Shorter time to reperfusion is associated with better survival for patients with ST-segment elevation myocardial infarction (STEMI), whether they receive fibrinolytic therapy (1- 2) or percutaneous coronary intervention (PCI) (3- 5). Recently published guidelines recommend goals of 30 min for presentation at the hospital to the administration of fibrinolytic therapy (door-to-needle time) and 90 min for presentation at the hospital to the inflation of an angioplasty balloon (door-to-balloon time) (6). The Centers for Medicare and Medicaid Services and the Joint Commission on Accreditation of Healthcare Organizations have identified time to reperfusion as an important quality indicator for acute myocardial infarction (AMI).

No contemporary information exists about hospital performance in achieving national benchmarks for time to reperfusion. Previous investigators determined that the average door-to-needle time decreased from 62 min in 1990 to 38 min in 1999 (7); however, this average time in 1999 was still longer than the guideline recommendations, and no information was provided on door-to-balloon time. Moreover, studies of time to reperfusion did not evaluate hospital-level performance.

Several related questions remain. First, at the hospital level, has time to fibrinolytic therapy decreased in more recent years and are hospitals reaching the guideline recommendations? Second, has the time to PCI improved? Third, what is the variation in individual hospital improvement in time to reperfusion? Finally, are there hospital characteristics associated with improvement? To address these questions, we used detailed patient-level and hospital-level longitudinal data from a national sample of patients with STEMI admitted from 1999 to 2002 from the National Registry of Myocardial Infarction (NRMI)-3 and -4 (8).

We used NRMI, a voluntary AMI registry sponsored by Genentech Inc. (South San Francisco, California), to obtain a cohort of patients with STEMI who received acute reperfusion therapy with either fibrinolytic therapy or primary PCI. The NRMI criteria (9- 10) include a diagnosis of AMI according to the International Classification of Diseases-Ninth Revision-Clinical Modification (code 410.X1) and any one of the following criteria: total creatine kinase or creatine kinase MB that was two or more times the upper limit of the normal range or elevations in alternative cardiac markers; electrocardiographic evidence of AMI; or nuclear medicine testing, echocardiography, or autopsy evidence of AMI. During our study period of January 1, 1999, to December 31, 2002, there were 830,473 AMI admissions in NRMI. We excluded patients who had neither ST-segment elevation (in 2 or more leads) nor left bundle branch block on the first electrocardiogram (ECG) (n = 535,993); who were transferred from another acute care institution (n = 72,756); whose AMI symptom onset was after the admit date and time (n = 5,088); who were without chest pain and in whom symptom onset time was not known (n = 20,415); whose first ECG obtained was not the diagnostic ECG (the ECG with ST-segment elevation or left bundle branch block; n = 17,066); and those with time of diagnostic ECG missing, more than 1 h before presentation, or more than 6 h after presentation (n = 6,335).

From the 172,820 remaining patients, we analyzed fibrinolytic therapy and PCI separately. We assigned patients who received both therapies to the group on the basis of the therapy they received first. From the patients who received fibrinolytic therapy (n = 73,422), we excluded those with door-to-needle time that was negative, more than 6 h, or missing (n = 512); those who received fibrinolytic therapy after PCI (n = 13); those in hospitals with fewer than 20 total patients (n = 4,436); and those from the single NRMI hospital outside of the 50 states (n = 22), leaving a final fibrinolytic therapy cohort of 68,439 patients from 1,015 hospitals. From the patients who received PCI (n = 36,763), we excluded those with door-to-balloon time that was negative, more than 6 h, or missing (n = 891); those who received PCI after fibrinolytic therapy (n = 507); and those in hospitals with fewer than 20 total patients (n = 1,718), leaving a final PCI cohort of 33,647 patients from 421 hospitals.

Because NRMI is a voluntary registry, many hospitals did not participate consistently over the entire time period. Some hospitals started submitting data later than 1999, some stopped submitting data during the study period, and some intermittently submitted data. We repeated analyses with hospitals that reported at least five cases for each year of the study period (as opposed to the original cohort with hospitals that reported an average of at least five cases/year). The results from these smaller cohorts (for both fibrinolytic therapy and PCI) were not substantially different from those of the more inclusive cohorts. Thus, we present only the analyses using the larger cohort.

Our principal outcome was the time in minutes between hospital arrival and the delivery of reperfusion therapy, as noted in the medical record and recorded in the NRMI case report form. Calendar time was measured in elapsed time from the beginning of the study period.

Hospital characteristics included U.S. Census division; ownership (government, nonprofit, for-profit); cardiac facilities (open heart surgery, cardiac catheterization laboratory only, other); annual reperfusion volume (<20, 20 to 40, >40 for fibrinolytic therapy and <15, 15 to 50, >50 for PCI); and reperfusion specialization (percentage of all reperfused STEMI patients in the hospital who received the given strategy). We classified hospitals on the basis of urban (residence in a county with a population of at least 50,000) versus rural location and teaching status (i.e., participation in a residency or fellowship training program accredited by the Liaison Committee on Medical Education). Data for hospital characteristics were obtained from the American Hospital Association Annual Survey of Hospitals (11) and the SMG data set (SMG Marketing Group Inc., Chicago, Illinois).

Patient demographic and clinical covariates included sociodemographic variables (gender, age, race/ethnicity, and payer type); medical history (smoker, chronic renal insufficiency, previous AMI, hypertension, family history of coronary artery disease, hypercholesterolemia, congestive heart failure, previous percutaneous transluminal coronary angioplasty, previous coronary artery bypass graft surgery, chronic obstructive pulmonary disease, stroke, angina, diabetes); presentation characteristics (whether a pre-hospital ECG was performed, chest pain at presentation, systolic blood pressure, pulse, heart failure); the results of the diagnostic ECG (number of leads with ST-segment elevation, AMI location, ST-segment depression, nonspecific ST-T-wave changes, Q-wave); the admission time of day (day, evening, or night); and admission day of week (weekday or weekend).

We reported the number of patients who were administered fibrinolytic therapy or PCI within the recommended time frame for each year of the study. We then examined differences in hospital performance (geometric mean time to reperfusion) and improvement (change over calendar time in geometric mean time to reperfusion) with multivariable hierarchical models (12). Hierarchical models were used to account for the non-independence of observations in this sample, in which patients were clustered within participating NRMI hospitals. In addition, hierarchical models allowed us to estimate variation in performance and improvement across hospitals by modeling both the intercept and calendar time as random effects. We identified hospital characteristics that were associated with greater improvement using cross-level interactions between calendar time and hospital characteristics.

To account for potential floor and ceiling effects (hospitals starting with very high or low treatment times having more or less opportunity to improve over time) and for regression to the mean (13), each model also included terms representing the interaction of calendar time with hospital baseline performance. Baseline performance was calculated as the median time to reperfusion of the first 10 cases reported in the study period for a given hospital. Time to reperfusion was treated as a continuous variable, and log transformation was used to correct for skewness (14). For easier interpretation, we transformed coefficients back into natural units with geometric means or simulation methods (15). Statistical analyses were performed with SAS versions 6.12 and 8.2 (SAS Inc., Cary, North Carolina), Stata version 8.0 (Stata Corp., College Station, Texas), and HLM version 5 (SSI, Lincolnwood, Illinois). The authors had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

The characteristics of the patients included in the fibrinolytic therapy cohort (n = 68,439) and the PCI cohort (n = 33,647) are displayed in (Table 1). The cohorts were predominantly male and white, with a significant proportion of patients having a prior diagnosis of coronary artery disease and/or traditional cardiac risk factors. Over 93% had chest pain, 11% were in overt heart failure, and about 2% had left bundle branch block. The characteristics of the hospitals in the study are shown in (Table 2). A greater proportion of hospitals in the fibrinolytic therapy cohort were rural teaching hospitals, whereas the PCI cohort included a greater proportion of urban teaching hospitals.

In 1999, 46% (11,107 of 24,024) of patients in the fibrinolytic therapy cohort were treated within 30 min, and 35% (3,104 of 8,798) of the patients in the PCI cohort were treated within 90 min, the recommended times (16- 17) ((Figure 1)A and Figure 1B). These proportions increased by 1% and 2%, respectively, over the next three years. In addition, the proportion of patients receiving fibrinolytic therapy after 40 min or PCI after 120 min did not change substantially.

With multivariable hierarchical modeling, the mean door-to-needle time over the study period, adjusted for both patient and hospital characteristics, was 34.3 min (95% confidence interval [CI] 33.9 to 34.7). Several hospital characteristics were significantly associated with performance (Table 2). Door-to-needle times differed by geographic location, with best times in the West North Central (Minnesota, Iowa, Missouri, North Dakota, South Dakota, Nebraska, and Kansas) and Mountain (Montana, Idaho, Wyoming, Nevada, Utah, Colorado, Arizona, and New Mexico) divisions. Urban teaching hospitals had longer times than nonteaching hospitals or rural teaching hospitals. Hospitals with higher volumes and those with higher proportions of fibrinolytic therapy to total reperfusion therapy (reperfusion specialization) had modestly shorter times.

The adjusted mean door-to-balloon time over the study period was 108.0 min (95% CI 106.5 to 109.4). Volume and reperfusion specialization were associated with shorter door-to-balloon times. Hospitals in the West North Central and New England (Maine, New Hampshire, Vermont, Massachusetts, Rhode Island, and Connecticut) divisions averaged the best times for the PCI cohort. Teaching and rural/urban status were not significantly associated with door-to-balloon time. Hospitals without cardiac surgery facilities had shorter times than those with cardiac surgery facilities (108.4 min vs. 102.4 min). Hospital ownership status did not affect performance for either the fibrinolytic therapy or PCI cohort.

The mean door-to-needle times for the fibrinolytic therapy cohort were stable over the four-year time period, with a nonsignificant yearly improvement of 0.01 min (p > 0.9). The decrease in door-to-balloon time for the PCI cohort also was not statistically significant (−0.57 min/year; p = 0.09) (Figure 2).

The change in mean time to reperfusion by hospital showed a wide distribution nearly centered on 0 for both cohorts ((Figure 3)A and Figure 3B), with ranges of −5.7 min to +7.4 min for door-to-needle times and −12.9 min to +10.1 min for door-to-balloon times. Only 33% (337 of 1,015) of hospitals in the fibrinolytic therapy cohort improved more than one min/year, and 26% (110 of 421) of PCI hospitals improved more than three min/year. Average door-to-needle and door-to-balloon times worsened by these amounts in 32% (325 of 1,015) and 18% (74 of 421) of hospitals, respectively.

No hospital characteristic significantly predicted improvement in door-to-needle time for the fibrinolytic therapy cohort (Table 3). Improvement in hospitals with a higher proportion of patients receiving fibrinolytic therapy (reperfusion specialization) nearly reached statistical significance (p = 0.09). Although total fibrinolytic therapy volume predicted overall door-to-needle times, fibrinolytic therapy volume did not predict improvement over the study period.

In the PCI cohort, PCI volume significantly predicted improvement, with high-volume hospitals (>50 PCIs/year) showing improvement in mean door-to-balloon time (−1.9 min/year), medium-volume hospitals (15 to 50 PCIs/year) showing nearly no change in mean time (−0.1 min/year), and low-volume hospitals (<15 PCIs/year) showing an increase in time (+1.0 min/year, p = 0.033). Thus, in addition to being associated with overall performance, PCI volume predicted improvement over the study period. Hospitals in New England showed the most improvement (−7.5 min/year). Proportion of PCIs to total reperfusion therapy (reperfusion specialization) did not predict improvement in door-to-balloon time.

Despite strong evidence that decreased time to reperfusion for patients with STEMI is related to improved morbidity and mortality, we found that less than one-half of these patients received reperfusion therapy within the recommended time in any year from 1999 to 2002. Furthermore, we found little evidence of substantial improvement in time to reperfusion over these years. Although the entire cohorts did not improve, the analysis of the individual hospitals demonstrated a notable hospital-level variation in change from 1999 to 2002. Some hospitals showed substantial improvement, whereas performance worsened in others. Traditional hospital characteristics generally were not good predictors of improvement, indicating a need to identify other, perhaps more subtle, reasons why some hospitals improve.

Our results extend the work on earlier trends in time to fibrinolytic therapy. In an earlier study of the NRMI database, the median door-to-needle time was shown to decrease from 62 min to 47 min from 1990 to 1994, with a subsequent decrease to 38 min by 1999 (7). This slowing of improvement over time is consistent with the lack of change in door-to-needle times we found from 1999 to 2002. In a review of care given to Medicare beneficiaries, a state-by-state comparison found the door-to-needle time in the median state increased by four min between 1998 and 1999 and 2000 and 2001 (18). In this same time period, the door-to-balloon time in the median state decreased by 19 min; however, the power of this study to address the specific question of time to reperfusion was limited; the median state contributed only 17 patients to the number of patients for the fibrinolytic therapy group and 6 to the PCI group (19). The size of the cohorts in our study and the focus only on time to reperfusion enabled us to determine more precise estimates and to investigate potential hospital-level predictors of improvement.

Previous studies demonstrated that hospital volume predicts lower door-to-needle (1) and door-to-balloon times (20). In addition to confirming these results for performance, we further demonstrated the association of higher number of procedures with improvement in door-to-balloon times. Thus, hospitals performing >50 PCIs/year not only started out with better times but managed to improve even more over the four years. In comparison, higher-volume fibrinolytic therapy hospitals did not improve more over the four-year study period than lower-volume hospitals. The difference in complexity between fibrinolytic therapy and PCI might account for these disparate results. The PCI strategy, with involvement of the interventional cardiologist and the cardiac catheterization laboratory in addition to the emergency department, is inherently a more complex process than the fibrinolytic strategy. A complex process such as PCI might benefit from more frequent use to identify more opportunities to improve. Also, in contrast to the greater than two-decade experience with the fibrinolytic strategy, the PCI strategy has only recently been widely adopted (7). More improvement is likely seen early after introduction of a new strategy than after one has become established.

Hospitals that predominantly treated patients with one strategy (>90% of the time) had better performance for that strategy than those hospitals that had a more balanced profile. One explanation could be that there is a time delay associated with the decision of whether to administer fibrinolytic therapy or perform PCI. Interestingly, improvement over the four years was not statistically greater in these hospitals with a predominant strategy compared with the balanced hospitals. More investigation into the relationship between the strategies might provide valuable insights.

Our findings suggest that factors other than traditional hospital characteristics are largely responsible for improvement in door-to-reperfusion times seen in some hospitals. These findings might not be surprising in that management of AMI involves many individuals, including nurses, physicians, technologists, and paramedics from different areas of the hospital, such as the emergency medical system, the emergency department, and cardiology. In this setting, achieving rapid reperfusion likely requires multifaceted, coordinated efforts at quality improvement specific to the institution. Qualitative approaches such as structured interviews and site visits might supplement quantitative outcome analysis by supplying in-depth insight into the complex processes of care.

One potential explanation for a lack of significant improvement in our analysis would be a “floor effect”—that no further improvement is possible. Although the low rates of guideline adherence would suggest ample potential room for improvement for most hospitals, we included the mean of the first 10 cases for each hospital as a baseline covariate in the model to account for this possibility and for potential regression to the mean. Even adjusting for baseline performance, many hospitals showed substantial improvement. Thus, a “floor effect” is not likely significantly limiting opportunities for improvement. In-depth qualitative investigation into hospitals with the best performance revealed valuable strategies common to many of them (21). These qualitative studies that identify optimal performance and disseminate those strategies that have proven positive outcomes complement the current analysis that identifies what is being done in actual practice.

Whereas NRMI is a large database of more than one million patients with AMI, certain issues are important to consider in interpreting our results. First, the database is voluntary and might not be representative of all patients presenting with STEMI. Second, the membership of the cohort was not constant. Some hospitals started enrolling in NRMI during the study period, others stopped; however, the results from a smaller cohort of patients from the hospitals that consistently participated throughout the study period were similar. Third, the NRMI database might contain some patients for whom there were legitimate delays in time to reperfusion. To decrease the effect of extreme delays on our results, we only evaluated patients treated with reperfusion therapy within 6 h and used geometric mean times. Finally, we evaluated the hospital characteristics most frequently shown to be associated with performance, but other unmeasured structural characteristics might exist that predict improvement.

In conclusion, the significant improvement in overall door-to-needle times seen in the 1990s seems to have ended. Despite national initiatives to measure and reduce these times, guideline recommendations were met less than one-half of the time, with no substantial trend toward improvement. Nevertheless, some individual hospitals experienced substantial improvement while others worsened. Structural features of hospitals such as volume and teaching status do not adequately predict change in performance. Other factors that are related to improvement, such as organizational culture, physician leadership, interrelationship between cardiology and emergency physicians and staff, emergency medical systems, and administrative support need to be identified and used by hospitals for performance-improvement programs to decrease door-to-needle and door-to-balloon times and decrease mortality from STEMI.