Please refer to the information below for corrections to the NIS data set. In some cases, there is a downloadable document that either provides more information or provides a program to correct the problem.

Contents:

• Correction for New Jersey Diagnoses
• Correction to 1993 Weights file
• Correction to SPSS Label Program
• Correction to Instruction on How to Calculate Variances (Standard Errors)
• Trend Weights for 1993-2011 HCUP NIS Data

Correction for New Jersey Diagnoses

Background

A programming error caused a problem with some of the diagnosis codes for the discharge abstracts from New Jersey hospitals in the 1993 Nationwide Inpatient Sample (NIS), Release 2. Some of the hospitals' data have three- or four-digit diagnoses with trailing zeros that render otherwise valid codes invalid. Therefore, we have constructed a program to fix the principal diagnosis on the NIS Core A file and the complete diagnosis vector on the diagnosis file (DX).

Downloading Information

There are two files you will need to download. The first is a self-extracting zipped file NJ93_NJDX.EXE (4.4 MB). When unzipped, it contains two SAS datasets: N93NJDXS.sd2, which contains the corrected secondary diagnoses for New Jersey hospital discharges (14.3 MB), and N93NJDX1.sd2, which contains the corrected principal diagnoses for New Jersey hospital discharges (3.6 MB).
The second file, a self-extracting zipped SAS program, will repair diagnoses for the 1993 NIS, Release 2, using the files N93NJDXS.sd2 and N93NJDX1.sd2.

Shift-click from this Web page:

Your browser may support loading the files from this Web page. To download the files from this Web page, hold down the "Shift" key on your keyboard while you click the left mouse button on the following links:

• Corrected Diagnoses for New Jersey Hospital Discharges (exe file, 4.4 MB)
• SAS Repair Program for 1993 NIS, Release 2 (exe file, 5 KB)

To expand the self-extracting files:

Use Windows® Explorer (Windows® 95) or File Manager (Windows® 3.x) to find the files (nj93_njdx.exe and fixnis93.exe), then double click on them. The inflated files will be in a new subfolder.

Repairing 1993 NIS Release 2

You will need to change the program's macro variables to direct the program to your input/output files:

• DIR_ORIG = subdirectory of original NIS files.
• DIR_CORR = subdirectory of the two NJ files with corrected diagnosis data to be used to repair NIS, N93NJDX1.SD2 and N93NJDXS.SD2.
• DIR_FIX = subdirectory of corrected NIS files, NIS93_A.SD2 and NIS93_DX.SD2 (OUTPUT).

If you have limited disc space, the program can be split to separately correct the NIS Core A and DX files. Once the program has been completed, the 1993 NIS, Release 2 will be repaired.

There were no problems with the diagnoses vectors from other states. The validity flags for the diagnoses (DXVn), Clinical Classifications for Health Policy Research (CCHPRs), and diagnosis-related groups (DRGs) for all States (including New Jersey) are correct.

If you have questions or problems with these files, please E-mail your questions to hcup@ahrq.gov.

MEDSTAT discovered the following errors in the weight file for the NIS, Release 2 (1993 data).

1. Discharge weights to the universe.
   In calculating discharge weights to the universe (DISCWT_U), slightly incorrect sample counts were used for the Pennsylvania data in the NIS. During HCUP-3 processing of the Pennsylvania data, 208 of 408,318 hospital discharges were discarded from the sample but not discarded from the weight calculations. Consequently, the total number of U.S. discharges is underestimated by about 2,900 (.008 %). The effect is negligible on regional- and U.S.-level analyses with the data (including the Comparison Report).
2. Discharge and hospital weights to the state.
   Four Arizona hospitals and seven Washington hospitals were omitted from both the sample and the universe in the calculation of state-level hospital weights (HOSPWT_S) and discharge weights (DISCHWT_S). In addition to the obvious undercount of hospitals for these two states, these omissions led to an undercount of 21,090 discharges in Arizona (4.5 %), and 135,152 discharges in Washington(24.0 %). Finally, the overcount of sample discharges in Pennsylvania (item 1, above) also led to a negligible undercount of 3,404 discharges in Pennsylvania (0.2 %).
3. Stratification variables.
   We misinterpreted security restrictions for South Carolina. Consequently, for two South Carolina hospitals the weight file incorrectly contains missing values for the NIS stratification variables which describe hospital ownership, bedsize, teaching status, and urban/rural location (H_CONTROL, H_BEDSZ, H_TCH, H_LOC, and H_LOCTCH).

We have an updated version of the weight file that corrects the above errors. In addition, this new file increases the precision of the weights from four digits to five digits. Please contact the HCUP Central Distributor at hcup@ahrq.gov for the updated file.

Included on the Documentation CD for the 1988-1997 NIS is a program (VLABEL) to label the data elements. This program contains errors and should not be used. Value labels in SPSS can be entered into the data interactively via the data editor or entered and contained in pivot tables.

Click to view the Correction to Instruction on How to Calculate Variances (Standard Errors) (PDF file, 76 KB)

Click to view Trend Weights for 1993-2011 HCUP NIS Data