Data Harmonization Methods and Data Quality Assessment for Integrating Routinely Collected Data Across Multiple Sources: A Systematic Review.

Background: Integrating routinely collected health data across heterogeneous systems is essential for improving patient care, disease surveillance, and programmatic decision-making. However, heterogeneity in data formats, coding systems, and semantic definitions create challenges affecting consistent data interpretation and reuse across systems. Despite its importance, harmonization methods and their relationship to data quality have not been systematically examined. This review analyzes harmonization approaches, reported data quality attributes, associated benefits, and relationships among them. Methods: We conducted a systematic review following Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. PubMed, Scopus, CINAHL, and ProQuest were searched for studies describing harmonization of routinely collected health data, integrate data from multiple digital systems and reported data quality attributes. Covidence software facilitated screening, full-text review, and data extraction. Descriptive statistics summarized study characteristics, and Spearman correlations assessed associations among number of harmonization methods used, number of systems used, and number of data quality attributes reported. Results: Of 2,309 records, 69 studies met inclusion criteria. Most (88%) were published between 2013 and 2024, predominantly in high-income countries (93%), with the United States accounting for the majority (64%). Harmonization approaches grouped into 6 categories: established data standards; national/global standards; database/schema mapping; extract-transform-load standardization; manual terminology mapping; and local data dictionaries. Established data standards were most common (40 studies [58%]), followed by national/global standards (26 studies [38%]) and database/schema mapping (20 studies [29%]). Most studies applied 1 or 2 methods (58 [84%]), while 11 (16%) used 3 methods. Frequently reported data quality attributes included completeness (43 [62%]), accuracy (29 [42%]), and timeliness (24 [35%]), while relevance (2 [3%]) and usefulness (1 [1%]) were rarely reported. Overall, 46 studies (67%) reported at least 1 benefit. Data reuse and reduced work burden each reported by 20 studies (29%), and improved decision-making in 18 studies (26%). Correlations indicated negligible associations between number of harmonization methods and number of systems (ρ = 0.028) or number of data quality attributes (ρ = 0.074). Conclusions: Data element harmonization is essential for integrating heterogeneous health data and enabling interoperability. Standardized approaches dominate, though manual and localized methods remain relevant where global standards are not fully adopted. More consistent reporting of harmonization practices, inclusion of underreported data quality attributes, and examination of contextual factors are needed to strengthen global evidence.