Using latent class analysis to identify clinical features of patients with occlusive myocardial infarction: Preangiogram prediction remains difficult

Abstract Background Treatment decisions in myocardial infarction (MI) are currently stratified by ST elevation (ST‐elevation myocardial infarction [STEMI]) or lack of ST elevation (non‐ST elevation myocardial infarction [NSTEMI]) on the electrocardiogram. This arose from the assumption that ST elevation indicated acute coronary artery occlusion (OMI). However, one‐quarter of all NSTEMI cases are an OMI, and have a higher mortality. The purpose of this study was to identify features that could help identify OMI. Methods Prospectively collected data from patients undergoing percutaneous coronary intervention (PCI) was analyzed. Data included presentation characteristics, comorbidities, treatments, and outcomes. Latent class analysis was undertaken, to determine patterns of presentation and history associated with OMI. Results A total of 1412 patients underwent PCI for acute MI, and 263 were diagnosed as OMI. Compared to nonocclusive MI, OMI patients are more likely to have fewer comorbidities but no difference in cerebrovascular disease and increased acute mortality (4.2% vs. 1.1%; p < .001). Of OMI, 29.5% had delays to their treatment such as immediate reperfusion therapy. With latent class analysis, while clusters of similar patients are observed in the data set, the data available did not usefully identify patients with OMI compared to non‐OMI. Conclusion Features between OMI and STEMI are broadly very similar. However, there was no difference in age and risk of cerebrovascular disease in the OMI/non‐OMI group. There are no reliable characteristics therefore for identifying OMI versus non‐OMI. Delays to treatment also suggest that OMI patients are still missing out on optimal treatment. An alternative strategy is required to improve the identification of OMI patients.

optimal treatment. An alternative strategy is required to improve the identification of OMI patients. Scotland, ACS is the cause of 6600 deaths making it the leading cause of death. 1 The most severe form of ACS is when a coronary artery is occluded, commonly presenting as a ST-elevation myocardial infarction (STEMI) with a high short-term mortality (9.7% of all hospital patients). Interestingly, non-ST elevation myocardial infarction (NSTEMI), which during the acute phase are less fatal but have a higher 1-year mortality (18.7% vs. 8.4% of hospital survivors). 2 The traditional explanation of differentiation and mortality rates between STEMI and NSTEMI, was that ST elevation on the electrocardiogram (ECG) represents acute coronary artery occlusion with a large area of cardiac myocardium with no blood flow and therefore an increased short-term mortality. 3 In contrast, NSTEMI patients tend to be older with multivessel disease and increased premyocardial infarction (MI) comorbidities but only have partial coronary artery occlusion, giving a potential explanation to why they have increased long-term mortality but lower short-term mortality than STEMI. 4 However, there is increasing evidence that there is a subset of NSTEMI patients who do have acute coronary artery occlusion. [5][6][7][8] Meta-analysis of studies looking at angiographic data of NSTEMI patients found that 25.5%-39% of NSTEMI patients have coronary artery occlusion and this is associated with increased short and long-term mortality. 5,9 There is also an increase in mortality in comparison to STEMI patients, as due to the lack of ST-elevation on the ECG, these patients may be mis-triaged and do not receive timely reperfusion therapy such as percutaneous coronary intervention (PCI) or thrombolysis. 10 There is no clear way to clinically distinguish between occlusive MI and nonocclusive MI before angiography as ST-elevation appears nonspecific for coronary artery occlusion and troponin is raised in any cause of myocardial necrosis or turnover regardless of coronary artery occlusion. 11 Clearly, there is increased imperative to classify ACS as occlusive MI and nonocclusive MI. In turn, there is an obvious need to identify features that distinguish between occlusive MI and nonocclusive MI. 12 The aim of this study was to use a form of unsupervised learning called latent class analysis to analyze the demographics of patients presenting with ACS to identify if there were, differing features in patients with occlusive and nonocclusive MI that may in turn, improve preangiogram triage.
This was a single-center, retrospective case-control study in a PCI unit based at a rural regional center in a hospital in the North of

| Study design and data set collection
In this study, data from 2015 to 2019 were analyzed after the removal of identifiable data such as names, address, dates of procedure, and date of birth. To determine whether a patient had acute coronary artery occlusion, the recorded stenosis status of the coronary artery before and after PCI was compared. In the data set, the stenosis pre-and post-PCI of the coronary arteries left main stem, left anterior descending artery (LAD-proximal and distal), right coronary artery (RCA), and left circumflex (LCx) were recorded. Acute coronary artery occlusion was identified on angiogram if the pre-PCI stenosis was 100% and post-PCI stenosis was 0%-49%.

| Statistical analysis
The data were entered onto SPSS™ version 25 (IBM) for statistical analysis. For initial descriptive and inferential statistical analysis, crosstabs with Pearson χ 2 testing were used to determine for categorical variables. Binary logistical regression was used if the independent variable had more than two levels. Fisher's exact test was used when there were categories that had values less than five and an independent sample t-test was used to determine statistical significance for continuous variables. A p-value less than .05 was considered significant. Given the relatively large number of indicator variables compared to cases, the data set was rendered more tractable for analysis by dichotomizing certain variables (LVEF and New York Heart Association symptoms). An initial assessment of the contribution of the various indicator variables to the model also identified smoking status as being of low significance in partitioning the classes, and it was therefore removed.
Class membership is considered a "latent" or unobserved variable, that may capture underlying phenotypes not accessible through more traditional analysis. 13 Overall goodness of fit of a particular number of classes to the data set is assessed by various measures and statistical tests, and robust confidence intervals (CIs) for indicators can be calculated. Further details of the analysis conducted are provided in the Supporting Information. As an omnibus test, the Wald test does not identify between which classes the significant difference arises; therefore, z tests with a Holm-Bonferroni correction were used to establish where significant differences occur.

| Ethics
Ethical permission for the research was obtained from the NHS Highland Caldicott committee. As the data had already been collected for audit purposes and this study did not involve any patients contact or intervention, full ethical permission was not required.

| RESULTS
A total of 1412 underwent PCI for acute MI, and 263 had occlusive MI on angiogram (Table 1). Of these, 510 (36.1%) patients were classified as a STEMI compared to 902 (63.9%) who were classified as NSTEMI. There were 263 (18.6%) patients with an occlusive MI and 1149 (81.4%) patients with nonocclusive MI. Table 2 lists the demographics and outcomes of the occlusive MI and nonocclusive MI cohort as well as the outcomes of the NSTEMI occlusive MI and STEMI occlusive MI cohorts.
In the initial approach, a latent class analysis model was derived for all the indicators together, determining that a model with three classes fit the data best ( Figure 1). Thus, for example, a member of latent class 2 has an 80% chance of having a history of hypertension, but only a 24% chance of having a STEMI. Latent class 2 is characterized as an older, comorbid cluster, while latent class 1 has significantly higher rates of STEMI and acute occlusion.
Considering those NSTEMI patients found to be acutely occluded, the majority were assigned to Class  However, patients with nonocclusive MI were more likely to receive thrombolysis (15% vs. 6.8%). This finding would make sense as thrombolysis would increase the likelihood of reperfusion of an occluded artery and for the vessel to appear nonocclusive when the patient undergoes PCI and thus initial appearances may have been of occlusive MI. Occlusive MI patients were also significantly less likely to have comorbidities. A lower prevalence of diabetes and hypertension has already been described in occlusive MI but unlike our results they also found occlusive MI patients younger and smokers. 14 Conversely, diabetes and hypertension have been found to be independent risk factors for occlusion as part of the CHA2DS2-VASc scoring system along with previous stroke and vascular disease. 15 Although numbers in the study were low, further research is required to clarify the risk factors for occlusion.
As around one-quarter of all NSTEMI have acute occlusion with associated increased mortality, we would expect the same in our data. 5 In the cohort analyzed in this study, 36.5% (n = 96) of OMI were classified as NSTEMI according to ECG and only 32.4% (n = 31) received immediate reperfusion therapy compared to 94.0% (n = 157) of occlusive STEMI (p < .001). Apart from a higher rate of hypercholesteremia (8.3% vs. 1.8%; p = .02), these patients did not have any differing characteristics compared to occlusive STEMI. There was also no significant difference in mortality and the occlusive STEMI patients were significantly more likely to have a reduced LVEF. This would be expected as occlusive NSTEMI is more associated with LCx and RCA occlusion and thus less likely to cause LV dysfunction. 5,16 When comparing similarities and differences between occlusive MI and known STEMI patient cohort, they were broadly similar with a higher percentage likely to be unstable, less comorbidities, and higher risk of death and morbidity. 2

| Limitations
This study had several limitations. Importantly, it is data from a single center and under the influence of regional and local population var-