RESULTS
A total of 1193 patients were included in the STEMI database provided by the coordinators at the six hospitals. As shown in Table 2, the number of patients per hospital ranged from 122 to 322. As expected, males exceeded females in the overall sample (66.3% versus 33.5%, respectively). The mean age of patients in the database was 62.0 ± 13.7 years. Of the 1193 patients in the databases, D2B times were available for 818 patients. Reasons provided for incomplete D2B data of the other 375 patients included missing data (37.3%), cardiopulmonary arrest/death before reaching CCL (16.5%), normal coronary arteries (13.6%), catheterization canceled by ED physician or cardiologist (13.3%), non-STEMI (6.4%), atypical presentation/STEMI missed (5.1%), radiology imaging for an alternative diagnosis (3.2%), patient refusal (2.1%), inability to cross lesion (1.1%), needing coronary artery bypass graft (1.1%) and unavailable CCL (0.3%).
There were significant differences in the patient populations in the six hospitals (Table 3). For example, mean patient age differed significantly among the six facilities (p=0.009). While 68.2% of the total patients were admitted via ambulance, this form of admission was highest in hospital #1 and lowest in hospital #6. Similarly, scenarios for admission differed significantly among the hospitals (p<0.001). Overall, Scenario A was the most frequent type of admission (54.8%), followed by Scenario C (39.5%). Although Scenario B occurred infrequently overall (4.9%), its highest frequency was observed at hospital #5 (18.5%).
At least one patient-related confounder variable was present in 23.4% of the total sample (Table 3) and was distributed differently among the six hospitals (p<0.001). Patients at hospital #3 had the highest incidence of at least one patient-related confounder (34.4%), while hospital #4 had the lowest incidence (10.6%). Four patient-related confounders were found to differ significantly among the six hospitals including cardiac arrest (p=0.015), intubation (p=0.009), need for MRI or CT (p=0.017) and inconclusive ECG findings (p=0.05). At least one non-patient-related confounder was present in 4.4% of the population and differed significantly among the facilities (p<0.001). There was not enough data to examine the specific non-patient related confounders by hospitals. The mortality rate of patients in the database was 9.2% and there was significant difference in the mortality across all six hospitals (p<0.001). The highest mortality was observed in hospitals #1 (13.9%) and #2 (15.5%) and the lowest mortality was in hospital #5 (2.8%).
As shown in Table 4, the median (IQR) overall time from the ED to the first ECG was 4 (1-8) minutes. All hospitals had median times less than 10 minutes; however, 25% of the patients in hospitals #2, #3 and #5 had median times above 9 minutes. The range in median values among the six hospitals differed significantly (-7.5 to 5 minutes, p<0.001). When categorized as ≤10 minutes or >10 minutes, hospital #4 met the requirement of ≤10 minutes 88% of the time, while hospital #2 only met it 79% of the time, p=0.025. The overall median (IQR) ECG to activation time was 3 (-9 to 9) minutes and the shortest one was -5 (-16 to 4) minutes (hospital #4). Although there is no guideline recommendation for ECG to activation time, studies suggest activation should occur within 5 minutes after the diagnostic ECG is obtained. Since D2B time is a time sensitive medical situation, the time to activation is an important factor that contributes to overall D2B time. The overall median (IQR) time of activation to the CCL table in this study was 40 (31-48) minutes.
As shown in Table 5, the overall median (IQR) D2B time for the entire sample was 63 (50-78) minutes. The difference between D2B time among hospitals was significant (p<0.001); the majority of the hospitals had a median D2B time of 60-62 minutes and the worst median D2B time was 72 (58-87) minutes at hospital #5. When D2B times were categorized as ≤90 or >90 minutes the highest percent of adherence to the guideline of ≤90 minutes was 97.1% at hospital #4 and lowest was 79.1% at hospital #5, p<0.001. The overall percent of adherence was higher than the minimal adherence score recommended by the American Heart Association (87.8% vs 75% respectively, p <0.001).3
The two hospitals with the highest mortality (hospitals #1 and #2) had longer average D2B times (72.1+ 51.5 minutes for hospital #1 and 73.0+ .6 minutes for hospital #2) than in the total patients documented (69.2+ 38.9 minutes). These findings indicate that more patients in hospitals #1 and #2 did not get recanalization of the occluded vessel as timely as others. The effect of D2B times< 90 minutes on mortality was very substantial; patients with D2B times < 90 minutes had a significantly lower mortality than those with D2B times >90 minutes (5.3% vs 19.0% respectively, p<0.001). The effect of D2B times on mortality was not completely consistent and there were some unexplainable findings. Hospital #5 had the highest average D2B time (80.3+ 45.7 minutes) and lowest percentage of attaining D2B times< 90 minutes (79.1%) but had the lowest mortality rate (2.8%). Hospital #6 which had a lower mortality rate (7.8%), had a slightly lower percentage of patient achieving D2B times< 90 minutes (86.8%) but had substantially lower average D2B times (66.4+ 44.1 minutes) than the total population.
As shown in Table 6, significant predictors of D2B time included: hospital, number of steps from ED to CCL, mode of arrival, scenario of arrival, and patient-related confounders: Significant differences in D2B time were found in some of the hospitals, even after adjusting for other variables. When compared with hospital #1, hospital #3 had a 14.4-minute lower D2B time (p=0.01) and hospital #6 had a 11.6-minute lower D2B time (p=0.01) after adjusting for other variables. Although hospital #4 had a substantially lower unadjusted time (p=0.049), it was not statistically significant (p=0.08) after adjustment. Hospitals #2 and 5 also did not differ significantly from hospital #1 in D2B time. While the unadjusted number of steps from the ED to the CCL was statistically significant (p=0.02), the difference was no longer statistically significant (p=0.31) after adjusting for other variables.
Compared to ‘walk-in’ as a mode of arrival, EMS arrival was associated with a 12.5-minute shorter D2B time, p<0.001; however, after adjusting for other variables it was no longer significant. After adjusting for other variables, compared to Scenario A, Scenario B was associated with a 27.3-minute lower D2B time (p<0.001) and Scenario C was associated with a 22-minute lower DTB time (p=0.001). Field ECG/activation with or without stopping in the ED was superior to ED ECG/activation (either walking in or EMS transport). Presence of one or more patient-related confounders was associated with about a 21.5-minute greater D2B time in both adjusted models compared to someone with no patient-related confounders (p<0.001 for both). Presence of a non-patient-related confounder was found to have a 9 to 10-minute greater D2B time in both adjusted models; however, it was not a significant predictor of D2B time in either model (model 1: p=0.13 and model 2: p=0.17). Non-significant predictors of D2B in the unadjusted and either adjusted models were age and day of week.