For the present study, all cases of AMI (ICD code, 10 I.21) reported between 2001 and 2007 in the national database for hospital admissions and in the GEMI hospital registry over the same period were selected.

The nationwide hospital database records information on all admissions to hospital, whether public or private, in Chile. The data is digitalized by trained staff, and then processed by the Department of Health Statistics and Information of the Chilean Health Ministry. Information on diagnosis, sex, age at the time of infarction, condition on discharge, and type of hospital (public or private) was collected from this database. In Chile, healthcare institutions are classified as public or private: public establishments provide care to those with least resources and receive a subsidy from the state. This variable was therefore used as a proxy measure of socioeconomic level.

The GEMI registry was designed as a prospective, multicenter registry of patients with diagnosis of AMI with or without ST elevation, admitted to public or private hospitals in the different regions of Chile. The patients included in the registry have elevated creatine kinase and an MB fraction twice or more the normal value or troponin elevation above the normal value. They also must have at least one of the following criteria: a) typical chest pain lasting at least 30min, and b) abnormal electrocardiogram characteristic of AMI (ST elevation or depression≥0.1mV at 2 or more adjacent leads, pathologic Q waves, or new left bundle branch block). The registry began collecting data in 1993 and has incorporated patients continuously through to the present day. Participation by the hospitals is voluntary, and so this is a convenience sample of patients who, for the most part, come from bigger hospitals, many of which are reference hospitals in their respective regions. In each case, a form is completed that collects information on demographic and socioeconomic characteristics, medical history, treatment, and outcome of hospitalization. The information is then digitalized and processed in a central facility of the Chilean Society of Cardiology and Cardiovascular Surgery.

Thus, the GEMI database is a sample of the cases of AMI registered in Chile, and provides additional information that is not included in other national databases.

In the study period, 9909 patients from 29 centers with different catchment populations were included in the GEMI registry. Of this total, 55 patients with no information on sex or age were excluded (n=9854). Subsequently, reference hospitals for their respective regions or health areas with a constant participation in the registry (at least 5 of the 7 years studied) were selected. The final sample was therefore reduced to 16 centers and 9278 patients (94% of the initial sample).

An exploratory analysis of the data was undertaken, checking the normal distribution of the continuous variables with the Shapiro-Wilk test. Quantitative variables were described as means (standard deviation) or quartiles, depending on their distribution, and qualitative variables as absolute and relative frequencies.

In-hospital mortality was calculated, defined as the ratio of deaths due to AMI during hospitalization and total of cases diagnosed and hospitalized with AMI. Mortality was calculated by sex and age group, divided into 10-year intervals from<45 years to≥75 years. To compare these variables according to group and between databases, the Student t test or the Wilcoxon test were used for continuous variables, the χ2 test for categorical variables. In addition, the trend towards in-hospital mortality in the different age groups was assessed using the χ2 trend test.

The odds ratios (ORs) were estimated along with the 95% confidence interval (CI95%), both raw and adjusted for socioeconomic status, for in-hospital mortality of women at a national level and in the GEMI registry for the different age groups (taking men as reference). From the information in the GEMI registry, multiple logistic regression models were constructed to estimate the risk of in-hospital mortality in women, stratified according to type of infarction (with and without ST elevation) and age group. The models were adjusted by risk factors (diabetes mellitus, hypertension, dyslipidemia, obesity, and tobacco use), prior infarction, Killip functional class on admission (Killip I and II vs III and IV), treatments during hospitalization (beta blockers, angiotensin-converting enzyme inhibitors, acetylsalicylic acid, and statins), and reperfusion procedures (primary angioplasty and thrombolysis). To achieve groups of comparable size, only 4 age groups were used:<55, 55-64, 65-74, and≥75 years. In total, 353 patients were excluded because they did not have a diagnostic electrocardiogram, 239 because there was no electrocardiographic information, and 360 for the absence of information on Killip class on admission. The final sample for the multivariate model was composed of 8464 patients, 91.2% of the initial sample. To assess the impact of these losses on the estimates, a sensitivity analysis was done taking the best possible case for the Killip variable. The final model included only those variables that showed a statistically significant association in the univariate analysis. After each model, the predictive capacity was assessed by calculating the specificity and sensitivity for providing the correct classification. The STATA 11.219 statistical package was used and statistical significance set to 5%.

In Chile, 49 287 patients were admitted to hospital for AMI between 2001 and 2007. Of these, 31.3% (n=15 408) were women, with a mean age of 68.6 (13.6) years, whereas the mean age of men was 61.5 (13.2) years (P<.001). In the GEMI registry, in the same period, 9278 patients were included, with women accounting for 27.1% (n=2517). The mean age of women was 67.8 (12.7) years. The mean age of men in this registry was 61.0 (12.5) years (P<.001). On comparison of the 2 registries, both men and women in the GEMI database were slightly younger (P<.01). Most of the men were aged 55-64 years, but most women were older than 75 years. The distribution of cases was similar in both databases (Fig. 1). The proportion of patients admitted to public hospitals was similar (75.7% in the national database and 79.3% in the GEMI registry).

Figure 1.

Distribution by sex and age of patients with acute myocardial infarction, for the entire country and in the GEMI registry, 2001-2007. GEMI, acute myocardial infarction multicenter study group.

(0.11MB).

In-hospital mortality at a national level was 14.2%, and significantly higher among women than among men (20.4% vs 11.3%; P<.001). In the GEMI registry, in-hospital mortality was lower (9.2%), and was also greater among women (14.2% vs 7.3%; P<.001).

Analysis of the mortality trend by age group showed that it increased significantly with age in men and in women and in both registries analyzed. At a national level, in-hospital mortality among women ranged from 6.2% in those under 45 years to 30.9% in those aged ≥75 years; among men of the same groups, it ranged from 2.8% to 25.4% (P<.0001 for the trend). In-hospital mortality among women was significantly greater than that of men for all age groups in both databases (P<.01) (Fig. 2). There was no interaction between sex and age (P=.30) but it was significant between sex and Killip class (I-II vs III-IV; P=.001).

Figure 2.

In-hospital mortality due to myocardial infarction by sex and age group, for the entire country and in the GEMI registry, 2001-2007. GEMI, acute myocardial infarction multicenter study group.

(0.09MB).

On evaluating the estimated risk of death (OR), young women were at greatest risk compared to men of the same age group. As age increased, this difference decreased, but the risk was always significantly higher in women. These findings were similar in both databases analyzed, although the variation between age groups was lower in the national registry; the risk decreased from OR=2.3 (1.5-3.3) in those under 45 years to OR=1.3 (1.2-2.4) in those aged 75 years or older, whereas the decrease was from OR=2.7 (1.1-6.8) to OR=1.5 (1.2-1.9) in the GEMI registry. After adjusting for socioeconomic level, the same trend persisted (Fig. 3). Although mortality was higher in the national registry than in the GEMI registry, given that the study aimed to assess the estimated relative risks and that the OR confidence intervals derived from the national database encompassed the point estimates of the OR derived from the GEMI registry, we proposed to using information from the GEMI registry to explore possible explanations for the difference in risk between men and women in the different age groups in large hospitals.

Figure 3.

Estimated risk of in-hospital death due to myocardial infarction in women (compared to men) by age group, for the entire country and in the GEMI registry, 2001-2007. Odds ratio are adjusted for socioeconomic level. Men are taken as reference. GEMI, acute myocardial infarction multicenter study group.

(0.09MB).

No significant differences were observed between men and women in the time between symptoms onset and seeking medical attention.

Table 1 shows the estimated risks in women for clinical characteristics, reperfusion procedures, and in-hospital treatment according to age. In all age groups, diabetes mellitus and hypertension were reported more frequently among women. Dyslipidemia and obesity were also more frequent among women, except in the youngest age groups. Tobacco use was higher among men in all age groups, as was prior infarction, although this difference was only significant above 65 years. With regard to the clinical condition on admission, in all age groups women were more likely to suffer severe heart failure (Killip class III-IV), and this difference was greater in the youngest groups (OR=1.98; 95%CI, 1.29-3.02). With regard to in-hospital pharmacological treatment, younger women had a lower probability of receiving treatments of proven clinical efficacy (statins, OR=0.68; 95%CI, 0.55-0.84; acetylsalicylic acid, OR=0.38; 95%CI, 0.24-0.62; and oral beta blockers, OR=0.76; 95%CI, 0.59-0.97). Likewise, the use of thrombolysis was lower (OR=0.62; 95%CI, 0.48-0.80). This difference remained when patients who were admitted with Killip class IV (OR=0.63; 95%CI, 0.48-0.81) were excluded from the analysis. In addition, in all age groups fewer coronary angiograms were undertaken in women.

After adjusting for clinical characteristics, reperfusion procedures, and pharmacological treatment in the multivariate analysis, the risk of death persisted in the oldest and youngest groups (age<55 years and≥75 years), but only in women with ST-segment elevation myocardial infarction (STEMI). The effect was larger in the younger women (Table 2). After stratifying for Killip classes I-II vs III-IV and considering the interaction, the greater risk of death was only found in women aged <55 years with STEMI and in Killip class I-II (OR=4.27; 95%CI, 2.10-8.87), while a trend was observed in the group aged≥75 years (OR=1.50; 95%CI, 0.97-2.33).

The sensitivity analysis did not show statistically significant differences in the estimated risk of any of the age groups (results not shown).

The biggest strength of this study is that it included all cases of AMI in Chile during the study period, thereby ensuring that the findings of the in-hospital mortality analysis by age and sex were reliable. The GEMI registry was used to search for possible factors that might explain the greater mortality observed among women; however, given the characteristics of the registry and because it was a non-probabilistic sample, it was not possible to make firm inferences of the results for the entire population. This is one of the limitations of the study. In addition, this was a sample of the GEMI registry. Thus, only data from large hospitals were used (those with coronary units and many of them with catheterization laboratories) and mortality was lower. In addition, patients of lower clinical risk with satisfactory progress in smaller hospitals would not be sufficiently represented. This may generate a selection bias that overestimates the higher risk of death among women in this sample. However, in normal clinical practice, most patients are admitted to large hospitals or, after receiving acute therapy, are transferred to smaller hospitals to complete treatment and subsequent coronary angiography.

Another limitation is the lack of continuity in the participation of the hospitals during the study period. This could also introduce a selection bias, even though more than 80% of the sample was obtained from hospitals that contributed to the registry for the 7 years analyzed.