Attitude, Training and Trust in Data Drive Use of Ethiopia’s Digital Health Records

In the public health facilities of eastern Ethiopia, a digital revolution has quietly been underway for years. The District Health Information System 2, known widely as DHIS2, is an open-source software platform designed to collect, store, analyze and distribute health data from the most remote clinics to the highest levels of health administration. On paper, it promises a transformation: health managers should be able to monitor disease trends, track service performance and make evidence-based decisions in near real time. In practice, however, a new study from Haramaya University reveals that the promise is only half fulfilled. Nearly half of the health workers surveyed in the Harari Regional State and the Dire Dawa City Administration are not actually using the data their own facilities generate, and the reasons why turn out to be as much about human psychology and workplace culture as about technology.

The research, published in BMC Health Services Research, was led by Daniel Gudina, Gudeta Ayele, Behailu Hawulte and Ibsa Mussa from the School of Public Health at Haramaya University’s College of Health and Medical Sciences. The team set out to answer a deceptively simple question: what determines whether health workers in public facilities actually use the data flowing through DHIS2? This question matters because data utilization is one of the key functions of any health information system. When performance monitoring teams can read and interpret their own data, they can identify gaps in immunization coverage, spot outbreaks earlier, allocate staff and supplies more rationally and ultimately improve the quality of health service delivery. When they cannot, or do not, the entire investment in digital infrastructure risks becoming an expensive exercise in data entry without insight.

To investigate, the researchers conducted an institutional-based cross-sectional study between June 1, 2020 and July 31, 2020. They used a stratified sampling technique to select 220 participants working in public health facilities across the two administrative areas of eastern Ethiopia. Data collection relied on a structured questionnaire complemented by an observational checklist, allowing the team to capture both what health workers reported about their practices and what could be verified on the ground. The completed questionnaires were thoroughly checked, coded and entered into Epi-data 3.1 before being transferred to Stata 14 for statistical analysis. To identify the factors associated with data utilization, the researchers applied a binary logistic regression model, treating results as statistically significant when the p value fell below 0.05 with 95 percent confidence intervals.

The headline finding is sobering. Overall utilization of data from DHIS2 stood at about 45 percent, with a 95 percent confidence interval ranging from 39 to 50 percent. That figure falls well below the national recommended threshold, which calls for more than two-thirds of health facilities to be actively using their routine health information data for decision making. In other words, even in a system where the software has been deployed and staff have been trained to feed it, a majority of facilities are not translating the numbers into action. The study also notes that prior research across Ethiopia has reported significant variability in DHIS2 data utilization, suggesting that this is not a local anomaly but a systemic challenge with roots that vary from place to place.

What, then, separates the facilities that use their data from those that do not? The regression analysis identified four independent determinants, and each one is revealing. The first is attitude. Health workers who held a favorable attitude toward data use were roughly three times more likely to utilize DHIS2 data, with an adjusted odds ratio of 3.0 and a 95 percent confidence interval of 1.90 to 8.62. Attitude, in this context, reflects whether staff believe that reviewing data is worth their time, whether they see it as relevant to their daily work and whether they feel that acting on evidence is part of their professional identity. A health worker who views monthly data review as a bureaucratic chore will behave very differently from one who sees it as a diagnostic tool for improving services.

The second determinant is perceived self-competence. Workers who felt confident in their ability to interpret and apply the data were nearly three times more likely to use it, with an adjusted odds ratio of 2.9 and a confidence interval of 1.14 to 7.38. This finding speaks to a well-documented problem in health information systems across low- and middle-income countries: staff may be trained to enter data but not to analyze it. DHIS2 offers dashboards, pivot tables and visualization tools, but these features are only useful to people who understand what the outputs mean and how they connect to programmatic decisions. Self-doubt, in this environment, becomes a silent barrier. A nurse or health officer who feels incompetent with data will avoid opening the very reports that could guide their work, and the avoidance reinforces the incompetence in a self-perpetuating cycle.

The third and strongest single factor was perceived data quality. Health workers who believed the data in the system were accurate, complete and timely were more than four times as likely to use them, with an adjusted odds ratio of 4.4 and a confidence interval of 1.76 to 10.9. This is perhaps the most intuitive of the findings, and also the most troubling. If staff suspect that the numbers in DHIS2 are riddled with errors, duplicates or gaps, they will reasonably distrust any conclusion drawn from them. Data quality and data use are locked in a feedback loop: poor quality suppresses use, and without use there is little incentive or feedback mechanism to correct quality problems. Breaking that loop requires deliberate investment in data verification, feedback to data enterers and a culture in which accuracy is valued and rewarded rather than assumed.

The fourth determinant was supportive supervision. Facilities whose staff received supportive supervision were more than four times as likely to use DHIS2 data, with an adjusted odds ratio of 4.3 and a confidence interval of 1.48 to 12.45. Supportive supervision, in the language of the Performance of Routine Health Information System framework, known as PRISM, means supervisors who do more than inspect forms. They review data with frontline staff, help troubleshoot technical problems, encourage discussion of trends and connect the numbers to concrete service improvements. The PRISM framework, which underpins much of the conceptual thinking in this field, holds that technical, behavioral and organizational determinants together shape whether routine health information systems deliver value. The Ethiopian findings map neatly onto that framework: attitude and self-competence are behavioral determinants, perceived data quality is a technical one, and supportive supervision is organizational.

The implications for policy are direct. Ethiopia’s Federal Ministry of Health has invested substantially in DHIS2 as the backbone of its routine health information system, and the platform is expected to increase the utilization of health data nationwide. But the study’s results suggest that software deployment alone does not close the gap between data availability and data use. Interventions should target the four determinants identified: building favorable attitudes toward evidence-based practice, strengthening the analytical self-competence of health workers through practical mentorship rather than one-off training, assuring and communicating data quality so that staff trust what they see, and institutionalizing supportive supervision so that every facility benefits from regular, constructive engagement around its own numbers. The performance monitoring teams that exist in Ethiopian facilities could become the natural vehicle for this work, provided they are equipped and encouraged to function as intended.

There is also a broader lesson for the global health informatics community. DHIS2 is now used in dozens of countries, and the dream of a digital health information backbone is closer to reality than ever. Yet the eastern Ethiopia study is a reminder that the last mile of any information system is human. A dashboard nobody opens is indistinguishable from a filing cabinet nobody opens. The researchers, whose work was financially supported by the Doris Duke Charitable Foundation as part of the Capacity Building and Mentorship Program project, with no funder role in study design or interpretation, obtained ethical clearance from Haramaya University’s Institutional Health Research Ethics Review Committee in accordance with the Helsinki II declaration. Their message to health systems everywhere is clear: to unlock the value of digital health data, invest as much in confidence, trust and supervision as in servers and software. Until the people closest to the data believe in its quality and in their own ability to act on it, the numbers will keep flowing, and the decisions will keep waiting.

Subject of Research: Determinants of District Health Information System 2 data utilization among health workers in public health facilities in eastern Ethiopia

Article Title: Determinants of district health information system 2 data utilization in public health facilities in Harari Regional States and Dire Dawa City Administration, Eastern Ethiopia

Article References: Determinants of district health information system 2 data utilization in public health facilities in Harari Regional States and Dire Dawa City Administration, Eastern Ethiopia. (n.d.). https://doi.org/10.1186/s12913-026-15620-w

Image Credits: AI Generated

DOI: 10.1186/s12913-026-15620-w

Keywords: DHIS2, health information systems, data utilization, Ethiopia, public health facilities, health informatics, supportive supervision, data quality, evidence-based decision making, routine health information systems, Haramaya University, performance monitoring