Our results demonstrate that an increasing proportion of older patients with AMI have medical conditions that could lead to their exclusion from publicly-reported process of care measures. Indeed, of the patients that could be included in a given quality measure (that is, of those that are not uniformly excluded) up to 69% were in the discretionary category based on chart-abstracted data in 2000–01; they did not have a specified contraindication that would automatically lead to their exclusion from the measure, nor were they ideal for the given treatment. We found, additionally, that rates of treatment with medications for which these patients had potential contraindications also increased. These results highlight the uncertainties surrounding the best care for older patients with relative contraindications to treatment: it is unclear whether inclusion and treatment or discretionary exclusion represents the better care.
Our results build upon prior work that described the growing proportion of older AMI patients with coexisting conditions and potential contraindications to treatment for AMI.6, 9
A study by Masoudi et. al. indicated, for example, that the proportion of patients ideal for aspirin at admission dropped from 67% to 47% over 10 years, with similar drops found for other measured drug therapies. In our study, less than 1% of Medicare patients were ideal candidates for all 5 process of care measures in 2001, which is to say that for nearly every Medicare beneficiary presenting with an AMI, a complex decision has to be made about whether to provide at least one standard medical treatment.
These findings echo concerns about the evidence-base for current quality measures for older patient groups.17, 18
Although the CMS process measures for AMI are based upon substantial clinical evidence, older and sicker patients are rarely included in clinical trials that established standards of care. A number of observational studies have supported the use of aspirin, beta-blocker and ACE-I in older patient populations,19–21
but these generally have also excluded patients with potential contraindications to care. Without the inclusion of such patients in treatment studies it is difficult to judge what treatment decisions are in the patients’ best interest, thus leaving clinicians with challenging medical decisions.
Our findings also raise questions about how best to account for patients with relative contraindications when measuring quality of care. An earlier approach delineated a comprehensive list of potential contraindications for each therapy and excluded all such patients, whether or not they received treatment.9
In more recent efforts, CMS and the Joint Commission, recognizing the potential overriding benefit of treatment for many patients with relative contraindications, now specify a much narrower set of absolute exclusion criteria. This approach supports more individualized decision-making about care, but the allowance for discretionary exclusions complicates interpretation of publicly reported data. First, the use of discretionary exclusions are invisible to the health care consumer, so the public can not discern to what extent the quality measures are representative of the full population of patients seen at the hospital. Second, use of discretionary exclusions may vary greatly between hospitals and thus limit the comparability of measures. Furthermore, the combined factors of 1) discretion about whether to include patients with contraindications and 2) the lack of evidence about what is best for such patients create a situation that may give hospitals an incentive to treat patients despite relative contraindications, and thus hospitals could seemingly receive credit for care whether or not it is in the patient’s best interest. Indeed, we found rates of treatment for patients with potential contraindications have increased over time.
Finally, the exclusion of large numbers of patients from quality indicators raises broader questions about quality measurement. If a substantial proportion of patients are not represented in quality measures, because they are excluded or ineligible, we cannot provide any definitive assurances regarding the care they receive. This is particularly disconcerting because exclusion and ineligibility for process of care measures cluster in older, sicker patients who are more medically vulnerable and are being missed by quality of care measurement. This also has implications for our ability to ascertain quality at institutions when a notable proportion of patients, typically a sicker cohort, are not included in their overall assessment of quality.
There are a number of potential implications of our work. The first, as described, is the need for more evidence upon which to base treatment decisions for older patient groups with multiple coexisting conditions. Second, quality reporting for older patients may be improved by reporting outcomes or quality of life, as opposed to processes of care. Clinical outcomes could include all patients after risk-adjustment for clinical differences between populations and may be more meaningful to patients. Finally, more detailed information on the portion and characteristics of included patients should be reported for currently reported process measures.
A number of factors must be taken into consideration when interpreting of our work. First, we examined older data and cannot determine what course the observed trends in treatment have taken in more recent years. However, these data permitted detailed analysis of coexisting illnesses and treatment. We know of no other nationally representative source of chart-review data on AMI care. Second, we cannot be sure that all of the increases in discretionary exclusions are due to changes in the AMI population; it is possible that some of these changes represent changes in documentation. However, data collection was done prior to the era of public reporting and we know of no national effort to better document relative contraindications to care at that time. Third, our study is based on applying current measure criteria to patient populations prior to the era of public reporting. Thus, although we illuminate important changes in the populations of AMI patients that could be excluded from the measures, we do not know how this would translate into actual practice. The goal of this work was to highlight the growing population of AMI patients that could be excluded and the lack of transparency around these exclusions. Finally, our categorizations of patients were based on variables selected for prior quality improvement projects and do not precisely match current CMS/Joint Commission criteria. However, it is unlikely that this would dramatically change the trends described.
Important progress has been made in the last decade toward making care provided by hospitals to AMI patients more transparent. Most indications suggest that there has been simultaneous improvement in the quality of care provided to AMI patients. Our work identifies ongoing challenges with performance measurement in this population by revealing potential limitations of process measurements that incorporate discretionary exclusion of patients. Despite allowing for patient-specific decision-making, discretionary exclusion may lead to variability in patient populations included in measures across hospitals. Public quality reports, by failing to indicate who is excluded from measures, do not reflect the care provided to a large group of older patients whose inclusion or discretionary exclusion is invisible to the healthcare consumer.