Assessing care of patients with severe mental illness in the era of primary angioplasty pathways for ST elevation myocardial infarction using artificial intelligence methods

Abstract

Background<br/>People with severe mental illness (SMI) experience significant health disparities, including in cardiovascular care. This study used natural language processing (NLP), an artificial intelligence method, to investigate care pathways for SMI patients within a highly protocolized urban primary angioplasty network for ST-elevation myocardial infarction (STEMI).<br/>Methods<br/>This cohort study used NLP to extract SMI diagnoses and clinical data from free-text electronic health records, including STEMI and non-STEMI audit data. Primary outcomes included 28-day-and-1-year mortality, with secondary outcomes examining pathway performance metrics and procedural data. Exploratory outcomes assessed presenting symptoms and comorbidities.<br/>Results<br/>2,292 adults presented with first-time STEMI from 2012-2019, of whom 51 (2.2%) had recorded SMI diagnoses. No significant differences were found in mortality between SMI and non-SMI patients at 28-day and 1-year, though a non-significant trend toward higher crude mortality in SMI patients emerged at 3 years (23.53% vs 16.06%, p=0.152). Procedure metrics and timings were similar between groups. SMI patients were more likely to have presented with cardiac arrest (adjusted OR 2.48 [1.18-5.22],p=0.017) and had higher rates of chronic obstructive pulmonary disease (adjusted OR 2.20 [1.00-4.83], p=0.049). PCI rates in SMI patients with a non-STEMI were less than their representation in the overall non-STEMI population (2.19% vs 3.32%, p=0.042).<br/>Conclusion<br/>Highly protocolized STEMI pathways appear to provide equitable acute care for patients with SMI without excess short-term mortality. However, potential disparities emerge in longer-term outcomes and in non-STEMI treatment pathways, suggesting a need for improved secondary prevention and targeted interventions for this vulnerable population.

Ähnliche Arbeiten