Alaska randomly assigns a re-identification number to reported person numbers. The number is not based on any systematically applied set of rules. The patient retains his/her re-identification number permanently, which allows for longitudinal analysis.

In Arkansas, the re-identified person number (PNUM_R) is based on encrypted social security numbers.

California reports encrypted social security numbers as person numbers.

Beginning in 2017, Delaware recycles person number (PNUM) from the prior year. Since VisitLink is derived from verified person numbers, it cannot be used to track patients consistently over time.

Florida bases their person number on the patient's social security number. The supplied identifiers are encrypted during HCUP processing (PNUM_R). Because of changes in source-provided information, longitudinal analyses using PNUM_R across years are not always possible.

Beginning in 2005, Florida used a different masking method from the one in 2004. From 1998-2004 the coding of the person number is consistent across years. In 1997, source documentation indicated that the Florida encryption routine used to create PNUM from social security numbers was different than previous years. Source-supplied values in the 1996 inpatient data were the same length as prior years, but looked very different.

Staring in 2020, Georgia changed to use Cross Reference Code for HCUP PNUM. Therefore, this Identifiers prior to 2020, and 2020 and after are not consistent. Longitudinal analyses that cross 2020 are not recommended.

Beginning in 2010, Iowa provides a patient identification number. Social security number (SSN), birth date, and gender are used to create a patient identifier. A patient identifier is not assigned if any of the three fields are missing or invalid (e.g., the SSN is not nine digits or is filled with a recurring character).

Beginning in 2013, Iowa provides a new unique ID that is develop using probabilistic matching based on name, gender and date of birth.

Beginning 2017, Iowa changed their coding scheme for person numbers. Therefore, they cannot be used to track patients in prior years.

In 2013, Maryland began providing a new statewide encrypted unique patient identification number (CRISP_EID) to link patients across all hospitals.

In 2012, Maryland provided HCUP with an Enterprise Master Patient Identifier (EMPI). This is a hospital system enterprise identifier, also known as the master patient identifier, that produces is a unique patient identification number using a probabilistic matching algorithm and is used to link patients across system hospitals only. For 2012, the EMPI value is populated for 19% of the inpatient records. The supplied identifiers are encrypted during HCUP processing.

Beginning in 2020, Massachusetts created a new patient identifier (PNUM) for the SID. Since VisitLink is derived from verified person numbers, it cannot be used to track patients consistently across years prior to 2020 versus 2020 and later for the SID, or across database types beginning in 2020.

Massachusetts supplied encrypted social security numbers. The supplied identifiers are encrypted again during HCUP processing (PNUM_R).

Because of the timing of HCUP data processing for the 1999 NIS, the Massachusetts source file provided to HCUP was an interim file that included records that had failed edit checks. The percent of failed records is very small, ranging from 0.0% to 1.5% (with a mean of 0.4%) for most hospitals. A handful of hospitals had a large percent of failed records. Failed records have one or more of the following errors:

• Invalid diagnosis code *
• Invalid procedure code *
• Invalid or missing birth weight
• Invalid claim certificate number **
• Invalid or inconsistent UB-92 revenue code
• Invalid medical record or person number
• Invalid type of payer *
• Inconsistent primary and secondary payer
• Invalid physician identifier
• Invalid patient or employer ZIP Code. **

* These errors would have been handled during HCUP data processing.

** These data elements are not included in the HCUP data files.

Beginning in 2013, Mississippi provides a new unique ID that is developed using probabilistic matching based on name, gender and date of birth.

Beginning in 2010, Mississippi provides a patient identification number, which are supplied as encrypted social security numbers. The supplied identifiers are encrypted again during HCUP processing (PNUM_R).

The Missouri person number (PNUM_R) cannot be used to track patients consistently over time because the coding scheme changed in 2002. Beginning in 2002, Missouri randomly assigns a re-identification number to reported person numbers. The number is not based on any systematically applied set of rules. Nonetheless, the patient retains his/her re-identification number permanently, which allows for longitudinal analysis from 2002 forward.

Prior to 2002, the Missouri PNUM_R is based on the patient's social security number. Missouri person numbers were encrypted for HCUP. Nonetheless, the patient retains the same identification number up through 2001, which allows for readmission studies within years and longitudinal analysis across years through 2001.

Missouri law prohibits the release of person identifiers for drug- and alcohol-related discharges. Therefore, Missouri set person numbers to missing in the data supplied for HCUP. Prior to 2002, social security numbers were set to asterisks; beginning in 2002, patient re-identification numbers were set to 9-filled values. These codes are included in the HCUP Missouri files as missing (PNUM_R = blank). The supplied identifiers are encrypted during HCUP processing (PNUM_R).

Beginning with the 2020 data, Nebraska used probabilistic matching based on patient name, gender, and date of birth to generate the patient linkage number. Therefore, the patient linkage numbers prior to 2020 and 2020 and after are not consistent. Longitudinal analyses that cross data year 2020 are not recommended.

From 2002 through 2019, Nebraska provides a patient re-identification number. The Nebraska person number (PNUM/PNUM_S) cannot be used to track patients consistently over time because the coding scheme changed in 2002. Prior to 2002, Nebraska created their unique patient identifier through a 256 byte one-way hashing of the patient's first name, patient's last name, date of birth, and social security number. Beginning in 2002, Nebraska randomly assigns a re-identification number to the individual. The number is not based on any systematically applied set of rules. Nonetheless, the patient retains his/her re-identification number permanently, which allows for longitudinal analysis from 2002 forward.

New Mexico changed their coding for person numbers in 2013; they cannot be used to track patients between 2013 and prior years.

The New York person number (PNUM/PNUM_R) cannot be used to track patients consistently over time because the coding scheme changed. Beginning in 2005, New York started using a new encryption key which is 22 characters in length. In 2004, New York used a hexadecimal format to mask person numbers. Prior to 2004, a 10-digit encrypted person number (PNUM) was created by New York using the first 2 characters of the last name, the last 2 characters of the last name, the first 2 character of the first name, and the last 4 digits of the person's SSN.

Beginning with the 2008 data, the HCUP data element PNUM_R is missing (.) for AIDS/HIV patients. New York identifies AIDS/HIV records by ICD-9-CM diagnosis code, DRG, or MS-DRG:

• An admitting, principal or any secondary diagnosis of "042", "043", "044", "7958", "27910", "27919", "2793", "1363", "79571", "07951", "07952", "07953", "V017" or "V08".
• A DRG of 488 through 490 (prior to October 2007) or MS-DRG of 700 through 716 (beginning in October 2007).

Please note that the admitting diagnosis is not retained in the HCUP databases prior to 2012.

North Carolina provides an encrypted social security number. Reporting of the patient's social security number is optional for hospitals in North Carolina. Beginning in the 2000 data, this data element is frequently missing. During HCUP processing, this identifier is re-encrypted.

Beginning in 2022, Utah changed their encryption process for PNUM. Since VisitLink is derived from verified person numbers, it is not consistent with data from prior years.

Utah supplied source-encrypted social security numbers as person numbers. These identifiers are encrypted again during HCUP processing (PNUM_R). Three-digit codes may indicate:

• newborns who have not obtained a SSN
• patients who have no SSN
• patients who choose not to provide his/her SSN

The 2011 and 2012 files that WA sent to us had the same visitlink values because their master link file was overwritten. A new format was used for 2013 therefore visitlink will not be able to link across the 2011-2013 files.

Washington had changed their formats for person numbers in 2009 so they will not be able to be linked with previous years. Starting in 2010, Washington only provides the data element visitLink as synthetic patient identifier.

The Washington person number (PNUM_R) cannot be used to track patients consistently over time because the coding scheme changed.

In 2008, Washington changed their format again for PNUM_R and it is being loaded from Washington's Revisit file. They are using a probabilistic linking method to unduplicate the file (LinkPlus for UB04 records and Automatch for UB92 records). They assign a random number to the unique patient groups with a SAS program, sort by the random number, and then assign each unique patient a number beginning with 1. They discovered a very large number of cases where hospitals are reporting the following names for infants: BABY, BABY BOY, BABY GIRL, BABYBOY, BABYGIRL, BB-, BBABY, BG-, and INFANT. Since infants from multiple births will have the same last name and the same date of birth, identifying revisits may incorrectly assign infants of multiple births as revisits. They also link their hospitalization data to our birth data and they plan to identify revisits for infants by replacing the infant name from the linked birth record with the name reported in the hospitalization data. This means that infants will not be linked until a later date. Since discharges for some of the infants does not occur until the first quarter of next year, their birth linkages will not be done until after they release the first quarter data for a year.

In 2007, Washington changed their format for PNUM_R. Patient's first 4 letters of the last name, first 3 letters of the first name, middle initial, and last 4 digits of SSN. If part of a name or SSN is missing, then it is filled in with dashes (-).

Prior to 2007, Washington used the first two characters of the patient's last name, the first two characters of the first name, and the birth date to create an encrypted person identifier. People with similar names and the same birth date may have the same identifier.

Prior to 1990, one person may have the two different values of the encrypted person number across time. The state reports that before 1990 some hospitals did not follow the patient number convention and assigned this identifier based on the last two letters of patient's first and last names, rather than the first two letters. Starting in 1990, all hospitals followed the same conventions.

After 7/1/90, it is possible that more than one person could have the same PNUM_R value. Wisconsin derives the identifier from the patient first and last names, making it possible for people with similar names to have the same identifier. This data element, however, can be used in combination with other data elements to track transfers and readmissions. Prior to 7/1/1990 PNUM_R was not available.

Beginning in 2013, Wisconsin provides a unique patient identifier. The field is a 64-character encrypted code submitted by the facility based on the patient's name, date of birth, and gender.