Impact of record-linkage methodology on performance indicators and multivariate relationships

Area of Science:

• Public Health
• Health Services Research
• Data Science

Background:

• Program evaluation necessitates linking diverse data systems, such as substance abuse treatment and criminal justice records.
• Data inaccuracies, including name misspellings and digit transposition, pose significant challenges to accurate record linkage.

Purpose of the Study:

• To evaluate the impact of different record-linkage algorithms on program evaluation outcomes.
• To assess how algorithm choice influences prevalence and risk estimates in substance abuse treatment research.

Main Methods:

• Three record-linkage algorithms (match-merge, common patient identifier, probabilistic) were employed.
• These algorithms linked outpatient substance abuse treatment recipients to statewide arrest and death data.
• Multivariate survival analysis was used to predict arrest risk.

Main Results:

• Algorithm choice substantially altered estimates of arrest rates before and after treatment admission.
• The range of arrest rates varied significantly based on the linkage algorithm used.
• The predictive power of prior arrests on subsequent arrests was also influenced by the chosen algorithm.

Conclusions:

• The selection of a record-linkage algorithm critically impacts the validity of findings in program evaluation and public health research.
• Researchers must carefully consider algorithm choice to ensure accurate data linkage and reliable outcome estimation.
• This study highlights the sensitivity of epidemiological and public health metrics to data linkage methodologies.