Beginning in 2008, Arizona reports six "cause of injury" E-codes in separate fields. From 2004-2007 Arizona reports five "cause of injury" E-codes in separate variables.

Arizona reports some diagnosis codes with an explicit decimal point. The decimal point was removed during HCUP processing.

Beginning with 1995 discharges, Arizona reports two "cause of injury" E-codes in separate variables. During processing of the 1995 to 2002 data, these E-codes are placed after the last non-missing diagnosis code if they are not already recorded as a secondary diagnosis. Beginning in 2003, the separately reported E-codes, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

HIV Test Result Diagnoses

California law prohibits the release of HIV test results in patient-identifiable form to any outside party without the patient's consent. Therefore, records that include certain ICD-9-CM codes that indicate HIV test results were not included in the data supplied for HCUP. California eliminated all occurrences of these codes from the diagnosis fields and packed the diagnosis vectors to cover gaps from such removals.

The following ICD-9-CM codes were affected:

• From January 1988 to October 1, 1994, diagnosis codes of 044.x or 795.8 were removed by the data source prior to submitting data to HCUP.
• Beginning October 1, 1994, diagnosis codes of 795.71 or V08 were removed by the data source prior to submitting data to HCUP. These ICD-9-CM codes replaced the earlier codes.

HIV-related diagnoses 042.x and 043.x were unaffected.

The number of such diagnoses eliminated from the principal diagnosis position will be smaller than it otherwise might have been due to a practice in California that actively discourages the reporting of codes for HIV test results (044.x, 795.8, 795.71, and V08) as a principal diagnosis. During data editing, California flags discharges reporting one of these codes in the principal diagnosis position and then calls the submitting hospital to ask if the principal diagnosis should be changed. Hospitals have the option of deleting the code, changing it, or leaving it in place.

Missing Total Charges

Some hospitals do not report total charges.

Psychiatric Diagnoses

Prior to 1995, some hospitals reported psychiatric diagnoses in DSM III which California then converted into ICD-9-CM diagnosis codes. The ICD-9-CM diagnosis codes are included in the HCUP database.

From 1995-1998, some psychiatric hospitals began submitting data for primary diagnosis according to DSM IV criteria. DSM IV codes are indistinguishable in appearance from ICD-9-CM codes but have substantially different meanings. Because of similarities in the coding structure, the source was unable to convert the DSM IV codes to ICD-9-CM codes. DSM IV codes may occur in the HCUP data. Psychiatric hospitals may be included in the California data; no documentation was available on the use of DSM IV codes in psychiatric units of acute care hospitals.

Beginning in 1999, DSM psychiatric codes are not accepted by OSHPD and are not present in the HCUP databases.

E-Codes

Beginning with 1990 discharges, the source reports five "cause of injury" E-codes as separate variables. During processing of the 1990 to 2002 data, E-codes were placed after the last non-missing diagnosis code. Beginning in 2003, the five separately reported E-codes, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

Prior to data year 2010, California did not require the reporting of E-codes in the range E870-E879 (misadventures and abnormal reactions).

District of Columbia does not report "cause of injury" E-codes separately. During HCUP processing, any E-code encountered in the diagnosis variables are placed in a separate array specific to E codes (ECODEn).

Hawaii reports one "cause of injury" E code as a separate data element. Prior to 2003, During HCUP processing, this E code was placed after the last non-missing diagnosis code during HCUP processing. Beginning with 2003 HCUP processing, this E code is placed at the beginning of the separate HCUP E-code array (ECODEn).

For 2013-2017, we retained only 50 DX fields. This truncated data for less than 25 records.

Beginning in data year 2001, the Iowa Hospital Association prohibits the release of two types of discharges:

• HIV Infections (defined by MDC of 25) and
• Behavioral Health including chemical dependency care or psychiatric care (defined by a service code of BHV).

These discharges were not included in the source file provided to HCUP and are therefore not included in the HCUP files.

Beginning in 1994, Iowa reports one "cause of injury" E-code. Beginning in 1998, Iowa added one "place of injury" E-code. During processing of the 1994 to 2002 data, these separately reported E-code variables are placed at the end of the diagnosis vector; since the vector is packed during processing to remove blanks, the position of the E-code for a specific discharge depends on the number of diagnoses reported. Beginning in 2003, the two separately reported E-codes, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

Prior to 2002, Kentucky reports one "cause of injury" E-codes as a separate variable. Beginning in June 2002, Kentucky reports two separate E-code fields. Prior to 2003, during HCUP processing, these separately reported E-codes were placed after the last non-missing secondary diagnosis. Beginning in 2003, the two separately reported E-codes, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn). Beginning 2008, a third E-code was provided and placed into the E-cdoe arrays.

Kentucky supplied diagnosis codes in a field length of 7. Only the first five characters contained in the left-justified source field were used to assign the HCUP diagnosis codes.

Maryland reports one "cause of injury" E-code as a separate variable. Prior to 2003, during HCUP processing, this separately reported E-code was placed after the last non-missing secondary diagnosis. Beginning in 2003, the separately reported E-code, and any E-codes encountered in the diagnosis variables are placed in a separate array specific to E codes (ECODEn).

Maryland supplied diagnosis codes in a field of length 7. Only the first five characters contained in the left-justified source field were used to assign the HCUP diagnosis codes.

The last secondary diagnosis field on the source data was 9-filled instead of blank when no diagnosis was coded. During HCUP processing, the 9-filled diagnosis was set to blank.

Beginning in 1993, Massachusetts reported one "cause of injury" E-code. Prior to 2003, during HCUP processing, the separately reported E-code was placed after the last non-missing secondary diagnosis. E-codes can appear in other secondary diagnosis codes. Beginning in 2003, the separately reported E-code, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

Minnesota reports one "cause of injury" E-codes in a separate variable. Prior to 2003, during HCUP processing, this E-code was placed after the last non-missing diagnosis code. Beginning in 2003, the separately reported E-codes, and any E-code encountered in the diagnosis variables are placed in a separate array specific to E codes (ECODEn).

The Nebraska Hospital Association prohibits the release of discharge records for patients with HIV diagnoses. These discharges were not included in the source file provided to HCUP and are therefore not included in the HCUP files.

Nebraska reports one "cause of injury" E-code in a separate variable. Prior to 2003, during HCUP processing, this E-code was placed after the last non-missing diagnosis code. Beginning in 2003, the separately reported E-code, and any E-code encountered in the diagnosis variables are placed in a separate array specific to E codes (ECODEn).

Beginning with 1993 discharges, New Jersey reports one "cause of injury" E-codes as a separate variable. Prior to 2003, during HCUP processing, this E-code was placed after the last non-missing diagnosis code. Beginning in 2003, the separately reported E-code, and any E-codes encountered in the diagnosis variables are placed in a separate array specific to E codes (ECODEn).

Before 1994, the diagnosis codes provided by the state were right-padded with zeros (e.g., the diagnosis code '436' was supplied as '43600'). For the HCUP database the following algorithm was used to validate the diagnosis codes:

Check the five-digit code for validity (using a six-month window for coding changes, 3 months before and 3 months after October of each year when ICD-9-CM coding changes occur).

1. If the five-digit code is valid, set DXn to the five-digit code and set DXVn = 0.
2. If the five-digit code is invalid and the fifth digit is a zero, create a four-digit code by deleting the trailing zero and re-check for validity (using six-month window for coding changes). If the four-digit code is valid, set DXn to the four-digit code and set DXVn = 0.
3. If the four-digit code is invalid and the fourth digit is a zero, create a three-digit code by deleting the trailing zero and re-check for validity (using six-month window for coding changes). If the three-digit code is valid, set DXn to the three-digit code and set DXVn = 0.
4. If the five-, four- and three-digit codes are invalid, save the original five-digit code and set the validity flag to indicate an invalid code (DXVn = 1).

In 1993 only. An error in HCUP processing caused invalid five-digit codes that ended in non-zeros, as well as zeros, to be processed by the above algorithm. If deleting the rightmost non-zero digits created a valid code, then

• DXn was set to the original invalid five digit code,
• DXVn was set 0 to indicate a valid code,
• DCCHPR was set based on the stripped valid code, and
• DRG, MDC, DRG10, MDC10, NEOMAT and edit check variables ED100, ED1nn, ED3nn, ED4nn, ED600, and ED601 may have been incorrectly assigned based on the stripped valid code.

Beginning in 2008, New York suppressed the identifier for the hospital (DSHOSPID) on records with an indication of induced abortion. These records are retained in the HCUP SID with the DSHOSPID set to "BLNK". New York identifies an indication of induced abortion by ICD-9-CM diagnosis or procedure code:

• An admitting, principal, or secondary diagnosis of "6350" through "6399", or "7796".
• A principal or secondary procedure of "690", "695", "696", "6993", "738", "7491", "750", or "9649".

Beginning in 1993, New York reports "cause of injury" and "place of injury" E-codes. Prior to 2003, during HCUP processing, these separately reported E-codes were placed after the last nonmissing secondary diagnosis. When a "cause of injury" E-code in the range of E850.0-E869.9 or E880.0-E928.9 was reported, then a "place of injury" E-code was also reported. If the hospital stay involved the possibility of classifying more than one situation or event, only the single cause of injury, poisoning, or adverse effect that was most severe was reported. Beginning in 2003, the separately reported E-codes, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

North Carolina supplied diagnosis codes in a field length of 6. Only the first five characters contained in the left-justified source field were used to assign the HCUP diagnosis codes.

Prior to 1998, Oregon reports one "cause of injury" E-code as a separate variable. Between 1998 and 2002, Oregon reported two "cause of injury" E-codes. During HCUP processing, these separately reported E-codes are placed after the last non-missing secondary diagnosis.

Beginning in 2003, the two separately reported E-codes, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

Oregon supplied diagnosis codes in a field of length 6; beginning in 2008, the field length became 7. Only the first five characters contained the diagnosis code and were used to assign the HCUP diagnosis codes.

Rhode Island reports one "cause of injury" E-code in a separate variable. Prior to 2003, during HCUP processing, this E-code was placed after the last non-missing diagnosis code. Beginning in 2003, the separately reported E-code, and any E-codes encountered in the diagnosis variables are placed in a separate array specific to E codes (ECODEn).

Prior to 2000 data, a small number of discharges explicitly included decimals in the diagnosis field, usually the decimal is implicit. This is problematic because South Carolina supplied diagnoses in a field of length 5. If decimals were included, then a valid 5-digit code would be truncated. For example, the diagnosis for unspecified sickle cell anemia "28260" would be incorrectly reported as "262.6". Prior to 1998, invalid diagnosis codes are marked by a validity flag (DXVn = 1). Beginning in 1998, invalid diagnosis codes are masked (Dxn = "invl").

Beginning in 2000 data this was no longer a problem; explicit decimals were not included in the diagnosis codes.

E-codes
Prior to data year 2000, one cause of injury E-code and one place of injury E-code may be missing from South Carolina discharges even though E-codes are present in the secondary diagnosis fields. Separate E-code fields that are collected by the data organization in South Carolina were not obtained for HCUP. Beginning in 2000, these separate E-code fields were obtained by HCUP. Beginning in 2003, the separately reported E-codes, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

South Carolina does not require the reporting of E-codes in the range E870-E879 (misadventures and abnormal reactions). Beginning with the 2003 databases, these E-codes were removed from the HCUP data, however they may be present on earlier years of HCUP data. AHRQ researchers and project staff may not include these diagnosis codes in any analyses (effective May 2004).

Prior to data year 2000, one cause of injury E-code and one place of injury E-code may be missing from South Carolina discharges even though E-codes are present in the secondary diagnosis fields. Separate E-code fields that are collected by the data organization in South Carolina were not obtained for HCUP. Beginning in 2000, these separate E-code fields were obtained by HCUP. Beginning in 2003, the separately reported E-codes, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

South Carolina does not require the reporting of E-codes in the range E870-E876 (misadventures and abnormal reactions). Beginning with the 2003 Intramural databases, these E-codes were removed from the HCUP data, however they may be present on earlier years of HCUP data. AHRQ researchers and project staff may not include these diagnosis codes in any analyses (effective May 2004).

South Dakota separately reports one "cause of injury" E-code and one "place of injury" E-code. Prior to 2003, during HCUP processing, these E-codes were placed after the last non-missing diagnosis code. Beginning in 2003, the separately reported E-code, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

Utah reports one "cause of injury" E-code as a separate variable. Prior to 2003, during HCUP processing, this E-code was placed after the last non-missing diagnosis code. Beginning in 2003, the separately reported E-code, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

Vermont reports one "cause of injury" E-code as a separate variable. Prior to 2003, during HCUP processing, this separately reported E-code was placed after the last non-missing secondary diagnosis. Beginning in 2003, the separately reported E-code and any E-codes encountered in the diagnosis variables are placed in a separate array specific to E codes (ECODEn).

Vermont supplied diagnosis codes in a field of length 6. Only the first five characters contained in the left-justified source field were used to assign the HCUP diagnosis codes.

After the 2002 Vermont SID file was created, the data source reported that a small number of discharges have incorrect information. This affects about 10 hospitals and less than 100 discharges.

Washington reported diagnosis codes in a field of length 6 for 1988-1992 and, beginning in 1993, in a field of length 7. Only the first five characters contain the diagnosis code and were used to assign the HCUP diagnosis code.

In 1988, Washington did not report "cause of injury" E-codes. From 1989-1992, Washington reports two "cause of injury" E-codes. Beginning in 1993, Washington reports only one "cause of injury" E-code. Prior to 2003, during HCUP processing, any separately reported E-code was placed after the last non-missing secondary diagnosis. Beginning in 2003, the separately reported E-code, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).

West Virginia reports one "cause of injury" E-codes as a separate variable. Prior to 2003, during HCUP processing, are placed in a separate array specific to E codes (ECODEn). Beginning in 2003, the separately reported E-codes and any E-codes encountered in the diagnosis variables are placed in a separate array specific to E codes (ECODEn).

West Virginia supplied diagnosis codes in a field length of 6. Only the first five characters contained in the left-justified source field were used to assign the HCUP diagnosis codes.

To comply with statutory requirements, Wisconsin modified diagnosis and procedure codes that explicitly referenced induced termination of pregnancy to eliminate distinctions between induced and spontaneous termination. The following codes were modified:

• Diagnoses with the first three digit of 634, 635, 636, 637, 638 were recoded to 637, while retaining the reported fourth digit,
• Procedure 6901 was changed to 6902,
• Procedure 6951 was changed to 6952,
• Procedure 6993 was changed to 6999,
• Procedure 7491 was changed to 7499,
• Procedure 750 was changed to 7599, and
• Procedures 9641-9649 were changed to 964 (which would be flagged as invalid, PRV=1).

Beginning Q4 2015, the following codes were modified:

• Diagnosis Z33.2 was changed to Z31.9,
• Diagnoses O04x and O07x were changed to O03.9,
• Procedure 10A07ZZ was changed to 10D17ZZ,
• Procedure 10A08ZZ was changed to 10D18ZZ,
• Procedure 10A07ZZ was changed to 10D17ZZ,
• Procedure 10A08ZZ was changed to 10D18ZZ,
• Procedure 10A07ZW was changed to 0UCB7ZZ,
• Procedure 10A00ZZ was changed to 10D00Z0,
• Procedure 10A03ZZ was changed to 0UC93ZZ,
• Procedure 10A04ZZ was changed to 0UC94ZZ,
• Procedure 10A07ZX was changed to 3E0P7GC.

Wisconsin reports one "cause of injury" E-code. Prior to 2003, during HCUP processing, this separately reported E-code was placed after the last non-missing secondary diagnosis. Beginning in 2003, the separately reported E-code, and any E-codes encountered in the diagnosis variables, are placed in a separate array specific to E codes (ECODEn).