause mortality gives an overall assessment of survival for a cohort of patients. Relative survival provides a more accurate and informed estimation of the impact of an index cardiovascular event and, if necessary, patient characteristics on survival. - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Introduction
The increasing global burden of cardiovascular survivorship, as a
consequence of an aging, and more intensively treated
population,1 – 4 has resulted in a need to prioritise survival studies as a
means of evaluating quality of care and clinical outcomes.5,6 Over
the last 30 years, there has been a shift from predominantly
cardiovascular to non-cardiovascular deaths after percutaneous coronary
intervention.7 To date, however, estimates of survival after a
cardiovascular event are typically reported using all-cause mortality. Such
data also include the risk of death from causes other than
cardiovascular disease, and cause-specific risks of death cannot be
determined explicitly from these analyses.8,9
Specifically, studies that use all-cause mortality to investigate the
effect of age on survival may overestimate the effects of increasing
mortality rates with age regardless of whether the patient has
cardiovascular disease or not. Similarly, a study describing temporal
improvements in all-cause survival may not be able to accurately
determine whether this was associated with clinical factors (such
as improvements in the application of evidence based treatments)
or was a reflection of decreasing mortality rates over time in the
general population. Moreover, all-cause survival rates limit
international comparisons of care—due to potential differences in the
underlying mortality rates between populations with different
health profiles and provision.
One solution is the use of cause-specific mortality data, which is
the proportion of patients who died from the disease being studied;
arguably more useful to patients, clinicians, and service providers
than all-cause mortality rates.10 However, cause-specific mortality
data are often unavailable and where it is available, it has been shown
to be unreliable for cancer11 – 14 and cardiovascular-related15 – 17
causes of death. Inherent problems with cause-specific death data
arise due to disagreements between clinicians or recording of
mechanisms of death rather than underlying causes.15,17 In particular,
there is evidence to show that death from coronary heart disease
tends to be over-represented within death certificate data.16
In contrast, relative survival techniques provide an estimate of the
likelihood that a patient will not die from causes associated with the
disease under study, without the need for cause-specific death data
provided the assumption that factors affecting disease-specific and
other cause survival are controlled for by comparing with the
mortality in the background population.18 Such methods are well
established and are the gold standard in cancer survival (...truncated)