ERROR QUANTIFICATION TOOL OF HEALTH RECORDS FOR DEVELOPING COUNTRIES Rose Nakasi 1 , Ernest Mwebaze 2 , Aminah Zawedde 1 and Gilbert Maiga 1 1 Makerere University, Uganda 2 Google AI ABSTRACT In developing countries, the greatest information sources for disease prevalence are official Electronic Health Records (EHR). However, data is usually affected by human error which flaws decisions by stakeholders. The previous reviews considered data quality findings at only an aggregated level, however, this method is vulnerable to errors if wrong data is aggregated and therefore hard to trace the error. This exploratory study though limited, utilized a statistical approach for computing differences between paper and EHR records. A sample of selected health facilities was used as a base study and results indicated varying errors at each level of reporting tool. KEYWORDS Paper Health Records, Online Health Records, Error Quantification 1. INTRODUCTION Electronic Health Records provide statistics about disease distribution cases which can adjust accuracies of disease modeling in a certain location at a given period of time. Therefore, for stakeholders to rely on such data, it must be reliable which isn’t the case now (WHO, 2O11) and the fact that EHR is aggregated data, it is hard to track the errors from the source. Taking a Case for Uganda, a developing country, the National eHMIS (electronic Health Management Information system) aggregates diagnosis cases monthly from health facilities that are in the system. The records are captured at health facilities daily and aggregated to weekly and then sent to the District Health officers who electronically capture these records into the system using data entrants. We believe this kind of system is prone to errors and must be investigated. Error quantification in health records can contribute to a reliable sour ce of health record data. It’s even more useful if one wants to use the eHMIS data but has no idea of the error to account for in their analysis. The Uganda Ministry of Health (MOH) recognizes that improper data quality checks in the eHMIS system, unrealistic analysis using such data will affect true representation of the output from the analysis. Crucial in dealing with limitations in eHMIS, error quantification process must start from the health facility where the records are captured from (the laboratory) and then monitor captured and aggregated records on weekly and monthly basis both at facility and district level. In this research, we believe that proper tracking of daily records at health facility level provides a more realistic audit track and trail of the health records before and after they are aggregated in the weekly and monthly records. Error quantification will undoubtedly increase the level of confidence to be placed on health data for effective disease distribution modeling. Here, we describe a simple statistical method based on the difference measure that makes a comparison of records at each reporting level implemented through a developed software tool that quantifies the error in reporting records. This method is utilized from the source of data at the health facility and spans through at the aggregated level at district level. We further incorporate the analysis tool into a web based platform that reliably helps to monitor and trace the error with a user interface. International Conference e-Health 2019 201