Advance Data No. 359 + July 8, 2005
that relate to the personal characteristics of the patient. These items include birth date (converted to age), sex, race, ethnicity, marital status, ZIP Code, and expected sources of payment. Administrative items such as admission and discharge dates, admission type and source, and discharge status were also included. The medical information about patients includes up to seven diagnoses, as many as four surgical and nonsurgical operations and procedures, and dates of surgery. Medical data are coded according to the International Classification of Diseases, 9th Revision, Clinical Modification (ICD–9–CM) (33). In 2002, the ICD–9–CM Coordination and Maintenance Committee created ‘‘Chapter 00, Procedures and Interventions Not Elsewhere Classified’’ to handle space limitations in the existing hierarchical structure. This chapter, used for the first time to code the 2003 NHDS data, is included in the miscellaneous diagnostic and therapeutic procedures and new technologies category shown in tables 8–11.
For the manual data collection, an ongoing quality control program is conducted on the coding and entering of data from abstracts to machine-readable form. Approximately 10 percent of the abstracts are independently re-coded by an NHDS coder, with discrepancies resolved by a chief coder. The overall error rate for records manually coded for the 2003 data year was 0.1 percent for medical (ICD–9–CM) coding and keying and 0.1 percent for demographic coding and keying.
Because of the complex multistage design of the NHDS, the survey data must be inflated or weighted to produce national estimates. The estimation procedure produces essentially unbiased national estimates and has three basic components: inflation by reciprocals of the probabilities of sample selection; adjustment for nonresponse; and population weighting ratio adjustments. These three components of the final weight are described in more detail in another report (31).
The standard error is primarily a measure of sampling variability that
occurs by chance because only a sample rather than the entire universe is surveyed. Estimates of the sampling variability for this report were calculated with SUDAAN software, which takes into account the complex sample design. A description of the software and the approach it uses has been published (34). The standard errors of statistics presented in this report are included in each of the tables.
Use of tables
Discharges are reported by first- listed diagnosis, which is the one specified as the principal diagnosis on the face sheet or discharge summary of the medical record or, if a principal diagnosis is not specified, the first one listed on the face sheet or discharge summary of the medical record. It is usually the main cause of the hospitalization. The number of first- listed diagnoses is the same as the number of discharges.
Estimates of procedures include surgical or nonsurgical operations, diagnostic procedures, and special treatments reported on the medical record. Up to four procedures are coded for each discharge. All-listed procedures include all occurrences of the procedure coded regardless of the order on the medical record. Definitions of the terms used in this report have been published (31).
The diagnoses and procedures appear in separate tables of this report, presented by ICD–9–CM chapter. Within these chapters, subcategories of diagnoses or procedures are shown. These specific categories were selected primarily because of the large number of discharges or because they are of special interest. Data for newborn infants, defined as patients admitted to a hospital by birth, are excluded from this report.
Because of low reliability, estimates with a relative standard error of more than 30 percent or those based on a sample of fewer than 30 records are replaced by asterisks (*). The estimates based on 30–59 patient records are preceded by an asterisk to indicate that they also have low reliability.
Estimates are rounded to the nearest thousand. Therefore, figures within tables do not always add to the totals. Rates and average lengths of stay are calculated from unrounded figures and may not precisely agree with rates or average lengths of stay calculated from rounded data.
Rates are computed using estimates of the civilian population of the United States as of July 1, 2003, based on the 2000 census. The data are from unpublished tabulations provided by the U.S. Census Bureau. Civilian population estimates for 2003 are available at http://www.cdc.gov/nchs/about/major/ hdasd/nhds.htm.
This is the third year that the NHDS publications used rates calculated using the 2000 census. The rates for NHDS reports on 1991–2000 data were computed using postcensual estimates of the civilian population based on the 1990 census. Population estimates for the civilian population from the 2000 census were not available until after the 2000 NHDS reports were prepared.
DeFrances CJ, Hall MJ. 2002 National Hospital Discharge Survey. Advance data from vital and health statistics; no 342. Hyattsville, MD: National Center for Health Statistics.
Pokras R, Kozak LJ, McCarthy E, Graves EJ. Trends in hospital utilization: United States, 1965–86. National Center for Health Statistics. Vital Health Stat 13(101). 1989. http://www.cdc.gov/nchs/data/series/ sr_13/sr13_101.pdf.
Gillum BS, Graves EJ, Kozak LJ. Trends in hospital utilization: United States, 1988–92. National Center for Health Statistics. Vital Health Stat 13(124). 1996. http://www.cdc.gov/ nchs/data/series/sr_13/sr13_124.pdf.
Bernstein AB, Hing E, Moss AJ, et al. Health care in America: Trends in utilization. Hyattsville, MD: National Center for Health Statistics. 2003. http://www.cdc.gov/nchs/data/misc/ healthcare.pdf.
U.S. Department of Health and Human Services, National Institutes of Health, National Heart, Lung, and Blood Institute. Chronic obstructive