BACKGROUND
Electronic Health Records (EHRs) play an increasingly important role in the delivery of HIV care in low-and-middle-income countries. The data collected is used for direct clinical care, quality improvement activities, reporting and clinical research. Despite widespread EHR use for HIV care in East Africa, and countries like Nigeria, Mozambique and Haiti, challenges remain especially with achieving and maintaining high quality data.
OBJECTIVE
To determine factors influencing data quality in a large scale EHR deployment and potential strategies to improve quality.
METHODS
We carried out a data quality evaluation in 50 health facilities in Rwanda using the OpenMRS EHR systems for HIV care. The sites were part of a larger randomized controlled trial and half of the sites had an enhanced version of the EHR with clinical decision support alerts. Trained data collectors visited 50 health facilities to collect 28 variables from both paper charts and the EHR system using the ODK app. We measured the data completeness, timeliness, the degree of matching of paper and EHR data including use of concordance scores. Factors potentially effecting data quality were drawn from a previous survey of users in the 50 sites.
RESULTS
We randomly selected 3467 records, reviewed from the paper charts and EHR, collecting a total of 194,152 data items. Data completeness was above the 85% threshold for all variables except viral load (VL) results and second and third line drug regimens. Matching scores for data values were close to or above 85%, but lower for dates particularly for drug pickups and VL. Data concordance had a mean of 10.2/15 (68%) for the 15 variables. Site and user factors (years of EHR use, technology experience, EHR availability/uptime, intervention status) were tested for correlation with data quality measures. EHR system availability/uptime was positively correlated with concordance, and users experiences of technology was negatively correlated. The alerts for missing viral load results implemented in 11 intervention sites showed clear evidence of improving timeliness and completeness of initially low matching of VL results in the EHR and paper records (11.9% to 26.7%, P<0.001). Similar effects were seen on completeness of recording of medication pickups (18.7% to 32.6%, P<0.001).
CONCLUSIONS
The EHR records in the 50 health facilities generally had a high level of completeness. Matching results were within 3% of the 85% threshold or above for non-date variables, but lower for dates. Higher EHR stability and uptime was associated with better data quality. Alerts and reminders for entering VL were shown to have a strong effect in improving data quality. Overall, most data was considered fit for purpose, but more regular data quality assessments, training, mentoring and technical improvements in forms and data reports would be beneficial particularly for dates and key lab data.
CLINICALTRIAL
N/A