Hospital Inpatient Statistics, 1996

Contents

Foreword
Overview
HCUP Databases
Software Tools
Data

HCUP Nationwide Inpatient Sample
Study sample
Description of the classification systems
Descriptive statistics
Using the tables
Conclusion
References
Tables

Select to access ICD Code Assignments to Clinical Classifications Software (CCS, formerly called CCHPR).


This Research Note provides descriptive statistics for U.S. hospital inpatient stays in 1996 using the Healthcare Cost and Utilization Project's Nationwide Inpatient Sample. National estimates are provided for all discharges by principal diagnosis and by principal procedure. Statistics are presented on the number of discharges, mean length of stay, mean charges, charges in quartiles (25th, 50th, and 75th percentiles), percent who died in the hospital, percent male, and mean age. Diagnoses and procedures are categorized using the Clinical Classification Software (CCS), formerly called Clinical Classifications for Health Policy Research (CCHPR). CCS is a system for collapsing diagnosis and procedure codes into clinically meaningful categories. This update of the CCS incorporates International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) coding changes through September 1998.

CCS categories can be employed in many types of projects analyzing data on diagnoses and procedures, such as identifying populations for disease- or procedure-specific studies; providing statistical information (such as charges and length of stay) about relatively specific conditions; defining comorbidities; and cross-classifying procedures by diagnoses to provide insight into the variety of procedures performed for particular diagnoses.


By Anne Elixhauser, Ph.D., and Claudia A. Steiner, M.D., M.P.H., Agency for Health Care Policy and Research

Foreword

The unprecedented volume, pace, and variation of change in the U.S. health care system requires a new information paradigm in which State, Federal, and private-sector policymakers have timely and direct access to standardized databases and the tools for using them.

Through the Healthcare Cost and Utilization Project (HCUP), a Federal-State-Industry partnership to build a standardized, multi-State health data system, the Agency for Health Care Policy and Research (AHCPR) has taken the lead in developing databases, Web-based products, software tools, and statistical reports and in making them publicly available to policymakers, health system leaders, and researchers.

This report, Hospital Inpatient Statistics, 1996, is based on data from the latest year of the HCUP Nationwide Inpatient Sample (NIS). Hospital Inpatient Statistics, 1996 updates previously published statistical reports that presented data for 1992 and 1995. This report and its predecessors provide answers to the most common questions about hospital care, such as:

The data presented in Hospital Inpatient Statistics, 1996, are organized using the Clinical Classifications Software (CCS), formerly called the Clinical Classifications for Health Policy Research (CCHPR). CCS collapses about 12,000 diagnosis codes and 3,500 procedure codes from the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) into a smaller number of clinically meaningful, relatively homogenous clusters. Without the CCS tool, researchers have two relatively unsatisfactory alternatives: They may work directly with the ICD-9-CM coded data, but 12,000 diagnosis codes and 3,500 procedure codes are cumbersome and often too detailed to be useful. Or they may use software that was designed to bundle or aggregate procedures for purposes of payment, such as diagnosis-related groups (DRGs). However, these "reimbursement groupers" mask clinically important details about procedures that were performed. CCS, a "clinical grouper," makes ICD-9-CM data more amenable to clinically focused statistical analyses.

Select NIS data are now available through HCUPnet, an interactive data repository which can accommodate specific queries on number of discharges, length of stay, charges, and inhospital mortality for diagnoses and procedures. NIS database releases can be purchased in CD-ROM form from the National Technical Information Service by calling 800-553-6847, or online at https://www.ntis.gov/ . Select to access more information on NIS and other HCUP databases, tools, and publications.

We invite you to tell us how you are using HCUP data and tools and to share suggestions of how these products might be improved to better meet your needs. Please E-mail us at hcup@ahrq.gov or send a letter to the address below:

Irene Fraser, Ph.D.
Director
Center for Organization and Delivery Studies
Agency for Health Care Policy and Research
2101 East Jefferson Street, Suite 605
Rockville, MD 20852

Return to Contents

Overview

This publication provides descriptive statistics for hospital inpatient stays in 1996 using the Healthcare Cost and Utilization Project (HCUP) Nationwide Inpatient Sample (NIS). The NIS approximates a 20-percent sample of U.S. community hospitals and collects all inpatient stays from these institutions. National estimates are provided for all discharges by principal diagnosis and by principal procedure, categorized using the Clinical Classifications Software (CCS), formerly referred to as the Clinical Classifications for Health Policy Research (CCHPR). CCS is a system for collapsing diagnosis and procedure codes into clinically meaningful categories that aggregate similar conditions or procedures. Previous reports have provided inpatient statistics for 1992 (Elixhauser and McCarthy, 1996) and 1995 (Elixhauser, Steiner, Whittington et al., 1998).

Return to Contents

HCUP Databases

HCUP, which comprises both nationwide and State databases, contains over 100 variables, including:

State Inpatient Databases (SID). Individual data sets from 22 participating States comprise the SID; each data set contains the universe of that State's non-Federal hospital discharge abstracts. The data have been translated into a uniform format to facilitate cross-State comparisons. The SID represent more than half of all U.S. hospital discharges, and State participation is growing. The SID are particularly well suited for policy inquiries unique to a specific State, studies comparing two or more States, and small area variation analyses. Effective summer 1999, some HCUP partner States are making their inpatient databases for recent years available on CD-ROM through an AHCPR-designated central distributor.

Nationwide Inpatient Sample (NIS). The NIS is a stratified probability sample of hospitals drawn from the SID. The NIS is designed to approximate a 20-percent sample of U.S. community hospitals, including roughly 6.5 million discharges from about 900 hospitals. NIS is the largest all-payer inpatient database in the United States, and data are now available from 1988 to 1996. The NIS is useful for developing national estimates, for analyzing national trends, and for research that requires a large sample size (e.g., care patterns for rare conditions such as congenital anomalies, frequency and distribution of uncommon procedures such as organ transplantations, and hospitalization utilization for population subgroups such as children). NIS releases are available in CD-ROM form, and select data from NIS can be accessed interactively through HCUPnet.

Data files for HCUP were constructed under the technical direction of AHCPR by The MEDSTAT Group (formerly SysteMetrics, Inc.), Santa Barbara, CA, and its subcontractor, the National Association of Health Data Organizations.

Return to Contents

Software Tools

AHCPR has developed two software tools from the Healthcare Cost and Utilization Project that can be used on HCUP and other administrative databases.

Quality Indicators (QIs). This set of 33 clinical performance measures can be used with SID, NIS, and other hospital discharge abstract data as a framework for assessing quality in hospitals. The QIs assess three dimensions of care: 1) potentially avoidable adverse hospital outcomes, such as inhospital mortality following common elective procedures or complications that occur in the hospital; 2) potentially inappropriate use of hospital procedures, such as cesarean sections; and 3) potentially avoidable hospital admissions, such as immunization-preventable influenza among the elderly. The QIs can be used by individual hospitals (e.g., to monitor performance over time or to compare their performance with that of other hospitals) and by States and communities (e.g., to track aggregate quality of care in hospitals or measure access to primary care).

Clinical Classifications Software (CCS). The CCS program aggregates about 12,000 diagnosis codes and 3,500 procedure codes from the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM) into a smaller number of clinically oriented, relatively homogenous clusters. ICD-9-CM codes are the standard codes used in all institutionally based records (e.g., hospitals, outpatient surgery centers) and in insurance claims data. CCS can be applied to all ICD-9-CM data from 1980 to date, with simple adjustments needed for data prior to 1993. CCS, a "clinical grouper," makes ICD-9-CM data more amenable to clinically focused statistical analyses than Diagnosis Related Groups (DRGs) or other "reimbursement groupers" that may mask important clinical details. CCS can help users examine ICD-9-CM data from a clinical perspective, develop clinically based profiles of resource use, and study patterns of diagnoses and procedures. CCS was formerly called the Clinical Classifications for Health Policy Research (CCHPR).

The Quality Indicators and the Clinical Classifications Software, with user instructions, can be downloaded from the AHCPR Web site.

Return to Contents

Data

HCUP Nationwide Inpatient Sample

The statistics presented in this report are based on data pertaining to hospital inpatient stays from HCUP, a Federal-State-industry partnership in health care data. The HCUP Nationwide Inpatient Sample contains resource use information included in a typical discharge abstract. NIS Release 5 covers 19 States (Arizona, California, Colorado, Connecticut, Florida, Illinois, Iowa, Kansas, Maryland, Massachusetts, Missouri, New Jersey, New York, Oregon, Pennsylvania, South Carolina, Tennessee, Washington, and Wisconsin) and includes 906 hospitals and over 6.5 million discharges for 1996.

The NIS is designed to be a 20-percent sample of U.S. "community" hospitals, as defined by the American Hospital Association (AHA). The AHA defines community hospitals as "all nonfederal, short-term, general and other specialty hospitals, excluding hospital units of institutions" (American Hospital Association, 1993). The HCUP sample is a stratified probability sample of hospitals in the frame, with sampling probabilities proportional to the number of U.S. community hospitals in each stratum.

The hospital universe is defined using the AHA Annual Survey of Hospitals. This universe of hospitals is divided into strata using five hospital characteristics: ownership/control, bedsize, teaching status, rural/urban location, and geographic region. Hospitals from HCUP participating States (the sampling frame) are selected to represent these strata, and all discharges from sampled hospitals are included in the database. Weights indicate the number of discharges that the sample discharge represents in the universe of discharges from U.S. hospitals for that year in that stratum. The total number of discharges in the universe from that stratum is taken from the AHA Annual Survey of Hospitals.

Because administrative data on inpatient stays were not created for research purposes, there may be problems with the reliability and validity of certain data elements. Green and Wintfield (1993) summarized the literature on coding errors for hospital administrative data and described a decline in error rates during the 1970s and 1980s. Fisher, Whaley, Krushat et al. (1992) reported that the accuracy of principal diagnosis and procedure has improved since 1983, when such information became important for determining reimbursement by Medicare and other payers. Green and Wintfield (1993) reported the results of a reabstraction study using records from the California Office of Statewide Health Planning and Development. Information on age and sex was most reliable (error rates less than 1 percent), and principal diagnosis was inaccurate in 9 percent of records. Whittle, Steinberg, Anderson et al. (1991) reported that estimates of cancer incidence rates based on Medicare claims data were within 6 percent of estimates using the National Cancer Institute's Surveillance, Epidemiology, and End Results (SEER) data. Other problems inherent in hospital inpatient data include missing data, underreporting of socially stigmatized conditions such as alcoholism and drug abuse, and underreporting of minor procedures.

Analyses limited to principal diagnoses and procedures will produce an underestimate of diagnoses that tend to appear in secondary positions such as hypertension, osteoporosis, and Alzheimer's disease (May, Kelly, Mendlein et al., 1991). Diagnostic and minor therapeutic procedures, which usually appear as secondary procedures, will likewise be underrepresented when the focus is on principal procedures. Because the unit of analysis for this study is the inpatient stay rather than the patient, principal diagnoses and procedures are employed to reduce double-counting of stays. Furthermore, the principal diagnosis is of greater interest because it represents the diagnosis which, after evaluation, was the primary reason for admission to the hospital. Similarly, the principal procedure is of interest because it should be the primary therapeutic procedure received by the patient during the stay. Despite these definitions, other diagnoses and procedures are coded into the principal position. For example, diagnostic procedures are often coded into the principal procedure field.

Study sample

The entire sample of NIS discharges for 1996 was used for this study (N = 6,542,069). These discharges were weighted to obtain estimates that are representative of hospital inpatient discharges in the United States. The estimated total number of discharges represented in these analyses is 34,872,474. This estimate compares favorably with the estimate of 34.4 million discharges (including newborns) based on the National Hospital Discharge Survey (Graves and Kozak, 1999).

The unit of analysis is the discharge, or hospital stay, rather than the patient. Because the NIS is limited to inpatient hospital data, conditions treated and procedures performed on an ambulatory basis are not represented here.

Return to Contents

Description of the classification systems

Diagnoses and procedures for hospital inpatient stays are coded using the International Classification of Diseases, 9th Revision, Clinical Modification (ICD-9-CM), Fifth Edition (Public Health Service and Health Care Financing Administration, 1994). The ICD-9-CM consists of over 12,000 diagnosis codes and 3,500 procedure codes. Although it is possible to present descriptive statistics for individual ICD-9-CM codes, it is often helpful to aggregate codes into clinically meaningful categories that group similar conditions or procedures.

The CCS consists of two related classification systems. The "single-level CCS" is a system of mutually exclusive categories for both diagnoses and procedures. These categories were developed to provide a convenient way to report hospital statistics by diagnosis or procedure.

The second system, or "multi-level CCS," expands the mutually exclusive categories into a hierarchical system. The multi-level categories of the expanded system were developed to support the production of aggregate statistics for larger groupings of CCS categories as well as more detailed statistics about subgroups of diagnoses and procedures within these categories.

The development of the CCS categories was described in Elixhauser and McCarthy (1996) and Elixhauser (1996). Both CCS systems are available online. The categories in this report reflect ICD-9-CM coding changes in effect through September 1999.

Return to Contents

Descriptive statistics

For all analyses, only the principal diagnosis and principal procedure were used. The statistics include:

Both mean and median charges are presented because the mean may be strongly influenced by extreme values. Charge information for procedures is not the charge for the procedure itself. Instead, it refers to the total charge for the hospitalization in which this procedure was listed as the principal procedure. All charge data are charges for the hospitalization. Charges do not necessarily reflect costs nor are they synonymous with reimbursements. In the past, missing charge data often presented a problem. In this study, charge data were present for 98 percent of all discharges.

During HCUP processing, length of stay and total charges are submitted to edit checks; values are recorded as "missing" if: (1) the length of stay is over 365 days and is not justified by a long-term care diagnosis, a perinatal diagnosis, discharge to another facility, or the patient's death; (2) charges per day are less than $100 and are not justified by discharge to another facility or by the patient's death; or (3) charges per day are more than $20,000 and are not justified by discharge to another facility or by the patient's death. An additional 355 cases were dropped from this analysis because of anomalous lengths of stay or total charges that passed the edit checks but were apparently miscoded.

Results are not presented for any diagnosis or procedure category for which the unweighted number of discharges is less than 70. Using a generalized variance technique for proportions, it was determined that a sample of at least 70 discharges is required to assure, with 95-percent confidence, that the reported proportions had a relative error of less than 30 percent (i.e., if the reported value is p, the error is <.3p).

Return to Contents

Using the tables

For the most part, the diagnosis and procedure categories listed in the Tables 1 and 2 follow the order determined by the ICD-9-CM system; chapter headings have been maintained and the individual categories are arranged by numeric ICD-9-CM code. To locate a general condition, body system, or procedure, scan the major headings provided in the contents page preceding each table. Then go to that section of the table and browse through the specific categories until the appropriate condition or procedure is found. To view the ICD-9-CM codes assigned to each CCS category, see Elixhauser, Steiner, Whittington et al. (1998) or http://www.hcup-us.ahrq.gov/reports/natstats/his95/clinclas.htm#app.

The information summarized in Tables 1 and 2 can be used to evaluate the costs, length of stay, patient characteristics, and outcomes associated with hospitalizations for particular conditions and procedures. In addition, prior inpatient statistics reports can be used to assess trends over time (Elixhauser and McCarthy, 1996; Elixhauser, Steiner, Whittington et al., 1998). Although numerous issues warranting further study can be identified using these tables, some of the potential uses of this information include the following:

Assessing frequent diagnoses and procedures in U.S. hospitals. For example, using CCS categories, among the most frequent reasons for hospitalization are coronary atherosclerosis and other heart disease (category 7.2.4) with over 1.4 million hospital stays and pneumonia (8.1.1) with over 1.2 million stays. The single most common principal diagnosis is uncomplicated liveborn (15.1) with over 3.8 million discharges. Circumcision (11.3) is the most commonly performed single procedure accounting for nearly 1.1 million principal procedures. Cesarean sections (13.2) account for about 791,000 principal procedures while diagnostic cardiac catheterization and coronary arteriography (7.5) account for 651,000 principal procedures.

Assessing hospital stays with high levels of resource utilization. The most expensive hospitalizations include stays during which the following procedures were listed as principal procedures: other organ transplantation (category 16.1), with mean total charges of about $191,000; tracheostomy (6.1), $147,000; and bone marrow transplantation (8.1), $137,000. Average lengths of stay were longest for the following conditions: paralysis (6.3), 16 days; transposition of great vessels (14.1.1), 17 days; short gestational age, low birth weight, and fetal growth retardation (15.2), 23 days; respiratory distress syndrome (15.4), a condition affecting infants, 22 days; and spinal cord injury (16.3), 16 days. These figures compare with mean charges across all hospitalizations of about $10,700 and a mean length of stay of about 5 days.

Examining those conditions and procedures with high rates of inhospital mortality. For example, conditions with the highest rates of in-hospital death are cardiac arrest and ventricular fibrillation (category 7.2.10), with a mortality rate of 50.0 percent; shock (17.1.5), 49.8 percent; and malignant neoplasm without specification of site (2.13), 29.6 percent. Overall, 2.5 percent of hospitalized patients died in the hospital.

Evaluating the variability in resource use and outcomes among similar conditions or procedures. For example, among the various types of septicemia listed in Table 1, staphylococcal septicemia (category 1.1.2.2) appears to have the highest associated charges (mean = $24,878) compared with mean charges of $16,149 to $22,173 for other types of septicemia.

Comparing the patient characteristics associated with similar conditions or procedures. For example, among patients hospitalized for diabetes mellitus with complications (category 3.3), patients with circulatory manifestations (3.3.5) were, on average, older and more often male (66 years, 58 percent male) than patients with ophthalmic manifestations (3.3.3) (55 years, 45 percent male). Similarly, patients who received an open cholecystectomy (9.16.1) were older (60 years) and more likely to be male (41 percent) than patients who received laparoscopic cholecystecomy (9.16.2) (52 years, 27 percent male).

Examining trends in hospitalization. For example, from 1992 to 1996, hospitalizations for septicemia (category 1.1.2) increased by nearly 35 percent from 311,450 in 1992 to 378,060 in 1995 to 419,175 in 1996. This is compared with slight declines in the total number of hospitalizations (from 35.0 million in 1992 to 34.9 million in 1996). On the other hand, the rates of many procedures continue to decline, often because these procedures are more frequently performed on an outpatient basis. For example, in 1996, 51,849 inguinal and femoral hernia repairs (9.17) were performed, compared with 97,431 in 1992, a decline of nearly 47 percent.

Return to Contents

Conclusion

This Research Note provides descriptive statistics on inpatient stays in U.S. community hospitals for 1996 using the HCUP Nationwide Inpatient Sample. These statistics should prove useful to health policy analysts, health services researchers, and health care administrators in need of information on hospital discharges by principal diagnosis and principal procedure. Additional statistics can be found in the AHCPR Web site's Data & Surveys section.

Return to Contents

References

American Hospital Asociation hospital statistics. 1993-94 edition. Chicago, IL: American Hospital Association; 1993.

Elixhauser A, Steiner CA, Whittington CA, McCarthy E. Clinical classifications for health policy research: Hospital inpatient statistics, 1995. Healthcare Cost and Utilization Project, HCUP-3 Research Note. Rockville, MD: Agency for Health Care Policy and Research; 1998. AHCPR Pub. No. 98-0049.

Elixhauser A, McCarthy E. Clinical classifications for health policy research, version 2: Hospital inpatient statistics. Healthcare Cost and Utilization Project (HCUP-3) Research Note 1. Rockville, MD: Agency for Health Care Policy and Research; 1996. AHCPR Publication No. 96-0017.

Elixhauser A. Clinical classifications for health policy research, version 2: Software and user's guide. Healthcare Cost and Utilization Project (HCUP-3) Research Note 2. Rockville, MD: Agency for Health Care Policy and Research; 1996. AHCPR Publication No. 96-0046.

Fisher ES, Whaley FS, Krushat WM, et al. The accuracy of Medicare's hospital claims data: Progress has been made, but problems remain. American Journal of Public Health 1992; 82: 243-248.

Graves EJ, Kozak LJ. National Hospital Discharge Survey, Annual Summary, 1996. Vital Health Stat 13(140). National Center for Health Statistics; 1999.

Green J, Wintfield N. How accurate are hospital discharge data for evaluating effectiveness of care? Medical Care 1993; 31(8): 719-731.

May DS, Kelly JJ, Mendlein JM, Garbe PL. Surveillance of major causes of hospitalization among the elderly, 1988. Morbidity and Mortality Weekly Reports 1991 40(SS-1): 7-17.

Public Health Service and Health Care Financing Administration. International classification of diseases, 9th revision, clinical modification. Vols. 1, 2, and 3; fifth edition. Washington, DC: Public Health Service; 1994. HHS Publication No. (PHS) 94-1260.

Whittle J, Steinberg E, Anderson G, Herbert R. Accuracy of Medicare claims data for estimation of cancer incidence and resection rates among elderly Americans. Medical Care 1991; 29(12): 1226-1236.

Tables

Tables
  1. 1996 Statistics by Multi-Level CCS Diagnosis—Principal Diagnosis (PDF File, 165 KB)
  2. 1996 Statistics by Multi-Level CCS Procedure—Principal Procedure (PDF File, 103 KB)

Return to Contents

AHCPR Pub. No. 99-0034
Current as of September 1999

Send Questions & Comments to: hcup@ahrq.gov


Internet Citation:

Hospital Inpatient Statistics, 1996. HCUP-3 Research Note. Agency for Health Care Policy and Research, Rockville, MD. http://www.hcup-us.ahrq.gov/reports/natstats/his96/clinclas.htm


Return to Summary, Hospital Inpatient Statistics, 1996
HCUP Index
AHRQ Home Page
Department of Health and Human Services