strengthening in low resource settings. Investments in electronic medical record (EMR) systems offer promise for improved information for patient and public health program management. Strong data quality (DQ) is a precursor to strong data use. Routine DQ assessment (DQA) within EMR systems can be resource-intensive when using typical methods of audit and chart review. Automated database queries on the completeness, accuracy and timeliness of data offer an efficient alternative. This DQA focuses on Haiti’s national EMR iSante a system deployed in 119 health care facilities and containing longitudinal data for more than 400,000 patients. Methods: This mixed-methods evaluation focused on data related to HIV care and treatment. We first used a qualitative Delphi process to identify DQ priorities among local HIV experts, followed by a quantitative DQA on these priority data elements. The quantitative DQA examined 13 indicators of completeness, accuracy and timeliness of data using retrospective data from HIV patients, collected during more than 3.5 million encounters from 2005e June 2013. We described levels of DQ for each indicator over time, and examined the consistency of within-site performance. Using generalized linear models (GLM) with logit link and binomial errors for each of the 13 DQ indicators, we examined associations between DQ results and site and system characteristics, such as facility type, urban vs. rural location, and number of iSante system users. Findings: Ninety-five sites using the iSante data system were included in the evaluation. On average, completeness was high for demographic data but low for clinical data, accuracy of age data was low, and timeliness of data entry was low. For most indicators, DQ tended to improve over time, both overall and within specific sites. DQ was highly variable across sites, and sites which performed strongly in some indicators performed weakly in others. In adjusted analyses, site and system factors with generally favorable and statistically significant associations with DQ indicators were University hospital type, private sector governance, presence of more advanced IT infrastructure, greater site experience, greater maturity of the iSante system, having more overall system users but fewer new users. Interpretation: The heterogeneity in performance on various priority DQ indicators across sites indicates that excellent DQ is achievable in Haiti, but that many sites have much work to do to improve their DQ performance. A dynamic, interactive “data quality dashboard” within iSante could bring transparency and motivate improvement. Further investment in the IT infrastructure supporting iSante, including assuring stable power supply, and in on-going training for new users, is also warranted. Funding: Health Resources and Services Administration (#U91HA0680); US Centers for Disease Control and Prevention (#5U2GGH000549-03). Abstract #: 02ITIS025
Read full abstract