Delay From Symptom Onset to Hospital Presentation for Patients With Non–ST-Segment Elevation Myocardial Infarction FREE

American College of Cardiology/American Heart Association (ACC/AHA) guidelines recommend that patients who experience symptoms suggestive of acute coronary syndrome should call 9-1-1 if those symptoms are unimproved or worsening after 5 minutes (class I).¹ Since patients typically cannot differentiate if their symptoms represent ST-segment elevation myocardial infarction (STEMI) or non-STEMI, immediate presentation to the hospital after symptom onset is critical in both instances. For patients with STEMI, studies have documented that the average delay time from symptom onset to hospital presentation is 2 hours and has not decreased substantially despite multiple public education campaigns.^2- 9 Furthermore, longer delays are associated with higher in-hospital mortality in patients with STEMI.³

While delays from symptom onset to hospital presentation have been linked to worse outcomes in patients with STEMI, the impact of such delays in patients with non-STEMI is unknown. Secular trends and factors associated with delay time have also not been evaluated for a contemporary, nationally representative cohort of patients with non-STEMI. To address this, we undertook a study to evaluate delay time from the contemporary, community-based, national registry Can Rapid Risk Stratification of Unstable Angina Patients Suppress Adverse Outcomes With Early Implementation of the ACC/AHA Guidelines (CRUSADE) National Quality Improvement Initiative. We specifically examined (1) the overall distribution of delay time, (2) secular trends in delay time, (3) factors associated with longer delays time, and (4) the association of delay time with in-hospital mortality.

The details of the CRUSADE National Quality Improvement Initiative have been previously published.^10- 12 Patients are eligible for the CRUSADE registry if they have ischemic symptoms lasting 10 or more minutes within 24 hours of hospital presentation in combination with high-risk features including ST-segment depression, transient ST-segment elevation, and/or elevated cardiac biomarkers (creatine kinase–MB or troponin I or T). Between January 1, 2001, and December 31, 2006, there were 168 191 patients enrolled in CRUSADE at 568 hospitals. The following patients were excluded sequentially: those who had received a diagnosis of unstable angina (n = 13 177), those who had an unknown date or time of symptom onset (n = 26 604), and those who were transferred in from another hospital (n = 23 788). Patients who were transferred out to another hospital were included in this analysis since the delay time from symptom onset to hospital presentation was available. The remaining 104 622 patients with non-STEMI comprised the study population for this analysis.

Hospitals participating in CRUSADE collected data on patient demographic and clinical information, physician and hospital characteristics, medical histories, immediate use of medications within 24 hours of presentation, use and timing of invasive cardiac procedures, in-hospital outcomes, and discharge medications. Each hospital's institutional review board reviews and approves participation in CRUSADE, and because data are collected anonymously and without unique patient identifiers, informed consent is not required. Delay from symptom onset to hospital presentation was calculated from the documented date and time of symptom onset to the documented date and time of hospital arrival and was expressed as median and interquartile range (hours). Candidate factors evaluated for association with longer delay times included age; sex; race/ethnicity (categorized as white, black, Hispanic, Asian, or other/unknown); payer type (categorized as private/health maintenance organization, Medicare, Medicaid, or self-pay/none); medical history (family history of coronary artery disease, hypertension, diabetes, current smoker, hyperlipidemia, prior myocardial infarction [MI], prior percutaneous coronary intervention [PCI], prior coronary artery bypass graft surgery [CABG], prior congestive heart failure [CHF], prior stroke, renal insufficiency, CHF at presentation); body mass index (BMI) (categorized as underweight <18.5, normal 18.5 to <25, overweight 25 to <30, obese 30 to <40, or extremely obese ≥40 [calculated as weight in kilograms divided by height in meters squared]); heart rate and systolic blood pressure at presentation; electrocardiogram (ECG) findings (categorized as ST-segment depression, transient ST-segment elevation, both, or neither); time of day and day of week at presentation (weekdays were defined as Monday to Friday and included daytime from >8 AM to 4 PM, evening from >4 PM to 12 AM, and night from >12 AM to 8 AM; weekends were defined as Saturday and Sunday and included daytime from >8 AM to 4 PM, evening from >4 PM to 12 AM, and night from >12 AM to 8 AM); year of presentation; attending physician (categorized as cardiologist or noncardiologist); hospital region (West, South, Midwest, Northeast); type of hospital (no catheterization laboratory; catheterization laboratory only, no PCI; PCI only, no on-site surgery; or PCI with on-site surgery); teaching hospital; and number of hospital beds. Holidays were categorized as weekdays if they fell on Monday to Friday and weekends if they fell on Saturday or Sunday.

We plotted the distribution of the delay from symptom onset to hospital presentation for the study population and performed tests of linear trend of the median delay times from calendar year 2001 through 2006. Continuous variables were presented as medians and 25th and 75th percentiles, while categorical variables were presented as frequencies. We used the Wilcoxon rank sum (2-sample) or Kruskal-Wallis (comparing more than 2-sample) test for continuous variables.

To determine factors associated with longer delay from symptom onset to hospital presentation, continuous variables (such as age, BMI, and hospital beds) were investigated for nonlinear relationship with respect to delay time and indicated the use of categorical cut points. Because delay from symptom onset to hospital presentation is a skewed continuous variable, a log-transformation was applied to normalize the distribution. Plots of standardized residuals against predicted values of delay time demonstrated no evidence of heteroscedasticity. The estimated coefficients were transformed to their natural units to facilitate interpretation.¹³ In determining the factors associated with delay from symptom onset to hospital presentation, linear generalized estimating equations with exchangeable working correlation matrix were used to account for within-hospital clustering because patients at the same hospital are more likely to have similar responses relative to patients at other hospitals (ie, within-center correlation for responses). This method produces estimates similar to those from ordinary linear regression, but their variances are adjusted for the correlation of outcomes within a hospital.¹⁴

The association between delay time and in-hospital mortality was explored using the logistic generalized estimating equations method with exchangeable working correlation matrix to account for within-hospital clustering. The mortality model adjusted for age, sex, race/ethnicity (white vs nonwhite), family history of coronary artery disease, hypertension, diabetes, current smoker, hyperlipidemia, prior MI, prior PCI, prior CABG, prior CHF, prior stroke, renal insufficiency, CHF at presentation, BMI, heart rate and systolic blood pressure at presentation, electrocardiogram findings (categorized as ST-segment depression, transient ST-segment elevation, both, or neither), and home medications (aspirin, angiotensin-converting enzyme inhibitor, statin, β-blocker, and clopidogrel). The relationship between delay times and in-hospital mortality was not linear; therefore, we categorized delay time into 5 categories (0 to ≤1 hour, >1 to 2 hours, >2 to 3 hours, >3 to 6 hours, and >6 hours) as clinically meaningful groupings. For modeling in-hospital mortality, patients who were transferred out were excluded since their survival status was not documented or available.

The average percentage of missing data was less than 2% for the baseline covariates. For modeling, missing values for categorical variables were imputed to the lower-risk group (for instance, if diabetes status was missing, it was imputed as no diabetes), and missing values for continuous variables were replaced with sex-specific medians of nonmissing data. Patients with missing variables for age, sex, race, hospital region, and teaching hospital were excluded from the models because the clinicians did not believe that these covariates were appropriate for imputation. The impact of these imputations is an anticonservative estimation of the standard errors of the covariates being estimated. With the size of this data set, we had adequate statistical power.

We performed 2 sensitivity analyses. First, to minimize confounding from patients who are transferred-in from another hospital (n = 23 788) as well as to determine the stability of our findings, we performed the analyses for the factors associated with delay time and the association of delay and outcomes by first excluding (main analysis) and then adding in patients who were transferred-in from another hospital. Second, we explored the impact of missing data imputation by carrying out a complete case analysis (n = 91 316). Essentially, the results of these sensitivity analyses were similar (eg, direction and magnitude of the effects) to the main analyses, thus the results were not reported. P < .05 was considered statistically significant for all tests, and all tests of statistical significance were 2-tailed. All analyses were performed using SAS software version 9.1 (SAS Institute Inc, Cary, North Carolina) by the Duke Clinical Research Institute.

The overall study population included 104 622 patients with non-STEMI presenting at 568 hospitals who participated in CRUSADE between January 2001 and December 2006. Table 1 provides the baseline demographic, clinical, and hospital characteristics of the study population as well as the unadjusted presentation delay times observed with these subgroups. Time of symptom onset was not available for 26 604 patients who were excluded from this analysis. The characteristics of the 26 604 excluded patients compared with the study population are given in Table 2. Excluded patients had a slightly higher prevalence for patients with older age, female sex, and diabetes. Furthermore, those excluded patients had a lower prevalence of prior MI, PCI, or CABG, and a lower rate of hospital presentation during weekday or weekend nights (>12 AM to 8 AM).

The median delay time for the study population was 2.6 hours with an interquartile range of 1.3 to 6.0 hours during the study period (Figure 1). Approximately 59% of patients had delay times from symptom onset to hospital presentation greater than 2 hours, and 11% of patients presented more than 12 hours after onset of symptoms. Secular trend in delay from symptom onset to hospital presentation is shown Figure 2. There was no significant change in delay times from 2001 to 2006 (P = .16).

Multivariable analysis of factors associated with delay times from symptom onset to hospital presentation is given in Table 3. Compared with patients younger than 55 years, adjusted delay times were 3.9%, 8.1%, and 7.8% longer in age groups 65 to 74 years, 75 to 84 years, and 85 years or older, respectively (P < .001). Men had 3.4% shorter delay times than women (P < .001) and white patients had 2.2% shorter delay times than nonwhite patients (P = .03). Patients who have had prior MI or prior PCI had 3.6% and 3.9% shorter adjusted delay times compared with respective reference groups (P < .001).

Patients who presented during weekday daytime (>8 AM to 4 PM) had the longest adjusted delay from symptom onset to hospital presentation. Compared with patients who presented during weekday daytime, those who presented during weekday evenings (>4 PM to 12 AM), weekday nights (>12 AM to 8 AM), weekend evenings (>4 PM to 12 AM), and weekend nights (>12 AM to 8 AM) had 10.9%, 24.7%, 15.8%, and 24.3% shorter delay times, respectively (P < .001).

After adjusting for covariates and compared with the reference group of patients with delay times of more than 6 hours, the in-hospital mortality odds ratio (OR) was 1.19 (95% confidence interval [CI], 1.08-1.30) for delay times of 0 to 1 hour or less, 0.91 (95% CI, 0.83-1.00) for delay times of more than 1 to 2 hours, 0.77 (95% CI, 0.69-0.88) for delay times more than 2 to 3 hours, and 0.90 (95% CI, 0.81-1.00) for delay times more than 3 to 6 hours. The overall relationship between in-hospital mortality and delay time was not linear and generally not strong for patients with non-STEMI.

Patients who were taking the following home medications had lower adjusted in-hospital mortality rates: aspirin (OR, 0.91; 95% CI, 0.84-0.99), angiotensin-converting enzyme inhibitor (OR, 0.87; 95% CI, 0.80-0.93), and statin (OR, 0.88; 95% CI, 0.80-0.96). However, home use of β-blockers (OR, 0.95; 95% CI, 0.88-1.03) or clopidogrel (OR, 1.01; 95% CI, 0.91-1.11) was not associated with lower in-hospital mortality.

In our study of 104 622 patients with non-STEMI, we found that median delay time from symptom onset to hospital presentation was 2.6 hours, and 59% of patients had delay times greater than 2 hours. More importantly, delay time has not changed significantly from 2001 to 2006 among a contemporary, nationally representative cohort of patients with non-STEMI. Factors associated with longer delay from symptom onset to hospital presentation included patients with older age, female sex, nonwhite race, diabetes, and current smoking. Patients who presented during weekday or weekend nights (defined as >12 AM to 8 AM) had delay times 25% shorter than those who presented during weekday daytime (defined as >8 AM to 4 PM). Thus, identifying novel strategies for improving patient responsiveness to seek care is important in all patients who develop symptoms suggestive of myocardial ischemia.

The factors associated with delay from symptom onset to hospital presentation have been well studied among patients with STEMI. Among a cohort of 482 327 patients with STEMI enrolled in the National Registry of Myocardial Infarction from 1995 to 2004, geometric mean for delay time was 114 minutes.² Patient subgroups with a combination of risk factors (older age, women, Hispanic or black race, and/or diabetes) have particularly long delay times, which were 60 or more minutes longer than subgroups without those characteristics. Longer delay times were associated with a reduced likelihood of receiving any reperfusion therapy, and even among those treated, late presenters had significantly longer door-to-balloon and door-to-drug times.³

However, secular trends in delay time and impact of longer delays on mortality are less understood in patients with non-STEMI. Population-based studies that included all patients with acute MI (both STEMI and non-STEMI) found that longer delay from symptom onset to hospital presentation were more common among elderly patients, women, nonwhite patients, diabetic patients, and those with atypical symptoms.^7,9,15- 16 Two prior studies have specifically evaluated delay time among patients with non-STEMI. Goldberg and colleagues⁸ found that median delay time was 3.0 hours among 2935 patients with non-STEMI from the Global Registry of Acute Coronary Events (GRACE) registry. Factors associated with shorter delay time were male sex, presence of diaphoresis, ambulance transport, and symptom onset from 6 AM to before 12 AM (compared with 12 AM to <6 AM). In a study of 1219 patients with non–ST-segment elevation acute coronary syndrome from the Canadian Acute Coronary Syndrome II registry, Elbarouni and colleagues¹⁷ found that early presenters were more likely to be older, have prior angina or stroke, and have Killip class 1 compared with late presenters (defined as >6 hours from symptom onset). No differences were observed in use of medications, PCI, CABG, and in-hospital or 1-year mortality between early presenters (defined ≤6 hours from symptom onset) vs late presenters (>6 hours from symptom onset).

Our study is the largest, contemporary, nationally representative cohort of patients with non-STEMI, and it advances the existing research in several respects. We found that delay time has not improved from 2001 to 2006 for patients with non-STEMI. We also found that older patients, female patients, nonwhite patients, diabetic patients, and those with current smoking had longer delay times; however, the magnitude of effect (<10%) on delay time from each factor was overshadowed by the overall duration of delay (median delay time, 2.6 hours). Therefore, interventions aimed at improving patient awareness of symptoms and responsiveness to seek care will likely need to target all patients at risk for MI, and not just those who have individual risk factors (age, sex, or diabetes) for longer delay time.^18- 20 Our study also showed that patients with prior MI or PCI had modestly shorter delay times and those with prior CABG had similar delay times compared with respective reference groups. Efforts to improve patient responsiveness to seek care should target these subgroups with prior events or procedures. A potential educational intervention to improve delay time would be to routinely provide this during discharge planning or as a standard part of the office visit. The minimal reductions in delay times we observed suggest that current practice is failing to effectively educate and activate our patients.

In addition, we found that delay times were 25% shorter during weekday or weekend nights from after 12 AM to 8 AM compared with weekday daytime from after 8 AM to 4 PM. This finding differs from previous studies that showed longer delays during the night.^8,15,21 While we cannot determine why patients decided to seek care more quickly at night, potential hypotheses include heightened fear during the night when patients may be alone at home, higher tolerance of symptoms during the daytime when a patient is active or at work, or a perception of shorter waiting times and less crowding in emergency departments during the night.^22- 23

We found that the relationship between delay time and in-hospital mortality for patients with non-STEMI was not linear and generally not strong. First, patients with the shortest delay time (0-1 hour) had a higher risk of in-hospital mortality compared with the reference group (delay time >6 hours), whereas patients with delay times of more than 1 to 2 hours, more than 2 to 3 hours, and more than 3 to 6 hours had similar or slightly lower mortality compared with the reference group. Although we adjusted for many clinical variables, patients with the shortest delay time likely had greater disease severity at presentation that we did not measure and could not adequately adjust for. Patients with greater disease severity are likely to seek care early²⁴ and we have previously found that patients with STEMI with the shortest delay time also exhibited higher adjusted in-hospital mortality rate.^2- 3 Second, the overall relationship between delay times and in-hospital mortality was generally not strong for patients with non-STEMI. Patient presentation can vary and evolve from unstable angina, to transient occlusion, and to complete occlusion of a coronary artery. Furthermore, in the absence of ST-segment elevation, there is no therapy that can provide a large benefit such as immediate reperfusion with primary PCI or fibrinolytics. Interestingly, home use of medications including aspirin, angiotensin-converting enzyme inhibitor, and statin were associated with lower adjusted in-hospital mortality.

This study calculated delay using the documented time from symptom onset to the time of hospital arrival. The time of symptom onset is subject to patient recall bias, language barriers, and socioeconomic status and relies on accurate documentation by each hospital participating in this registry. There is also the potential for survival bias because patients who died out-of-hospital and never presented are not included in this registry. Also, the type and nature of symptoms (continuous vs intermittent), socioeconomic status, marital status, and depression were not captured or included in the multivariable model. The mode of transport to the hospital (emergency medical services vs self-transport), number of hospitals per population density, distance to the nearest hospital, and traffic patterns were not included in the analysis because these variables were not collected in the CRUSADE registry. Patients who were transferred out were excluded from the in-hospital mortality model because their survival status was not available.

In conclusion, the median delay time from symptom onset to hospital presentation among patients with non-STEMI was 2.6 hours and has not changed significantly from 2001 to 2006. Time of day had the largest impact on delay time; patients who presented to the hospital during weekday or weekend nights (>12 AM to 8 AM) have 25% shorter delay times compared with those presenting during weekday daytime (>8 AM to 4 PM). Novel strategies to improve patient responsiveness to seek care are critical and important for both patients with STEMI or non-STEMI.

Correspondence: Henry H. Ting, MD, MBA, Division of Cardiovascular Diseases, Mayo Clinic, 200 First St SW, Rochester, MN 55905 (ting.henry@mayo.edu).

Accepted for Publication: May 16, 2010.

Author Contributions:Study concept and design: Ting, Roe, and Spertus. Acquisition of data: Ting, Chen, Roe, and Peterson. Analysis and interpretation of data: Ting, Chen, Roe, Chan, Spertus, Nallamothu, Sullivan, DeLong, Bradley, Krumholz, and Peterson. Drafting of the manuscript: Ting. Critical revision of the manuscript for important intellectual content: Ting, Chen, Roe, Chan, Spertus, Nallamothu, Sullivan, DeLong, Bradley, Krumholz, and Peterson. Statistical analysis: Ting, Chen, and DeLong. Obtained funding: Roe. Administrative, technical, and material support: Ting. Study supervision: Ting and Roe.

Financial Disclosure: Dr Peterson has received research grants from Bristol-Myers Squibb/sanofi-aventis Pharmaceuticals Partnership, Merck/Schering, and Schering-Plough Corporation. Dr Roe was an investigator for BMS, Eli Lilly, Portola Pharmaceuticals, Schering-Plough, and sanofi-aventis and a consultant for Adolor, Astra Zeneca, BMS, Daiichi-Sankyo, Eli Lilly, Merck, Novartis, sanofi-aventis, and Schering-Plough.

Funding/Support: CRUSADE is a national quality-improvement initiative of the Duke Clinical Research Institute. CRUSADE is funded by the Schering-Plough Corporation. Bristol-Myers Squibb/sanofi-aventis Pharmaceutical Partnership provides additional funding support. Millennium Pharmaceuticals Inc also provided funding for this work.