Eliciting and documenting potentially ischaemic symptoms are important events in the early treatment of an evolving acute myocardial infarction (MI).1 Indeed, prior studies have shown lower quality of care and worse in-hospital survival when ischaemic symptoms were not noted on admission.2^–4 However, these studies focused solely upon the presence or absence of chest pain and did not consider the full spectrum of ischaemic symptoms (eg, dyspnoea or referred pain). More importantly, these studies relied entirely upon chart abstraction of documented symptoms at the time of presentation, and did not validate the findings of chart abstraction with patient interviews to quantify patients’ perspectives of why they presented for care. It therefore remains unclear whether these studies dealt with unrecognised, or absent, ischaemic symptoms.

These distinctions are important to resolve if quality improvement efforts to increase symptom recognition are to have the potential to improve care and outcomes. If patients truly lack symptoms of ischaemia at the time of presentation, then it is difficult to imagine how recognition could be improved. Conversely, if patients do present with potentially ischaemic symptoms that providers fail to recognise and document, then quality improvement efforts targeting improved interviewing and sensitivity to possible ischaemic symptoms might be an important opportunity to improve care and outcomes. Disentangling these possibilities requires collecting data from both clinicians’ and patients’ perspectives. Underlining the importance of collecting both sources of data, prior studies have noted that clinicians sometimes fail to elicit and document atypical symptoms from patients presenting with an acute coronary syndrome.5^–7

To examine these problems, we studied the presentation, care, and outcomes of patients with MI in the multicentre, prospective PREMIER (Prospective Registry Evaluating Myocardial Infarction: Event and Recovery) study. First, we examined the association of documented ischaemic symptoms with initial care, inpatient mortality and the quality of care at the time of hospital discharge. Then, we studied prospectively collected patient interviews designed to capture the symptoms that prompted patients to seek medical care initially and compared patient-reported symptoms with chart documentation of ischaemic symptoms. Together, these analyses can clarify the importance of recognising ischaemic symptoms in patients with MI and suggest whether improvements in symptom recognition may translate into improved care and outcomes.

PATIENTS AND METHODS

Patient sample

The rationale and methods for the PREMIER study have been previously described.8 The study recruited consecutive patients presenting with an MI at 19 US hospitals (Appendix) between 1 January 2003 and 28 June 2004. All patients with a positive troponin (or CK-MB) test during the initial 24 hours of admission were screened for possible inclusion. Patients were eligible if they were ⩾18 years and only if there was other evidence supporting an MI diagnosis, such as prolonged (>20 minutes) ischaemic signs/symptoms and/or electrocardiographic ST changes. In addition, all patients in the cohort had a final discharge diagnosis of MI. Thus, raised troponins alone were insufficient for inclusion. Incarcerated patients and those having raised cardiac enzymes as a complication of elective coronary revascularisation were not eligible. Institutional review board approval was obtained at each participating institution.

Of the 10 911 patients screened, 3953 patients were eligible and enrolled in the PREMIER study. We excluded those who presented with non-cardiac, potentially life-threatening acute illness that might confound the relationship between lack of ischaemia recognition and in-hospital mortality, such as concurrent trauma (n = 6), stroke (n = 24), severe gastrointestinal bleed/anaemia (n = 42) and hip fracture (n = 4). We also excluded those with dementia (n = 52) since inaccurate reporting of symptoms could confound the analyses. Lastly, we excluded those whose data for ischaemic symptoms were missing (n = 46) and those who were transferred from other facilities (n = 1685). We assumed that inclusion of transferred patients would probably introduce a selection bias in favour of patients with recognised symptoms. The final cohort for this study contained 2094 patients.

A principal goal of PREMIER was to interview patients for their health status and psychosocial characteristics. For those institutions whose data collection teams could not interview all patients with MI, subjects were randomly sampled (using a site-specific, predetermined sampling scheme) to avoid bias introduced by convenience sampling. A final subset of 1356 (65%) patients were approached and consented to interviews.

Data elements

Documenting ischaemic symptoms on presentation

We assumed that documentation of ischaemic symptoms represented recognition of ischaemic symptoms. Ischaemic symptoms were broadly defined in order to capture clinicians’ documentation of the various potential manifestations of myocardial ischaemia. Patients were classified as having had recognised symptoms of ischaemia if any of the emergency room notes, paramedic transcript records, or initial admitting notes documented any ischaemic symptoms, such as chest pain or pressure, arm, neck, shoulder or jaw pain, dyspnoea, nausea/vomiting, weakness, syncope or cardiac arrest.

Patient interview

Among the subset of 1356 patients interviewed, specific symptoms were elicited and recorded by the interviewers only if patients themselves first indicated recollection of their presenting symptom complex to avoid leading questions. Patients indicating symptom recollection were asked whether they experienced chest pain or pressure, arm, neck, shoulder or jaw pain, dyspnoea, nausea/vomiting and weakness or syncope upon presentation. If they reported chest pain, severity of their pain was rated on a scale of 1–10, with higher scores for more severe symptoms. Median time from admission to interview was 2 days.

Outcome variables

To assess the association of symptom documentation and outcome, we compared in-hospital mortality according to documented ischaemic symptoms on presentation. To compare the quality of care between patients with MI with and without documented ischaemic symptoms, we used the same core performance measures used by the Centers for Medicare and Medicaid Services (CMS).9

Specifically, we compared the use of β-blockers and aspirin within the first 24 hours, and use of reperfusion (fibrinolytic therapy or percutaneous coronary intervention) for ST-elevation MI.9 Among in-hospital survivors, we examined care at discharge by comparing rates of aspirin and β-blocker use, ACE/ARB use for left ventricular systolic dysfunction, and smoking cessation counselling. The CMS inclusion and exclusion eligibility criteria were applied for all performance measures to ensure that we studied patient treatment variation only in those eligible for treatment, without documented contraindications.9 Additionally, the frequency of statin use was compared with documentation of ischaemic symptoms. The denominator for this analysis included both those with raised low-density lipoprotein cholesterol (>2.59 mmol/l) or >1.81 mmol/l) as well as those who were receiving statins at admission.

Statistical analyses

We compared patient characteristics between patients with and without documented ischaemic symptoms using χ² tests for categorical variables and t tests for continuous variables. Logistic regression models identified whether demographic and clinical characteristics were associated with the absence of documented ischaemic symptoms. We specifically evaluated age, gender, race, chronic renal insufficiency, heart failure, chronic lung disease, diabetes, alcohol abuse and absence of prior evidence of atherosclerotic disease (MI, stroke, peripheral arterial disease, angina, or prior coronary revascularisation). These variables conditions were chosen a priori on clinical grounds based on their potential to mimic (eg, chronic lung disease) or mask (eg, diabetes) myocardial ischaemia.

We also defined the risk-adjusted association between documented ischaemic symptoms and in-hospital mortality, using logistic regression analyses. We adjusted for variables related to demographics (age, gender, race); prior comorbidity (cancer, chronic renal insufficiency, diabetes, prior MI, prior angina); disease severity (ST elevation, initial heart rate and blood pressure, initial creatinine, new onset atrial fibrillation); and in-hospital processes of care (reperfusion therapy and use of aspirin and β-blockers within the first 24 hours of admission). An interaction term for documentation of ischaemic symptoms and chronic renal insufficiency assessed whether the mortality risk associated with undocumented ischaemic symptoms varied according to the presence or absence of chronic renal insufficiency. To account for variation across hospitals, we used a generalised linear mixed model with a logit-link function10 that incorporated random effects for hospital intercepts.

In a secondary analysis of those 1356 patients with interviews during their index hospitalisation, we compared their description of presenting symptoms with chart documentation of ischaemic symptoms. Because more than one symptom can be experienced by a patient, we sequentially calculated the cumulative proportion of patients who indicated that they experienced chest pain (with or without radiation to neck, shoulder or jaw); isolated referred pain (to neck, shoulder, jaw, arm or back); shortness of breath; nausea or diaphoresis; weakness or lightheadedness. These last symptoms may all be early symptoms of MI as described in current guidelines.1

Missing data resulted in the exclusion of only 14 patients (0.4% of the sample) and therefore imputation and/or propensity analyses were not performed. Model discrimination for logistic regression models was determined using the c statistic.11 Statistical analysis was performed using the SAS software system for Windows, version 8 (SAS Institute, Cary, NC. USA).

RESULTS

Patient characteristics

The mean (SD) age of the cohort was 62 (14) years. Women accounted for 36% (753/2094) of the sample and 58% (1221/2094) were Caucasian. Of the 2094 patients studied, 217 (10%) patients had no documentation of ischaemic symptoms on presentation. Table 1 shows bivariate comparisons between patients with and without documented ischaemic symptoms. Patients without documented ischaemic symptoms were more likely to be female, of non-white race and have a greater burden of comorbidity. In multivariable analysis, patient characteristics associated (p<0.05) with an absence of documented ischaemic symptoms were non-ST-elevation MI, older age, a history of diabetes, chronic renal insufficiency, prior stroke and a lack of prior coronary artery disease (table 2); the c statistic was 0.85.

Adherence to acute care quality indicators

Among eligible patients, those without documented ischaemic symptoms were less likely to be treated with aspirin (89% vs 96%, p<0.001) or beta-blockers (77% vs 90%, p<0.001) within the first 24 hours of their admission. In the subset of eligible patients with ST elevation MI, patients without documented ischaemic symptoms at presentation were significantly less likely to receive reperfusion therapy (7% vs 58%, p<0.001; table 3).

Association with inpatient mortality

Crude in-hospital mortality rates were significantly higher in those without documented ischemic symptoms (15 vs 3%, p<0.001). Multivariable analyses demonstrated that failure to document ischaemic symptoms on presentation remained a strong predictor of in-hospital mortality (odds ratio = 3.21, 95% CI 1.76 to 5.84) after adjustment for age, gender, non-Caucasian race, cancer, diabetes, chronic renal insufficiency, chronic heart failure, initial heart rate (per 10-unit increase), initial systolic blood pressure (per 10-unit decrease), initial creatinine, Killip class 4, acute reperfusion therapy (for ST-elevation MI), aspirin and/or β-blocker within 24 hours of admission, and clustering of patients within hospitals (table 4). The c statistic was 0.83. The interaction term between chronic renal insufficiency and documentation of ischaemic symptoms was not significant (p = 0.99), indicating that the excess mortality in those with undocumented ischaemic symptoms did not differ significantly according to the presence or absence of chronic renal insufficiency.

Adherence to discharge quality indicators

At the time of discharge (table 3), eligible patients without documented ischaemic symptoms were less likely to be discharged with aspirin (87% vs 93%, p = 0.009), β-blockers (81% vs 91%, p<0.001), or ACE inhibitors (or angiotensin receptor blockers) for left ventricular dysfunction (67% vs 80%). Patients without documented ischaemic symptoms were also less likely to receive smoking cessation counselling (46% vs 66%, p = 0.004) or statin treatment if their low-density lipoprotein was >2.59 mmol/l or >1.81 mmol/l (66% vs 87%, 67% vs 90%, respectively; p<0.01 for both).

Patients’ perspectives of their presenting symptoms

A major goal of this study was to establish whether the absence of ischaemic symptom documentation was due to the true absence of symptoms, or a failure of their healthcare team to recognise and record those symptoms. In an analysis of those patients who had an independent, prospectively acquired interview of their presenting symptoms, the majority of patients without documented ischaemic symptoms actually reported symptoms compatible with ischaemia (table 5). Among the 37% of undocumented patients who reported chest discomfort, there were no differences in the severity of their presenting pain as compared with those who had documented ischaemic symptoms (p = 0.19). In addition, roughly half of the patients without documented ischaemic symptoms had either chest discomfort or referred discomfort to the arm, shoulder, neck or jaw. An additional 19% of patients had symptoms including shortness of breath, nausea or diaphoresis that might have been consistent with ischaemia. Thus, a total of three-quarters of the patients without documented ischaemic symptoms reported symptoms compatible with ischaemia as responsible for their presentation to the hospital.

DISCUSSION

Our analyses of the documentation of ischaemic symptoms in acute MI reveal several important insights. Despite using a broad range of symptoms to capture manifestations of myocardial ischaemia, ischaemic symptoms were not documented in 10% of patients with confirmed acute MI. Such patients typically were older and had more coexisting non-cardiac comorbidity. Patients whose symptoms were not documented received lower quality of care throughout their hospitalisation. Moreover, after risk adjustment, patients without documentation of their ischaemic symptoms were more than three times as likely to die during their hospitalisation. An important finding of this study is that almost three-quarters of patients with MI without documented ischaemic symptoms reported symptoms during the study interview that were consistent with myocardial ischaemia. In light of these observations, the data suggest that in most cases where ischaemic symptoms were not documented, improved recognition has the potential to be associated with better treatment and outcomes.

Establishing that most patients experienced symptoms consistent with myocardial ischaemia, despite the absence of documentation of these symptoms, is an important extension of previous knowledge. Canto and colleagues examined data from the National Registry of Myocardial Infarction and described lower quality of care and higher subsequent in-hospital mortality in patients presenting without chest pain.2 Similarly, other investigators have noted higher in-hospital mortality and undertreatment both at admission and at discharge in those who presented without chest pain.3 4 These prior studies, however, have relied exclusively upon chart documentation of patients’ presenting symptoms. By interviewing a majority of patients, our study demonstrates that in most cases in which ischaemic symptoms were not documented, patients actually experienced ischaemic symptoms that the clinician failed to elicit and/or record in the chart.

Importantly, our study also suggests that the initial failure to recognise myocardial ischaemia influences care well beyond the initial encounter. Despite clear biochemical evidence confirming that patients sustained a recent MI (along with other supporting evidence such as ECG changes), many patients failed to receive important evidence-based treatments at discharge. While early treatment differences arising from under-recognition of ischaemia may be understandable, the failure to initiate aggressive secondary prevention strategies after positive troponin blood tests have established myocardial necrosis warrants further investigation. Numerous publications12^–14 and guidelines15 16 have established that raised troponins herald increased short- and long-term risk, regardless of the presenting symptom complex experienced by the patient. Thus, most clinicians would agree that raised troponins warrant aggressive treatment at discharge. Since undocumented ischaemic symptoms in the medical record were strongly associated with poorer quality of care, as well as higher in-hospital mortality, our study identifies a potentially modifiable target to improve MI care—namely, the improved acquisition and documentation of patients’ history. Given the annual incidence of myocardial infarction (∼565 000 in the USA), a 10% prevalence of unrecognised ischaemic symptoms within this group could represent roughly 56 500 patients.17 The absolute number of patients likely to have unrecognised ischaemic symptoms (and an opportunity to improve their recognition and treatment) is therefore substantial.

Furthermore, with the emergence of public reporting of MI performance measures in the USA (eg, http://www.hospitalcompare.hhs.gov (accessed 23 November 2007)),18^–21 failure to recognise presenting symptoms as ischaemic may be the reason for some of the observed variability in performance among providers. Thus, doctors and hospitals may need to develop additional strategies to insure better identification of patients’ presenting symptoms through education and/or structural changes in the delivery of MI care. Programmes such as “Guidelines Applied in Practice” from the American College of Cardiology22 and “Get with the Guidelines” from the American Heart Association23 24 are potential strategies for dealing with this current gap in care. Such interventions have the potential to improve both patients’ prognosis and providers’ publicly reported performance.

Our study has several potential limitations. First, we did not interview 35% of the study patients. While interviewing a subset of the patients may have introduced a bias, great care was taken to ensure that interviewed subjects were randomly sampled when the capacity to interview the caseload of subjects was exceeded. Although previous analyses have suggested small differences between interviewed patients and the total MI population,8 it is not possible to know how this might have influenced our findings. A second potential limitation is that patients were interviewed 2 days after their admission and their description of symptoms might have changed from that originally presenting to the admitting team. Although interviews were designed to avoid leading questions, we cannot exclude this possible explanation for our findings. A third concern is that among those without documented ischaemic symptoms, we had limited power to compare differences in the care and outcomes between those who truly did not experience ischaemic symptoms and those who did experience ischaemic symptoms that went undocumented. Yet, given that these patients had unequivocal evidence of infarction including positive troponins, the poorer quality of care at discharge is concerning. A fourth limitation is that our study was observational, and our analysis might not have adjusted for potential confounders (eg, socioeconomic status, education level); nonetheless, risk adjustment was extensive, and we believe that quality of care would be unlikely to vary according to these unmeasured factors. Finally, we did not examine the longer-term outcomes related to poorer quality of care at the time of discharge. Further study of unrecognised patients’ subsequent care, health status and survival is also needed.

In summary, we found that patients with MI whose clinical presentation is not documented as being consistent with myocardial ischaemia tend to receive lower-quality care, from admission through discharge, and that these patients are at increased risk for in-hospital mortality. In light of these findings, and the large numbers of patients who might benefit from improved recognition and care, we believe that identifying opportunities to improve the recognition and treatment of MI should be a research priority for quality improvement.