Relationship Between Time of Day, Day of Week, Timeliness of Reperfusion, and In-Hospital Mortality for Patients With Acute ST-Segment Elevation Myocardial Infarction

Context Understanding how door-to-drug and door-to-balloon times vary by time of day and day of week can inform the design of interventions to improve the timeliness of reperfusion therapy.

Objective To determine the pattern of door-to-drug and door-to-balloon times by time of day and day of week and whether this pattern may affect mortality.

Design, Setting, and Participants Cohort study of 68 439 patients with ST-segment elevation myocardial infarction (STEMI) treated with fibrinolytic therapy and 33 647 treated with percutaneous coronary intervention (PCI) from 1999 through 2002. We classified patient hospital arrival period into regular hours (weekdays, 7 AM-5 PM) and off-hours (weekdays 5 PM-7 AM and weekends).

Main Outcome Measures Geometric mean door-to-drug time for fibrinolytic therapy and door-to-balloon time for PCI and all-cause in-hospital mortality. All outcomes were adjusted for patient and hospital characteristics.

Results Most fibrinolytic therapy (67.9%) and PCI patients (54.2%) were treated during off-hours. Door-to-drug times were slightly longer during off-hours (34.3 minutes) than regular hours (33.2 minutes; difference, 1.0 minute; 95% confidence interval [CI], 0.7-1.4; P<.001). In contrast, door-to-balloon times were substantially longer during off-hours (116.1 minutes) than regular hours (94.8 minutes; difference, 21.3 minutes; 95% CI, 20.5-22.2; P<.001). A lower percentage of patients met guideline recommended times for door-to-balloon during off-hours (25.7%) than regular hours (47%; P<.001). Door-to-balloon times exceeding 120 minutes occurred much more commonly during off-hours (41.5%) than regular hours (27.7%; P<.001). Longer off-hours door-to-balloon times were primarily due to a longer interval between obtaining the electrocardiogram and patient arrival at the catheterization laboratory (off-hours, 69.8 minutes vs regular hours, 49.1 minutes; P<.001). This pattern was consistent across all hospital subgroups examined. Furthermore, patients presenting during off-hours had significantly higher adjusted in-hospital mortality than patients presenting during regular hours (odds ratio, 1.07; 95% CI, 1.01-1.14; P = .02).

Conclusions Presentation during off-hours was common and was associated with substantially longer times to treatment for PCI but not for fibrinolytic therapy. To achieve the best outcomes, hospitals providing PCI during off-hours should commit to doing so in a timely manner.

Reperfusion therapy with either fibrinolytic therapy or percutaneous coronary intervention (PCI) reduces mortality for eligible ST-segment elevation myocardial infarction (STEMI) patients.^1-8 The shorter the time from symptom onset to treatment, the greater the survival benefit with either therapy.^3,9-14 The choice between therapies should take into account reperfusion treatment times.¹⁵ Although prior studies^16-18 have shown that door-to-balloon times for PCI are longer on evenings, nights, and weekends than on weekday days, several important questions remain. Prior studies neither assessed whether such variation was common to all types of hospitals nor did they determine where in the PCI process the delays occurred. In addition, these studies did not evaluate the impact of delayed door-to-balloon times on adherence to guideline recommended treatment times. Finally, previous studies focused on PCI and did not evaluate whether door-to-drug times for fibrinolytic therapy also varied by patient arrival period. Understanding the reasons for variation in reperfusion treatment times by patient arrival period, and whether such variation is common to all hospitals and to both fibrinolytic therapy and PCI, can inform the design and targeting of interventions to improve timely reperfusion.

To determine whether patient hospital arrival period is associated with door-to-drug and door-to-balloon times, we examined the relationship between time of day and day of week and reperfusion treatment times for STEMI patients treated with fibrinolytic therapy or PCI. Because fibrinolytic therapy is generally administered in emergency departments, which are continuously staffed, we hypothesized that door-to-drug times would not vary substantially by patient arrival period. In contrast, because most hospitals do not staff their cardiac catheterization laboratory around-the-clock, we hypothesized that door-to-balloon times would be longer during evenings, nights, and weekends than on weekday days. In addition, we examined whether the pattern of door-to-drug and door-to-balloon times by time of day and day of week varied according to hospital characteristics.

Our primary data source was the National Registry of Myocardial Infarction (NRMI), a voluntary prospective database of patients admitted with acute myocardial infarction (AMI). Characteristics of the NRMI registry, data-gathering procedures, and reliability have been described.¹⁹ This study used data for patients enrolled in the NRMI 3 or NRMI 4 registries from January 1, 1999, through December 31, 2002. All patient-level data and hospital-specific annual reperfusion volumes were obtained from the NRMI registry. Additional hospital characteristics were obtained from the American Hospital Association Annual Survey of Hospitals and the Health Facility Master File (SMG Inc, Chicago, Ill).^20,21

We conducted a retrospective cohort study of patients who presented acutely to the hospital within 12 hours of symptom onset who had an initial electrocardiogram (ECG) demonstrating ST-segment elevation or presumably new left bundle-branch block. To focus on primary reperfusion, we only included patients administered fibrinolytic therapy or PCI within 6 hours after hospital arrival.

The NRMI registry contains data on 830 473 AMI hospitalizations during the study period. We excluded 177 468 transfer patients who received their initial care at another hospital and 431 281 patients with neither ST-elevation nor left bundle-branch block on their initial ECG. We also excluded patients whose AMI symptoms developed after admission (n = 5088); patients without chest pain and whose symptom onset time was unknown (n = 20 415); patients whose first ECG was nondiagnostic (n = 17 066); patients whose diagnostic ECG time was missing, invalid, or more than 6 hours after hospital arrival (n = 6335); and patients who were not administered fibrinolytic therapy or PCI (n = 62 635).

With the remaining 110 185 patients, we developed 2 cohorts based on the reperfusion therapy administered (fibrinolytic therapy, n = 73 422 or PCI, n = 36 763). We excluded 512 patients from the fibrinolytic cohort and 891 patients from the PCI cohort because reperfusion treatment time could not be determined. Patients were assigned to the group of the first therapy they received. We excluded 13 patients from the fibrinolytic cohort initially treated with PCI and 507 patients from the PCI cohort initially treated with fibrinolytic therapy. We excluded 4436 patients from the fibrinolytic cohort admitted to hospitals administering fibrinolytic therapy in fewer than 20 cases and 1718 patients from the PCI cohort admitted to hospitals performing fewer than 20 PCI procedures over the 4-year study period. From the fibrinolytic cohort, we excluded an additional 22 patients who were admitted to non-US hospitals. The final cohorts included 68 439 patients administered fibrinolytic therapy at 1015 hospitals and 33 647 patients treated with PCI at 421 hospitals.

The primary outcome was reperfusion treatment time. For the fibrinolytic cohort, treatment time was defined as the time from hospital arrival to the initiation of fibrinolytic therapy (door-to-drug time). For the PCI cohort, treatment time was defined as the time from hospital arrival to first balloon inflation (door-to-balloon time). We also determined the proportion of patients administered fibrinolytic therapy within the American College of Cardiology/American Heart Association (ACC/AHA) guideline–recommended 30-minute door-to-drug time¹⁵ and the proportion with prolonged door-to-drug times (>45 minutes). Similarly, we assessed the proportion of patients receiving PCI within the ACC/AHA guideline–recommended 90-minute door-to-balloon time¹⁵ and the proportion with prolonged door-to-balloon times (>120 minutes).

The primary independent variable, patient arrival period, was defined by categorizing the week into 2 workday groups: weekday (Monday-Friday) and weekend (Saturday-Sunday), and 3 daily shifts: day (7 AM-5 PM), evening (5 PM-12 AM) and night (12 AM-7 AM). Shift times were determined by consideration of typical emergency department and catheterization laboratory workshifts and inspection of the distribution of the dependent variable by hour. In the primary analysis, patient arrival period was dichotomized into regular hours (weekdays, 7 AM-5 PM) and off-hours (weekdays, 5 PM-7 AM and weekends). In secondary analyses, we categorized off-hours into 5 periods: weekday-evening, weekday-night, weekend-day, weekend-evening, and weekend-night.

Additional patient covariates included baseline sociodemographic and medical history factors and characteristics of the clinical presentation. Hospital covariates included US Census division, ownership status, cardiac interventional capabilities, teaching status, reperfusion volume, reperfusion specialization, and urban vs rural location (Table 1, Table 2, Table 3, and Table 4).

In this study, race was used to control for potential confounding in multivariable regression analysis. Patient race or ethnicity was coded as a set of dummy variables indicating patients’ racial or ethnic group, which was abstracted from the medical records using the following categories: white, African American/black, Hispanic, Asian/Pacific Islander, American Indian or Alaska native, and other or unknown race or ethnicity. Because the number of patients categorized as Asian/Pacific Islander or American Indian or Alaska native was small, these patients were grouped into the “other” category during the analysis. Admission or triage staff recorded race or ethnicity as the patient was registered, and in NRMI, patients were assigned to only 1 racial or ethnic category.

For both the fibrinolytic and PCI cohorts, time to reperfusion was treated as a continuous variable, log transformed to correct for skewness.²² The geometric means for door-to-drug and door-to-balloon times were reported among the categories of patient arrival period. We constructed multivariable hierarchical models for each cohort to assess the relationship between patient arrival period and reperfusion time adjusted for all patient and hospital characteristics.²³ The outcome variable for all models was logarithm of reperfusion treatment time. We also included calendar time in all models to account for secular trends and staggered reporting periods by hospitals; random effects were specified for the main intercept and the coefficient of calendar time in all models.²⁴ To facilitate clinical interpretation, we retransformed model results to natural units (ie, minutes) using simulation.²⁵ We also assessed the relationship between patient arrival period and reperfusion treatment time in subgroups defined by hospital characteristics.

We examined reperfusion treatment time subintervals for fibrinolytic therapy and PCI using the same analytic approach. The door-to-drug time was divided into 2 subintervals: door-to-data (hospital arrival to ECG completion), and data-to-drug (ECG completion to fibrinolytic therapy administration). The door-to-balloon time was divided into 3 subintervals: door-to-data, data-to-catheterization laboratory (ECG completion to catheterization laboratory arrival), and catheterization-laboratory-to-balloon (catheterization laboratory arrival to balloon inflation).

To determine whether mortality differed between patients presenting during off-hours vs regular hours, we constructed a multivariable hierarchical model adjusted for all patient characteristics. To address the potential selection bias arising from differential use of therapies by patient arrival period, we compared the in-hospital mortality for all patients administered reperfusion (PCI plus fibrinolytic therapy) during regular hours and off-hours. To assess the degree to which any mortality differences could be explained by variation in reperfusion treatment time, we subsequently added reperfusion treatment time to the multivariable model. Since this model included patients treated with either PCI or fibrinolytic therapy, we standardized reperfusion treatment times based on the therapy administered. We used the same analytic approach to determine whether mortality differed among patients presenting during off-hours vs regular hours in the PCI cohort and the fibrinolytic therapy cohort, respectively. Mortality data for the entire hospitalization episode was not available for transfer-out patients (defined as patients who present initially to one hospital but who are subsequently transferred to a second institution prior to hospital discharge). Because mortality data were available only from the first hospital for these transfer-out patients they were assumed to be alive in the analysis.

Statistical analyses were performed using SAS version 8.2 (SAS Institute Inc, Cary, NC), Stata version 8.0 (Stata Corp, College Station, Tex), and HLM version 5 (SSI, Lincolnwood, Ill). P <.05 was considered statistically significant. The Yale University School of Medicine Institutional Review Board determined that this protocol was exempt from review because we used secondary data that had no patient identifiers.

The characteristics of the fibrinolytic and PCI cohorts stratified by patient arrival period are presented in Table 1 and Table 3. Patient and hospital characteristics were generally similar for different arrival periods. However, patients in the fibrinolytic and PCI cohorts who came to the hospital during off-hours were younger, more likely to be smokers, and less likely to have had a prehospital ECG than patients who came to the hospital during regular hours. Compared with patients who presented during regular hours, patients in the fibrinolytic cohort who presented during off-hours were more likely to be admitted to hospitals with cardiac surgery capability but less likely to be admitted to hospitals that used fibrinolysis for more than 90% of reperfusion cases. Finally, patients in the PCI cohort presenting during off-hours were more likely to be admitted to high-volume PCI centers and hospitals that used PCI for more than 90% of reperfusion cases than patients presenting during regular hours.

Overall 46 450 patients (67.9%) administered fibrinolytic therapy presented during off-hours while 21 989 (32.1%) presented during regular hours. The adjusted mean door-to-drug times were longer during off-hours (34.3 minutes) than regular hours (33.2 minutes), but the absolute difference was small (1.0 minutes 95% confidence interval [CI], 0.7-1.4; P<.001). Door-to-drug times ranged from 0.3 to 2.0 minutes longer for each off-hour period compared with regular hours: weekday evening (35.2 minutes), weekday night (33.5 minutes), weekend daytime (33.8 minutes), weekend evening (34.6 minutes), and weekend night (34.0 minutes). This pattern was consistent across all hospital subgroups (Table 5).

The proportion of patients with door-to-drug times less than 30 minutes and more than 45 minutes did not vary substantially by patient arrival period (Figure 1). A slightly lower proportion of patients were administered fibrinolytic therapy within 30 minutes during off-hours (41.2%) compared with regular hours (43.9%; difference, 2.7%; 95% CI, 1.9%-3.5%; P<.001). A similar proportion of patients had door-to-drug times more than 45 minutes during off-hours (30.3%) compared with regular hours (28.8%; difference, 1.5%; 95% CI, 0.8%-2.3%; P<.001).

Fibrinolytic treatment-time subintervals did not differ appreciably by patient arrival period (Figure 2). The geometric mean door-to-data subinterval was comparable for off-hours (7.5 minutes) and regular hours (7.4 minutes; difference, 0.1 minutes; 95% CI, 0-0.2; P = .04). The geometric mean data-to-drug subinterval was also comparable for off-hours (25.5 minutes) and regular hours (24.8 minutes; difference, 0.7 minutes; 95% CI, 0.4-1.0; P<.001).

Overall 18 228 patients (54.2%) received PCI during off-hours while 15 419 (45.8%) received PCI during regular hours. The adjusted mean door-to-balloon time was longer during off-hours (116.1 minutes) than regular hours (94.8 minutes), and the absolute difference was large (21.3 minutes; 95% CI, 20.5-22.2; P<.001). Door-to-balloon times ranged from 17.0 to 30.7 minutes longer for each of the off-hour time periods compared with regular hours: weekday evening (111.8 minutes), weekday night (118.3 minutes), weekend daytime (115.9 minutes), weekend evening (116.4 minutes) and weekend night (125.5 minutes). This pattern was consistent across all hospital subgroups (Table 5).

The proportion of patients with door-to-balloon times less than 90 minutes and more120 minutes varied markedly by patient arrival period (Figure 1). Fewer patients received PCI within 90 minutes during off-hours (25.7%) compared with regular hours (47.0%; difference, 21.3%; 95% CI, 20.3%-22.3%; P<.001). Substantially more patients experienced prolonged door-to-balloon times (>120 minutes) during off-hours (41.5%) compared with regular hours (27.7%; difference, 13.8%; 95% CI, 12.8%-14.9%; P<.001).

To better understand the source of the variation in PCI treatment times, we examined the relationship between PCI treatment time subintervals and patient arrival period (Figure 2). The door-to-data subinterval was the same for off-hours (7.9 minutes) and regular hours (7.9 minutes; difference, 0 minutes; 95% CI, −0.1 to 0.2; P = .6). The catheterization-laboratory-to-balloon subinterval was also similar for off-hours (33.8 minutes) and regular hours (33.7 minutes; difference, 0.1 minutes; 95% CI, −0.3 to 0.4; P = .7). In contrast, the data-to-catheterization-laboratory subinterval was longer during off-hours (69.8 minutes) than regular hours (49.1 minutes). This increase in the time from ECG completion to catheterization laboratory arrival (20.8 minutes; 95% CI, 20.0-21.5; P<.001) accounted for nearly all of the increased door-to-balloon times during off-hours.

To determine whether variation in reperfusion treatment times was also associated with differences in mortality, we examined the in-hospital mortality for patients presenting during off-hours compared with those presenting during regular hours. Among 21 989 patients treated with fibrinolytic therapy during regular hours, 963 (4.4%) died while among 46 450 treated during off-hours, 2043 (4.4%) died. Among 15 419 patients who underwent primary PCI during regular hours, 728 (4.7%) died while among 18 228 treated during off-hours 859 (4.7%) died.

Overall, after adjusting for all patient covariates, patients presenting during off-hours had significantly higher in-hospital mortality than patients presenting during regular hours (OR, 1.07; 95% CI, 1.01-1.14; P  = .02). When we also adjusted for reperfusion treatment time in the regression model, the difference in mortality between the regular hours and off-hours periods was attenuated and was no longer statistically significant (OR, 1.04; 95% CI, 0.98-1.11; P  = .18). After adjusting for all patient covariates, those treated with PCI presenting during off-hours had higher in-hospital mortality than patients presenting during regular hours but the difference was not statistically significant (OR, 1.05; 95% CI, 0.95-1.16; P  = .30). When we also adjusted for reperfusion treatment time in the regression model, the difference in mortality between the regular hours and off-hours periods was attenuated (OR, 1.02; 95% CI, 0.92-1.13; P  = .74). After adjusting for all patient covariates, patients treated with fibrinolytic therapy presenting during off-hours had higher in-hospital mortality than patients presenting during regular hours but the difference was not statistically significant (OR, 1.06; 95% CI, 0.98-1.15; P  = .13). Adjusting for reperfusion treatment time in the regression model did not alter the magnitude of the difference in mortality between the regular hours and off-hours period (OR, 1.06; 95% CI, 0.98-1.14; P  = .15).

Almost two thirds of the 102 086 STEMI patients administered reperfusion therapy in this study were treated during off-hours. Door-to-balloon times for PCI were longer during off-hours than regular hours while door-to-drug times for fibrinolytic therapy did not differ appreciably by patient arrival period. Patients who received PCI during off-hours had a 21-minute longer door-to-balloon times and were treated within the guideline-recommended time 45% less often than patients seen during regular hours. Increases in the time from ECG completion to catheterization laboratory arrival accounted for nearly all of the delay in door-to-balloon times during off-hours. Patients treated with reperfusion therapy during off-hours had a higher mortality rate than patients treated during regular hours. This mortality difference was attenuated by 43% when we adjusted for differences in reperfusion treatment times, suggesting that the higher off-hours mortality was due in part to longer reperfusion treatment times.

Although differences in PCI treatment times by patient arrival period have been previously noted,^16-18 our study advances existing research in several respects. This study is the first to demonstrate that PCI reperfusion times are greater during off-hours than regular hours although fibrinolytic reperfusion times are similar for all times of the day and days of the week. In addition, this study demonstrates that delays to PCI during off-hours are common to all types of hospitals, including high-volume PCI centers. Finally, this is the first study to suggest that higher off-hours in-hospital mortality rates may be partly due to longer off-hours reperfusion times.

We found that increases in the time interval from ECG completion to catheterization laboratory arrival accounted for nearly all of the increases in door-to-balloon times during off-hours. The major clinical processes that occur during this subinterval are (1) the diagnosis of the patient with a reperfusion-eligible STEMI and (2) activation of the cardiac catheterization laboratory. Because the diagnosis of a reperfusion-eligible STEMI is the same whether the patient is treated with fibrinolytic therapy or PCI and because door-to-drug times for fibrinolysis do not vary appreciably by patient arrival period, it is unlikely that variations in the time to diagnosis contribute significantly to the prolonged PCI treatment times during off-hours. However, activation of the catheterization laboratory is likely to be longer during off-hours, because catheterization laboratory personnel are frequently off site.

In contrast to our findings for PCI, door-to-drug times for fibrinolysis did not vary appreciably during off-hours. One explanation is that the administration of fibrinolytic therapy requires only emergency department personnel and the emergency departments of most US acute care hospitals are staffed around the clock.²⁶ In contrast, the administration of PCI requires a multidisciplinary team of emergency department and cardiac catheterization laboratory personnel, and very few US hospitals provide onsite staffing of their cardiac catheterization laboratory continuously.

We acknowledge that several issues merit consideration in the interpretation of our study. Though the NRMI registry includes geographically diverse hospitals, our results may not be generalizable to all US hospitals. The NRMI hospitals tend to have greater AMI volumes and are more likely to be nonprofit than nonparticipating hospitals. Nevertheless, NRMI is the largest and most current database of AMI patients available, and our results were consistent in multiple hospital subgroups. The quality of the NRMI time data collected during regular hours vs off-hours was not assessed. We believe it is unlikely that variation in documentation during off-hours would result in a differential bias and doubt that such variation could account for the substantial increase in door-to-balloon times noted during off-hours. A physician might choose to treat patients with fibrinolytic therapy instead of PCI during off-hours when anticipated door-to-balloon times are longer, potentially confounding the relationship between patient arrival period and reperfusion treatment times. However, the effect of such confounding would be to underestimate the association between off-hours and longer door-to-balloon times. Finally, because mortality data for the entire hospitalization episode were not available for transfer-out patients, they were assumed to have survived to hospital discharge. Since patients treated with fibrinolytic therapy are much more likely to be transferred than patients treated with PCI, this assumption introduces a bias that likely underestimates the mortality of patients treated with fibrinolytic therapy vs patients undergoing PCI. Such a comparison was not an objective of our study, and it would be inappropriate to draw any inferences about the mortality following PCI vs fibrinolytic therapy from this analysis. However, the transfer-out issue does not likely introduce a substantial bias for the regular hours vs off-hours comparisons made separately for the PCI cohort and fibrinolytic therapy cohorts. Furthermore, in the primary mortality analysis that included all patients treated with reperfusion, the transfer-out issue introduces a small but conservative bias, which underestimates the mortality during off-hours relative to regular hours because a higher proportion of patients treated during off-hours than during regular hours received fibrinolytic therapy.

Our study has implications for the delivery of reperfusion therapy during off-hours. Because delays to PCI can result in lower survival rates for STEMI patients,¹⁴ institutions providing PCI during off-hours should commit to doing so in a timely manner. One way to improve the timeliness of PCI during off-hours would be to provide onsite staffing of the cardiac catheterization laboratory around-the-clock. However, the clinical benefits of providing continuous in-house staffing of the cardiac catheterization laboratory must be weighed against the extra cost of providing such coverage. Another possible solution is to cross-train noncardiac catheterization laboratory staff to assist with PCI during off-hours. However, the benefits of cross-training staff may not be realized unless rapid access to interventional cardiologists is also available. Still another approach would be to regionalize interventional cardiac care, transporting off-hour patients to institutions with continuous cardiac catheterization laboratory staffing and rapid door-to-balloon times. However, this approach would only affect patients transported by emergency medical services and the faster door-to-balloon times at regional centers might be offset by prolonged transport times to these hospitals.

In conclusion, door-to-balloon times were longer during off-hours than regular hours while door-to-drug times did not differ appreciably by time of day or day of week. Increases in the time from ECG completion to catheterization laboratory arrival mainly accounted for the longer door-to-balloon times during off-hours. To achieve the best outcomes, hospitals providing PCI during off-hours should commit to doing so in a timely manner.

Corresponding Author: Harlan M. Krumholz, MD, SM, Cardiovascular Section, Yale University School of Medicine, 333 Cedar St, PO Box 208088, New Haven, CT 06520-8088 (harlan.krumholz@yale.edu).

Author Contributions: Dr Krumholz had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study concept and design: Magid, Wang, Herrin, McNamara, Pollack, Krumholz.

Analysis and interpretation of data: Wang, Herrin, Magid, McNamara, Bradley, Curtis, Pollack, French, Blaney, Krumholz.

Drafting of the manuscript: Magid, Herrin.

Critical revision of the manuscript for important intellectual content: Magid, Wang, McNamara, Bradley, Curtis, Pollack, French, Blaney, Krumholz.

Statistical analysis: Wang, Herrin

Obtained funding: Krumholz.

Administrative, technical, or material support: French, Blaney

Study supervision: Krumholz.

Funding/Support: This research was supported by grant R01 HL072575 of the National Heart, Lung, and Blood Institute.

Role of the Sponsor: The National Heart, Lung, and Blood Institute had no involvement in the design or conduct of the study, data management or analysis, or manuscript preparation, review, or authorization for submission. Genentech approved the study and provided access to the NRMI database at no charge; however, Genentech did not provide any direct support for the study and was not involved in the study design and conduct, analysis and interpretation of the data, and preparation of the manuscript.