The China PEACE-Retrospective AMI Study protocol has been published earlier [13, 14]. Briefly, a two-stage random sampling procedure was used to draw nationally representative cases hospitalized for acute myocardial infarction (AMI) in 2001, 2006, and 2011. In the first stage, a simple random sampling process was used to include representative hospitals from five economic-geographic strata: eastern-rural, central-rural, western-rural, eastern-urban, and central/western-urban regions. These strata were used because socioeconomic levels and healthcare resources vary across categories (Supplementary Fig. 1). In the second stage, AMI patients from the sampled hospitals were drawn using systematic random sampling methods. AMI hospitalizations were identified by International Classification of Diseases (ICD) codes, Ninth Revision (410.xx) and Tenth Revision (I21.xx), if available, or by the principal discharge diagnosis. This study also included patients admitted in 2015 using the same method.

Trained staff retrieved data from medical charts using clear abstraction approaches and standardized data definitions. Each medical record was copied by the participating hospital and transmitted to the central abstraction center. All abstractors received central training for two weeks. Those who could extract 10 sample medical records with more than 98% accuracy after training received certification. Rigorous monitoring was employed at each stage to ensure the accuracy of abstraction. Data quality was monitored by randomly auditing 5% of the abstracted records. The overall accuracy exceeded 98% [13, 14].

The central ethics committee of the Chinese National Center for Cardiovascular Diseases approved this study. Given the retrospective nature, written patient consent was not required. Ethics approval was also obtained from all participating hospitals.

Patients with a discharge diagnosis of STEMI were included, which was determined by combining the diagnosis at discharge with the results of electrocardiograms (ECGs). In cases where there was no definitive diagnosis from the local hospital, cardiologists from the coordinating center reviewed the medical logs and ECGs to determine the STEMI diagnosis. An independent cardiologist who did not take part in abstraction validated the AMI type by reviewing ECGs in randomly selected records. The concordance rate for the selected cases was 94.7%.

Patients were excluded if they had an established myocardial infarction, percutaneous coronary intervention (PCI), or coronary artery bypass graft before admission, or if their STEMI occurred during hospitalization. Also excluded were patients with missing baseline data on SMuRFs, patients transferred in or out because their hospitalization was truncated, and patients discharged alive within the first 24 hours, given that they were very likely to have left against medical advice and had very little time to receive treatment (Supplementary Fig. 2).

Similar to a previous study [1], we defined SMuRFs as having at least one of the following modifiable risk factors: current smoking, hypercholesterolemia, diabetes, or hypertension. The definition of current smoking is having smoked regularly (at least one cigarette per day) during the last six months. Hypercholesterolemia was defined as having a low-density lipoprotein cholesterol concentration $\ge$3.4 mmol/L, or a total cholesterol concentration $\ge$5.2 mmol/L during the index admission, or an established or new diagnosis of hypercholesterolemia. Diabetes was defined as having an established or new diagnosis of diabetes, and hypertension was defined as having an established or new diagnosis of hypertension. Blood pressure in the acute phase and fasting glucose were excluded from the definitions, as both are influenced by the neurohormonal response to AMI, which was consistent with the previous study [1].

The use of in-hospital therapies recommended by the guidelines for the management of STEMI was assessed, which included aspirin, P2Y${}_{12}$ inhibitors, dual antiplatelet therapy (DAPT), $\beta$-blockers, angiotensin-converting enzyme (ACE) inhibitors/angiotensin receptor blockers (ARBs), statins, and reperfusion therapy (primary PCI or fibrinolytic therapy) [15]. We assessed the usage rates of therapy merely in the eligible patients (i.e., those with no documented contraindications) after excluding those who died within the first 24 hours, since these may not have sufficient opportunity to be treated (Supplementary Method 1). To assess PCI and diagnostic catheterization, the study population was limited to the patients admitted to the hospitals with PCI capacity.

In-hospital mortality was defined as death or treatment withdrawal due to a terminal condition. The Chinese Government uses death or withdrawal of treatment as an indicator of hospital quality [16]. At the coordinating hospitals, cardiologists judged the clinical status of patients withdrawn from treatment based on their medical records. Composite complications included death, treatment withdrawal due to a terminal condition, congestive heart failure, re-infarction, ischemic stroke, or cardiogenic shock.

Continuous variables were analyzed using t-tests or non-parametric equivalent tests and presented as medians with interquartile ranges. Categorical variables were listed as percentages, and analysis was performed using $\chi$${}^2$ tests.

To explore the link between SMuRF-less status and therapies received, mixed models with hospitals as a random intercept were used to account for age, sex, medical histories, clinical profiles at admission, and admission ward.

Survival curves and hazard functions for in-hospital mortality were plotted. Mixed effect models were also used to assess whether SMuRF-less was independently associated with mortality, accounting for all explanatory variables stepwise. The models used in the study were: an unadjusted model; model 1 (adjusted for age and sex); model 2 (adjusted for model 1, medical history [stroke, atrial fibrillation, chronic renal disease, heart failure, peripheral arterial disease], and clinical characteristics at admission [systolic blood pressure [SBP], heart rate, chest discomfort, cardiac arrest, cardiogenic shock, stroke]); and model 3 (adjusted for model 2, pharmacotherapies during hospitalization [aspirin, P2Y${}_{12}$ inhibitors, DAPT, $\beta$-blockers, statins, ACE inhibitors/ARBs], as well as reperfusion therapy [primary PCI and fibrinolysis]). Stratified analyses were performed according to sex. Given that there were small differences in length of hospital stay between patients with and without SMuRFs, we also compared the adjusted 7-day mortality as a sensitivity analysis.

To explore the possible contribution of each treatment to the disparities in mortality, the effect of each treatment was investigated using age- and sex-adjusted analyses, taking the SMuRF-less status into consideration. Formal mediation analyses were also performed to examine the extent to which specific variables (including all clinical profiles and treatments) might contribute to in-hospital mortality in SMuRF-less patients. A mediator was defined as a variable that lies along the causal chain connecting the predictor and mortality. Traditionally, mediators are often adjusted in the assessment of causal association. Meaningful associations between mortality and the predictor could thus be removed, leading to incorrect conclusions of no association. Therefore, formal mediation analyses would facilitate identifying potential factors to explain the higher mortality among SMuRF-less and SMuRF patients. We calculated the percent mediation by dividing the indirect effect with the total effect and presented the proportion of the total effect attributable to the mediator. These analyses were performed with the mma package, as detailed elsewhere [17].

The present study did not impute missing values for SBP and heart rate in the models, since the missing data was minimal (0.2% of patients).

All statistical analyses were performed by SAS (version 9.4, SAS Institute Inc., Cary, NC, USA) and software R (version 4.2.1, R Foundation for Statistical Computing, Vienna, Austria). A two-sided p value 0.05 was considered statistically significant.

We included 16,541 patients (4970 women [30.0%], median age 66 years [56–74]), of whom 3288 (19.9%) had no documented SMuRFs and were henceforth referred to as SMuRF-less.

Baseline characteristics are presented in Table 1. SMuRF-less individuals were older (69 years [59–76] vs. 65 years [55–74], p 0.001) and more often female (33.6% vs. 29.2%, p 0.001) compared to those with SMuRFs. The duration from the onset of symptoms to hospital admission did not differ between the two groups (15 h [4–72] vs. 16 h [4–72], p = 0.970). SMuRF-less individuals were less likely to have chest discomfort or to be admitted to the cardiac ward. Despite similar median troponin concentrations (37.7-fold vs. 37.5-fold the upper limit of normal), we observed a longer delay in measuring cardiac enzymes in SMuRF-less patients (107 min [12–630] vs. 93 min [7–340], p 0.001). SMuRF-less patients also had a significantly lower SBP (120 mmHg [100–130] vs. 130 mmHg [111–150], p 0.001) and a greater proportion of SBP 90 mmHg (9.7% vs. 4.4%, p 0.001) and cardiogenic shock (9.4% vs. 6.1%, p 0.001).

The proportion of eligible patients for aspirin, P2Y${}_{12}$ inhibitors, DAPT, and statins was not different between patients with and without SMuRFs after excluding those who stayed in hospital for $\le$24 hours. However, the SMuRF-less group was less likely to be eligible for $\beta$-blockers (74.6% vs. 80.3%, p 0.001), ACE inhibitors/ARBs (93.5% vs. 96.2%, p 0.001) and reperfusion therapy (51.5% vs. 54.5%, p = 0.004) (Supplementary Table 1).

Among eligible patients, those in the SMuRF-less group were less likely to be treated with medications, including aspirin (89.0% vs. 94.7%, p 0.001), P2Y${}_{12}$ inhibitors (61.1% vs. 78.6%, p 0.001), DAPT (59.4% vs. 77.0%, p 0.001), $\beta$-blockers (78.3% vs. 85.7%, p 0.001), ACE inhibitors/ARBs (49.9% vs. 68.1%, p 0.001), and statins (69.9% vs. 85.1%, p 0.001) (Table 2). Additionally, the SMuRF-less group had lower utilization of primary PCI (17.3% vs. 28.8%, p 0.001), but similar use of fibrinolytic therapy (35.0% vs. 32.9%, p = 0.100). These differences persisted after adjusting for age, sex, medical history, clinical characteristics at admission, and admission ward. The most marked differences were observed for the use of ACE inhibitors/ARBs (odds ratio [OR] 0.56; 95% confidence interval [CI] 0.51–0.61) and statins (OR 0.61; 95% CI 0.53–0.70) (Table 2).

Individuals without SMuRFs experienced significantly higher crude in-hospital mortality (18.5% vs. 10.5%, p 0.001) and composite complications (26.0% vs. 19.0%, p 0.001) (Supplementary Table 2). 56.1% of deaths occurred during the first 24 hours in SMuRF-less patients compared to 38.9% in SMuRF patients. The 24-hour mortality rate was more than 2-fold higher in the SMuRF-less group than in patients with at least one SMuRF (10.4% vs. 4.1%, p 0.001). For patients who survived the first 24 hours of admission, the disparity in mortality between the two groups narrowed but remained significant (9.1% vs. 6.7%, p 0.001) (Supplementary Table 2). The length of hospital stay was shorter in SMuRF-less patients (9 days [IQR 5–14] vs. 11 days [7–15], p 0.001). Fig. 1A shows the survival curves for the two patient groups. In both, the highest hazard function (instantaneous risk) for death was within the first 24 hours (Fig. 1B). SMuRF-less individuals had a consistently higher risk of in-hospital mortality in all subgroups examined (Supplementary Table 3).

Fig. 1.

Survival curves and hazard function curves based on SMuRF status. (A) Kaplan-Meier survival curves for in-hospital mortality until discharge. (B) Hazard function for mortality during hospitalization. SMuRF, standard modifiable cardiovascular risk factor.

After adjustment for age and sex, the SMuRF-less group had a 68% greater risk of in-hospital death (OR 1.68; 95% CI 1.50–1.88) (Fig. 2). This difference was reduced after adjusting for clinical profiles (OR 1.41; 95% CI 1.25–1.59). After further adjustment for in-hospital treatment, the difference was no longer significant (OR 1.05; 95% CI 0.92–1.20). Among all of the individual treatments, the use of ACE inhibitors/ARBs (OR 1.29; 95% CI 1.15–1.45) or statins (OR 1.36; 95% CI 1.21–1.54) resulted in the largest reduction in the OR of mortality for SMuRF-less patients (Supplementary Fig. 3). The results of these analyses agreed with those for sex stratification (Supplementary Figs. 4,5) and for the use of a 7-day timeframe (Supplementary Figs. 6,7). However, after excluding individuals who died within 24 hours of admission and after adjusting for age and sex, the disparity in in-hospital mortality was only marginally significant (OR 1.16; 95% CI 1.00–1.35) (Supplementary Fig. 8).

Fig. 2.

Adjusted odds ratio for mortality during hospitalization between patients with and without SMuRFs. Model 1: adjusted for age and sex; Model 2: adjusted for the variables in model 1 and for the clinical profiles (including previous stroke, previous atrial fibrillation, previous chronic renal disease, previous heart failure, previous peripheral arterial disease, chest discomfort, cardiac arrest at admission, cardiogenic shock at admission, stroke at admission, heart rate, and systolic blood pressure); Model 3: adjusted for the variables in model 2 and for in-hospital pharmacotherapies (including aspirin, P2Y${}_{12}$ inhibitor, dual antiplatelet therapy, $\beta$-blocker, angiotensin-converting enzyme inhibitor/angiotensin receptor blocker, statin) and reperfusion therapy (fibrinolytic therapy, primary percutaneous coronary intervention). SMuRF, standard modifiable cardiovascular risk factor; OR, odds ratio.

Table 3 lists the mediating factors and their percent mediation in the overall population. Mediating factors were estimated to account for 92.5% of the excess in-hospital mortality observed in SMuRF-less patients compared to those who had SMuRFs. Although 23.0% of the excess mortality in SMuRF-less patients was mediated by worse clinical profiles, the majority (69.1%) was mediated by suboptimal in-hospital treatment. The contributions from in-hospital statin and ACE inhibitor/ARB treatments were 32.5% and 22.4%, respectively, accounting for the largest proportion of difference in mortality. The mediation analyses were repeated in the eligible patients, with similar results obtained. Among the eligible patients for ACE inhibitors/ARBs, 23.4% of the excess in-hospital mortality was due to the underuse of this treatment, while among the eligible ones for statins, 30.6% of excess mortality was due to treatment underuse.

After excluding patients who died within 24 hours of hospitalization, mediation analyses produced similar results to those of the overall population, i.e., 30.3% of the difference in mortality between SMuRF-less and SMuRF patients could be explained by the underuse of ACE inhibitors, and 13.8% by the underuse of statins.

Given that the most prominent in-hospital mortality difference was within the first 24 hours of hospitalization, we performed a sensitivity analysis using mediation analysis to examine the potential contributors to excess mortality in SMuRF-less patients within this period (Supplementary Table 4). The clinical profiles and treatments within 24 hours jointly accounted for 57.9% of the relationship between the SMuRF-less status and mortality within 24 hours. The underuse of P2Y${}_{12}$ inhibitor therapy within 24 hours accounted for the largest proportion (13.0%) (Supplementary Table 4).