Acute myocardial infarction is responsible for most of the deaths in developed countries and for a very large number of hospital admissions. Specifically in Spain, each year about 140 000 deaths and 5 million hospital stays are due to acute myocardial infarction, corresponding to health care costs reaching 15% of total expenditure. Therefore, this paper presents an exhaustive analysis of acute myocardial infarction and the related prognosis, such as recurrence and mortality.
MethodsThis observational study was carried out in Spain. Data were obtained using the Hospital Discharge Administrative Database from 2000 through 2007, inclusive. Specifically, 12 096 cases of acute myocardial infarction (8606 women and 3490 men) were reported during this period, with 2395 readmissions for this diagnosis. Readmissions were analyzed for frequency and duration using logistic regression and the Wang survival model. Mortality was analyzed using logistic regression.
ResultsReadmission rates were 50% for patients younger than 45 years and 38% for those older than 75 years (P<.001). Men were readmitted more frequently than women throughout the follow-up period. Variables related to hospital mortality from acute myocardial infarction were the presence of diabetes, previous ischemic heart disease, and cerebrovascular disease.
ConclusionsMid-term hospital readmissions are highly frequent in acute myocardial infarction survivors. Male sex, previous coronary heart disease, and the number of classical cardiovascular risk factors are the major risk predictors of this readmission. Our results highlight the need for improved medical care during acute myocardial infarction admission, integrated into secondary prevention programs.
Keywords
Cardiovascular disease (CVD), especially acute myocardial infarction (AMI), is the leading cause of death in developed countries and the second leading cause of hospital mortality.1, 2 Readmission after AMI is common, ranging from 8% to 20% in the first year.3, 4 Therefore, readmission rates should be a relevant outcome in clinical trials. Cardiovascular risk factors (CVRF) can currently explain most of the cases of AMI.1, 5, 6 However, little is known about the association between these CVRF and readmission or hospital mortality after an AMI. Diabetes and hypertension are the most frequently studied CVRF in relation to AMI readmissions, but few studies on predictors of readmission for AMI have included both.7
In order to reduce readmission rates after AMI, accurate risk estimates are needed. This information would allow clinicians caring for patients to implement secondary prevention, and will also serve as performance measure, which would enable the policy makers to allocate limited healthcare resources.
The aim of this study was to analyze the contribution of CVRF to readmission rates and mortality after an AMI in a general population cohort.
Methods SubjectsThis observational study includes all patients aged 15 years or older, admitted in all emergency services in Aragon (a northern Spanish region, with a population of 1 269 027, representing 2.88% of the Spanish population) between 2000 and 2007 with the principal diagnosis of AMI. Excluded were patients not residing in the region. AMI diagnosis, coded as 410.xx, was based on International Classification of Disease, 9th Clinical Modification (ICD 9-CM). Also included were variables such as sex, age, duration of hospitalization, and hospital mortality. Other diagnostic codes included as CVRF in the database were smoking use (305.1, v1582), diabetes mellitus (250.*), lipid disorder (272.*), obesity (278.*), and hypertension (401.*, 405.*). We also considered other previous CVD: angina pectoris (413.*), previous coronary artery disease (425.4, 425.5, 425.9, 428.0, 428.9, 402.*), and cerebrovascular disease (435.*-438.*).
Information about hospitalization episodes was obtained from the Minimum Basic Data Set (MBDS) on Hospital Discharges. The MBDS is a registry of all admissions, which includes administrative data (age, sex, city, family doctor, type of admission, etc.) and clinical data (principal diagnosis and other secondary diagnoses, diagnostic and therapeutic procedures); all public and private hospitals are required to report these data to a central state database. Diagnoses and procedures in the MBDS are also coded in accordance with the ICD 9-CM.
Statistical AnalysisPatients with or without readmission for recurrent AMI were compared. Continuous variables are presented as average and standard deviation; for categorical variables the frequencies distribution was selected. The differences between groups were tested by the nonparametric Mann Whitney U test due to their non-normal distribution. For qualitative variables the differences were analyzed by the chi-square test.
The AMI mortality rates were analyzed using a logistic regression model (with entry of variables with univariate significance level of 0.2 and stepwise rejection of variables at the 0.05 level of significance). Regression model was validated using the Hosmer-Lemeshow test and receiver operating characteristic curves.
As patients can be readmitted more than once, which can cause correlations between the events tested, classical statistical estimators are biased and inefficient. Statistical methodology must take this problem into account. In the case of independent data, the interoccurrence survival function can be estimated by the generalization of the limit-product estimator (Kaplan-Meier).8 However, if data are correlated, as is the case for readmission rates in a forward follow-up longitudinal data set, other models should be used such as the estimator proposed by Wang and Chang that considers whether interoccurrence times are correlated or not.9
For testing correlation we used a graphic method proposed by González et al., drawing the survival function using first a survival function estimator based on correlation between events and next a survival function estimator based on independence between events. In case of similarity, we assume independence. In our case, both graphs were very different.
A value of P<.05 was considered statistically significant for all analyses. Statistical analyses were conducted using STATA 10.0/SE (Stata Corp, College Station, Texas, United States) and the survival and “survrec” libraries implemented in the R 2.8.1 software package.
ResultsA total of 11 062 patients (7870 men [71.1%] and 3192 women [28.9%]) was discharged with a diagnosis of AMI between 2000 and 2007. Figure 1 summarizes the outcomes of patients throughout the observation period.
Figure 1. Sample Size Description: Flowchart.AMI, acute myocardial infarction.
Table 1 describes baseline characteristics of patients with or without AMI recurrence according to sex.
Table 1. Demographics, Cardiovascular Risk Profile and Characteristics According to Sex in Patients With Acute Myocardial Infarction Diagnosis, Patients With More Than One Discharge and Patients Without Readmission.
All AMI patients | Patients without recurrent AMI | Patients with recurrent AMI | |||||||
Men (n=7870) | Women (n=3192) | P-value | Men (n=6070) | Women (n=2597) | P-value | Men (n=1800) | Women (n=595) | P-value | |
Age, years | 67±13.35 | 75.93±11.14 | <.001 | 66.45±13.56 | 75.88±11.31 | <.001 | 68.84±12.44 | 76.80±10.37 | <.001 |
<45 | 543 (6.31) | 66 (1.89) | 2490 (41.02) | 370 (14.25) | 595 (33.06) | 57 (9.58) | |||
45-65 | 3000 (34.86) | 440 (12.61) | 1616 (26.62) | 604 (23.26) | 516 (28.67) | 145 (24.37) | |||
65-75 | 2419 (28.11) | 916 (26.25) | 192 (3.16) | 88 (3.39) | 55 (23.61) | 21 (3.53) | |||
>75 | 2644 (30.72) | 2068 (59.26) | 1772 (29.19) | 1535 (59.11) | 634 (35.22) | 372 (62.52) | |||
Acute phase mortality | 898 (10.43) | 679 (16.46) | <.001 | 611 (10.07) | 533 (20.52) | <.001 | 286 (15.89) | 156 (24.54) | <.001 |
Diabetes | 1968 (22.87) | 1285 (36.82) | <.001 | 1286 (21.19) | 892 (34.35) | <.001 | 507 (28.17) | 264 (44.37) | <.001 |
Dyslipemia | 3317 (38.54) | 1112 (31.86) | <.001 | 2217 (36.52) | 806 (31.04) | <.001 | 837 (46.50) | 231 (38.82) | .010 |
Hypertension | 3601 (41.84) | 1994 (57.13) | <.001 | 2519 (41.50) | 1480 (56.99) | <.001 | 838 (46.56) | 381 (64.03) | <.001 |
Obesity | 866 (10.06) | 492 (10.14) | <.001 | 614 (10.12) | 355 (13.67) | <.001 | 202 (11.22) | 105 (17.65) | .001 |
Current smoker | 3100 (36.02) | 217 (6.22) | <.001 | 2341 (38.57) | 180 (6.93) | <.001 | 568 (31.56) | 29 (4.87) | <.001 |
No. of CVRF | .012 | .043 | .176 | ||||||
0 | 1490 (18.93) | 624 (19.55) | 1176 (19.37) | 538 (20.72) | 314 (17.44) | 86 (14.45) | |||
1 | 2607 (33.13) | 1097 (34.37) | 2084 (34.33) | 916 (35.27) | 523 (29.06) | 181 (30.42) | |||
2 | 2379 (30.23) | 906 (28.38) | 1808 (29.79) | 721 (27.76) | 571 (31.72) | 185 (31.09) | |||
3 | 1059 (13.46) | 451 (14.13) | 763 (12.57) | 337 (12.98) | 296 (16.44) | 114 (19.16) | |||
4 | 288 (3.66) | 109 (3.41) | 207 (3.41) | 81 (3.12) | 81 (4.50) | 28 (4.71) | |||
5 | 47 (0.60) | 5 (0.16) | 32 (0.53) | 4 (0.15) | 15 (0.83) | 1 (0.17) | |||
Previous ischemic cardiopathy | 866 (10.06) | 602 (17.25) | <.001 | 557 (9.18) | 413 (15.90) | <.001 | 272 (15.11) | 160 (26.89) | <.001 |
Cerebral vascular disease | 275 (3.20) | 158 (4.53) | <.001 | 169 (2.78) | 116 (4.47) | <.001 | 91 (5.06) | 35 (5.88) | .434 |
Angina | 131 (1.52) | 53 (1.52) | .988 | 86 (1.42) | 37 (1.42) | .977 | 41 (2.28) | 16 (2.69) | .132 |
AMI, acute myocardial infarction; CVRF, cardiovascular risk factors: obesity, dyslipemia, diabetes, hypertension, current smoker.
Comparison uses a chi-square test for categorical variables and Mann Whitney U test for continuous variables.
Data are expressed as mean±standard deviation or no. (%).
Patients who died during follow-up were older and had more prevalence of diabetes and previous coronary and cerebrovascular disease, in both men and women (Table 2).
Table 2. Demographics Cardiovascular Risk Profile and Characteristics According to Sex in Patients With Acute Myocardial Infarction Diagnosis Who Died, Compared to Survivors With the Same Diagnosis.
No survivors | Survivors | |||||
Men (n=897) | Women (n=679) | P-value | Men (n=6973) | Women (n=2513) | P-value | |
Age, years | 75.98±9.97 | 80.83±9.20 | <.001 | 65.84±13.29 | 74.76±11.27 | <.001 |
<45 | 97 (10.81) | 34 (5.01) | <.001 | 2988 (42.85) | 393 (15.64) | <.001 |
45-65 | 248 (27.65) | 87 (12.81) | 1884 (27.02) | 662 (26.34) | ||
65-75 | 35 (3.90) | 13 (1.91) | 212 (3.04) | 96 (3.82) | ||
>75 | 517 (57.64) | 545 (80.27) | 1889 (27.09) | 1362 (54.20) | ||
Diabetes | 262 (29.21) | 274 (40.35) | <.001 | 1531 (21.96) | 882 (35.10) | <.001 |
Disliypemia | 190 (21.18) | 148 (21.80) | .768 | 2864 (41.07) | 889 (35.38) | <.001 |
Hypertension | 367 (40.91) | 369 (54.34) | <.001 | 2990 (42.88) | 1492 (59.37) | <.001 |
Obesity | 32 (3.57) | 56 (8.25) | <.001 | 784 (11.24) | 404 (16.08) | <.001 |
Current smoker | 144 (16.05) | 10 (1.47) | <.001 | 2765 (39.65) | 199 (7.92) | <.001 |
No. of CVRF | .036 | .023 | ||||
0 | 291 (32.44) | 171 (25.18) | 1199 (17.19) | 453 (18.03) | ||
1 | 316 (35.23) | 249 (36.67) | 2291 (32.86) | 848 (33.74) | ||
2 | 203 (22.63) | 179 (26.36) | 2176 (31.21) | 727 (28.93) | ||
3 | 76 (8.47) | 70 (10.31) | 983 (14.10) | 381 (15.16) | ||
4 | 10 (1.11) | 10 (1.47) | 278 (3.99) | 99 (3.94) | ||
5 | 1 (0.11) | 0 | 46 (0.66) | 5 (0.20) | ||
Previous coronary disease | 262 (29.21) | 122 (17.97) | <.001 | 1230 (17.64) | 346 (13.77) | <.001 |
Cerebrovascular disease | 75 (8.36) | 40 (5.85) | .062 | 185 (2.65) | 111 (4.42) | <.001 |
Angina | 7 (0.78) | 10 (1.47) | .188 | 120 (1.72) | 43 (1.71) | .974 |
CVRF, cardiovascular risk factors: obesity, dyslipemia, diabetes, hypertension, current smoker.
Comparison uses a chi-square test for categorical variables and Mann Whitney U test for continuous variables.
Data are expressed as mean±standard deviation or no. (%).
Table 3 shows the main clinical and demographic variables related with the acute phase mortality for AMI. Variables directly related to hospital mortality from myocardial infarction were the presence of diabetes and previous ischemic heart disease and cerebrovascular disease. An inverse correlation was fond between acute phase mortality and percutaneous coronary intervention (PCI) during index admission. Women and older patients showed a higher risk of mortality for AMI.
Table 3. Logistic Regression Model. Acute Phase Mortality as Dependent Variable. Variables Related to Demographic Characteristic, Cardiovascular Risk Factor and Intervention Considered as Independent.
Variable | Mortality, OR (95%CI) | P-value |
Age | 1.05 (1.04-1.06) | <.001 |
Sex (female) | 1.29 (1.14-1.46) | <.001 |
Diabetes | 1.32 (1.16-1.49) | <.001 |
Atherosclerosis | 0.61 (0.54-0.70) | <.001 |
Hypertension | 0.83 (0.74-0.93) | .002 |
Obesity | 0.56 (0.44-0.71) | <.001 |
Current smoker | 0.61 (0.50-0.74) | <.001 |
Ischemic heart disease | 1.69 (1.47-1.93) | <.001 |
Cerebrovascular disease | 1.60 (1.26-2.02) | <.001 |
PCI | 0.50 (0.38-0.66) | <.001 |
Coronary angiography | 0.60 (0.48-0.75) | <.001 |
95%CI, 95% confidence interval; PCI, percutaneous coronary intervention; OR, odds ratio.
R2 Negelkerke=0.186. c statistic=0.765, P-value (Hosmer Lemerhow test), P=.812.
Estimated readmission rates (Table 4) were 50% for patients younger than 45 years and 38% for those older than 75 (P<.001). Men were readmitted more frequently than women throughout the follow-up period. Although CVRF did not influence the probability of being admitted for AMI, they clearly influenced median time to readmission (809 days for patients with no CVRF vs 35 days for those having all 5). Probability estimation of readmission in the first year for patients without previous CVD was 85.6% compared to 35.7% of patients with previous CVD.
Table 4. Percentage for Acute Myocardial Infarction Readmission Obtained by Wang Model.
Readmission for AMI, % | Median time to readmission, days | ||
1 year | 3 year | ||
Age | |||
<45 | 49.97 | 75.56 | 378 |
45-65 | 44.19 | 73.90 | 467 |
65-75 | 38.60 | 71.70 | 632 |
>75 | 38 | 72.34 | 583 |
Sex | |||
Male | 9.60 | 21.50 | 2010 |
Female | 7.30 | 12.90 | 2436 |
Diabetes | |||
No | 41.08 | 74.36 | 511 |
Yes | 41.30 | 73.46 | 546 |
Hypertension | |||
No | 41.78 | 74.77 | 502 |
Yes | 40.63 | 73.70 | 556 |
Dyslipemia | |||
No | 39.92 | 73.24 | 543 |
Yes | 41.88 | 76.17 | 524 |
Obesity | |||
No | 41.58 | 75.86 | 511 |
Yes | 41.92 | 75.94 | 506 |
Current smoker | |||
No | 42.31 | 72.33 | 477 |
Yes | 41.52 | 74.53 | 516 |
Number of CVRF present | |||
0 | 34.43 | 67.09 | 809 |
1 | 40.17 | 74.18 | 1437 |
2 | 41.37 | 73 | 1552 |
3 | 37.96 | 68.55 | 845 |
4 | 43 | 78.15 | 237 |
5 | 35.10 | 66.40 | 35 |
Previous CVD | |||
No | 85.56 | 94.53 | 34 |
Yes | 35.71 | 74.07 | 606 |
AMI, acute myocardial infarction; CVD, cardiovascular disease; CVRF, cardiovascular risk factors.
The main result of our study is that long-term readmission for recurrent AMI is common in patients discharged for an AMI and is related to the presence of CVRF. General characteristics and hospital events of our population are similar to other national2, 3, 10, 11 and international4, 6, 12, 13 cohorts, which might make our results representative of current clinical practice in the real-world post-AMI clinical setting and could also help to identify patients at high risk for long-term complications.
One of the major strengths of our study is the large sample size, which allows us to identify risk factors associated with clinical endpoints. Although some of these associations were quite weak, our results are in concordance with previous data that clearly identified age, diabetes, and previous CVD as major determinants of hospital mortality in AMI patients. We have also found that existing hypertension or current smoking is associated with lower mortality rates. These apparently paradoxical associations have been largely described previously.14, 15, 16, 17 The “protective” effect of hypertension on cardiovascular outcomes has been attributed to the especially bad prognosis associated with hypotension in the acute phase of AMI; in the case of current smokers, the reason could be that these patients are usually much younger and apparently healthier than nonsmokers (ie, the “sick quitters” bias). Moreover, patients with hypertension and current smokers usually receive more medical treatment and coronary revascularization.14, 15, 16, 17
Previous Cardiovascular Disease and MortalityAnother relevant result of our study is that previous CVD is the main predictor of mortality. The reduction in hospital mortality of patients with acute coronary syndromes has lead to an increase in the number of patients with chronic coronary heart disease that are prone to suffer new cardiovascular events.4, 11, 18 In fact, patients with established CVD are considered as high-risk in the risk stratification of patients with hypertension,19 chest pain, or acute coronary syndromes.20 Moreover, the presence of atherosclerotic lesions in different vascular territories has an additive effect on cardiovascular events and mortality. The 3-year follow-up of the REACH (REduction of Atherothrombosis for Continued Health) registry showed that all cardiovascular events increase from 25.5% to 40.5% and cardiovascular mortality from 4.7% to 8.8% if more than one vascular territory was affected.21 Similar results have also been reported in Spain in the setting of post-AMI patients.11 Our results emphasize the critical role of previous CVD on long-term prognosis of post-AMI patients.
Readmissions and Quality of CareHospital readmission is considered for many health providers as a measure of health care quality. As CVD is the leading cause of mortality, especially due to AMI1, all clinical events related to its morbidity and mortality should be carefully described. Moreover, efforts for their reduction are likely to be endorsed by clinicians and administrators.
Several studies have analyzed 6-month or 1-year readmission for AMI in hospital survivors after the index AMI. Our study is the first in which the relationship between all main CVRF and mid-term hospital readmission has been correlated. Previous reports have attempted to relate the majority of readmissions to the quality of the care provided during the index admission, the patient education protocol, or the outpatient care after discharge, but findings have been varied greatly. In all clinical conditions studied by Thomas,22 readmission rates of patients who received poor-quality care and patients whose care was acceptable were very similar. Roe et al. assessed risk-adjusted outcomes after different diagnoses, such as stroke, AMI, and heart failure, and concluded that the length of hospitalization, the mortality rate, and unplanned readmissions were predicted mainly by factors such as age, severity of the disease, and comorbidity.23 Finally, CVD is known to significantly alter quality of life, mainly when patients develop symptoms of angina or heart failure1; the long-term follow-up of our study also highlights the high rate of hospital readmissions and recurrent AMI in these patients that may, eventually, contribute to worsen both the quality of life and the prognosis.
Comprehensive discharge planning plus outpatient support may reduce readmission rates and improve health outcomes for patients with AMI. Current management of postdischarge AMI patients that include ambulatory care and multidisciplinary teams are highly recommended due to the beneficial effects; the efficacy of programs incorporating discharge planning, transitional care, and postdischarge management for these patients has been clearly established.3, 24, 25
It is important to emphasize the effect of PCI in reducing recurrent readmissions and mortality rates. Thrombolysis has revolutionized emergency treatment of AMI,20, 26, 27 but primary PCI is currently considered the best reperfusion treatment and is successful in over 90% of cases.20, 26, 27, 28, 29, 30, 31, 32 Nevertheless, thrombolysis is still performed in many countries because PCI requires a complex organization of health resources which many countries cannot afford.27 The GRACE registry has confirmed that trial results obtained in carefully selected patients treated by experienced cardiologists in well-equipped centers may not be so easily obtained in the average patient in day-to-day practice of busy hospitals.21 More recently, a randomized trial demonstrated that complete revascularization of AMI patients before hospital discharge reduces mortality and readmissions during the following 3 years.30 Nevertheless, PCI-treated patients still have a high risk of future cardiovascular events if new coronary lesions develop or previous nonsignificant plaques become unstable.32 In the study by Lemesle et al., the need for a new PCI on a nonculprit coronary lesion was as high as 50% in the first year after an acute coronary syndrome; diabetes, previous coronary bypass, and peripheral artery disease were identified as major risk factors for this event.12 Our results agree with those of Lemesle et al. and provide longer follow-up with similar trends in the need for future revascularization.
LimitationsThe use of an administrative database for data collection has both advantages and disadvantages. First, all patients diagnosed with AMI in the emergency room in public hospitals were recruited in this study. Nevertheless, the quality of diagnostic characterization can be biased due to the suspicion of a certain pathology that can or cannot be confirmed afterwards. This possibility is low due to the specificity of the clinical data in AMI, but should not be overlooked.
ConclusionsMid-term hospital readmissions are highly frequent in AMI survivors. Male sex, previous coronary heart disease, and number of classic CVRF are the major risk predictors of this event. In our wide sample of current clinical practice, hospital readmission is effectively reduced by PCI. Our results highlight the need for improved medical care during AMI admission, integrated into secondary prevention programs.
FundingThis project has been supported by the Carlos III Institute of Health of Spain, (RETICS D06/0018; PS09/01936; PI09/1378).
Conflicts of interestNone declared.
Received 31 March 2011
Accepted 3 September 2011
Corresponding author: Unidad de Investigación-Epidemiología Clínica, Hospital Universitario 12 de Octubre, Instituto de Investigación 12 de Octubre (i+12), CIBER de Epidemiología y Salud Pública, Avda. de Córdoba s/n, 28041 Madrid, Spain. eandres@h12o.es