A retrospective observational study was performed in patients admitted for ACS in SNHS hospitals between January 1, 2003, and December 31, 2018. The data source was the Minimum Data Set (MDS) of the SNHS, which includes data on the demographic characteristics of hospitalized patients, as well as administrative variables regarding the care process and clinical variables related to diagnoses and procedures, coded using the International Classification of Diseases (ICD) 9th (2003-2005) and 10th (2016-2018) editions.15 The ACS population was divided into 2 groups: ST-segment elevation acute myocardial infarction (STEMI) and non–ST-segment elevation acute coronary syndrome (NSTEACS) (non–ST-segment elevation acute myocardial infarction and unstable angina). The ICD-9 and -10 (Clinical Modification) codes used to identify diagnoses and procedures are detailed in table 1 of the supplementary data. PCI performance and in-hospital mortality were compared between WHA patients (considering national holidays alone) and those admitted on weekdays (from Monday to Friday, except national holidays).

To improve data consistency, transfers between hospitals were consolidated in a single episode that was attributed to the more complex hospital if the destination hospital was identified. In addition, episodes corresponding to patients younger than 18 years were excluded, as well as those corresponding to patients who were discharged to home and had hospital stays of 1 day or less and to patients registered as voluntary discharges with an unknown discharge location or with discharges to another hospital that could not be resolved after concatenation.

Quantitative variables are presented as mean ± standard deviation and categorical variables as frequency and percentage.

For adjustment by the risk of in-hospital mortality, models were based on the methodology of the Centers for Medicare and Medicaid Services. The independent variables were those included in the 30-day acute myocardial infarction mortality model,16 with adaptation of the Centers for Medicare and Medicaid Services model to the structure of the MDS data. The secondary diagnoses were then grouped according to the clinical condition categories developed by Pope et al.17 and updated annually by the Agency for Healthcare Research and Quality.16

To analyze the impact of the type of hospital on the association between WHA or weekday admission and in-hospital mortality, hospitals were classified by the availability of cardiology-related resources using RECALCAR criteria.18 To differentiate the impact of weekend admission from national holiday admission, the independent variable WHA was separated into 2 components: a) national holiday admission (dichotomous); and b) weekday of hospital admission (categorical, with Tuesday as reference, to clarify the impact of Friday and Monday, the days immediately before and after the weekend).

In-hospital mortality was adjusted using multilevel logistic regression models.19 All adjustment models specifically considered comorbidities with statistical significance and an odds ratio (OR) < 1.0. The adjustment models were estimated using backward elimination; the significance thresholds for the selection and elimination of factors were P < .05 and P ≥ .10, respectively. Using the specified models, we calculated risk-adjusted in-hospital mortality rates (RAMRs).20 Calibration was analyzed graphically after patients were grouped in deciles with regard to predicted probabilities, and predicted vs observed mean probabilities were tabulated, as well as discrimination using the area under the receiver operating characteristic (ROC) curve.

Quantitative variables were compared using the t test for 2 categories. Categorical variables were compared using the chi-square test or Fisher exact test.

Temporal trends in mortality and PCI performance during the study period were assessed using a Poisson regression model with year as the sole independent variable. In all models, we calculated the incidence rate ratios (IRRs) and their 95% confidence intervals (95%CIs).

As sensitivity analysis, we evaluated the changes over time in the frequency of PCI performance considering, in both groups (STEMI and NSTEACS), the impact of WHA vs weekday admission; 2 periods were defined, taking as cutoff the year in which a significant change was detected, for which the multilevel model was adjusted by including PCI performance among the independent variables.

All comparisons were 2-sided and differences were considered significant at P < .05. Statistical analyses were performed with STATA 16 and SPSS v21.0.

In total, 475 362 STEMI episodes and 548 176 NSTEACS episodes were identified from 2003 to 2008; after exclusions (figure 1 of the supplementary data), we obtained a study population of 438 987 (92.3%) and 486 565 (88.8%) episodes, respectively. Of these, 126 528 (28.8%) and 127 188 (26.1%) were WHAs, without significant variations in these proportions during the study period. The profiles of the patients admitted on weekdays and on weekends or public holidays were notably homogeneous in the 2 groups (table 1 and table 2), with the exception of a slightly higher rate of cardiogenic shock and cardiorespiratory failure in patients with NSTEACS admitted at the weekend.

The crude mortality rate during the study period was 11.7% in the STEMI group and 5.5% in the NSTEACS group. Both groups exhibited slightly but significantly higher mortality between weekday admission and WHA (STEMI, 11.6% vs 11.9%, P = .008; NSTEACS, 5.3% vs 5.9%, P < .001). In STEMI, the crude mortality rate significantly fell during the study period (IRR = 0.998; 95%CI, 0.998-0.999; P < .001) in both patients admitted on weekdays (IRR = 0.999; 95%CI, 0.999-0.999; P < .001) and those admitted on public holidays (IRR = 0.998; 95%CI, 0.998-0.999; P < .001). In contrast, in the NSTEACS group, the crude mortality rate increased during the study period (IRR = 1.004; 95%CI, 1.000-1.010; P = .002) in both patients admitted on weekdays (IRR = 1.003; 95%CI, 1.000-1.010; P = .049) and those admitted on public holidays (IRR = 1.004; 95%CI, 1.002-1.007; P = .003).

The adjusted in-hospital mortality risk models for both STEMI and NSTEACS showed very good discrimination (ROC, 0.88 and 0.87, respectively) and calibration (table 2 of the supplementary data, table 3 of the supplementary data, figure 2 of the supplementary data, and figure 3 of the supplementary data). Both groups exhibited a significant trend for a decrease in the RAMR for both weekday admissions and WHAs (total RAMR of each group, IRR = 0.999; 95%CI, 0.999-0.999; and IRR = 0.996; 95%CI, 0.995-0.997; both P < .001). The WHA variable was found to be a risk factor in risk-adjusted models for STEMI (OR = 1.05; 95%CI, 1.03-1.08; P < .001) and NSTEACS (OR = 1.08; 95%CI, 1.05-1.12; P < .001).

When we analyzed the day of admission and admission on national holidays separately instead of together in WHA (table 4 of the supplementary data and table 5 of the supplementary data), there was a greater effect in both STEMI and NSTEACS of national holiday admission (OR = 1.11; 95%CI, 1.03-1.19; P < .001; and OR = 1.15; 95%CI, 1.05-1.26; P < .001) than admission on Saturday or Sunday (STEMI: OR = 1.04; 95%CI, 1.00-1.09; P = .04; and OR = 1.03; 95%CI, 0.99-1.08; P = .13; NSTEACS: OR = 1.08; 95%CI, 1.03-1.14; P < .001; and OR = 1.08; 95%CI, 1.02-1.14; P = .01).

In STEMI, hospitals with a catheterization laboratory were protective factors (RECALCAR types 3, 4, and 5, OR = 0.86, OR = 0.80, and OR = 0.64, respectively; P < .001); the same effect was found in NSTEACS, although the results were only significant for type 5 hospitals (with catheterization laboratories but without beds specifically assigned to cardiology) (type 3, OR = 0.95; type 4, OR = 0.93 [P > .05]; type 5, OR = 0.72; P < .001); in both cases, WHA remained a risk factor (STEMI: OR = 1.05; NSTEACS: OR = 1.08; P < .001).

From 2003 to 2018, the PCI performance rate was significantly lower in patients with WHA than in those admitted on weekdays in both STEMI (46.8% vs 48.7%; P < .001) and NSTEACS (28.2% vs 29.1%; P < .001). However, the differences seemed to have no clinical relevance. In contrast, the rate was significantly increased in STEMI in both patients admitted on weekdays (IRR = 1.069; 95%CI, 1.055-1.082, P < .001) and WHA patients (IRR = 1.078; 95%CI, 1.063-1.092; P < .001), and the same tendency was found in NSTEACS (weekday admission, IRR = 1.072; 95%CI, 1.061-1.083; P < .001; WHA, IRR = 1.077; 95%CI, 1.066-1.087; P < .001) (figure 1).

Figure 1.

Changes over time in the rate of percutaneous coronary intervention (PCI) performance. A: in ST-segment elevation acute myocardial infarction (STEMI). B: in non–ST-segment acute coronary syndrome (NSTEACS) (weekdays [week] and public holidays [hol]).

(0.08MB).

Through analysis of the changes over time in the frequency of PCI performance, a significant change was observed in 2012 in both groups (STEMI and NSTEACS) in patients admitted on weekdays or weekends/public holidays, and 2 study subperiods were identified: 2003 to 2011 and 2012 to 2018. The rate of PCI performance in STEMI was more than 2 percentage points higher in patients admitted on weekdays from 2003 to 2011 and similar or even lower from 2012 to 2018 (figure 1), with a significant interaction between the year and a lower difference in the rate of PCI performance (the later the period, the lower the difference) (P = .006). In contrast, this interaction was not found between the year and the difference in PCI performance in NSTEACS (P = .253).

The performance of PCI in STEMI had a protective effect against in-hospital mortality at all admission times and in both periods (2003-2011, OR = 0.25; 95%CI, 0.24-0.26; P < .001; 2012-2018, OR = 0.29; 95%CI, 0.27-0.30; P < .001). The same effect was observed in NSTEACS (2003-2011, OR = 0.36; 95%CI, 0.33-0.39; P < .001; 2012-2018, OR = 0.34; 95%CI, 0.32-0.37; P < .001).

The effect of WHA was identified as a significant risk factor in both periods in both STEMI and NSTEACS, even when PCI performance was considered in the multilevel adjustment models for in-hospital mortality (table 3).

Our study has several limitations. Due to its observational and retrospective design, we cannot exclude the possible existence of a selection bias and a possible effect of unmeasured confounding factors. In addition, given the data source, there may be biases because the MDS does not collect information on certain clinically relevant variables—such as date and time of PCI performance, the artery affected, multivessel disease, and left ventricular ejection fraction—and the quality of the data on diagnoses and procedures performed in patients during their hospitalization depends on the coding adequacy in each hospital. Moreover, the MDS does not provide information allowing a comparison of outcomes between WHA patients and outcomes obtained on weekdays by the timing of catheterization alert activation. However, in contrast, one of the strengths is that the use of administrative databases for investigating health outcomes has been validated through their comparison with data extracted from medical records,35 as well as that the MDS provides considerable statistical power because it includes a nationwide population over a large time period and with established validity for the study of ACS in the SNHS.36

In contrast to the models developed by the Centers for Medicare & Medicaid Services, the MDS did not record comorbidities present at admission until 2016, which is why the secondary diagnoses used as adjustment variables could be comorbidities or complications that occasionally reflect inadequate management.34 Nonetheless, the models used in this study are comparable with other published models regarding their discrimination capacity.16 In addition, it must be remembered that the adjustment models include confounding factors that are impossible to consider and may have a significant impact.