Incorporating the Last Four Digits of Social Security Numbers Substantially Improves Linking Patient Data from...

Area of Science:

• Health Informatics
• Data Linkage
• Patient Confidentiality

Background:

• Linking de-identified patient data is crucial for health research.
• Existing methods may not ensure complete accuracy or confidentiality.
• Developing robust algorithms is essential for reliable health data analysis.

Purpose of the Study:

• To assess algorithms for linking patients across de-identified databases.
• To evaluate linkage accuracy without compromising patient confidentiality.
• To determine the effectiveness of SSNL4 in patient data linkage.

Main Methods:

• Cross-sectional assessment of four linking algorithms using hospital discharge data.
• Utilized de-identified claims including gender, birthdate, and zip code.
• Matched assessment records with institutional sources including SSNL4.

Main Results:

• Gender, birthdate, and zip code achieved 98.0% sensitivity for readmissions.
• SSNL4 inclusion resulted in nearly perfect identification of readmissions and deaths.
• Statewide application showed lower rates in border regions with incomplete data.

Conclusions:

• SSNL4 addition to administrative data enables highly accurate patient linkage.
• Appropriate data use policies are vital for maintaining confidentiality.
• Centralized de-identified databases should incorporate SSNL4 for improved data linkage.