Discussion
Our study of a large, contemporary cohort of NSTEMI patients in the ACTION Registry–GWTG demonstrates wide variation in clinical practice across different risk profiles of in-hospital bleeding and mortality. We also observed that 40% of patients were at higher than average risk for death as well as major bleeding. These challenging dual high-risk patients were treated less aggressively than others, both in terms of acute and secondary preventive pharmacotherapy as well as invasive angiography. Ironically, we found that the high-risk group was paradoxically the most likely to receive excessive doses of antithrombotic medications likely to cause bleeding.
In our observational registry cohort, we found wide variation in treatment patterns across risk profiles. For example, subjects at low predicted risk for both bleeding and in-hospital mortality received the most aggressive adjunctive antithrombotic therapy with higher use of clopidogrel and glycoprotein IIb/IIIa inhibitors, were most likely to receive guideline indicated secondary preventive pharmacotherapy, and were most likely to undergo early invasive management. Paradoxically, we also observed that despite significant differences in demographic characteristics (including age, sex, and diabetes) and clinical presentation, subjects in the two discordant groups (low:high and high:low) received almost identical adjuvant antithrombotic therapy and had similar rates of invasive angiography within 48 hours though this may be an instance where the risk-benefit ratio may be optimized by managing low:high patients more conservatively but high:low patients more aggressively.
The dissociation between treatment intensity and patient risk was most apparent when examining treatment patterns among patients within the same stratum of bleeding but across different estimated risks of mortality. For instance, we observed that among subjects with low estimated risk for bleeding, those with high predicted mortality were significantly less likely to receive clopidogrel or glycoprotein IIb/IIIa inhibitors, to undergo invasive angiography, or to be prescribed clopidogrel or statins at the time of discharge as compared with subjects with low predicted mortality. A similar pattern was observed among patients with a high estimated risk for bleeding with treatment intensity inversely related to predicted mortality.
In contrast to paradoxical management patterns across different risk groups for mortality, in this registry analysis, we found that provider decision-making consistently avoided anti-thrombotics and invasive procedures in those at higher estimated risk for bleeding. Specifically, we noted that among individuals with either high or low risk for mortality, those with high predicted bleeding were significantly less likely to receive clopidogrel or glycoprotein IIb/IIIa inhibitors, to undergo invasive angiography, or to be prescribed clopidogrel at the time of discharge as compared with subjects with low predicted bleeding.
Our findings demonstrating inverse relationships between treatment intensity and patient risk suggest opportunities for improvement in the assessment of major bleeding and mortality in NSTEMI patients. These findings are consistent with a recent study which demonstrated relatively low correlation between patient risk assessment performed by treating physicians and risk assessment from objective tools such as the TIMI, PURSUIT, and GRACE Risk Scores. Given that subjects at high predicted risk for in-hospital mortality and major bleeding were also less likely to receive therapy that is unlikely to increase the risk of bleeding, i.e. statins, practitioners may be reflexively undertreating such patients.
The observation that patients with higher baseline risk of adverse events who are likely to derive the greatest benefit from aggressive interventions are amongst the least likely to receive them is consistent with prior analyses which have examined patients at high risk for adverse events, specifically those with renal dysfunction, the elderly, and female patients. Although specific reasons underlying this paradoxical care or "treatment-risk mismatch" were not captured in the ACTION Registry–GWTG, it is likely that underutilization derives in part from by concerns about 'attributable risk.' Physician decision-making seems to be most concerned about the care they provide resulting in a negative event or outcome (eg, bleeding with an antithrombotic agent) compared to an outcome that they perceive as being out of their control (eg, mortality). Our finding of paradoxical treatment parallels findings in numerous other studies.
In addition to underutilization of evidence based therapies, our study also ironically demonstrates that subjects at high predicted risk for both bleeding and in-hospital mortality who were least likely to receive guideline-indicated adjunctive antithrombotic therapy were most likely to receive inappropriate, excess doses of such medications—significantly increasing the risk of a major bleeding event. These findings are concordant with a prior analysis from the CRUSADE Registry which found that not only did 42% of patients with NSTEMI receive at least one excess dose of an anti-thrombotic agent but excess dosing was particularly prevalent among patients whom practitioners were least likely to administer such medications given fears over bleeding. Our findings demonstrate that excess dosing of antithrombotic therapy remains an important area for performance improvement initiatives.
There are several limitations to our analysis that should be considered. First, hospital participation in the ACTION Registry–GWTG is voluntary and participating sites may have a greater interest in quality improvement than non-participating sites and, therefore, may be more likely to follow evidence-based guidelines thus overestimating or underestimating our results compared to broader clinical practice. Another consideration is unmeasured confounders or details about physician or patient decision-making that are not captured by the registry that could account for the treatment paradoxes observed. While the risk models for in-hospital major bleeding and mortality have been derived and validated using population-based clinical registries, they do not take into account other important, patient-level factors such as frailty, compliance, and personal choices which may influence treatment decisions. Furthermore, our analysis uses observational, non-randomized data, and thus, associations between various treatments and outcomes may be confounded by unmeasured variables. Also, there may have been appropriate contraindications to adjunctive pharmacotherapy or invasive angiography that were not collected. In-hospital clinical outcomes in ACTION Registry–GWTG are not independently verified or adjudicated and the reliability of this data is dependent on the site submitting this information to the registry which could introduce reporting bias. Finally, because long-term outcomes are not captured in ACTION Registry–GWTG, we were limited in examining in-hospital events only. However, to the best of our knowledge, our study is the first to stratify NSTEMI patients based on dual risks of adverse ischemic and bleeding events and examine patterns of care. Much more work needs to be done and randomized data are needed to identify novel approaches for integrated risk assessment to optimize patient outcomes.