IWeb Administrator Guide

Deterministic (Default) Rules/Logic

The SIIS deduplication logic uses rule-based algorithms to determine matches in the following order (each is discussed in more detail later):

Likely Matches
Matches
Possible Matches
Separate Rules/Special Cases

The following are frequently asked questions about the process of determining a match:

Question	Answer
Can two different people have the same name and birthdate?	Yes. Since it is possible for two different people to have the same name and birthdate, these records should not be merged. However, if additional fields do match (e.g., phone number), it may be safe to merge the records.
Can two different people have the same social security number but no other matching fields?	No. More than likely, there is a typo (typographical error) in the social security number.
Can two different people have the same first name, middle name, last name, birthdate and address, but have different social security numbers?	Yes. This is probably the same person, but the social security number was probably entered incorrectly in one of the records.
What are other factors to consider?	Last names can change due to marriage, divorce, custody changes, adoption, etc. Addresses change when patients move. Twins can have all fields in common except for first names and (possible) middle names. Sons can have the same name as their fathers, where the only record difference would be the birthdate. Orphanages can have several children located at the same address, so addresses and phone numbers may not be significant for record comparisons. Watch for swapped entries, such as a first name placed in the last name field, a middle name in the first name field, etc. Watch for systemic bad data. For example, all patients might have the same social security number. The number of records listed for a manual deduplication might be enormous. If this happens, the sending site must be contacted and instructed to fix the data before re-sending it.

Rules for Likely Matches

The first step of deduplication is to obtain a subset of likely matches. These are the existing patients in the database that match on any of the following field values:

Birthdate and First Initial
Birthdate and Last Initial
Street Address, State, and Zip Code
Street Address and City
7-digit Phone Number (excluding area code)
Guardian Social Security Number
Social Security Number

The next step is to compare each patient in that subset to the "new" patient.

If a match is found at any point, the program immediately exists and goes to the next patient.
If a match is not found, the records are flagged as a possible match or as separate patients.

The following values are pre-processed:

Addresses are pre-processed to conform to postal standards and to remove spaces.
Names are pre-processed to remove spaces and special characters (such as dashes in hyphenated names).
Names are checked to see if First and Middle names have been swapped, or if Middle and Last names have been swapped.
Names are checked to see if First and Middle names were both entered into the First Name field.

The deduplication logic examples the following fields:

Medicaid Number
Birth File Number
SSN
First Name
Middle Name
Last Name
Birth Date
Street Address, State, and Zip Code
Street Address, City
7-digit Phone Number (excluding area code)
Guardian SSN
Guardian First Name
Guardian Last Name
Mother Maiden Name
Sex
Second Guardian First Name
Second Guardian Last Name
Second Guardian SSN
Birth Order
Multiple Birth Count

Additional information is determined as follows:

If Field	Quantifier	Field/Value	Then Field	Result
First Name	EQUALS	BABY BABY BOY Babyboy BABY GIRL Babygirl Babynb BB BB1 BB2 Boy BOY FC1 FC2 FC3 Female GIRL Male NEWBORN Newborn Newborn Boy Newborn Girl MC1 MC2 MC3 Twin A TwinA Twin B TwinB	First Name	Is identified as a special baby name match
Street Address State Zip Code	EQUALS	Street Address City	Address	Is identified as an address match
Guardian Last Name or Mother's Maiden Name	EQUALS	Patient's Last Name	Last Name	Additional weight is given to unique last names
Patient 1's Last Name	EQUALS	Patient 2's Last Name Patient 2's Guardian's Last Name Patient 2's Mother's Maiden Name	Last Name	Unique Last Names is incremented by 1
Patient 1's Guardian's Last Name	EQUALS	Patient 2's Last Name Patient 2's Guardian's Last Name Patient 2's Mother's Maiden Name	Last Name	Unique Last Names is incremented by 1
Patient 1's Mother's Maiden Name	EQUALS	Patient 2's Last Name Patient 2's Mother's Maiden Name Patient 2's Guardian's Last Name	Last Name	Unique Last Names is incremented by 1
Patient 1's Last Name	EQUALS	Patient 1's Guardian's Last Name Patient 1's Mother's Maiden Name	Last Name	Unique Last Name is incremented by 1
Patient 1's Guardian's Last Name	EQUALS	Patient 1's Mother's Maiden Name	Last Name	Unique Last Name is decremented by 1

Since the unique last name is calculated in this manner, its value can be anywhere from 0 to 3.

Approximate Birth Date signifies the birthdate is close. It is calculated as follows:

If the birthdates are less than nine months apart, Birth Date is approximate if any of the following are true:
- If the day and month are swapped (i.e., 10/11/2003 and 11/10/2003)
- If the year and day are the same, but the months are different
- If the year and month are the same, but the days are different
If the birthdates are less than ten years apart, Birth Date is approximate if the day and month match.
First Guardian Name and Second Guardian Name are examined to check for swaps (i.e., if the father was the first guardian on one record and the second guardian on the other record).
If both Birth Order and Multiple Birth Count are present and different, and the record would have otherwise matched, the record is sent to Manual Deduplication.
If Mother Maiden Name is "ADULT" or "SELF," these values are ignored.

Rules for Matches

The SIIS Deduplication logic uses a rule-based algorithm to identify a match. These merge (match) rules are identified by a rule number. The rules, descriptions, and fields are listed below:

Rule #	Description	Fields: Col. 1	Fields: Col. 2
100	Medicaid Number and File Number	N/A	N/A
110	SSN and Birth File Number	N/A	N/A
120	First Name and at least of the fields from columns 1 and 2	SSN Birth File Number Medicaid Number Guardian First Name	Birth Date Middle Name Unique Last Name Address Phone
121	Nickname and Sex and at least one of the fields from columns 1 and 2	SSN Birth File Number Medicaid Number	Birth Date Middle Name Unique Last Name Guardian First Name Address Phone
122	First Name and approximate Birth Date, and at least one of the fields from column 1	SSN Birth File Number Medicaid Number
130	Special Baby Name, Birth Date, and at least two of the fields from column 1	Unique Last Name Address Phone Guardian First Name Medicaid Number Birth File Number SSN Guardian SSN Middle Name
160	Birth Date and First Name, and at least one of the fields from column 1	Address Phone Guardian SSN
161	Birth Date and Nickname and Sex, and at least one of the fields from column 1	Address Phone Guardian SSN
190	Birth Date and First Name and at least two fields from column 1	Middle Name Guardian First Name Unique Last Name
191	Birth Date and Nickname and Sex, and at least one of the fields from column 1	Middle Name Guardian First Name Unique Last Name
200	First Name, Middle Initial, Last Name, Approximate Birth Date, and at least one of the fields from column 1	Address Phone Guardian SSN
201	First Name, Middle Initial, Last Name, Approximate Birth Date, and at least one field from column 1	Guardian First Name Unique Last Name
202	First Name, Last Name, Approximate Birth Date, and at least one field from column 1 and 2	Address Phone Guardian SSN	Guardian First Name Unique Last Name
203	First Name, Middle Name, Approximate Birth Date, and at least one field from column 1 and 2	Address Phone Guardian SSN	Guardian First Name Unique Last Name
204	First Name, Last Name, Approximate Birth Date, and at least two of the fields from column 1	Address Phone Guardian SSN
205	First Name, Last Name, Approximate Birth Date, and Unique Last Name has at least two matches

Additionally, there are rules to determine if the patients in the subset are possible matches. The Possible Match rules are not examined until all of the Merge rules for definite matches have been examined. Any patients remaining after that are then examined to see if they are possible matches.

For Louisiana users: As of IWeb version 5.17.5.2, manually entered vaccinations are not overwritten by HL7-sourced vaccinations. If a manually entered vaccination exists, it is automatically considered the best match.

Rules for Possible Matches

The Possible Match rules are not examined until all of the Match (merge) rules for definite matches have been examined. (Note that these rules do not have rule numbers.)

States can optionally define a list of Organizations (IRMS) whose patients should never be sent to manual deduplication. If the patient is a possible match with any patient that is currently in the registry, the incoming patient record is deleted rather than sent to manual deduplication. The list of Organizations (IRMS) must be provided and sent to STC's Help Desk.

Description	Fields: Col. 1	Fields: Col. 2
First Name, Middle Initial, Last Name, and Birth Date	N/A	N/A
First Name, Middle Name, Last Initial, Birth Date	N/A	N/A
SSN	N/A	N/A
Medicaid Number	N/A	N/A
Birth File Number	N/A	N/A
At least one field from column 1 and 2	First Name Birth Date	Address Phone Guardian SSN
First Names are like-sounding and Approximate Birth Date matches, and at least one field from column 1	Address Phone Guardian SSN
Birth Date and at least two fields from column 1	First Name Middle Name Guardian First Name Unique Last Name
Birth Date and First Names are like-sounding, and last names are like-sounding	N/A	N/A

Finally, it is possible that rules for possible matches have flagged too many records, so additional separate rules are applied for special cases. These apply only to the records that have been flagged as possible matches.

Separate Rules / Special Cases

If the rules for possible matches have flagged too many records, these additional, separate rules apply for special cases. There are no rule numbers for these.

Description	Fields: Col. 1	Fields: Col. 2
Birth Dates are very different: Day, Month, and Year do not match and Birth Dates are more than ten years apart.	N/A	N/A
Twins (Birth Dates are the same and Family/Address information is the same). If First Names do not sound alike and First/Middle/Last Names were not swapped, the records are "twins" if any of the fields in column 1 are true.	Sexes are opposite SSNs are not null and not equal Provider's ID Number is sequential First Initials do not match and First Names, excluding First Initials, do not sound alike Both the First and Middle Names are populated and the first three characters of both Names are different
Birth Date does not match and it is not an Approximate Birth Date, and the First Name does not match and the First Name does not match as a Nickname	N/A	N/A
Middle Initials are present and different in both names and there are no swapped initials	N/A	N/A
Birth Date does not match and it is not an Approximate Birth Date, and the First Initial is different and the Last Name does not match	N/A	N/A
Birth Date and Phone Number match but SSN, Address, and Last Initial do not match, and neither First Name or Nickname match, and First Names do not sound similar (this handles the special case of a Phone Number being reissued)	N/A	N/A
To handle twins with the same First Name, if the records otherwise match and the Middle Initial is different, or the Middle Name is populated and sounds different, the record is sent to manual review	N/A	N/A
If First Name, Middle Name, and Birth Date match, but none of these match: Last Name, SSN, Mother Maiden Name, Guardian First Name, Address, the records are not a match	N/A	N/A
If Phone Number and First Name match, but Birth Date, Approximate Birth Date, SSN, Address, Last Initial, Guardian First Name, and Middle Initial do not, the records are not a match (it otherwise would have been sent to manual). This addresses a corner-case of a Phone Number being reissued to someone with the same First Name	N/A	N/A
If Address and First Name match, but Birth Date, Approximate Birth Date, SSN, Address, Last Initial, Guardian First Name, and Middle Initial do not, the records are not a match (it otherwise would have been sent to manual). This addresses a corner-case of a person with the same First Name moving to an address that someone else had previously	N/A	N/A
If a record is a possible match and 3+ vaccinations match, the record is automatically processed as an exact match