Decision Making Analysis of Patients ‘Turned Down’ for Primary Percutaneous Coronary Interventions Due to Inapplicable Referrals: A Need for Enhanced Performance in Human and Machine ECG Interpretation

Introduction: Primary percutaneous coronary intervention (PPCI) establishes flow in an obstructed coronary artery. Before this, the electrocardiogram (ECG) is used to signpost patients to PPCI. We evaluate the factors that may have influenced the decision pathway. Methods: Case reports from patients referred for PPCI were evaluated as part of a compulsory audit. Demographics, symptoms, computer ECG interpretation, activator interpretation, time of referral, 30 day and 12 month mortalities were recorded. Computer and activator interpretations were simplified to being suggestive of acute myocardial infarction (MI). Results: 942 patients (68% male), mean age 68 + 17 years referred for PPCI were appropriately turned down for catheterisation. 62% (584) of these referrals were ‘out of hours’, with less referrals made on Saturdays and Sundays. Peak referral times occurred at 11am (75/941[8%]), 3pm (57/941 [6%]) and 8pm (63/941 [7%]). 12 month mortality rates were highest if patients were referred at 3am (7/22 [32%]). 76 (8.1%) patients died within 30 days and 108 (11.5%) within 12 months. 373 (39.6%) had chest pain, 142 (15.1%) had resolved chest pain leaving 427 (45.4%) who did not experience chest pain. 394 (42%) computerised diagnoses were suggestive of acute MI whereas 327 (35%) activator diagnoses were suggestive of acute MI. There were 182 (19%) cases where the human and computer agreed that there was an acute MI. They agreed more often when there was not an acute MI (403 [43% of cases]). Overall agreement rate was 62%. There were 145 (15.4%) cases where the computer did not suggest an acute MI but the activator did. There were 212 (22.5%) cases where the computer suggested acute MI but the activator did not. Age (p<0.001), a computerised acute MI diagnosis and being an out of hour referral increased the odds of mortality after one year (Table 1). Activator and computer acute MI agreement and having chest pain (p<0.01) reduced the odds of mortality after one year. The odds of mortality within 12 months of referral was lower in patients with chest pain compared to those patients without chest pain. Conclusion: Mortality in patients appropriately turned down for PPCI is higher than the reported mortality for STEMI patients at 12 months. Agreement between computerised and human interpretation is poor perhaps leading to a high inappropriate referral rate. Work is needed to improve machine and human decision making to ensure that patients are signposted to the correct treatment facility for time critical therapy.