HEALTHCARE COST & UTILIZATION PROJECT

User Support

Do Your own analysis
Explore Expert Research & Limited Datasets

2001 HCUP Nationwide Inpatient Sample (NIS) Comparison Report



HCUP Methods Series 2001 Nationwide Inpatient Sample (NIS) Comparison Report

Report #2004-04

Contact Information:


Healthcare Cost and Utilization Project (HCUP)
Agency for Healthcare Research and Quality
http://www.hcup-us.ahrq.gov


For Technical Assistance with HCUP Products:

Email: hcup@ahrq.gov

or

Phone: 1-866-290-HCUP


Recommended Citation: Whalen D, Houchens R, Elixhauser A, Final 2001 Comparison Report. 2004. HCUP Method Series Report # 2004-04. ONLINE October 22, 2004. U.S. Agency for Healthcare Research and Quality.
Available: http://www.hcup-us.ahrq.gov/reports/methods/NIS_Comparison_Report_2001.pdf



TABLE OF CONTENTS

INDEX OF TABLES

INDEX OF FIGURES 


EXECUTIVE SUMMARY

This report compares statistics calculated from the 2001 Nationwide Inpatient Sample (NIS) with estimates from two comparable databases – the National Hospital Discharge Survey (NHDS) and the Medicare Provider Analysis and Review (MedPAR) – with the objective of assessing potential biases. The report focuses on important inpatient outcomes. Outcomes examined include: total discharges, length of stay, in-hospital mortality rates, and total hospital charges. In addition to national statistics, these data were also evaluated across several categories, including procedure and diagnosis groupings, expected payer, patient demographics, region, and hospital characteristics.

NIS Background

The 2001 NIS was established as part of the Healthcare Cost and Utilization Project (HCUP) to provide data supporting analyses of hospital utilization across the United States. NIS data were selected using a stratified probability sample of hospitals, drawn from a frame of 33 states. Sampling probabilities were calculated to select 20 percent of the universe in each stratum defined by hospital characteristics (region, urban/rural location, number of beds, teaching status, and ownership/control). As a result, the NIS includes approximately 7.5 million discharges from 984 hospitals, with weights to facilitate national estimates. It is important to note that NIS data differed in scope from the two comparison databases in three ways:

NHDS Background

In 2001, the National Center for Health Statistics drew a sample of more than 330,000 short-stay discharges from 448 hospitals, including both general-specialty and children’s hospitals for the NHDS data set. Statistics from the NHDS are considered geographically representative because the NHDS sampling frame was relatively unrestricted.

MedPAR Background

Obtained from the Centers for Medicare and Medicaid Services (CMS, formerly the Health Care Financing Administration), MedPAR data include all paid fee-for-service Medicare discharges from Medicare-certified, short-stay U.S. hospitals. For calendar year 2001, a total of 12.0 million discharges from U.S. community hospitals were included. Of special importance is the fact that MedPAR data underreported total Medicare discharges by omitting most managed care discharges (approximately 15% of Medicare patients). This particular omission has significant implications for the various comparisons between the MedPAR and NIS data files.

Return to Contents

Methods

Outcome statistics compared in the NIS, NHDS, and MedPAR databases included:

These measures of utilization and outcomes were selected because they are common in health services research and serve important roles in health policy and resource planning analyses.

Both the NIS and NHDS are samples, and statistics derived from them are estimates. Comparisons between NIS and NHDS estimates utilized two-sample z-tests. MedPAR data, in contrast, are not a sample. The NIS-MedPAR comparisons employed one-sample z-tests, which are useful in comparing an entire population (MedPAR) with sample estimates (NIS).

The report cautions that estimates cannot be expected to be identical when two different samples are taken. When viewing results, readers should note that statistically significant differences between the NIS and the NHDS can be expected for a number of reasons; these include:

Considering all of these possible reasons for encountering significant differences among the samples, data analyses revealed remarkable similarity among the estimates.

Major Findings

NIS estimates of essential healthcare policy variables (i.e., in-hospital mortality, inpatient population size, length of stay, and costs) are accurate and precise. The estimates were drawn from states that encompass 76 percent of all short-stay hospitals and more than 81 percent of U.S. discharges. The large NIS sample allows for the study of relatively uncommon disorders, procedures, and hospital types; in fact, NIS estimates can be calculated for any number of special sub-populations. In addition, the NIS contains hospital charges and all payers.

A summary of overall and regional comparisons:

Comparisons by hospital characteristics:

Comparisons by patient characteristics:

Comparisons by diagnosis and procedure categories:

Return to Contents

Conclusion

Each data source possesses distinct strengths and weaknesses and may be regarded as the optimum choice for answering different research questions. In general, NIS estimates of essential healthcare policy variables are accurate and precise. The NIS offers a large sample that might allow for the study of disorders, procedures, and hospital types that occur with low frequency in other databases. NIS estimates can be calculated for thousands of special sub-populations that may be of interest to researchers. The NHDS sample and MedPAR data were drawn from all 50 states, while only 33 states were included in the NIS database. However, for 2001, NIS states encompassed 76 percent of all short-stay hospitals and more than 81 percent of all U.S. discharges. The NIS contains charges for each hospital stay, all payers, and a large sample of discharges. In contrast, the NHDS has a smaller number of discharges, does not contain charges, but does sample from all 50 states. The MedPAR database is limited to Medicare discharges and contains all Medicare patients covered by the fee-for-service program, but excludes Medicare patients enrolled in managed care plans.

INTRODUCTION

This report compares statistics estimated from the Nationwide Inpatient Sample (NIS), a database containing patient-level information from a sample of hospital discharges in the year 2001, with estimates from two other data sources. These comparisons will interest researchers who intend to make inferences about hospital outcomes using the 2001 NIS. This report is the seventh in a series; the six previous reports compared the NIS with other data sources for the years 1991, 1993, 1995, 1997, 1999, and 2000, respectively. These data years correspond to NIS releases that expanded the number of states contributing data – the first release sampled discharges from only eight states, while this latest release sampled discharges from the 33 states shown in Figure 1:

Figure 1. States Participating in the NIS, 2001 (text version)

Figure 1. States Participating in the NIS, 2001

Although NIS coverage of U.S. discharges is impressive (these states include more than 81 percent of all discharges from community hospitals nationwide during 2001), the possibility remains that hospital outcomes from these states may differ from hospital outcomes in the states not covered by the NIS.

Created as a part of the Healthcare Cost and Utilization Project (HCUP) and funded by the Agency for Healthcare Research and Quality (AHRQ), the NIS contains all discharges from a sample of community short-stay hospitals stratified by geographic region, urban vs. rural characteristics, teaching status, bed size, and type of ownership. The hospital sample was drawn from the participating states indicated in Figure 1. The final sample contained 7.4 million discharges from 986 hospitals. We compared outcomes from this sample with outcomes from two other hospital discharge databases: 1) the 2001 National Hospital Discharge Survey (NHDS), and 2) the 2001 Medicare Provider Analysis and Review (MedPAR) file.

The 2001 NHDS was created under the auspices of the National Center for Health Statistics (NCHS). Compared with the 2001 NIS, the 2001 NHDS featured a much smaller sample containing only 330,210 discharges from 438 hospitals. However, the sample was drawn from a frame that included nearly all hospitals in each of the 50 states. The NHDS sampled non-federal short-stay hospitals in the United States, and then sampled discharges from each of the sampled hospitals. Although the smaller sample size rendered the NHDS estimates less precise than the NIS estimates, the complete coverage of states and the NHDS sampling design minimized the potential bias for national estimates of hospital outcomes. This characteristic is the reason it was used as a comparative database in this study.

The 2001 MedPAR, obtained from the Centers for Medicare & Medicaid Services (CMS), included about 11.3 million fee-for-service Medicare discharges from more than 5,000 Medicare-certified, short-stay United States hospitals. It was not a sample of Medicare discharges. The MedPAR was nearly ideal for comparing NIS estimates of Medicare inpatient outcomes because it represented close to the entire population of Medicare discharges. As a comparative database, its main weakness was that it excluded Medicare managed care enrollees; these individuals accounted for 15.4 percent of the Medicare inpatient experience in 2001.

We compared the estimates from the 2001 NIS with estimates from the 2001 NHDS and the 2001 MedPAR on the following inpatient outcomes:

While many other statistics can be estimated from these data, hospital research commonly focuses on these outcomes. To the extent that the NIS generates reasonable estimates for these outcomes, it is likely that estimates for other, similar outcomes will also be reasonable.

Estimates from the three data sources were compared at the national level, as well as within hospital groups and patient categories. We grouped hospitals and evaluated estimates by geographic region, bed size, ownership, urban vs. rural location, and teaching status. We also categorized patients and compared estimates within age group, gender, race, primary payer, diagnosis category, and procedure category.

In addition, we compared weighted and unweighted frequencies between the 2001 NIS sample and the 2001 Hospital Survey of the American Hospital Association (AHA). These comparisons are purely descriptive because the NIS sample weights were derived from the AHA survey. Consequently, there was close agreement between the two sources by construction.

This report is divided into four sections. The first section describes the NIS and recent changes in the sampling strategy. The second section provides a discussion of the NHDS, the MedPAR file, and the methodology used in the analysis. The third section presents the results, and the final section includes a discussion and posits some conclusions.

Return to Contents

BACKGROUND INFORMATION ON HCUP AND THE NIS

HCUP is a Federal-State-Industry partnership formed to build a standardized, multi-state health data system. In September 2000, AHRQ provided funding to the HCUP project for Medstat to continue developing and expanding this health data system through data year 2003. The 2001 NIS was established as part of HCUP to provide analyses of hospital utilization across the United States.

The 2001 NIS universe included all acute-care discharges from all community hospitals in the United States. It comprised all discharges from a sample of hospitals in this target universe. However, the NIS sampling frame was constructed from the subset of universe hospitals that released discharge data for research use. For the 2001 NIS, AHRQ had agreements with 33 Partner organizations that maintain statewide, all-payer discharge data files. The 2001 NIS contains data from each of these states; this participation reflects an increase of five more states than the previous release and 25 more states than the first release.

Table 1 indicates how the NIS sampling frame has grown. It lists the states included in each NIS release, for data years 1988 through 2001.

Table 1. States in the Frame for NIS Releases
Years States in the Frame
1988 California, Colorado, Florida, Iowa, Illinois, Massachusetts, New Jersey, and Washington
1989-1992 Add Arizona, Pennsylvania, and Wisconsin
1993 Add Connecticut, Kansas, Maryland, New York, Oregon, and South Carolina
1994 No new additions
1995 Add Missouri and Tennessee
1996 No new additions
1997 Add Georgia, Hawaii, and Utah
1998 No new additions
1999 Add Maine and Virginia
2000 Add Kentucky, North Carolina, Texas, and West Virginia
2001 Add Michigan, Minnesota, Nebraska, Rhode Island, and Vermont

As with previous releases of the NIS, the 2001 NIS sampling frame was subject to further restrictions.

NIS Design

The NIS is a stratified probability sample of hospitals in the frame, with sampling probabilities calculated to select 20 percent of the universe contained in each stratum. Beginning in 1998, NIS databases differed from previous years of the NIS because of a sampling redesign. Therefore, longitudinal comparisons of the NIS might indicate differences that can be attributed to the following changes in the sampling design:

  1. Prior to 1998, the NIS design ensured that hospitals drawn for the sample in one year had a high probability of being drawn for the sample in the following year. Including the same hospitals across years improved the precision of trend analyses, although it may have introduced some form of bias into one or more years of the hospital sample. Medstat and AHRQ decided to discontinue any sampling scheme that increased the chance that hospitals would be included in successive years of the NIS.


  2. We found that patients treated in rehabilitation hospitals tend to have lower mortality rates and longer lengths of stay than patients in other community hospitals. In addition, the completeness of reporting for rehabilitation hospitals is uneven across the states. Therefore, we decided to eliminate rehabilitation hospitals from the NIS (and from the target universe).


  3. In previous NIS designs, we employed strata for geographic region, hospital ownership, urban/rural location, and teaching status. We identified strata that could be nested or collapsed, in order to avoid small cells in the final sample. This process reduced the number of NIS strata from 108 to 60, beginning with the 1998 NIS.


  4. In the previous NIS, bed size categories were defined only within location/teaching status. However, even within these location/teaching categories, the bed size distributions still varied widely by geographic region. We decided to define small, medium, and large bed size categories nested within region and location/teaching category such that approximately one-third of the hospitals would be allocated to each category.


  5. Prior to 1998, we stratified all hospitals into one of three ownership categories: public, voluntary, and proprietary. In several geographic regions, however, some ownership categories rarely occurred. Therefore, we used all three ownership categories for rural hospitals in the South and for urban non-teaching hospitals in the South and West. However, in the West and Midwest regions, we collapsed the proprietary and voluntary hospitals into a new "private" ownership category.


  6. Finally, we redefined teaching hospitals. In prior versions of the NIS, a hospital was designated a teaching hospital only if it had some interns or residents, and it was either a member of the Council of Teaching Hospitals or had an AMA-approved residency program. The new definition still defines those same hospitals as teaching hospitals. However, it also includes all hospitals with a ratio of interns and residents to beds of 0.25 or higher. This intern-to-bed ratio is similar to a component of the Centers for Medicare & Medicaid Services’ (CMS, formerly the Health Care Financing Administration) definition of teaching hospitals for Medicare payments.

NIS Sampling

The overall sampling objective was to select a sample of hospitals that could be generalized to the target universe, including hospitals outside the frame (which had a zero probability of selection). To improve the generalizability of the NIS estimates, five hospital sampling strata were used:

  1. Geographic Region – Midwest, Northeast, West, and South.


  2. Ownership – public, private non-profit, and proprietary (private or investor-owned).


  3. Location – urban and rural.


  4. Teaching Status – teaching and non-teaching. (Rural hospitals were not split according to teaching status, because rural teaching hospitals were rare.)


  5. Bed Size – small, medium, and large. Bed size categories were based on hospital beds, and were specific to the hospital's location and teaching status, as shown in Table 2. Bed size cut points were chosen so that approximately one-third of the hospitals in a given region/location/teaching combination would appear in each bed size category. This approach creates different divisions – small, medium, and large – for rural, urban non-teaching, and urban teaching hospitals. For example, a medium-sized urban, teaching hospital would be considered a rather large rural hospital. Further, the size distribution was different among regions for each of the urban/teaching categories. Using differing cut points in this manner avoids strata containing small numbers of hospitals.
Table 2. Bed Size Categories
Location and Teaching Status Hospital Bed Size
Small Medium Large
Northeast
Rural 1-49 50-99 100+
Urban, non-teaching 1-124 125-199 200+
Urban, teaching 1-249 250-424 425+
Midwest
Rural 1-29 30-49 50+
Urban, non-teaching 1-74 75-174 175+
Urban, teaching 1-249 250-374 375+
South
Rural 1-39 40-74 75+
Urban, non-teaching 1-99 100-199 200+
Urban, teaching 1-249 250-449 450+
West
Rural 1-24 25-44 45+
Urban, non-teaching 1-99 100-174 175+
Urban, teaching 1-199 200-324 325+

To further improve proportional geographic representation, hospitals were sorted by state and by the first three digits of their ZIP Code prior to systematic sampling. Refer to Design Report: HCUP Nationwide Inpatient Sample 2001 for more details on the sampling design.

Return to Contents

NIS Weights

Sample weights were developed for the NIS to obtain national estimates of the hospital and inpatient parameters. For example, weights make estimates of diagnosis-specific average lengths of stay over all U.S. hospitals possible. Within each stratum, the discharge weight was set at the ratio of discharges in the universe (estimated from the 2001 AHA hospital survey) to discharges in the sample.

METHODS

NIS statistics were compared with those calculated from two other sources, each of which is described below.

National Hospital Discharge Survey (NHDS)

Conducted by the National Center for Health Statistics (NCHS), the 2001 NHDS included 330,210 discharges from 448 hospitals. The NHDS covered discharges from U.S. hospitals categorized as short-stay (hospitals with an average length of stay under 30 days), including both general-specialty (medical or surgical) and children’s hospitals. Federal, military, and Veteran’s Affairs hospitals were excluded from the survey.

The NHDS sample included with certainty the largest hospitals: those with a minimum of 1,000 beds, or at least 40,000 discharges. The remaining sample of hospitals was based on a stratified, three-stage design:

Medical Coding and Edits. The medical information that was recoded manually on the sample patient abstracts was coded centrally by NCHS staff. Up to seven diagnostic codes were assigned for each sample abstract. In addition, if the medical information included surgical or non-surgical procedures, up to four codes for these procedures were assigned. As with the NIS, the system currently used for coding the diagnoses and procedures on the medical abstract forms, as well as on the commercial abstracting services data files, is the International Classification of Diseases, 9th Revision, Clinical Modification, or ICD-9-CM.

NHDS usually presents diagnoses and procedures in the order they were listed on the abstract form or obtained from abstract services. However, there were exceptions. For women discharged after a delivery, a code of V27 from the supplemental classification was entered as the first-listed code, with a code designating either normal or abnormal delivery in the second-listed position. In another exception, a decision was made to reorder some acute myocardial infarction diagnoses. If an acute myocardial infarction was listed with other circulatory diagnoses and was other than the first entry, it was reordered to the first position. If a symptom appears as a first-listed code and a diagnosis appears as a secondary code, the diagnosis replaced the symptom, which was moved to appear after the diagnosis.

Table 3. Comparison of 2001 NIS and NHDS Data Files
Characteristics 2001 NIS 2001 NHDS
Number of hospitals 986 448
Number of discharges 7,452,727 330,210
Intended universe Discharges from community hospitals, as defined by AHA – non-federal, short-term general, or other specialty hospitals that were not a hospital unit of an institution. Short-term rehabilitation hospitals were excluded. Discharges from non-institutional hospitals (excludes federal, military, and VA hospitals) located in the 50 states and the District of Columbia. Only short-stay hospitals (ALOS < 30 days) or those whose specialty is general (medical or surgical) or children’s general hospitals are included in the survey.
Bed size No restriction was placed on bed size in creating the file, but no hospitals in the sample have fewer than six beds. Must have at least six beds staffed for patient use to be included.
Sample or universe Sample Sample
Sampling frame 33 states 50 states and the District of Columbia
Sample design – hospitals By geographic region, control/ownership, location, teaching status, and bed size. Includes all hospitals with > 1,000 beds or > 40,000 discharges annually, plus an additional sample of hospitals in two stages. A sample of 112 PSUs was selected. These PSUs were a probability sample of the counties or metropolitan areas used in the 1985-1994 National Health Interview Survey. A sample of hospitals was selected within these PSUs.
Sample design – discharges All discharges from sampled hospitals were included. A systematic random sample of discharges was selected from each hospital.
Reassignment of diagnosis codes None For women discharged after delivery, a code of V27 was entered as the first-listed code.

If a symptom appeared as a first-listed code and a diagnosis was listed as a secondary code, the diagnosis replaced the symptom.

If acute myocardial infarction was listed with other circulatory conditions, it was reordered to the first entry.

Table 3 summarizes some of the key differences in hospitals and discharges represented by the NIS and NHDS data files. Sampling error exists in both the NHDS and the NIS. However, the NIS includes nearly 25 times the number of NHDS discharges and more than twice the number of hospitals than the NHDS. Further, the NIS contains all discharges from sampled hospitals, whereas the NHDS contains a sample of discharges from sampled hospitals. As a result of these sampling differences, statistics calculated from the NIS usually have much smaller standard errors than those calculated from the NHDS. In addition, it was not always possible to calculate valid estimates of standard errors from the NHDS for statistics calculated from rare subpopulations. For example, mortality estimates for low frequency procedures and diagnoses might be based on fewer than a dozen cases in the NHDS, while the same subpopulations could contain hundreds of discharges in the NIS. Statistics from the NHDS were assumed to be representative geographically, because the sampling frame was relatively unrestricted, encompassing all federal, acute-care general U.S. hospitals with six or more beds. In contrast, the NIS sampling frame for 2001 was limited to the 33 states that made their data available for research purposes.

Medicare Provider Analysis and Review (MedPAR)

The MedPAR data obtained from the Centers for Medicare & Medicaid Services (CMS) include all records for each fee-for-service Medicare discharge from a Medicare-certified, short-stay U.S. hospital. Federal fiscal year records for 2001 and 2002 were used to create a calendar year 2001 MedPAR file with nearly 11.5 million discharge records. To ensure that the hospital composition of the MedPAR file was consistent with the NIS universe, only AHA-defined community hospitals – as defined by the American Hospital Association (AHA) – were kept in the MedPAR-derived file for this study. In the MedPAR data, same-day stays (admission and discharge on the same day) were assigned a length of stay of one day. Consequently, in comparisons of average lengths of stay between the NIS and MedPAR data, same-day stays in the NIS were recoded from zero to one for this analysis.

Table 4 summarizes some of the key differences in hospitals and discharges represented by the NIS and MedPAR data files. Medicare discharge statistics from MedPAR have no sampling error associated with them because this file represents a census of 2001 fee-for-service Medicare discharges. However, analyses suggest that the MedPAR data underreport total Medicare discharges by omitting most discharges for managed care. In 2001, 15.4 percent of Medicare enrollees were in managed care, including HMOs (HCFA, 2001). However, only 0.9 percent of calendar year 2001 MedPAR discharges were identified as managed care enrollees, suggesting that more than 14 percent of the Medicare population may have been excluded (15.4 percent in the population - 0.9 percent in the MEDPAR file = 14.5 percent). As will be discussed throughout the report, this omission has significant implications for the various uses of the MedPAR and NIS data files.

Table 4. Comparison of 2001 NIS Medicare Discharges and MedPAR Data Files
Characteristic 2001 NIS (Medicare Only) MedPAR
Number of hospitals 977 (with Medicare discharges) 6,2131
Number of discharges 2,749,788 12,035,6812
Intended universe Discharges from community hospitals, except rehabilitation hospitals, as defined by AHA – non-federal, short-term general, or other special hospitals that were not a hospital unit of an institution. All Medicare discharges. Only discharges from non-rehabilitation, community hospitals were included, for comparison purposes.
Bed size No restriction was placed on bed size in creating the file, but no hospitals in the sample have fewer than six beds. No restriction was placed on bed size in creating the file, but no hospitals in the sample have fewer than six beds.
Sample or universe Sample Universe
Sampling frame 33 states 50 states and the District of Columbia
Sample design – hospitals By geographic region, control/ownership, location, teaching status, and bed size. All hospitals included.
Sample design – discharges All discharges from sampled hospitals were included. All fee-for-service discharges were included.
Reassignment of diagnosis codes None None

1Short-term general and specialty community hospitals.
2Discharges from short-term general and specialty community hospitals.

Variables Compared

The following measures were chosen to compare the NIS to the NHDS and MedPAR databases:

These measures of utilization and outcomes were selected because they are common in health services research and important for health policy and resource planning analyses.

The NIS-MedPAR comparison included total hospital charges in addition to the three variables noted previously. When comparing NIS records to MedPAR, only the NIS discharges for which Medicare was the expected primary or secondary payer were used.

Return to Contents

Statistical Testing

Estimates derived from both the NIS and NHDS were based on weighted discharge records from stratified samples. The SAS software PROC SURVEYMEANS was used to compute standard errors for the NIS (see the NIS Variance Report for details). The stratifier variable included in the NIS (NIS_STRATUM) was specified as the stratum, and the unique hospital identifier (HOSP_ID) was specified as the cluster variable. A description of the method used for calculating standard errors for the NHDS is provided in Appendix D.

NIS-AHA Comparisons

Tables comparing characteristics from AHA universe hospitals and NIS hospitals (Table 7 - Table 8) appear in Appendix A. All numbers in these tables come from the AHA Annual Survey; no significance tests were performed for these tables.

Significance tests were conducted for the discharge comparisons of AHA counts and NIS estimates (Table 9 - Table 11). The AHA data are a population, based on the annual AHA survey, so a one-sample z-statistic was computed for these comparisons. AHA discharges represent the survey counts adjusted for the number of well newborns. An estimate of the average length of stay (ALOS) was obtained from the AHA by dividing the total number of days by the total number of discharges reported in the 2001 AHA survey of hospitals.

Same-day discharges from the NIS are recorded with length of stay equal to zero. However, for comparisons with AHA statistics, the length of stay measures for NIS same-day discharges was changed to one day. The standard error for the NIS estimates used in these calculations was generated by the SURVEYMEANS procedure.

NIS-NHDS Comparisons

For each NIS-NHDS comparison, a test was performed to determine whether the NIS and NHDS estimates differed significantly. Because the NIS and NHDS estimates were both based on samples, two-sample z-tests were used where valid estimates of the NHDS standard error could be made. Because of the limited sample size, valid estimates were not available for all breakdowns of the NHDS data. Please see Appendix D for a description of comparison tests and an explanation of restrictions on calculating NHDS sample errors. Differences were reported at the .01 and .05 significance levels.

Tables comparing NIS and NHDS statistics (Table 12 -Table 16) appear in Appendix B.

NIS-MedPAR Comparisons

Because the MedPAR data represent the population, and not a sample, a one-sample z-statistic was computed for these comparisons. The standard error for the NIS estimate used in these calculations was generated by the SURVEYMEANS procedure for the subset of NIS discharges with Medicare identified as the principal payer. Same-day discharges from the MedPAR are recorded with a length of stay equal to one day, while same-day discharges from the NIS are recorded with length of stay equal to zero. So for NIS-MedPAR comparisons, NIS length of stay measures for same-day discharges were changed to one day.

Tables comparing NIS and MedPAR statistics (Table 17 - Table 23) appear in Appendix C.

Comparisons by Diagnosis and Procedure Categories

NIS data were compared with both NHDS and MedPAR data across selected diagnosis and procedure groups. For NHDS comparisons, the 25 diagnoses and procedure groups observed most frequently in the NIS were selected. For MedPAR comparisons, the 25 diagnosis and procedure groups selected were those found most frequently on NIS discharges for which Medicare was the expected payer. The diagnosis and procedure groups represent a majority of pertinent discharges. For both the NHDS and MedPAR comparisons, more than one-half of all discharges were represented by the 25 diagnosis groups, while the 25 procedure groups represent nearly 60 percent of discharges that include procedure codes. In addition, MedPAR comparisons included the 25 most frequent Diagnosis Related Group (DRG) codes found for NIS Medicare discharges.

Grouping of diagnoses and procedures was done with Clinical Classifications Software (CCS). The CCS, formerly known as the Clinical Classifications for Health Policy Research (CCHPR), was developed as a means to categorize diagnoses and procedures into a limited number of clinically relevant categories. Developed for health policy analysis, the CCS can be used for aggregating the thousands of ICD-9-CM diagnoses and procedures into a manageable number of meaningful categories. CCS codes were assigned based on the principal, or first-listed, diagnosis and procedure for each discharge.

RESULTS

Estimates from different samples will not be identical because of sampling variation. Statistically significant differences can be expected for a variety of reasons, including different sampling strategies. In addition, recoding of certain conditions – as sometimes occurs in the NHDS – may lead to significant differences in the affected categories. Finally, the sheer number of tests (more than 800), will produce some statistically significant results purely by chance.1

1 While some type of correction for the number of tests could be applied, given the number of tests, this would greatly increase the risk of a Type II error.

NIS-AHA Comparisons

This section refers to tables in Appendix A (Table 7 - Table 11) comparing:

These tables suggest that NIS hospitals were quite similar to hospitals in the AHA universe; differences between NIS and AHA facilities were generally small. While NIS hospitals tend to have more admissions and discharges than those hospitals in the AHA universe, the average difference was small (approximately 1.5 percent). Median NIS counts, however, were 4.5 percent higher than the AHA numbers, suggesting that NIS hospitals tend to have more discharges than hospitals in the AHA universe. Also, NIS hospitals were slightly smaller than AHA hospitals, although these differences were minor (average: 1.1 percent; median: 2.6 percent). In addition, the average NIS hospital’s length of stay – not adjusted for hospital size or discharges counts – was on average seven percent shorter than the AHA average.

NIS hospitals also tend to have more Medicare discharges than the universe of U.S. community hospitals (Table 7). The difference was small (approximately 1.1 percent), but it may be a factor, albeit a minor one, in some other differences observed for NHDS and MedPAR comparisons to the NIS.

As shown in Table 8, NIS hospitals tend to perform more surgeries and spend more than hospitals in the AHA universe: both expenses and payroll were slightly higher (1.7 percent) at the NIS facilities. While the differences in averages were slight, the disparity in median values was quite noticeable (10.1 percent for expenses and 7.8 percent for payroll). Expenses and payroll per bed were all higher in NIS hospitals, as was full-time employment per bed.

National and regional NIS discharge estimates (Table 9 - Table 11) closely align with the discharge counts from the AHA survey. This was not surprising because NIS sampling stratum and NIS discharge weights were based on AHA annual survey results. As with the regional comparisons, the AHA-derived sampling weights in the NIS yield hospital counts consistent with AHA universe counts for various categories of hospital types.

For average length of stay (ALOS), however, the NIS sometimes differs from the AHA. While the NIS and AHA numbers were generally in agreement, the overall NIS estimate was 2.4 percent higher than the AHA average. Significant differences were also discovered for several specific hospital categories. NIS estimates for ALOS were longer – by 2.5 to 3.9 percent – than AHA statistics in

In contrast, for rural hospitals, the NIS estimate was significantly shorter – by 5.4 percent – than the average of AHA facilities. In addition, significant differences were found between the NIS estimates and the AHA numbers for three of the 15 bed size categories within hospital type. Private non-profit hospitals with 100-199 beds and 200-299 beds, and urban, non-teaching hospitals with 100-199 beds had slightly higher ALOS in the NIS.

Return to Contents

NIS-NHDS Comparisons

Generally, NIS and NHDS estimates agree. This holds true overall and across patient categories. It was also true for most hospital comparisons and specific diagnosis and procedure categories. Overall, agreements were observed for more than 72 percent of the discharge comparisons. Of the NIS-NHDS differences discovered, most occur in the diagnosis and procedure groupings. The degree of consistency for in-hospital mortality rates was high: estimates agreed for 95 percent of hospital category comparisons and more than 93 percent of comparisons by patient category. Nearly all average length of stay (ALOS) estimates were consistent between the NIS and NHDS; ALOS estimates agreed across all hospital and patient groups. Appendix B includes Table 12 through Table 16, comparing NIS and NHDS estimates. The following sections describe these tables in more detail.

Overall and Regional Comparisons

Overall and by region, no statistically significant differences emerged between the NIS and NHDS data for discharges, ALOS, or in-hospital mortality rates (Table 12). ALOS comparisons could not be made for the Northeast and Midwest, because a reliable standard error for the NHDS estimate could not be determined. However, the magnitudes of the differences between the NIS and NHDS estimates in these regions were small and appear consistent with the non-significant differences shown in other regions.

Comparisons by Hospital Characteristics

NIS and NHDS estimates were similar for each of the three hospital control/ownership categories. But some significant differences for discharge estimates were discovered between the NIS and NHDS in the bed size groupings within control/ownership categories (Table 13).

It is likely that these differences were caused by the composition of the two samples: the NIS has a greater proportion of its discharges from larger hospitals, while the NHDS has a greater proportion of its discharges from smaller hospitals. As a result, the NIS, relative to the NHDS, tends to overestimate discharges from large hospitals and underestimate discharges from small hospitals. Figure 2 and Figure 3 illustrate discharge numbers from the AHA, NIS, and NHDS for two categories in which this was particularly true: private non-profit hospitals and proprietary hospitals. These charts show that NIS discharge statistics closely agree with AHA numbers, except for proprietary hospitals with 300-499 beds.

Figure 2. Estimated Discharges from Private Non-Profit Hospitals, 2001 (text version)

Figure 2: Bar chart of the 2001 estimated discharges from private non-profit hospitals for the NIS, AHA, and NHDS

Figure 3. Estimated Discharges from Proprietary Hospitals, 2001 (text version)

Figure 3: Bar chart of the 2001 estimated discharges from proprietary hospitals for the NIS, AHA, and NHDS

Because of these hospital discrepancies, significant differences exist in discharge count comparisons by hospital bed size. Significant differences occur with seven of the 14 discharge comparisons by hospital bed size within control/ownership categories. The NIS estimate was lower than the NHDS figure in four cases (categories with fewer than 200 beds) and higher in three other instances (categories with more than 300 beds). The NIS estimate was also higher in a fourth case – proprietary hospitals with 500 or more beds – but no comparison was made because the NHDS estimated zero discharges and a valid estimate of standard error was not available. The NIS estimate for this category was 416,000 discharges. According to the AHA data, there were 406,000 discharges for this category (refer to Table 10), suggesting that the NIS estimate was the more accurate number.

ALOS and in-hospital mortality estimates were consistent, with only two exceptions. The NIS estimates for proprietary hospitals with 1-99 beds was higher than the NHDS statistic by 42 percent, but lower by 25 percent for 100-199 bed proprietary hospitals (Table 13). No comparison was possible for proprietary hospitals with more than 500 beds because no standard error estimate was available for the NHDS statistics (the NHDS reported no discharges from this type of hospital).

Comparisons by Patient Characteristics

For nearly all comparisons by patient categories (Table 14), there was agreement between the NIS and NHDS estimates. The NIS and NHDS samples closely agree across most age groups, gender, and payer categories. There were no differences in ALOS estimates, and only one in-hospital mortality rate disparity emerged: the NIS statistic was lower than the NHDS figure for patients in the "0-15 years" age group. Comparisons were not possible for the unknown/missing categories of age group, gender, and payer because NHDS standard error estimates were not available. However, there were few discharges in these groups.

Comparisons of discharge estimates differed in three categories. For "other payer," the NIS statistic was 41 percent lower than the NHDS estimate (a difference of approximately 752,000 discharges). In contrast, the NIS estimates one million more Medicare discharges than does the NHDS. Although this difference was not significant, it demonstrates that NIS hospitals tend to have slightly more Medicare discharges (1.1 percent) than the universe of hospitals (refer to Table 8).

Table 5. Racial Composition of the U.S., NIS Sample, and NHDS Sample
Race U.S. Population2 NIS Discharges with Race Information NHDS Discharges with Race Information
White 69% 69% 80%
Black 12% 13% 15%
Other 19% 18% 5%

2U.S. Census Bureau, Annual Estimates of the Population by Sex, Race and Hispanic or Latino Origin for the United States: April 1, 2000 to July 1, 2003 (NC-EST2003-03).

Two discrepancies occurred in race categories. The racial composition of the two samples differed greatly. The NHDS contains proportionately more discharges for white patients, while the NIS contains proportionately more discharges for "other" race patients. Both samples include large numbers of discharges without racial information; the information was missing for 26 percent of NIS discharges and 23 percent of NHDS discharges. Some states do not report race/ethnicity to HCUP, so for these states race is missing in the NIS.3 Because the NHDS does not include state information, it is not possible to determine if the pattern of missing information is similar. Looking only at discharges with race information, however, the NIS was more representative of the U.S. population than the NHDS, as shown in Table 5.

3 NIS states for which race was not available include Georgia, Illinois, Kentucky, Maine, Minnesota, Nebraska, Nevada, Ohio, Oregon, Washington, and West Virginia.

Comparisons by Diagnosis Category

Comparisons for diagnosis categories revealed a great deal of consistency between the NIS and NHDS samples (Table 15). The majority of comparisons in these categories show no significant differences. NIS discharge estimates differed significantly from NHDS estimates for nine of the 25 most common diagnosis categories; the NIS estimate was higher in four categories and lower in the remaining five groupings:

NIS Estimates Higher than NHDS NIS Estimates Lower than NHDS
  • "Nonspecific chest pain"


  • "Other complications of birth, puerperium affecting management of mother"


  • "Complications of device, implant/graft"


  • "Other complications of pregnancy"
  • "Affective disorders"


  • "Fluid and electrolyte disorders"


  • "Normal pregnancy and/or delivery"


  • "Urinary tract infections"


  • "Asthma"

Of these nine significant differences in the number of discharges, four can be attributed to code reordering in the NHDS ("nonspecific chest pain," and three pregnancy/delivery categories). In contrast to the NHDS, there was no reordering of diagnoses for NIS data: the first diagnosis listed for each discharge was assigned as the principal diagnosis. Diagnoses were reordered in the NHDS under certain conditions. For example, when a symptom appeared as the first-listed code, it was reassigned as a secondary diagnosis. This explains the dramatically higher figure for non-specific chest pain in the NIS sample, as compared with the NHDS (nearly 14 times higher).

Of the 25 most common diagnoses, four relate to pregnancy and delivery, including the category "normal pregnancy." Significant differences emerged for three of these categories. (No statistical comparison was possible for the fourth category, "trauma to the perineum and vulva," because a valid estimate of the NHDS standard error was not available.) Again, the differences between the NIS and the NHDS can be attributed to reordering of diagnosis codes in the NHDS data.

The NHDS assigns a code of V27 ("outcome of delivery" included in the CCS category of "normal delivery") as the principal diagnosis for all women discharged after delivery, regardless of the original principal diagnosis. As a result, the NHDS estimates 3.85 million "normal deliveries" – significantly higher than the NIS estimate. However, the NHDS estimates for the other three pregnancy/delivery classifications were much lower than the NIS estimates.

The "normal delivery" diagnosis category was also responsible for the one ALOS difference. In the NIS, the "normal delivery" category was listed as the principal diagnosis only when coded by the hospital. In contrast, deliveries in the NHDS "normal delivery" category include women who had episiotomies, as well as a variety of minor birth complications. It was not surprising, then, that the average length of stay would be longer for the NHDS "normal" category, as it includes higher risk populations.

Five of the nine significant discharge differences could not be attributed to coding differences. In four categories, the NIS estimates were lower than NHDS estimates ("affective disorders," "fluid and electrolyte disorders,"urinary tract infections," and "asthma"). In the other category ("complication of device, implants, or graft"), the NIS estimate was significantly higher than the NHDS estimate.

In contrast to the discharge and ALOS estimates, however, there were many in-hospital mortality differences; the NIS estimate was significantly higher than the NHDS estimate in nine cases and significantly lower in eight other instances. While four of these differences can be explained by the reordering that occurred for some NHDS discharges, unexplained differences remain for most of the available comparisons.

Return to Contents

Comparisons by Procedure Category

Table 16 provides results by procedure category. NIS discharge estimates differed significantly from the NHDS estimates for six of the twenty-five categories ("other procedures to assist delivery," "diagnostic cardiac catheterization," "percutaneous coronary angioplasty," "coronary artery bypass graft," "laminectomy," and "pacemaker or cardioverter/defibrillator"). In each case, the NIS estimate was significantly higher than the NHDS estimate.

Comparisons of ALOS estimates revealed only one significant difference: the NIS estimate for "other vascular catheterization, not heart" was shorter than the NHDS number. But NIS-NHDS differences were discovered for almost half of the in-hospital mortality comparisons. The NIS mortality estimate was lower than the NHDS statistic for seven procedures and higher than the NHDS estimate for five other procedures.

NIS-MedPAR Comparisons

With the notable exception of discharge counts, NIS estimates of Medicare measures were generally consistent with MedPAR statistics. NIS discharge estimates were uniformly higher than the MedPAR numbers by approximately 21 percent (Table 17). The foremost cause of this discrepancy seems to be the omission of most managed care clients from the MedPAR. While approximately 15.4 percent of Medicare patients were enrolled in managed care programs, managed care discharges constitute only 0.9 percent of MedPAR discharges.

File composition was another contributing factor. While the MedPAR represents actual fee-for-service claims paid by Medicare, the NIS-Medicare sample consists of discharges (both fee-for-service and managed care) for which Medicare was the expected payer (either primary or secondary). This may explain the higher NIS counts, because the expected payer was not always the actual payer. Finally, a minor factor may be the composition of the NIS. As noted in the discussion of NIS-AHA comparisons, NIS hospitals had more Medicare discharges than the average U.S. community hospital. The difference between NIS and U.S. hospitals, however, was small: approximately one percent.

Because the overall NIS estimate of Medicare discharges exceeds the actual number in the MedPAR data, it was not surprising to find that nearly all the NIS discharge estimates were also significantly higher than the corresponding MedPAR totals. This suggests the need for a more useful comparison of discharges, so we have included a test between proportions of patients in the various categories. And for most comparisons of discharge proportions, few meaningful differences were discovered; proportions were consistent for almost 80 percent of all comparisons.

NIS Medicare estimates were also consistent with MedPAR measures of ALOS, in-hospital mortality rates, and average total hospital charges. Consistency was discovered for:

Across hospital categories only a handful of meaningful differences were observed. The tables in Appendix C compare NIS Medicare estimates with MedPAR statistics. The following sections refer to these tables.

Overall and Regional Comparisons

Overall, the NIS estimate of Medicare discharges was 21 percent higher than the total number of MedPAR discharges (Table 17). By Census region, all NIS estimates were higher than MedPAR counts, although the difference was not significant for the Midwest. The magnitude of difference was greatest in the Northeast (31 percent higher) and West (39 percent higher), regions with the largest Medicare managed care penetration. When examined from the perspective of proportions (percentage of discharges), significant differences were discovered only in the Midwest (11 percent lower) and West (14 percent higher).

No significant NIS-MedPAR differences, either nationally or regionally, were found for ALOS, in-hospital mortality, or average total hospital charge measures. The similarities of these statistics suggest that no fundamental differences exist between the two databases in their description of patient outcomes.

Comparisons by Hospital Characteristics

Two sets of hospital characteristics were compared for Medicare discharges: first, hospital control and number of beds (categories used in the NHDS comparisons); and second, hospital location, teaching status, and size (NIS stratification variables). While NIS discharge estimates generally exceed MedPAR counts, most other statistics were quite similar between the two databases, including discharge proportions.

By hospital control, or ownership, NIS estimates of Medicare discharges were uniformly higher than MedPAR counts – on the order of 22 to 23 percent (Table 18). However, the percentage of discharges differed only for the category of private, non-profit hospitals (7 percent higher for the NIS). For all other measures (ALOS, in-hospital mortality, and average total charge), the NIS estimates were similar to the MedPAR numbers.

When hospital control was examined by number of beds (Table 18), many NIS discharge estimates were actually in agreement with Medicare counts; significant differences were observed for only six of the 15 discharge comparisons by number of beds. Differences in discharge counts include:

Discharge proportions, however, were similar between the NIS and MedPAR databases. Only two significant differences emerged for the hospital control and bed size comparisons. Both differences occurred for proprietary hospital categories. This may be a result of the NIS make-up:

ALOS, in-hospital mortality, and average total charge statistics were also quite similar when control was examined across bed size categories. Of the 15 comparisons, few meaningful differences were discovered.

Two significant differences emerged for average length of stay comparisons:

Analysis also revealed two differences for in-hospital mortality rates:

There were two significant differences for average total charge:

A second set of hospital comparisons examines NIS and MedPAR statistics by hospital type, location, and teaching status (Table 19). Most NIS discharge estimates, including statistics for all three hospital types, were significantly higher than the MedPAR counts. However, for discharge proportions, only two substantial differences were discovered. The estimated NIS proportion was 6.2 percent lower than the MedPAR proportion for all rural hospitals, but 18 percent higher for small rural hospitals. Comparisons of other measures also revealed consistency between the NIS and MedPAR databases.

In overall comparisons of location and teaching status, there were no significant differences for the ALOS, in-hospital mortality rates, or average charge measures. Only one meaningful ALOS difference was observed in comparisons by hospital type and size (for small, non-teaching urban hospitals). In addition, one difference for in-hospital mortality rates emerged (for mid-sized rural hospitals). But no significant differences were discovered for average total hospital charges.

Comparisons by Patient Characteristics

Comparisons by patient characteristics revealed significant differences for all discharge count comparisons, as well as most discharge proportions. Several differences also emerged in relation to in-hospital mortality rate comparisons, but nearly all ALOS and average total charge evaluations were consistent between the NIS and MedPAR.

Several significant differences were observed for race categories. These are most likely caused by the large percentage of NIS discharges without race information (as discussed previously in this report). One of every four NIS Medicare discharges lacks race information, while more than 99 percent of MedPAR discharges include race information. Hence, the NIS and MedPAR present different mixes of patient characteristics:

Relative to MedPAR numbers, the NIS tends to overestimate patients between 65 and 84 years of age (the age group responsible for approximately two-thirds of Medicare inpatient discharges) and to underestimate patients younger than 65 and older than 85. Comparing the percentage of discharges in each age group, the NIS overestimates the 65-74 age group by 2.1 percent and the 75-84 age group by 3.8 percent. On the other hand, the NIS underestimates the 0-64 group by 8.5 percent and the 85+ age group by 3.6 percent. There were no differences between the NIS and MedPAR when comparing genders for percentages of discharges, ALOS, in-hospital mortality, and average total charges.

ALOS estimates were generally in agreement between the two databases; in nearly every category, no meaningful differences emerged between the NIS and MedPAR numbers. The NIS estimate was lower than the MedPAR average where race was unknown.

Significant differences were observed for one-half of the race and age group comparisons of in-hospital mortality rates. NIS estimates for "other" race and "unknown race," as well as for patients 75-84 years of age, were higher than the corresponding MedPAR statistic. For patients 65-74 years of age, the NIS estimate was lower than the MedPAR rate. Relative in-hospital mortality rate differences were:

Return to Contents

Comparisons by DRG

In comparisons of diagnosis related group (DRG) categories (Table 21), most NIS estimates, with the exception of discharge counts, were consistent with corresponding MedPAR statistics. However, significant differences were found for all 25 DRG comparisons of discharge counts. The NIS estimate was higher than the MedPAR count in every case, ranging from 12 percent higher ("psychosis") to 29 percent higher ("rehabilitation"). The median difference in number of discharges was 22 percent.

For DRG comparisons of discharge proportions, ALOS, in-hospital mortality, and average hospital charge, NIS and MedPAR statistics were quite similar. Differences for these measures include:

For most DRGs, significant differences were observed only for discharge counts. If other discrepancies arose, inconsistencies were generally limited to only one outcome. However, there were three exceptions:

Comparisons by Diagnosis Category

Significant differences were observed between NIS estimates of Medicare discharges and MedPAR discharges by count for 24 of the 25 principal diagnosis categories (Table 22). These differences ranged from 15 percent higher to 29 percent higher ("affective disorders" and "rehabilitation care, fitting of prostheses," respectively). The median difference was 21 percent.

Comparisons for other measures indicated a high degree of consistency between the NIS and MedPAR statistics. The 25 diagnosis category comparisons revealed only a handful of significant differences for any other measure (discharge proportions, ALOS, in-hospital mortality rates, and total charges):

Comparisons for most diagnosis categories revealed discrepancies only on discharge counts. When other differences were observed, inconsistency was generally limited to one other measure. There were, however, two exceptions:

Comparisons by Procedure Category

In contrast to the diagnosis and DRG evaluations, comparisons by procedure groups revealed greater variability in discharge count comparisons (Table 23). The range in differences was wider than that observed for diagnosis or DRG categories. All but three NIS discharge estimates by procedure exceeded the corresponding MedPAR total; the median difference was 20 percent. NIS discharge estimates were higher than MedPAR counts, ranging from 13 percent higher ("laminectomy, excision of intervertebral disc") to 39 percent higher ("physical therapy, exercises, manipulation").

For the majority of other measures, the NIS estimates were consistent with MedPAR statistics. Only a handful of differences were observed across the 25 most frequent procedure categories:

All NIS average hospital charge estimates were consistent with MedPAR averages.

Finally, only two procedure categories revealed more than one significant difference among the outcome measurements: discharge proportion, ALOS, in-hospital mortality rate, and average total hospital charge. These categories included:

Return to Contents

DISCUSSION

These results indicate that estimates from the 2001 NIS were generally in line with statistics from the 2001 NHDS and the 2001 MedPAR. Most NIS estimates were consistent with NHDS estimates for discharges, average length of stay, and in-hospital mortality rates. Nearly all of the average length of stay and most in-hospital mortality rate estimates were consistent between the two samples. Differences occurred primarily when comparing estimates for specific diagnosis or procedure groups. A critical difference between the 2001 NIS and 2001 NHDS data was that the NHDS reordered some diagnosis codes (in an effort to achieve more consistency within that sample). As a result of these coding alterations, some significant differences appear in the findings related to diagnosis categories.

While most NIS estimates were consistent with MedPAR statistics, NIS estimates of Medicare discharge counts were 21 percent higher than MedPAR estimates. The primary reason for this difference was the absence of most managed care discharges from the MedPAR data. This discrepancy was exaggerated because the NIS was drawn from states that have higher managed care penetration than the national average. Finally, most average length of stay, in-hospital mortality and average total charge estimates from the NIS were consistent with the corresponding MedPAR statistics.

The key difference between the NIS and the databases with which it was compared is geographic. Both the NHDS and the MedPAR are national in coverage; MedPAR data include all Medicare paid, fee-for-service discharges in the U.S., while NHDS data were gathered from a sampling frame of all 50 states plus the District of Columbia. In contrast, the 2001 NIS was drawn from the 33 states (as shown in Table 1); these states comprise more than 81 percent of all U.S. community hospital discharges. This difference may be a factor for researchers in cases where comprehensive geographic representation is important. Some significant differences between the states excluded and included in the NIS may offer explanations for several of the observed differences.

NIS states are disproportionately the more populous states. NIS states had an average population density of 124.5 persons per square mile in 2001, as compared with a national average of 80.6 persons per square mile. The average population density of non-NIS states was 28.8 persons per square mile. Of the ten states with the highest population density, all but two were included in the NIS. These states, and their rank in terms of population density order, are: New Jersey (1), Rhode Island (2), Massachusetts (3), Connecticut (4), Maryland (5), New York (7), Florida (8), and Pennsylvania (10). At the other end of the spectrum, only two of the ten least populous states were included in the NIS: Utah (41) and Nebraska (42).4 Given this difference in geographic sampling, the NIS sampling frame starts with few hospitals in sparsely populated areas. Even weighting the discharges from rural states does not adequately account for the remote areas of the country, which include a disproportionate number of the smallest hospitals. The most rural state included in the sample, Nebraska, has a population density of 22.4 persons per square mile, compared with population densities of 1.1 for Alaska, 5.1 for Wyoming, and 6.2 for Montana.5

4Source of state rankings: State and Metropolitan Area Data Book - 5th Edition and 2000 U.S. Census.

5None of these three states had all-payer hospital discharge data for the 2001 data year, so none were eligible for HCUP inclusion.

One impact of the specific subset of states selected for the NIS was an over representation of Medicare patients in managed care. In the 33 states included in the 2001 NIS, the market penetration of managed care providers for Medicare enrollees averaged 16.6 percent. In contrast, for the 17 states not included in the NIS, the mean market penetration of managed care providers was only 9.4 percent. Table 6 examines managed care penetration by region of NIS and non-NIS states. For 2001, Medicare managed care market penetration in the Northeast, South, and West regions was higher in NIS states than in non-NIS states; the greatest penetration discrepancies were observed in the West and Northeast.6 These were also the regions with the largest difference between MedPAR discharges and NIS estimates. This finding was consistent with the hypothesis that the MedPAR under represents total discharges by omitting most managed care discharges.7

6The NIS includes all Northeast states except New Hampshire.

7Source: Medicare Managed Care Market Penetration for All Medicare Plan Contractors - Quarterly State/County Data Files, June 2001 (http://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/HealthPlanRepFileData/Downloads/SC-2001.zip). (Accessed July 10, 2008)

Table 6. Medicare Managed Care Market Penetration by Region
  Non-NIS States NIS States All States in Region
Mean N Mean N Mean N
Northeast 0.8% 1 18.2% 8 17.9% 9
South 7.6% 6 10.0% 10 9.6% 16
Midwest 10.0% 4 8.2% 8 8.8% 12
West 14.2% 6 33.5% 7 31.2% 13

This exclusion by MedPAR was inconsequential in those areas with minimal market penetration by managed care providers, but greater for regions, particularly the West, in which managed care participation by Medicare patients was higher. Because the NIS includes discharges for all Medicare managed care patients and not just the fee-for-service discharges, it may be preferable to the MedPAR file for estimating the total Medicare discharges.

NIS Strengths

While the previous discussion focused on differences between the NIS and other data sources, it should be noted that these differences are only of concern when there is a reason to expect that geographic region might relate to the variable of interest. We must emphasize that the NIS provides a large sample size that tends to yield estimates with much smaller standard errors than does a sample such as the NHDS. Without a sample of several million, as provided by the NIS, estimates for less common procedures and diagnoses are unreliable. While the NIS may overemphasize urbanized areas, this emphasis on higher density states makes data available on atypical conditions that might rarely find inclusion in a smaller sample.

NIS discharge estimates were quite similar to numbers from the AHA, regardless of the hospital characteristics. NIS estimates were generally in line with NHDS estimates as well. When estimating ALOS and in-hospital mortality for the nation, or within any major categories, the NIS rates were consistent with the NHDS data. Because NIS estimates have greater precision – a result of the large sample size – it may be preferred for certain analyses based on relatively uncommon conditions. Furthermore, the NIS contains total hospital charges, while the NHDS does not. For analysis involving charges on all payers, the NIS is the only choice.

The NIS provides a large sample of Medicare discharges both in managed care and fee-for-service plans; it would therefore be the choice of researchers who wished to include all discharges regardless of payment type. Inclusion of Medicare managed care discharges leads to discrepancies in estimated discharge counts, but most other NIS Medicare estimates were similar to MedPAR statistics, particularly in comparisons by hospital characteristics.

NIS Weaknesses

NIS discharge estimates vary from NHDS estimates by hospital size; the NIS includes more discharges from large hospitals more than does the NHDS, although NIS discharge estimates were close to AHA survey results. Because the NHDS uses a more geographically complete sampling frame, however, it might be preferable for researchers in certain cases.

The NIS also contains significant numbers of discharges for which race was missing (26 percent). While the NHDS also suffers from this problem (23 percent of discharges without race), the MedPAR includes an insignificant number of discharges without race information.

Because of the states available for the sample, the NIS exaggerates the discrepancy between total Medicare discharges and the MedPAR’s primarily fee-for-service population. The MedPAR database provides no estimate for managed care participants, while the NIS database may overestimate the number of discharges in managed care.

Contrasting Findings from the 2000 and 2001 NIS Comparisons

NIS-NHDS Evaluations

Estimates of most outcome measurements from the 2001 NIS and NHDS data were consistent; this finding was similar to evaluations in previous years. Overall, the discharge and ALOS estimates from the two databases were similar for both years. NIS and NHDS estimates of ALOS were almost indistinguishable; there were few significant ALOS differences in either 2000 or 2001. More than 80 comparisons were made for each year of data: only four differences emerged for the 2000 data and only two significant differences emerged for the 2001 data. NIS and NHDS discharge estimates from 2001 and 2000 data were also similar, although for both years the data sources generate divergent statistics for large and small hospitals.

In-hospital mortality rate estimates for 2001 data were more consistent than the 2000 data across both hospital and patient categories, although 2001 comparisons revealed more discrepancies for the diagnosis and procedure classifications. Of all hospital comparisons, one significant mortality difference was observed, and a single meaningful mortality rate difference was discovered for patient categories as well. Both outcomes were improvements over 2000 assessments. For diagnosis and procedure comparisons, the 2001 evaluations revealed more differences than in previous years. No trend appears with these differences, however. Categories with lower NIS rates were as prevalent as those with higher NIS rates.

Differences in In-hospital mortality rate conflicts may be related to differences in the hospitals included in the two samples. The NIS tends to have better representation from larger hospitals. The NIS better captures less common diagnoses which tend to have higher mortality rates . In addition, because the NIS retains all discharges from a hospital, it was not possible to exclude some of the higher mortality cases that might have been treated in skilled-nursing facilities and other long-term care units within the hospital. There may also be differences with regard to a hospital’s teaching status or location, although this cannot be verified because the NHDS does not report hospital teaching status or urban/rural information.

8The average in-hospital mortality rate for discharges with the 50 most frequent diagnosis groups was 2.1 percent. This compares to an average of 4.8 percent for discharges with one of the 50 least frequently found diagnosis groups.

NIS-MedPAR Evaluations

As discussed earlier in this report, NIS Medicare discharge estimates were higher – overall, and for almost all categories – than MedPAR counts. Inconsistencies were noted for nearly all discharge counts. The overall discrepancy was 21 percent. This was also true for earlier years: the difference in 2000 was 22 percent and in 1999 it was 12 percent. The growth from 1999 to 2000 may have been caused by increases in Medicare managed care market penetration, particularly within NIS states.

While there were differences for discharge statistics, other estimates were similar between the two data sources. Most NIS estimates of discharge proportions, ALOS, in-hospital mortality rates, and average total hospital charge were comparable to MedPAR statistics. Mortality rates were quite similar in both years. Comparisons for 2001 data, however, did reflect improvement for most of these measures over prior years. In particular, estimates of discharge proportions improved in 2001, largely because diagnosis and procedure comparisons were more consistent.

ALOS comparisons were greatly improved as well. The overall NIS Medicare estimate of ALOS in 2000 was significantly shorter in duration than the MedPAR average. For the 2001 data, this was no longer the case; the NIS and MedPAR ALOS statistics were consistent. And ALOS evaluations for hospital and patient categories were also more consistent for 2001. Finally, average hospital charge comparisons revealed few differences in either 2001 or 2000.

Return to Contents

Conclusion

Each of the data sources discussed has its strengths and weaknesses, and each may be the preferred choice for different research questions. The NIS offers a large sample that enables study of low incidence disorders and less common procedures; NIS estimates can be calculated for literally thousands of special sub-populations that may be of interest to researchers. In addition, NIS hospitals accurately reflect the universe of U.S. hospitals, particularly the relative mix of large and small hospitals. So the NIS may be more appropriate when hospital type and size is an important consideration.

The NHDS and MedPAR, however, both offer data drawn from all 50 states, rather than the 33 states that make up the NIS. Where a comprehensive geographic representation is more important than a large sample size, and the question under study requires all age groups, the NHDS might be preferable. In the same situation, if only Medicare clients are of interest, the MedPAR data set might be preferable.

The NIS is not without bias. It does, however, provide a useful data source for answering many research questions. The source of the few differences that do exist between the NIS and NHDS are areas that warrant further investigation. The relationship between hospital size and treatment patterns is an example.

As for which of the data sources discussed is preferable or better, the answer depends on the needs of the researcher. The intended use of the data is the most critical factor in determining which data source will be most valuable. In general, the NIS estimates of variables essential to healthcare policy – including in-hospital mortality, inpatient population size, length of stay, and costs – are accurate and precise. Statistics can be calculated for large groups ranging from the inpatient population of the United States, as well as for small subsets featuring specific conditions. The characteristics documented in this report suggest that the 2001 NIS is a valuable tool for researchers and policy makers alike.

REFERENCES

Centers for Medicare & Medicaid Services (2004). Medicare Managed Care Market Penetration for All Medicare Plan Contractors - Quarterly State/County Data Files, June 2001. Washington, DC: CMS. http://www.cms.gov/Research-Statistics-Data-and-Systems/Statistics-Trends-and-Reports/HealthPlanRepFileData/Downloads/SC-2001.zip

Korn, E. L. & Graubard, B. I. (1999). Analysis of Health Surveys. New York: John Wiley & Sons.

National Center for Health Statistics (2001). National Hospital Discharge Survey Public Use Data File Documentation: 2001. Washington, DC: U.S. Department of Health and Human Services, National Center for Health Statistics.

U.S. Census Bureau (2004). Annual Estimates of the Population by Sex, Race and Hispanic or Latino Origin for the United States: April 1, 2000 to July 1, 2003. (NC-EST2003-03). Washington, DC: Population Division, U.S. Bureau of the Census.

U.S. Census Bureau (2004). Annual Estimates of the Population for the United States and States, and for Puerto Rico: April 1, 2000 to July 1, 2003. (NST-EST2003-01). Washington, DC: Population Division, U.S. Bureau of the Census.

U.S. Census Bureau (2004). State and Metropolitan Area Data Book – 5th Edition. Washington, DC: U.S. Bureau of the Census.

APPENDIX A: NIS-AHA TABLES

Table 7. Number of Hospitals in NIS Frame and Universe, 2001
  Hospital Counts
2001 AHA Universe 2001 NIS Frame1(Weighted) 2001 NIS Frame1(Unweighted)
U.S. 4,812 4,812 986
Region
Northeast 668 668 136
Midwest 1,392 1,392 284
South 1,848 1,848 379
West 904 904 187
Hospital Control
Public 1,158 1,140 235
Private, Non-Profit 2,953 2,956 604
Proprietary 701 717 147
Location/Teaching Status
Rural Hospitals 2,169 2,169 443
Urban, Non-Teaching 1,842 1,842 378
Urban, Teaching 801 801 165

Note: Significance tests were not performed because AHA numbers were not sample statistics.
1The 2001 frame contains 33 states.

Table 8. NIS Sampling Frame and AHA Universe Comparisons, 2001
  Mean Hospital Values Median Hospital Values
Universe NIS Frame1 Universe NIS Frame1
Hospital Admissions 6,930.91 7,029.04 3,747.00 3,914.50
Hospital Discharges 6,930.91 7,029.04 3,747.00 3,914.50
Hospital Discharges2 7,740.09 7,853.47 4,190.00 4,380.00
Hospital Beds 154.71 153.03 98.00 95.50
Occupancy Rate 0.50 0.51 0.52 0.52
Average Length of Stay 5.39 5.02 4.42 4.31
Average Length of Stay2 4.99 4.63 4.03 3.95
Births 809.18 824.43 322.00 320.50
Inpatient Surgeries 2,043.02 2,076.17 1,004.00 1,102.50
Total Hosp. Expenses [dollars] 79,118,593 80,483,684 35,598,000 39,185,347
Hosp. Expenses/Bed [dollars] 438,905 472,035 402,899 426,840
Total Hospital Payroll [dollars] 32,933,440 33,482,377 14,873,715 16,032,074
Hosp. Payroll per Bed [dollars] 182,638 195,357 165,590 176,007
Percent Medicare Days 53.34 54.28 53.94 54.96
Percent Medicare Discharges 47.48 47.99 47.56 47.84
Percent Medicare Discharges2 43.99 44.47 42.75 43.59
Percent Medicaid Days 14.01 14.18 11.94 12.12
Percent Medicaid Discharges 14.70 14.84 13.88 13.88
Percent Medicaid Discharges2 13.29 13.39 12.49 12.43
FTE3 823.00 823.03 406.50 426.75
FTE3 per Bed 4.96 5.21 4.55 4.78

Note: Significance tests were not performed because AHA numbers were not sample statistics.
1The 2001 frame contains 33 states.
2Adjusted for well newborns.
3Full-time equivalents.

Table 9. NIS and AHA Comparisons Overall and by Region, 2001
  Number of Discharges in Thousands (Standard Error) Average Length of Stay in Days (Standard Error)
NIS AHA NIS AHA
Overall 37,187
(593)
37,671 4.63
(0.03)
4.52**
Region
Northeast 7,408
(264)
7,407 5.33
(0.11)
5.25
Midwest 8,658
(248)
8,658 4.43
(0.06)
4.35
South 14,129
(393)
14,129 4.58
(0.05)
4.47*
West 6,990
(257)
6,990 4.23
(0.08)
4.08

Note: AHA discharges and length of stays were adjusted for well newborns.
*Significant at a 5 percent level.
**Significant at a 1 percent level.

Return to Contents

Table 10. NIS and AHA Comparisons by Hospital Control, 2001
Hospital Control Number of Discharges in Thousands (Standard Error) Average Length of Stay in Days (Standard Error)
NIS AHA NIS AHA
Public
Total 4,704
(472)
5,127 4.63
(0.13)
4.73
1-99 Beds 1,098
(63)
1,149 3.64
(0.06)
3.94**
100-199 Beds 917
(116)
1,044 4.16
(0.11)
4.29
200-299 Beds 714
(199)
606 4.72
(0.20)
4.55
300-499 Beds 1,262
(290)
1,108 5.11
(0.21)
5.32
500+ Beds 710
(245)
1,218* 5.86
(0.56)
5.41
Private Non-Profit
Total 27,640
(800)
27,469 4.64
(0.04)
4.49**
1-99 Beds 2,876
(149)
2,665 3.75
(0.08)
3.81
100-199 Beds 5,234
(312)
5,178 4.47
(0.08)
4.14**
200-299 Beds 5,307
(466)
5,303 4.72
(0.08)
4.46**
300-499 Beds 8,388
(709)
7,865 4.59
(0.07)
4.54
500+ Beds 5,833
(688)
6,458 5.21
(0.12)
5.03
Proprietary
Total 4,842
(355)
4,589 4.57
(0.09)
4.45
1-99 Beds 691
(71)
637 4.18
(0.26)
3.91
100-199 Beds 1,706
(128)
1,670 4.28
(0.13)
4.16
200-299 Beds 830
(145)
1,088 4.88
(0.20)
4.54
300-499 Beds 1,198
(204)
785* 4.71
(0.10)
4.67
500+ Beds 416
(114)
406 5.36
(0.33)
5.83

Note: AHA discharges and length of stays were adjusted for well newborns.
*Significant at a 5 percent level.
**Significant at a 1 percent level.

Table 11. NIS and AHA Comparisons by Hospital Characteristics, 2001
  Number of Discharges in Thousands (Standard Error) Average Length of Stay in Days (Standard Error)
NIS AHA NIS AHA
Location/Teaching Status
Rural - Total 5,800
(214)
5,800 3.92
(0.05)
4.13**
1-49 beds 1,253
(60)
1,248 3.53
(0.12)
3.77 *
50-99 beds 1,781
(136)
1,595 3.66
(0.04)
3.88**
100+ beds 2,765
(255)
2,956 4.27
(0.08)
4.41
Urban, Non-Teaching - Total 15,269
(373)
15,268 4.50
(0.05)
4.33**
1-99 beds 1,595
(116)
1,511 4.09
(0.16)
3.84
100-199 beds 4,793
(249)
4,954 4.41
(0.08)
4.10**
200+ beds 8,880
(390)
8,802 4.63
(0.06)
4.54
Urban, Teaching - Total 16,117
(408)
16,117 5.00
(0.06)
4.85*
1-299 beds 2,532
(307)
2,339 4.83
(0.12)
4.49**
300-499 beds 4,705
(540)
4,429 4.83
(0.11)
4.73
500+ beds 8,879
(658)
9,348 5.14
(0.10)
4.99

Note: AHA discharges and length of stays were adjusted for well newborns.
*Significant at a 5 percent level.
**Significant at a 1 percent level.

APPENDIX B: NIS-NHDS TABLES

Table 12. NIS and NHDS Comparisons Overall and by Region, 2001
  Number of Discharges in Thousands (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error)
NIS NHDS NIS NHDS NIS NHDS
United States 37,187
(593)
36,311
(1,470)
4.61
(0.03)
4.69
(0.30)
2.31
(0.03)
2.24
(0.12)
Region
Northeast 7,408
(264)
7,788
(632)
5.30
(0.11)
5.401
(c)
2.54
(0.08)
2.38
(0.27)
Midwest 8,658
(248)
8,206
(788)
4.41
(0.06)
4.171
(c)
2.11
(0.05)
2.13
(0.28)
South 14,129
(393)
14,138
(721)
4.56
(0.05)
4.76
(0.39)
2.41
(0.05)
2.31
(0.16)
West 6,990
(257)
6,177
(498)
4.21
(0.08)
4.34
(0.54)
2.12
(0.09)
2.07
(0.23)

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.
(a) Because of a limited sample, the NHDS estimate and standard error were unreliable and not reported.
(b) The NHDS estimate was reported but was not considered reliable; the standard error was not reported.
(c) A valid standard error could not be calculated.

Table 13. NIS and NHDS Comparisons by Hospital Control and Size, 2001
Hospital Control/Size Number of Discharges in Thousands (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error)
NIS NHDS NIS NHDS NIS NHDS
Total Public 4,704
(472)
4,585
(189)
4.61
(0.13)
4.67
(0.30)
2.28
(0.06)
2.19
(0.12)
1-99 Beds 1,098
(63)
1,414**
(61)
3.61
(0.06)
3.42
(0.23)
2.44
(0.08)
2.23
(0.13)
100-199 Beds 917
(116)
955
(42)
4.14
(0.11)
4.39
(0.30)
2.29
(0.12)
2.16
(0.13)
200-299 Beds 714
(199)
358
(18)
4.70
(0.20)
4.51
(0.34)
2.05
(0.25)
1.71
(0.12)
300-499 Beds 1,262
(290)
1,155
(50)
5.09
(0.21)
5.60
(0.37)
2.26
(0.12)
2.24
(0.13)
500+ Beds 710
(245)
701
(32)
5.84
(0.56)
6.11
(0.42)
2.30
(0.18)
2.30
(0.15)
 
Total Private Non-Profit 27,640
(800)
27,354
(1,109)
4.61
(0.04)
4.67
(0.29)
2.33
(0.04)
2.25
(0.12)
1-99 Beds 2,876
(149)
5,251**
(216)
3.72
(0.08)
4.04
(0.26)
2.28
(0.13)
1.96
(0.11)
100-199 Beds 5,234
(312)
7,884**
(322)
4.45
(0.08)
4.59
(0.29)
2.38
(0.08)
2.36
(0.13)
200-299 Beds 5,307
(466)
4,971
(205)
4.70
(0.08)
4.75
(0.30)
2.30
(0.08)
2.20
(0.12)
300-499 Beds 8,388
(709)
6,139**
(252)
4.57
(0.07)
4.86
(0.31)
2.31
(0.07)
2.23
(0.13)
500+ Beds 5,833
(688)
3,106**
(129)
5.19
(0.12)
5.43
(0.35)
2.34
(0.11)
2.55
(0.15)
Total Proprietary 4,842
(355)
4,371
(180)
4.55
(0.09)
4.86
(0.31)
2.26
(0.08)
2.28
(0.13)
1-99 Beds 691
(71)
1,345**
(58)
4.16
(0.26)
4.90
(0.33)
2.05
(0.17)
1.44**
(0.08)
100-199 Beds 1,706
(128)
1,625
(69)
4.26
(0.13)
4.88
(0.32)
2.21
(0.09)
2.95**
(0.17)
200-299 Beds 830
(145)
743
(34)
4.86
(0.20)
4.94
(0.34)
2.37
(0.11)
2.53
(0.16)
300-499 Beds 1,198
(204)
657**
(30)
4.69
(0.09)
4.65
(0.33)
2.29
(0.24)
2.10
(0.13)
500+ Beds 416
(114)
01
(a)
5.35
(0.33)
0.001
(a)
2.55
(0.39)
0.001
(a)

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.
(a) Because of a limited sample, the NHDS estimate and standard error were unreliable and not reported.
(b) The NHDS estimate was reported but was not considered reliable; the standard error was not reported.
(c) A valid standard error could not be calculated.

Return to Contents

Table 14. NIS and NHDS Comparisons by Patient Characteristics, 2001
  Number of Discharges in Thousands (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error)
NIS NHDS NIS NHDS NIS NHDS
Age Group
0-15 Years 5,968
(206)
6,386
(262)
3.44
(0.07)
3.77
(0.24)
0.37
(0.02)
0.50**
(0.02)
16-44 Years 10,225
(211)
10,174
(415)
3.56
(0.04)
3.67
(0.23)
0.44
(0.01)
0.39
(0.02)
45-64 Years 7,674
(143)
7,224
(296)
4.89
(0.04)
5.01
(0.32)
1.99
(0.03)
2.07
(0.11)
65+ Years 13,316
(266)
12,525
(510)
5.77
(0.04)
5.81
(0.37)
4.81
(0.05)
4.74
(0.27)
Unknown 2
(0)
01
(a)
4.86
(0.93)
0.001
(a)
0.88
(0.45)
0.001
(a)
Gender
Female 21,984
(369)
21,593
(876)
4.40
(0.03)
4.48
(0.28)
1.99
(0.03)
1.97
(0.11)
Male 15,197
(243)
14,717
(598)
4.90
(0.04)
5.01
(0.32)
2.77
(0.03)
2.65
(0.15)
Unknown 5
(1)
01
(a)
4.14
(0.38)
0.001
(a)
1.36
(0.69)
0.001
(a)
Race
White 18,998
(681)
22,351*
(1,271)
4.73
(0.04)
4.68
(0.40)
2.63
(0.04)
2.36
(0.19)
Black 3,553
(245)
4,333
(313)
5.33
(0.10)
5.40
(0.62)
2.17
(0.05)
1.91
(0.19)
Other 4,823
(312)
1,429**
(202)
4.27
(0.08)
4.741
(c)
1.53
(0.05)
2.821
(c)
Unknown 9,812
(701)
8,195
(1,212)
4.26
(0.06)
4.351
(c)
2.13
(0.04)
2.001
(c)
Principal Payer
Medicare 13,727
(269)
12,685
(560)
5.84
(0.04)
5.93
(0.42)
4.37
(0.04)
4.34
(0.27)
Medicaid 6,378
(245)
5,915
(425)
4.26
(0.07)
4.51
(0.53)
0.99
(0.04)
1.02
(0.10)
Private Insurance 14,124
(409)
14,145
(968)
3.70
(0.03)
3.80
(0.40)
1.08
(0.03)
1.05
(0.10)
Self Pay 1,676
(137)
1,613
(77)
3.81
(0.10)
3.88
(0.34)
1.38
(0.05)
1.41
(0.09)
No Charge 102
(34)
116
(21)
4.67
(0.21)
4.861
(c)
1.50
(0.17)
1.431
(c)
Other 1,082
(84)
1,834*
(326)
4.08
(0.09)
4.321
(c)
1.68
(0.13)
1.731
(c)
Missing 30
(8)
01
(a)
4.27
(0.37)
0.001
(a)
1.66
(0.50)
0.001
(a)

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.
(a) Because of a limited sample, the NHDS estimate and standard error were unreliable and not reported.
(b) The NHDS estimate was reported but was not considered reliable; the standard error was not reported.
(c) A valid standard error could not be calculated.

Table 15. NIS and NHDS Comparisons by Principal Diagnosis Category, 2001
Principal Diagnosis Number of Discharges in Thousands (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error)
NIS NHDS NIS NHDS NIS NHDS
218: Liveborn 3,999
(129)
3,668
(152)
3.01
(0.05)
3.27
(0.21)
0.30
(0.01)
0.36*
(0.02)
101: Coronary atherosclerosis and other heart disease 1,400
(58
1,255
(55)
3.63
(0.05)
3.35
(0.22)
0.75
(0.02)
0.68
(0.04)
122: Pneumonia (except that caused by tuberculosis or sexually transmitted disease) 1,222
(19)
1,317
(57)
5.88
(0.04)
5.74
(0.38)
5.83
(0.09)
5.41
(0.32)
108: Congestive heart failure, nonhypertensive 1,049
(21)
1,023
(45)
5.57
(0.04)
5.45
(0.37)
4.59
(0.08)
4.04*
(0.24)
102: Nonspecific chest pain 875
(25)
63**
(5)
1.79
(0.01)
1.291
(c)
0.05
(0.00)
0.00**
(0.00)
100: Acute myocardial infarction 773
(24)
794
(36)
5.40
(0.06)
5.74
(0.39)
8.12
(0.12)
9.92**
(0.61)
193: Trauma to perineum and vulva 763
(28)
--1
(a)
1.96
(0.01)
--1
(a)
0.00
(0.00)
--1
(a)
69: Affective disorders 708
(39)
949**
(42)
7.57
(0.16)
7.48
(0.51)
0.05
(0.00)
0.11**
(0.00)
106: Cardiac dysrhythmias 704
(18)
707
(32)
3.52
(0.03)
3.52
(0.25)
1.21
(0.03)
1.37
(0.08)
195: Other complications of birth, puerperium affecting management of mother 641
(24)
52**
(4)
2.55
(0.02)
2.891
(c)
0.03
(0.00)
0.07**
(0.00)
205: Spondylosis, intervertebral disc disorders, other back problems 639
(22)
573
(27)
3.10
(0.03)
3.25
(0.23)
0.18
(0.01)
0.12**
(0.00)
127: Chronic obstructive pulmonary disease and bronchiectasis 603
(11)
661
(30)
5.21
(0.04)
4.89
(0.34)
2.81
(0.07)
2.82
(0.17)
237: Complication of device, implant or graft 577
(18)
496**
(24)
5.69
(0.07)
6.04
(0.43)
2.00
(0.06)
1.99
(0.12)
109: Acute cerebrovascular disease 576
(11)
537
(25)
6.53
(0.07)
6.54
(0.46)
11.03
(0.17)
10.31
(0.65)
55: Fluid and electrolyte disorders 569
(10)
721**
(33)
4.02
(0.04)
3.81
(0.27)
2.84
(0.07)
2.06**
(0.12)
203: Osteoarthritis 501
(21)
496
(24)
4.26
(0.05)
4.42
(0.32)
0.15
(0.01)
0.07**
(0.00)
254: Rehabilitation care, fitting of prostheses, and adjustment of devices 482
(34)
461
(22)
12.52
(0.29)
12.47
(0.88)
0.66
(0.07)
0.99**
(0.06)
149: Biliary tract disease 464
(9)
472
(23)
4.08
(0.03)
4.02
(0.29)
0.84
(0.03)
0.45**
(0.02)
50: Diabetes mellitus with complications 461
(9)
456
(22)
5.58
(0.05)
5.39
(0.39)
1.38
(0.04)
1.62*
(0.10)
196: Normal pregnancy and/or delivery 453
(17)
3,851**
(159)
1.90
(0.01)
2.52**
(0.16)
0.00
(0.00)
0.01**
(0.00)
159: Urinary tract infections 445
(7)
499*
(24)
4.66
(0.05)
4.57
(0.33)
1.66
(0.05)
1.25**
(0.08)
181: Other complications of pregnancy 418
(13)
201**
(11)
2.46
(0.04)
2.66
(0.24)
0.02
(0.00)
0.09**
(0.00)
238: Complications of surgical procedures or medical care 414
(10)
405
(20)
6.17
(0.07)
6.18
(0.45)
1.70
(0.05)
2.19**
(0.14)
197: Skin and subcutaneous tissue infections 391
(7)
415
(20)
4.99
(0.04)
4.79
(0.35)
0.59
(0.02)
0.39**
(0.02)
128: Asthma 389
(16)
454*
(22)
3.29
(0.04)
3.20
(0.24)
0.34
(0.02)
0.18**
(0.01

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.
(a) Because of a limited sample, the NHDS estimate and standard error were unreliable and not reported.
(b) The NHDS estimate was reported but was not considered reliable; the standard error was not reported.
(c) A valid standard error could not be calculated.

Return to Contents

Table 16. NIS and NHDS Comparisons by Principal Procedure Category, 2001
Principal Procedure Number of Discharges in Thousands (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error)
NIS NHDS NIS NHDS NIS NHDS
137: Other procedures to assist delivery 1,341
(61)
931**
(41)
2.07
(0.01)
2.13
(0.15)
0.00
(0.00)
0.00
(0.00)
115: Circumcision 1,120
(42)
1,128
(49)
2.50
(0.02)
2.52
(0.17)
0.00
(0.00)
0.06**
(0.00)
134: Cesarean section 992
(35)
962
(43)
3.68
(0.03)
3.63
(0.25)
0.01
(0.00)
0.02
(0.00)
47: Diagnostic cardiac catheterization, coronary arteriography 719
(31)
567**
(27)
3.55
(0.05)
3.74
(0.27)
0.97
(0.03)
1.82**
(0.11)
45: Percutaneous transluminal coronary angioplasty (PTCA) 701
(46)
508**
(24)
2.81
(0.05)
2.79
(0.21)
0.85
(0.04)
0.74
(0.04)
70: Upper gastrointestinal endoscopy, biopsy 697
(17)
637
(29)
5.45
(0.08)
5.72
(0.40)
1.79
(0.05)
1.54*
(0.09)
140: Repair of current obstetric laceration 667
(32)
758
(34)
2.05
(0.01)
2.06
(0.15)
0.00
(0.00)
0.00
(0.00)
124: Hysterectomy, abdominal and vaginal 606
(17)
617
(29)
2.81
(0.02)
2.69
(0.19)
0.07
(0.00)
0.16**
(0.01)
216: Respiratory intubation and mechanical ventilation 542
(12)
520
(25)
11.16
(0.23)
11.80
(0.83)
29.99
(0.49)
28.95
(1.83)
133: Episiotomy 473
(23)
491
(23)
2.10
(0.01)
2.12
(0.16)
0.00
(0.00)
0.00
(0.00)
222: Blood transfusion 429
(15)
384
(19)
5.76
(0.05)
6.00
(0.44)
6.12
(0.13)
6.14
(0.40)
231: Other therapeutic procedures 419
(50)
430
(21)
5.17
(0.19)
4.83
(0.35)
2.30
(0.21)
3.08**
(0.20)
84: Cholecystectomy and common duct exploration 398
(8)
382
(19)
4.47
(0.04)
4.41
(0.33)
0.83
(0.03)
0.41**
(0.02)
219: Alcohol and drug rehabilitation/detoxification 379
(40)
309
(16)
5.89
(0.28)
6.24
(0.47)
0.12
(0.01)
0.11
(0.00)
228: Prophylactic vaccinations and inoculations 375
(52)
381
(19)
2.36
(0.05)
2.43
(0.19)
0.00
(0.00)
0.00
(0.00)
152: Arthroplasty knee 363
(14)
355
(18)
4.15
(0.04)
4.33
(0.33)
0.15
(0.01)
0.06**
(0.00)
54: Other vascular catheterization, not heart 345
(17)
342
(17)
9.61
(0.23)
11.52*
(0.85)
10.04
(0.42)
9.51
(0.63)
44: Coronary artery bypass graft (CABG) 344
(22)
264**
(14)
8.83
(0.11)
9.03
(0.69)
2.42
(0.08)
2.01*
(0.13)
153: Hip replacement, total and partial 329
(14)
314
(16)
5.36
(0.04)
5.41
(0.41)
1.19
(0.05)
1.08
(0.07)
3: Laminectomy, excision intervertebral disc 305
(13)
243**
(13)
2.72
(0.04)
2.63
(0.22)
0.14
(0.01)
0.20**
(0.01)
76: Colonoscopy and biopsy 297
(24)
267
(14)
5.19
(0.39)
5.82
(0.46)
1.08
(0.09)
1.48**
(0.10)
80: Appendectomy 282
(7)
289
(15)
3.00
(0.02)
3.09
(0.25)
0.10
(0.01)
0.03**
(0.00)
135: Forceps, vacuum, and breech delivery 278
(12)
266
(14)
2.26
(0.01)
2.39
(0.20)
0.00
(0.00)
0.02**
(0.00)
78: Colorectal resection 271
(7)
243
(13)
10.07
(0.06)
9.64
(0.75)
4.22
(0.11)
3.86
(0.27)
48: Insertion, revision, replacement, removal of cardiac pacemaker or cardioverter/defibrillator 267
(11)
214**
(12)
5.16
(0.07)
5.33
(0.44)
1.72
(0.08)
1.77
(0.12)

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.
(a) Because of a limited sample, the NHDS estimate and standard error were unreliable and not reported.
(b) The NHDS estimate was reported but was not considered reliable; the standard error was not reported.
(c) A valid standard error could not be calculated.

APPENDIX C: NIS-MEDPAR TABLES

Table 17. NIS and MedPAR Comparisons by Overall and by Region, 2001
  Number of Discharges in Thousands (Standard Error) Percentage of Discharges (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error) Average Total Hospital Charge (Standard Error)
NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR
U.S. 13,727
(266)
11,315** 100.00 100.00 5.85
(0.04)
5.90 4.37
(0.04)
4.33 $18,738
(363)
$18,507
Region
Northeast 2,823
(130)
2,150** 20.56
(0.82)
19.00 6.57
(0.13)
6.83 4.74
(0.12)
4.84 $21,026
(1,282)
$20,978
Midwest 3,325
(126)
3,087 24.22
(0.81)
27.28** 5.50
(0.08)
5.53 3.89
(0.07)
3.93 $16,008
(447)
$15,700
South 5,546
(166)
4,612** 40.40
(0.94)
40.76 5.80
(0.06)
5.84 4.37
(0.06)
4.37 $17,135
(374)
$17,270
West 2,032
(100)
1,465** 14.80
(0.68)
12.95** 5.61
(0.11)
5.53 4.61
(0.17)
4.28 $24,531
(1,119)
$24,689

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.

Table 18. NIS and MedPAR Comparisons by Control and Bed Size, 2001
Control / Bed Size Number of Discharges in Thousands (Standard Error) Percentage of Discharges (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error) Average Total Hospital Charge (Standard Error)
NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR
Total Public 1,769 (127) 1,439** 12.89
(0.93)
12.72 6.04
(0.13)
5.89 4.29
(0.11)
4.13 $23,525
(961)
$24,518
1-99 Beds 545
(28)
445** 32.27
(2.33)
30.90 4.49
(0.08)
4.48 4.22
(0.09)
3.93** $7,813
(239)
$7,978
100-199 Beds 350
(44)
334 20.72
(2.64)
23.19 5.44
(0.15)
5.61 4.43
(0.17)
4.49 $13,130
(576)
$13,413
200-299 Beds 210
(57)
179 12.48
(3.53)
12.46 6.36
(0.27)
6.31 4.53
(0.27)
4.55 $18,799
(2,403)
$16,569
300-499 Beds 360
(88)
238 21.31
(4.78)
16.56 6.51
(0.26)
6.59 4.72
(0.15)
4.47 $21,567
(1,416)
$22,825
500+ Beds 222
(87)
242 13.19
(4.87)
16.86 6.26
(0.66)
6.75 3.85
(0.47)
4.43 $20,766
(1,147)
$22,157
Total Private Non-Profit 10,268
(303)
8,434** 74.80
(1.39)
70.08** 5.87
(0.05)
5.94 4.38
(0.06)
4.36 $18,545
(452)
$18,052
1-99 Beds 1,246
(65)
990** 12.14
(0.68)
11.74 4.78
(0.14)
4.65 4.26
(0.20)
3.99 $9,934
(297)
$9,896
100-199 Beds 2,154
(130)
1,752** 20.98
(1.24)
20.77 5.65
(0.09)
5.63 4.37
(0.09)
4.34 $15,976
(812)
$14,553
200-299 Beds 1,952
(185)
1,661 19.00
(1.79)
19.69 6.00
(0.11)
5.94 4.45
(0.12)
4.38 $19,617
(904)
$17,499*
300-499 Beds 2,931
(265)
2,369* 28.54
(2.52)
28.08 5.96
(0.10)
6.18* 4.38
(0.12)
4.42 $21,090
(1,055)
$20,075
500+ Beds 1,984
(251)
1,660 19.32
(2.29)
19.69 6.52
(0.17)
6.69 4.40
(0.15)
4.51 $21,952
(1,291)
$24,277
Total Proprietary 1,769
(127)
1,439** 12.89
(0.93)
11.96 6.04
(0.13)
5.89 4.29
(0.11)
4.13 $23,525
(961)
$24,518
1-99 Beds 262
(25)
215 14.82
(1.47)
14.95 5.85
(0.49)
4.77* 4.11
(0.26)
3.49* $17,137
(1,886)
$14,325
100-199 Beds 651
(52)
535* 36.80
(2.58)
37.20 5.79
(0.17)
5.81 4.30
(0.15)
4.09 $22,045
(1,066)
$22,263
200-299 Beds 269
(60)
366 15.22
(3.47)
25.44** 6.54
(0.22)
6.12 4.68
(0.23)
4.34 $24,668
(2,313)
$28,496
300-499 Beds 422
(83)
208* 23.86
(4.36)
14.45* 6.05
(0.23)
6.42 3.98
(0.19)
4.35 $28,472
(3,108)
$32,109
500+ Beds 164
(42)
114 9.28
(2.42)
7.94 6.50
(0.29)
6.68 4.67
(0.27)
4.48 $25,002
(887)
$27,731**

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.

Return to Contents

Table 19. NIS and MedPAR Comparisons by Location, Teaching Status, and Size, 2001
Hospital Type / Size Number of Discharges in Thousands (Standard Error) Percentage of Discharges (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error) Average Total Hospital Charge (Standard Error)
NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR
Rural 2,702
(93)
2,373** 19.68
(0.65)
20.97* 4.92
(0.07)
5.00 4.20
(0.06)
4.08 $10,446
(307)
$10,677
1-49 beds 651
(27)
484** 24.09
(1.26)
20.41** 4.31
(0.21)
4.14 3.93
(0.10)
3.76 $7,169
(167)
$7,338
50-99 beds 809
(64)
668* 29.96
(2.61)
28.15 4.65
(0.07)
4.70 4.24
(0.10)
4.00* $9,925
(329)
$9,732
100+ beds 1,241
(111)
1,220 45.93
(2.94)
51.42 5.41
(0.10)
5.50 4.30
(0.10)
4.25 $12,503
(553)
$12,521
 
Urban, Non-Teaching 5,923
(170)
4,755** 43.14
(0.95)
42.02 5.91
(0.07)
5.90 4.37
(0.08)
4.37 $19,864 (569) $19,291
1-99 beds 584
(41)
475** 9.86
(0.72)
9.99 5.69
(0.28)
4.95** 4.56
(0.42)
3.98 $13,932
(911)
$12,689
100-199 beds 1,836
(92)
1,571* 31.00
(1.61)
33.03 5.87
(0.10)
5.84 4.49
(0.10)
4.39 $18,829
(868)
$17,692
200+ beds 3,502
(168)
2,709** 59.12
(1.65)
56.96 5.96
(0.09)
6.09 4.27
(0.10)
4.43 $21,395
(849)
$21,378
 
Urban, Teaching 5,102
(182)
4,186** 37.16
(0.98)
36.99 6.29
(0.08)
6.43 4.46
(0.08)
4.42 $21,876
(708)
$22,054
1-299 beds 699
(99)
556 13.70
(1.95)
13.28 6.20
(0.16)
6.02 4.47
(0.20)
4.31 $22,356
(2,066)
$19,470
300-499 Beds 1,477
(188)
1,159 28.95
(3.63)
27.70 6.15
(0.16)
6.35 4.49
(0.18)
4.36 $22,224
(1,469)
$21,210
500+ Beds 2,925
(227)
2,470* 57.33
(3.67)
59.01 6.39
(0.13)
6.55 4.44
(0.11)
4.47 $21,596
(897)
$23,032

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.

Table 20. NIS and MedPAR Comparisons by Patient Characteristics, 2001
  Number of Discharges in Thousands (Standard Error) Percentage of Discharges (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error) Average Total Hospital Charge (Standard Error)
NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR
Race
White 8,319
(306)
10,095** 60.60
(1.64)
83.87** 5.86
(0.05)
5.78 4.45
(0.06)
4.35 $18,931
(436)
$18,263
Black 1,059
(77)
1,394** 7.72
(0.56)
11.59** 6.95
(0.12)
6.80 4.44
(0.08)
4.33 $20,930
(1,162)
$20,120
Other 888
(78)
491** 6.47
(0.57)
4.08** 6.52
(0.12)
6.33 4.47
(0.10)
3.80** $23,934
(762)
$23,536
Unknown 3,459
(242)
53** 25.20
(1.77)
0.44** 5.33
(0.07)
5.85** 4.11
(0.07)
3.88** $16,287
(601)
$17,860**
Age Group
0-64 Years 1,939
(43)
1,858 14.13
(0.28)
15.44** 6.16
(0.06)
6.23 2.21
(0.04)
2.19 $18,396
(402)
$18,623
65-74 Years 4,256
(92)
3,651** 31.00
(0.19)
30.34** 5.54
(0.04)
5.62 3.35
(0.04)
3.44* $20,117
(396)
$20,058
75-84 Years 5,019
(110)
4,241** 36.56
(0.19)
35.23** 5.92
(0.04)
5.96 4.70
(0.05)
4.55** $19,044
(371)
$18,984
85+ Years 2,511
(55)
2,284** 18.29
(0.20)
18.97** 6.03
(0.06)
6.07 7.09
(0.08)
7.06 $16,059
(388)
$16,018
Gender
Female 7,840
(148)
6,845** 57.11
(0.16)
56.87 5.87
(0.04)
5.91 4.06
(0.04)
4.04 $17,754
(349)
$17,635
Male 5,887
(121)
5,189** 42.88
(0.16)
43.12 5.84
(0.04)
5.93 4.77
(0.05)
4.70 $20,050
(392)
$20,084

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1 A significance test was not performed because a valid standard error was not available.

Table 21. NIS and MedPAR Comparisons by DRG, 2001
DRG Number of Discharges in Thousands (Standard Error) Percentage of Discharges (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error) Average Total Hospital Charge (Standard Error)
NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR
127: Heart Failure & Shock 782
(16)
632** 5.70
(0.06)
5.59 5.19
(0.03)
5.24 4.53
(0.07)
4.67 $13,252
(339)
$12,863
89: Simple Pneumonia & Pleurisy Age >17 w/ CC 565
(10)
475** 4.12
(0.05)
4.19 5.75
(0.04)
5.81 6.10
(0.10)
6.07 $13,369
(265)
$13,011
88: Chronic Obstructive Pulmonary Disease 433
(8)
368** 3.15
(0.04)
3.25* 5.07
(0.04)
5.04 2.12
(0.06)
1.97* $11,968
(286)
$11,254*
209: Major Joint & Limb Reattachment Procedures Of Lower Extremity 418
(15)
356** 3.04
(0.09)
3.14 4.93
(0.04)
4.95 0.86
(0.04)
0.87 $26,237
(483)
$25,443
116: Oth Perm Card Pacemak Impl Or Ptca w/ Coronary Artery Stent Implnt 364
(20)
285** 2.65
(0.11)
2.51 3.51
(0.06)
3.60 0.92
(0.04)
0.92 $29,354
(718)
$29,107
14: Specific Cerebrovascular Disorders Except TIA 363
(7)
297** 2.65
(0.02)
2.62 5.78
(0.05)
5.81 10.84
(0.18)
10.99 $15,493
(331)
$15,253
462: Rehabilitation 341
(24)
264** 2.48
(0.17)
2.33 12.04
(0.24)
12.36 0.79
(0.08)
0.29** $18,759
(815)
$19,409
430: Psychoses 336
(18)
299* 2.44
(0.13)
2.64 10.80
(0.25)
10.97 0.14
(0.01)
0.13 $14,312
(582)
$13,789
143: Chest Pain 294
(8)
235** 2.14
(0.04)
2.07 2.01
(0.01)
2.08** 0.10
(0.01)
0.13* $7,080
(153)
$6,845
182: Esophagitis 290
(6)
245** 2.11
(0.02)
2.16 4.30
(0.03)
4.34 1.33
(0.05)
1.35 $10,605
(231)
$10,134*
296: Nutritional & Misc Metabolic Disorders Age >17 w/ CC 287
(5)
241** 2.09
(0.02)
2.13 4.99
(0.04)
5.07 4.56
(0.11)
4.59 $11,157
(260)
$10,871
174: G.I. Hemorrhage w/ CC 286
(6)
233** 2.08
(0.02)
2.06 4.70
(0.03)
4.77* 3.46
(0.08)
3.52 $13,074
(285)
$12,693
138: Cardiac Arrhythmia & Conduction Disorders w/ CC 233
(5)
191** 1.69
(0.01)
1.69 3.92
(0.03)
3.96 2.89
(0.09)
2.87 $10,757
(277)
$10,404
416: Septicemia Age >17 212
(5)
170** 1.54
(0.03)
1.50 7.31
(0.07)
7.39 19.67
(0.30)
20.03 $21,051
(766)
$20,621
320: Kidney & Urinary Tract Infections Age >17 w/ CC 208
(4)
179** 1.52
(0.02)
1.58** 5.22
(0.05)
5.26 2.77
(0.09)
2.89 $11,379
(239)
$11,065
79: Respiratory Infections & Inflammations Age >17 w/ CC 191
(5)
154** 1.39
(0.03)
1.36 8.48
(0.09)
8.44 15.63
(0.25)
15.19 $21,242
(687)
$20,731
121: Circulatory Disorders W Ami & Major Comp 190
(5)
153** 1.38
(0.02)
1.35 6.15
(0.05)
6.27* 0.00
(0.00)
0.001 $19,834
(492)
$19,373
132: Atherosclerosis w/ CC 172
(5)
140** 1.25
(0.03)
1.23 2.87
(0.03)
2.89 0.76
(0.05)
0.78 $8,387
(278)
$8,110
15: Transient Ischemic Attack & Precerebral Occlusions 171
(4)
140** 1.25
(0.02)
1.23 3.35
(0.03)
3.44** 0.49
(0.04)
0.49 $9,808
(240)
$9,523
124: Circulatory Disorders Except Ami 155
(7)
126** 1.13
(0.04)
1.12 4.23
(0.06)
4.32 1.02
(0.06)
0.99 $18,187
(496)
$18,318
148: Major Small & Large Bowel Procedures w/ CC 148
(3)
123** 1.08
(0.01)
1.09 12.02
(0.08)
12.27** 8.23
(0.18)
8.38 $44,914
(943)
$43,984
210: Hip & Femur Procedures Except Major Joint Age >17 w/ CC 134
(3)
114** 0.98
(0.01)
1.01* 6.76
(0.06)
6.85 2.93
(0.11)
3.12 $23,701
(460)
$22,790*
316: Renal Failure 132
(3)
109** 0.96
(0.01)
0.97 6.58
(0.05)
6.59 10.72
(0.27)
10.45 $17,933
(556)
$17,448
478: Other Vascular Procedures w/ CC 121
(4)
100** 0.88
(0.02)
0.89 7.24
(0.11)
7.36 3.26
(0.12)
3.39 $31,651
(752)
$31,477
141: Syncope & Collapse w/ CC 119
(3)
97** 0.87
(0.01)
0.86 3.50
(0.03)
3.56 0.49
(0.04)
0.52 $9,638
(273)
$9,393

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.

Return to Contents

Table 22. NIS and MedPAR Comparisons by Principal Diagnosis, 2001
Principal Diagnosis Number of Discharges in Thousands (Standard Error) Percentage of Discharges (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error) Average Total Hospital Charge (Standard Error)
NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR
108: Congestive heart failure, nonhypertensive 800
(17)
647** 5.83
(0.05)
5.72 5.62
(0.04)
5.70 4.96
(0.08)
5.13* $16,193
(411)
$15,996
101: Coronary atherosclerosis and other heart disease 774
(33)
607** 5.64
(0.18)
5.36 3.96
(0.06)
4.01 1.00
(0.03)
1.06 $25,218
(729)
$25,539
122: Pneumonia (except that caused by tuberculosis or sexually transmitted disease) 730
(13)
612** 5.32
(0.06)
5.41 6.51
(0.05)
6.59 7.86
(0.10)
7.80 $17,019
(346)
$16,740
106: Cardiac dysrhythmias 468
(12)
378** 3.41
(0.04)
3.34 3.87
(0.03)
3.92 1.46
(0.04)
1.52 $16,871
(419)
$16,340
100: Acute myocardial infarction 455
(14)
365** 3.31
(0.06)
3.23 6.05
(0.07)
6.04 11.01
(0.15)
11.23 $30,228
(731)
$29,437
127: Chronic obstructive pulmonary disease and bronchiectasis 422
(8)
361** 3.07
(0.04)
3.19* 5.43
(0.05)
5.40 3.19
(0.08)
2.97** $13,924
(330)
$13,227*
109: Acute cerebrovascular disease 395
(8)
324** 2.88
(0.03)
2.86 6.32
(0.06)
6.41 11.18
(0.19)
11.30 $18,564
(422)
$18,620
254: Rehabilitation care, fitting of prostheses, and adjustment of devices 346
(25)
268** 2.52
(0.17)
2.37 12.17
(0.24)
12.47 0.79
(0.08)
0.30** $19,127
(830)
$19,738
102: Nonspecific chest pain 339
(10)
275** 2.47
(0.05)
2.43 2.12
(0.02)
2.18** 0.10
(0.01)
0.13** $8,016
(156)
$7,855
237: Complication of device, implant or graft 328
(10)
278** 2.39
(0.05)
2.46 5.81
(0.07)
5.81 2.45
(0.07)
2.34 $26,740
(643)
$26,736
55: Fluid and electrolyte disorders 321
(6)
268** 2.34
(0.03)
2.37 4.80
(0.04)
4.90* 3.98
(0.10)
4.02 $10,988
(270)
$10,749
203: Osteoarthritis 308
(13)
256** 2.24
(0.08)
2.26 4.40
(0.07)
4.31 0.20
(0.01)
0.20 $24,194
(461)
$23,904
226: Fracture of neck of femur (hip) 257
(6)
213** 1.87
(0.02)
1.88 6.56
(0.07)
6.45 3.23
(0.09)
3.41* $22,661
(421)
$22,012
159: Urinary tract infections 257
(5)
219** 1.87
(0.02)
1.93* 5.27
(0.05)
5.30 2.49
(0.08)
2.61 $11,808
(250)
$11,548
2: Septicemia (except in labor) 235
(6)
188** 1.71
(0.03)
1.66 8.43
(0.09)
8.56 19.17
(0.29)
19.57 $25,991
(847)
$25,861
153: Gastrointestinal hemorrhage 220
(4)
181** 1.60
(0.01)
1.60 4.98
(0.03)
5.09** 4.75
(0.11)
4.75 $15,020
(320)
$14,751
50: Diabetes mellitus with complications 205
(4)
175** 1.49
(0.02)
1.55* 6.54
(0.07)
6.51 2.18
(0.08)
2.28 $17,985
(473)
$17,933
205: Spondylosis, intervertebral disc disorders, other back problems 205
(6)
178** 1.49
(0.03)
1.57* 4.15
(0.05)
4.05 0.41
(0.03)
0.39 $17,284
(427)
$17,036
69: Affective disorders 183
(10)
164 1.33
(0.07)
1.45 10.19
(0.25)
10.34 0.16
(0.02)
0.16 $13,912
(545)
$13,382
238: Complications of surgical procedures or medical care 182
(4)
149** 1.32
(0.02)
1.32 7.02
(0.08)
6.96 2.69
(0.09)
2.79 $21,025
(505)
$21,224
149: Biliary tract disease 174
(4)
146** 1.27
(0.01)
1.29 5.27
(0.05)
5.35 1.80
(0.07)
1.76 $20,258
(376)
$19,705
145: Intestinal obstruction without hernia 166
(3)
140** 1.21
(0.01)
1.23 6.76
(0.05)
6.92** 4.50
(0.12)
4.73 $19,050
(398)
$19,090
245: Syncope 161
(4)
129** 1.17
(0.02)
1.14 3.16
(0.03)
3.24* 0.36
(0.03)
0.40 $9,799
(256)
$9,692
146: Diverticulosis and diverticulitis 160
(3)
129** 1.17
(0.01)
1.14 5.68
(0.04)
5.83** 1.92
(0.08)
1.97 $17,162
(347)
$17,033
197: Skin and subcutaneous tissue infections 160
(3)
132** 1.17
(0.01)
1.16 5.87
(0.06)
5.84 1.07
(0.05)
1.04 $12,504
(365)
$12,152

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.

Table 23. NIS and MedPAR Comparisons by Principal Procedure, 2001
Principal Procedure Number of Discharges in Thousands (Standard Error) Percentage of Discharges (Standard Error) Average Length of Stay in Days (Standard Error) In-Hospital Mortality Rate Percent (Standard Error) Average Total Hospital Charge (Standard Error)
NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR NIS MedPAR
70: Upper gastrointestinal endoscopy, biopsy 402
(10)
334** 2.93
(0.04)
2.95 6.01
(0.05)
6.13* 2.29
(0.06)
2.38 $16,559
(392)
$16,201
47: Diagnostic cardiac catheterization, coronary arteriography 362
(17)
294** 2.63
(0.09)
2.60 4.09
(0.05)
4.18 1.43
(0.05)
1.42 $18,733
(496)
$18,640
45: Percutaneous transluminal coronary angioplasty (PTCA) 355
(24)
275** 2.58
(0.15)
2.43 3.13
(0.06)
3.21 1.28
(0.06)
1.28 $29,150
(888)
$29,393
222: Blood transfusion 270
(10)
209** 1.97
(0.07)
1.85 5.92
(0.06)
6.03 6.88
(0.16)
7.08 $16,048
(411)
$15,621
216: Respiratory intubation and mechanical ventilation 246
(5)
208** 1.79
(0.03)
1.84 9.49
(0.18)
9.29 42.10
(0.39)
41.47 $41,446
(1,002)
$39,711
153: Hip replacement, total and partial 223
(8)
188** 1.63
(0.05)
1.66 5.64
(0.05)
5.69 1.55
(0.07)
1.62 $27,953
(536)
$27,044
152: Arthroplasty knee 211
(8)
183** 1.54
(0.05)
1.61 4.35
(0.05)
4.35 0.21
(0.02)
0.19 $25,524
(497)
$24,965
48: Insertion, revision, replacement, removal of cardiac pacemaker or cardioverter/defibrillator 206
(9)
165** 1.50
(0.05)
1.46 5.22
(0.08)
5.32 1.86
(0.08)
1.84 $38,692
(974)
$37,676
146: Treatment, fracture or dislocation of hip and femur 183
(4)
156** 1.33
(0.02)
1.38* 6.28
(0.05)
6.38 2.39
(0.08)
2.56* $21,862
(426)
$21,072
44: Coronary artery bypass graft (CABG) 179
(12)
139** 1.31
(0.07)
1.22 9.73
(0.12)
9.66 3.37
(0.13)
3.51 $63,383
(1,980)
$65,181
76: Colonoscopy and biopsy 173
(7)
139** 1.26
(0.04)
1.23 5.84
(0.17)
6.13 1.48
(0.08)
1.56 $15,091
(616)
$15,180
58: Hemodialysis 169
(5)
150** 1.23
(0.03)
1.33* 5.60
(0.07)
5.43* 4.45
(0.15)
4.17 $16,402
(424)
$15,598
54: Other vascular catheterization, not heart 163
(9)
143* 1.19
(0.06)
1.26 9.50
(0.22)
9.47 15.14
(0.63)
15.15 $27,252
(1,253)
$26,703
78: Colorectal resection 145
(3)
119** 1.05
(0.01)
1.05 11.00
(0.07)
11.23** 6.36
(0.17)
6.54 $40,856
(854)
$40,016
84: Cholecystectomy and common duct exploration 141
(3)
120** 1.02
(0.01)
1.06* 5.95
(0.06)
6.05 1.85
(0.08)
1.91 $23,920
(447)
$23,263
61: Other OR procedures on vessels other than head and neck 136
(5)
115** 0.99
(0.02)
1.02 6.95
(0.13)
7.19 4.43
(0.15)
4.71 $33,176
(1,034)
$33,274
213: Physical therapy exercises, manipulation, and other procedures 129
(16)
93* 0.94
(0.11)
0.82 11.92
(0.50)
11.27 0.86
(0.11)
0.52** $20,414
(1,599)
$18,992
231: Other therapeutic procedures 126
(16)
108 0.92
(0.12)
0.96 5.54
(0.25)
5.49 5.48
(0.33)
5.66 $14,683
(685)
$14,200
193: Diagnostic ultrasound of heart (echocardiogram) 122
(11)
104 0.89
(0.08)
0.92 5.51
(0.10)
5.65 2.65
(0.14)
2.83 $15,768
(597)
$14,868
39: Incision of pleura, thoracentesis, chest drainage 103
(2)
85** 0.75
(0.01)
0.75 8.14
(0.07)
8.34** 8.62
(0.24)
8.86 $21,655
(501)
$21,348
51: Endarterectomy, vessel of head and neck 103
(4)
84** 0.75
(0.02)
0.75 2.93
(0.05)
3.08** 0.47
(0.04)
0.49 $16,945
(415)
$17,274
169: Debridement of wound, infection or burn 101
(2)
86** 0.73
(0.01)
0.76 11.64
(0.20)
11.39 4.79
(0.19)
4.90 $31,035
(857)
$30,339
177: Computerized axial tomography (CT) scan head 94
(11)
72 0.68
(0.08)
0.64 4.97
(0.14)
5.34** 4.41
(0.22)
4.78 $12,878
(942)
$13,631
113: Transurethral resection of prostate (TURP) 91
(3)
73** 0.66
(0.01)
0.64 3.34
(0.06)
3.44 0.34
(0.04)
0.40 $11,617
(304)
$11,121
3: Laminectomy, excision intervertebral disc 90 (4) 80* 0.65 (0.02) 0.71* 3.68 (0.07) 3.63 0.32 (0.04) 0.33 $16,922 (482) $16,023

*Significant at a 5 percent level.
**Significant at a 1 percent level.
1A significance test was not performed because a valid standard error was not available.

Return to Contents

APPENDIX D: ESTIMATES OF STANDARD ERROR FOR NHDS STATISTICS

Estimates of Standard Error for NHDS Statistics

A variety of statistics were estimated based on these NHDS data:

  1. Total number of discharges


  2. In-Hospital mortality


  3. Average length of stay (calculated as the difference between discharge and admission dates).

The standard errors were calculated as follows:

Total Numbers of Discharges

From the NHDS Documentation (National Center for Health Statistics, 2002), constants a and b were obtained for 2000. The relative standard error for the estimate of total discharges was approximated by:

Scientific equation

where WTD was the weighted sum of total discharges (i.e., the estimate of total discharges).

The standard error was then calculated as:

Scientific equation

Percent Mortality

Let p be the estimated proportion of in-hospital deaths (with the number of deaths estimated as the numerator and the discharge estimate as the denominator). The relative standard error of this proportion expressed as a percent was approximated by:

Scientific equation

The standard error was then calculated as:

Scientific equation

Where b was the parameter in the formula for approximated RSE(WTD) given by the NHDS documentation (i.e., the same used in the formula for calculating the standard error for number of discharges).

Average Length of Stay

Let average length of stay be the estimated average length of stay based on a weighted number of discharges equal to TD. If the weighted sum of patient length of stay was TLOS, and

Scientific equation

then the relative standard error is:

Scientific equation

The estimate of the relative standard error was valid only if:

  1. The relative standard error of the denominator (estimated discharges) was smaller than five percent.

    -or-

  2. Both the relative standard error of the numerator (estimated total stay days) and the denominator (estimated discharges) were smaller than ten percent.

For all parameter estimates, when values of a and b were available in the NHDS documentation (i.e., for procedures, gender, region, race, and diagnoses), the appropriate values for a and b were used. When a variable represented the sum of more than one NHDS category, as recommended by Korn and Graubard (1999, p.224), the standard error for each category was calculated, and the largest of these standard errors was reported and used in significance testing. For example, the NIS category of "private insurance" includes three NHDS categories: 1) Blue Cross/Blue Shield, 2) HMO/PPO, and 3) other private insurance. The standard error was calculated for all three categories, using the values of a and b provided in the NHDS documentation, and the largest value was used in computing the t-value to test for significant difference.

When no parameter estimates were available, the values of a and b for the total sample were used in calculating the standard errors. For example, in the hospital control X bed size comparisons, the values for the total sample were used in calculating standard errors, because the NHDS documentation provides parameter estimates by neither ownership nor bed size.

Tests of Statistical Significance

To test for a statistically significant difference between a NIS estimate, X, and a NHDS estimate, Y, the following procedure was used. The difference was significant if

Scientific equation

where SEx was the estimated standard error for the NIS estimate and SEy was the estimated standard error of the NHDS estimate.

Return to Contents


Internet Citation: HCUP NIS 2001 NIS Comparison Report. Healthcare Cost and Utilization Project (HCUP). November 2015. Agency for Healthcare Research and Quality, Rockville, MD. www.hcup-us.ahrq.gov/db/nation/nis/reports/2001niscomparisonrpt.jsp.
Are you having problems viewing or printing pages on this website?
If you have comments, suggestions, and/or questions, please contact hcup@ahrq.gov.
Privacy Notice, Viewers & Players
Last modified 11/2/15