Discussion
In this large, single centre cohort study examining the real-world use of cTn, we made several novel and confirmatory observations. Some of these observations suggest patterns of care that appear discordant with best clinical practices. While appropriate use criteria have established standards for use for many cardiovascular tests, the terms overuse and underuse remain difficult to apply. These terms may be relevant when our data are considered in the context of clinical guidelines and other data sources.
Some have suggested that the contemporary application of cardiac biomarker testing for ischaemia is an example of testing overuse.3 In our sample of 26 663 subjects with cTn testing performed, the proportion with elevation and the positive predictive value of cTn was low. Furthermore, during our study period there were 67 829 inpatient admissions. This corresponds to a total of 39.3% of all inpatients having their cTn measured at some point during their stay. Because of the pattern of pervasive cTn use, which results in a low positive predictive value for diagnosing MI, more judicious use of cTn has been encouraged in editorials and has been the subject of quality improvement initiatives seeking to reduce waste.7 8
The data on ECG use in our cohort raise multiple points for discussion. First, the discrepancy in documentation of ECG between the administrative and manual chart reviews is substantial. As we noted, this is often due to clinically available data (eg, prehospital ECG) not captured because our facility did not have the necessary billing infrastructure. For higher cost testing, such as echocardiography and nuclear stress testing, we did not observe any discrepancy. Even after manual review, however, some 5% of patients evaluated with cTn did not have an ECG performed. Given the simplicity and ubiquity of ECG in the hospital setting, it is perplexing why patients evaluated with cTn, presumably to diagnose myocardial damage, would not undergo an ECG as well. Guidelines give a class I recommendation for performing an ECG promptly in patients for whom acute coronary syndrome is suspected. This has a ‘C’ level of evidence, as there does not seem to be any reason why this recommendation should be empirically tested. The message worth communicating is that patients being evaluated for myocardial ischaemia should be evaluated clinically and augmented with both biomarkers and ECG.
The use of cardiology consultation differed nearly threefold between our cohorts. Unlike non-invasive testing, use of expert consultation is much more nuanced and does not lend itself to categorisation schema such as appropriate use criteria. Thus, we cannot comment as to whether consultation was overused or underused. Some patients with elevated cTn and without cardiology consult may have benefited from cardiology care; we cannot ascertain this from our dataset. We observed greater use of echocardiography with elevated cTn, which is reasonably expected. We saw less use of cardiac CT and SPECT with elevated cTn, which is also expected as patients without elevated cTn would be more likely to undergo a non-invasive ischaemia evaluation. These data would be stronger if we had been able to extract information on appropriateness of testing.9
A key question raised by our observations is: What is the likely impact of hscTn assays on demand for cardiovascular services?10 In 2001, a randomised trial compared the evaluation of emergency department patients with cTn with the then-standard-of-care, Creatine kinase-MB (CKMB).11 While subgroup analyses showed significant differences in hospitalisation and hospital costs, these were not different for the cohort as a whole. A more contemporary study from Spain compared conventional and hscTn finding that the newer assay resulted in some improvement of acute triage offset by higher use of invasive procedures and longer hospitalisations.12 A large registry study from Sweden compared use of cardiovascular services after adoption of a hscTn assay finding modest increases in demand for echocardiography and catheterisation, which did not persist after multivariate regression.13 Other studies evaluating adoption of hscTn found no difference in cardiovascular service demand.14 15 We could only speculate as to how demand for cardiovascular services will change with the transition to hscTn; however, our data on prevalence are informative for future study of this transition. As technology continues to advance rapidly, the medical community and industry partners should consider not just the accuracy of a new test, but also whether the procedure improves outcomes for patients.16
Our investigation adds to the body of literature documenting that elevation in cTn is clinically a poor prognostic indicator associated with mortality in both cardiac and non-cardiac conditions.17–19 While this mortality risk appears to differ based on the aetiology of the elevation,1 20 our data did not allow for this degree of detailed investigation. We did observe a decreased risk of mortality for African-American patients. A review of literature on inpatient mortality and race yields non-uniform findings; some finding African-Americans to have higher mortality, some with lower mortality and some where race was not a relevant factor.21–24 Another study evaluating disparities in care and mortality found that women were more likely to receive suboptimal care and that both African-American and female patients have higher AMI mortality than white men.25 We observed that use of cardiovascular services was less for women and African-Americans; however, we cannot determine if this represents underuse of services in these groups or overuse in white, male patients. It also remains unclear what impact race and sex will have on hscTn adoption given ongoing disagreement whether hscTn assays should have different cut-offs for different populations.26–28
Our study is not without limitations. We were not able to derive a statistical method to determine rise and fall of cTn, as would be preferred for verifying the diagnosis of AMI, within our data set. The data on pre-existing conditions are subject to the limitation of whether they were documented correctly by care providers; data on new hospital diagnoses are drawn from billing and coding. We did not have information on appropriateness of non-invasive imaging or catheterisation. Regression models are an imperfect method for determining risk associations between variables such as sex and ethnicity with outcomes such as consultation and catheterisation. We cannot ascertain any reasons for the disparities observed in women and African-American subjects; these findings should be considered hypothesis generating.