Health information systems rarely fail because the software does not work. They fail because the data going in is incomplete or late, because the reports coming out go unread, or because the system is quietly abandoned when the project that paid for it ends. In each case, the technology may be perfectly sound. What failed was the fit between the system and the people, routines and institutions around it.
Each of those failures is shaped by decisions taken at the design stage, often before a line of code is written. That is good news, because it means they can be avoided.
This explainer is for ministries of health, programme teams and funders planning a new reporting system, a dashboard, a mobile data collection tool or an extension to an existing platform. It sets out the design choices that most affect whether a system is adopted and sustained, illustrates them with a worked example, and suggests how to measure whether a system is genuinely in use.
Why routine health data matters
Routine health information systems collect data from health facilities and community health workers as part of normal service delivery: patients seen, children immunised, deliveries attended, tests performed, medicines dispensed. Unlike surveys, which are expensive and infrequent, routine data is available every month, for every facility and district.
When it is complete and reliable, routine data lets managers spot a fall in immunisation coverage before an outbreak, see which facilities are about to run out of essential medicines, allocate staff and supplies where they are needed, and track progress against national targets. When it is incomplete or unreliable, decisions fall back on guesswork, and the burden of reporting brings little benefit to anyone.
Why systems fail: a useful framework
A helpful way to think about failure comes from the PRISM framework (Performance of Routine Information System Management), developed by MEASURE Evaluation. It groups the factors that determine whether a routine information system performs well into three kinds:
- Technical factors: the design of forms, indicators, software and processes.
- Organisational factors: governance, resources, training, supervision and a culture of using information.
- Behavioural factors: the knowledge, skills, motivation and confidence of the people who collect and use the data.
Most system projects invest heavily in the first group and much less in the other two. Yet a well-built system will still fail if supervisors never look at the data, if staff do not see why it matters, or if there is no budget to keep it running. Good design addresses all three from the start.
Start with the person entering the data
In most health systems, data is entered by a facility nurse, a records officer or a community health worker, alongside clinical work and often after it has already been written in a paper register. Every extra form, field or password falls on that person. Their experience of the system determines the quality of everything downstream.
Watch the current process before designing a new one
Sit with the people who complete registers and monthly summaries. Follow a patient visit from arrival to the final tally. Note where data is copied twice, where the monthly summary form does not match the register, where staff have improvised their own tally sheets, and how long the monthly report actually takes. This kind of observation, done in a handful of facilities of different types and sizes, routinely reveals problems that no requirements workshop in a capital city would surface.
Collect less
Every indicator should have a named user and a decision it informs. Programmes tend to add indicators over time and rarely remove them, so forms grow until they are a burden. Indicators that nobody uses add work and lower the quality of the data that does matter, because tired staff fill in long forms carelessly.
A useful discipline is to ask, for every proposed data element, "who will look at this, how often, and what will they do differently as a result?". If there is no good answer, leave it out.
Follow the clinic's workflow
Forms should follow the order in which information arrives in the clinic, use the same terms and categories as the paper tools, and check entries as they are made. Validation that flags an impossible value at the moment of entry, such as more children vaccinated than were registered, is far more useful than a data quality review weeks later, when the person who entered the value can no longer remember why.
Plan for the connection people actually have
Unreliable power and mobile data are normal in many facilities, particularly in rural areas. Systems should let staff capture data offline and synchronise later, use pages that load quickly on slow connections, and show clear confirmation that data has been sent. Mobile tools should work on the devices staff actually have, not only on new ones. These are basic requirements, not extras.
Design for decisions, not just reports
A system that only moves data upwards, from facility to district to state to national level, gives the people producing the data little reason to care about its quality. Systems get used, and data improves, when every level gets something useful back.
Give facilities their own data
A simple view of a facility's own trends, coverage and stock levels, compared with its targets, turns reporting from a chore into a management tool. When a facility in-charge can see that antenatal attendance has dropped for three months running, the data has value to them, and they have a reason to make sure it is accurate.
Build dashboards around questions
Start from the questions a district health team or programme manager asks each month, and design the view that answers each one. For example:
- Which facilities have falling immunisation coverage?
- Which facilities are about to run out of a key commodity?
- Which facilities have not reported, or reported late?
- Where are outcomes improving, and what are those facilities doing differently?
A dashboard with fifty charts is usually a sign that this step was skipped. A dashboard built around five questions, each answered clearly, is far more likely to be opened.
Make data quality visible
Show completeness, timeliness and simple consistency checks alongside the data itself, so managers know how far to trust a figure and where to follow up. The World Health Organization's data quality review guidance offers a set of standard checks, including completeness of reporting, internal consistency and consistency over time, that can be built into routine dashboards.
Tie data review to existing routines
Data gets used when it has a place in meetings that already happen: monthly district reviews, supportive supervision visits, quarterly programme performance reviews and annual planning. Designing those routines, including who prepares what, who attends and what decisions are expected, is part of designing the system. A dashboard that is not linked to any meeting or decision will quickly be forgotten.
Build on what already exists
Many ministries already run a national platform for routine health data, and many have also seen a succession of parallel systems built by individual projects and funders. Each parallel system adds reporting work for facilities, often asking for the same data in different formats, and splits information across tools that do not talk to each other.
Extend before you build
Where a national platform such as DHIS2 is in place, adding a programme's indicators, a tracker for individual patients or a new module to it is usually cheaper and more sustainable than building a standalone tool. It also means facility staff use a system they already know, and the programme's data sits alongside everything else the ministry collects.
DHIS2 is open-source software developed by the HISP Centre at the University of Oslo, and is used by ministries of health in many countries for routine reporting, disease surveillance and programme tracking. It can be configured, rather than custom-coded, for most purposes, and its Android app supports offline data capture. Other national platforms exist, and the same principle applies to them: strengthen the shared system before adding a new one.
Use open standards to connect systems
Where separate systems are justified, for example electronic medical records in hospitals, laboratory information systems or logistics management systems for medicines, agreed standards make it possible to connect them. HL7 FHIR has become a widely used standard for exchanging health data. Shared master lists of facilities, administrative areas and indicators, maintained by the ministry, are just as important: without them, two systems cannot agree on what they are counting or where.
Architecture frameworks such as OpenHIE describe how these components can fit together nationally, with shared services for facility registries, terminology and interoperability.
Work within national governance
Ministries increasingly set architecture, standards and approval processes for digital health, often through a national digital health strategy and a technical working group. Working within these processes from the start, even when they slow things down, avoids building a system that the ministry will not adopt or cannot maintain. The Principles for Digital Development, endorsed by many governments and funders, make the same case: understand the existing ecosystem, design for scale and sustainability, and reuse and improve rather than starting again.
A worked example
The following is a hypothetical example, not a description of any real programme. A state immunisation programme wants a dashboard to track routine immunisation coverage and vaccine stock across its health facilities.
The first proposal is a standalone web application, built by a contractor, with its own monthly reporting form covering 40 data elements and a dashboard of 30 charts. Facilities would report into it in addition to the national platform.
After a short design review, the programme changes course:
- It observes the reporting process in eight facilities and finds that most of the 40 data elements are already reported monthly through the national platform, and that facility staff spend most of a day each month on reports.
- It reduces the new data elements to six, covering vaccine stock and cold chain status, and adds them to the existing monthly form on the national platform rather than creating a new one.
- It designs a dashboard around four questions agreed with district immunisation officers: which facilities are below target for key vaccines, which have stock below the minimum level, which have cold chain problems, and which have not reported.
- It adds a simple facility view so that each facility can see its own coverage and stock trends.
- It agrees that the dashboard will be reviewed at every monthly district meeting, and that supervisors will use it to plan visits.
- It agrees with the ministry which unit will own the configuration after the programme ends, and trains two ministry staff to maintain it.
The revised design costs less to build, places a lighter burden on facilities, and is far more likely to be used and sustained. None of the changes required new technology. They required asking the right questions before building.
Plan the handover from the first day
Many systems are built with project funding and expected to be run by government afterwards. Whether that happens depends on the transition being planned, costed and owned from the beginning, not negotiated in the last months of a project.
Agree ownership early
Decide which unit will own the system, who will administer it day to day, who approves changes, and who owns the data. Put it in writing before development starts. Ambiguous ownership is one of the most common reasons systems decline after a project ends.
Cost the whole life of the system
Hosting, support, software updates, refresher training for new staff, devices and connectivity continue long after launch. Estimate these costs over at least five years and agree who will pay them. A system whose running costs exceed what the ministry can budget for will not survive the end of the project, however well it was built.
Build skills inside government
Train ministry staff to configure, maintain and improve the system, not only to use it. Platforms that can be configured without custom code are far easier to sustain, because the skills needed are more widely available and changes do not depend on the original developer.
Protect personal data by design
Health data is among the most sensitive personal data there is. In Nigeria, the Nigeria Data Protection Act 2023 sets out obligations for processing personal data, including health data. Systems should apply role-based access so that users see only what they need, collect only the personal data required, store it securely, and be governed by clear data-sharing agreements. These should be designed in from the start, not added after an incident. Aggregate reporting systems carry lower risk than systems that track individual patients, and the design should reflect the difference.
Measure adoption, not just delivery
Projects often report that a system has been deployed to a certain number of facilities, or that a certain number of staff have been trained. Those figures say little about whether the system is used. More telling measures include:
- the share of facilities reporting completely and on time, and how that changes over time
- how long data takes to reach district and national level after it is entered
- how often dashboards are opened, and by whom
- whether data quality is improving on standard checks
- whether data is cited in review meetings, supervision reports, plans and budget requests
Tracking these from launch shows early where users are struggling, while there is still time to change the design. It also gives funders and ministries a much more honest picture of what an investment has achieved.
A checklist for programme teams
- Have we observed the current workflow in a sample of facilities?
- Does every indicator have a named user and a decision it informs?
- Will the system work offline, on the devices staff actually have?
- Does every level, including facilities, get useful information back?
- Is data review built into meetings that already happen?
- Are we extending the national platform or building something parallel, and why?
- Does the design follow national standards and governance?
- Who will own, run and pay for the system after the project ends?
- How are we protecting personal data?
- How will we know whether the system is being used?
Our founder Chisom Chima is a software engineer on DHIS2 at the HISP Centre, University of Oslo. Five Rivers Consulting brings that product and engineering discipline to the design, delivery and adoption of digital health systems. Talk to us about your system.