The National Inpatient Sample (NIS) from the Healthcare Cost and Utilization Project (HCUP) is a longitudinal hospital inpatient database containing all discharge data from over 1,000 hospitals located in 41 states, approximating a 20% stratified sample of U.S. community hospitals. The NIS captures both incident and recurrent AMI hospitalizations. Data include discharge-level data files with both trend weights and data elements consistently defined across data years (http://www.hcup-us.ahrq.gov/db/nation/nis/nistrends.jsp
). Using NIS data in HCUP from 2001–2007, we selected the following fields for analysis: age, gender, ethnicity, discharge status, length of stay, in-hospital death, principal diagnosis codes in order to identify AMI hospitalization, state of hospitalization, and the universal discharge weights which can be used to estimate the total number of events or admissions of the hospital. A hospitalization was classified as an AMI hospitalization if the primary discharge diagnosis was 410.xx, excluding 410.x2, based on International Classification of Diseases
, 9th revision, clinical modification. We included only primary discharge diagnoses of AMI since non-primary diagnoses may not reflect an acute process or does not reflect the reason for admission.
2000 Census data with annual, intercensal survey adjustments were used to calculate each subgroup population of subjects at risk for AMI in order to calculate population-based overall and AMI rates. (http://www.census.gov/popest/states/
). Corresponding numbers of people are calculated for the subgroups stratified by state, age groups, gender, and ethnicity in separate years from these data.
All discharges from hospitals in the HCUP-NIS from 2001–2007 were initially included (n=55,402,296). We then excluded discharges with missing data on patient age, gender, length of stay, and in-hospital death (n=194,331, 0.4%); discharges in which patients’ age is < 18 years old (n=9,735,028, 17.3%); discharges in which the patient discharged alive and on the same day of the admission because they were unlikely to be with acute disease (n=832,472, 1.8%); and discharges in which the patient was transferred in from another hospital (n=155,008, 3.5%), leaving a cohort of 43,272,788 discharges. To examine the trend stratifying by ethnicity in white and black patients, we limited to the 21 states that reported ethnicity data during each year in the study and excluded discharges in which the patients’ ethnicity was neither white nor black, which resulted in a study cohort of 22,713,429 discharges.
The AMI hospitalization rate was expressed as the number of AMI hospitalizations divided by the number of corresponding Census based persons within a given group. First we examined the distribution of AMI patients’ characteristics related to age (<45, 45 to <55, 55 to <65, 65 to <75, and ≥75 years), gender (male and female), and ethnicity (white and black). Then we examined the AMI hospitalization rates stratified by subgroups of age, gender, and ethnicity. We also calculated overall hospitalization rates stratified by subgroups of age, gender and ethnicity to provide a comparison for trends observed in AMI hospitalizations. Finally, we calculated the age-adjusted AMI hospitalization rates in the subgroups determined by a combination of gender and ethnicity. Linear trend of adjusted rates over different year was calculated using Poisson regression analyses. All the rates were calculated per 100,000 persons based on population information from the Census summary data. All statistical analyses of hospitalization rates were conducted with SAS version 9.2 (SAS Institute Inc, Cary, NC).