We used NRMI, a voluntary acute myocardial infarction (AMI) registry sponsored by Genentech, Inc. (South San Francisco, CA), to define a cohort of patients with STEMI who received acute fibrinolytic reperfusion therapy. The NRMI criteria (13
) include a diagnosis of AMI according to the International Classification of Diseases, Ninth Revision, Clinical Modification
(code 410.X1) and any of the following criteria: total creatine kinase or creatine kinase MB that was ≥2 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. Of those, 294,474 were diagnosed with ST elevation of 2+ leads or left bundle branch block. From this cohort, the following patients were excluded sequentially: patients who did not receive primary fibrinolytic therapy, including fibrinolytic therapy times that were negative, missing, or >6 hours (n=182,406); patients transferred from another acute care institution (n=31, 879); patients with time from symptom onset-to-diagnostic electrocardiogram (ECG) that was negative, >6 hours or missing (n=7,743); and patients with a diagnostic ECG time that was not the first ECG time, that was >1 hour before admission, or >6 hours after admission (n=5,189). In addition, patients treated in hospitals outside the US (n=22) or reporting <20 patients over the 4-year time period (n=4,765) were excluded. The final cohort included 62,470 patients from 973 hospitals. Mortality status at the time of discharge was known for all patients.
Our outcome was in-hospital mortality and the principal independent variable was the door-to-needle time, which is the time from hospital arrival to administration of fibrinolytic therapy, derived from the corresponding date/time noted in the medical record and recorded in the NRMI case report form. Patients who were transferred to another facility were counted as survivors. Other patient-level variables included age (<65, 65–79, ≥80 years), sex, race/ethnicity (white, black, Hispanic, other), insurance status, and clinical characteristics. Clinical characteristics consisted of medical history (current 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, the admission time of day [day, evening, or night], admission day of week [weekday or weekend], chest pain at presentation, systolic blood pressure, heart rate, heart failure); and the results of the diagnostic ECG (number of leads with ST-elevation, AMI location, ST depression, nonspecific ST/T-wave changes, Q-wave).
To assess the independent effect of door-to-balloon time on in-hospital mortality and of symptom onset-to-door time on in-hospital mortality, we employed a multivariable hierarchical logistic regression model using in-hospital death as the dependent variable. Because NRMI enrolls hospitals that then report information about their patients, we could not assume that patient observations were independent of hospital; assessment of intraclass correlations indicated that variation in both the logarithm of time to treatment (ρ = 0.0876, 95% Confidence Interval (CI) [0.0780, 0.0971]) and mortality (ρ = 0.0073, 95% CI [0.0052, 0.0935]) was partially explained by hospital. Thus, we used hierarchical models to account for clustering of patients within hospitals; a random effect was specified for the main intercept. We replicated the model in each of the 3 strata of symptom onset-to-door time. The stratification variable was not included in the corresponding subgroup model. Because 54% of the patients were transported to another hospital after receiving fibrinolytic therapy (median length of stay for patients transferred to another hospital was 1.1 days; 12% staying ≤2 hours), we conducted 3 secondary analyses to assess the robustness of our results. First, we re-estimated the main analysis using a cohort excluding patients transferred to another hospital; next, we repeated this analysis using only patients from hospitals with a transfer rate ≤15%. Finally, to account for any association between time to fibrinolytic therapy and length of stay that may affect in-hospital mortality, we repeated the analysis using survival models, with censoring for patients who were transferred, with standard errors adjusted for clustering by hospital. Statistical analyses were performed using HLM 6.02 for Windows (SSI, Lincolnwood, IL), and Stata version 9.2 (Stata Corp., College Station, TX).