Governments deploying a national EHR face a choice between open source and proprietary SaaS solutions. What if there were a third way?
Insight
The state of public sector digitization in LMICs today
Most low- and middle-income countries (LMICs) are yet to digitize their health systems. Africa CDC estimates that only 30% of the continent's health systems are digitalized, with primary health care significantly below this figure. Governments that are seeking to digitize increasingly want to own the system that holds their citizens' health records.
The events of 2025 strengthened this position. When US funding froze, USAID-funded EHR systems such as KenyaEMR became inaccessible. These systems were hosted with US-funded partners, and some clinics faced significant disruption, losing access to patient records.
Even though KenyaEMR is built on OpenMRS, the most widely used open-source EHR, many LMIC governments still see open source as the most effective way to own their national EHR and reduce their dependence on third parties.
Open source: the promise vs the reality
To an extent open source does provide the protection of "ownership". OpenMRS carries no license fee and gives the government full access to the code, so there is no risk of lock-in to a single vendor. Access to the code, however, does not necessarily mean that a government does not rely on others to run and maintain it. Most national deployments depend on donor funding and on implementing partners or contracted developers for hosting, upgrades and changes, as the KenyaEMR outage showed.
Open source is also widely seen as an affordable option, because it carries no license fee. However, the license fee is only a small part of what a national EHR costs. In Kenya, software made up only 9% of the cost of rolling out KenyaEMR, with the rest going on hardware, hosting, support and training. As a result, running costs ranged from $2.28 to $41.40 per patient per year depending on facility type. This includes recurring software costs as a result of the need for internal or contracted IT teams to maintain, secure and upgrade each deployment.
- 9%
- of KenyaEMR rollout costs went on software
- $2.28 to $41.40
- KenyaEMR running cost per patient per year
The perceptions of greater control and lower cost lead governments to accept the technical drawbacks of open source. OpenMRS usually works offline only through a server on a local network at each facility, which needs power backup and maintenance, adding a substantial hardware cost for a primary care clinic. Bahmni is slightly better, offering an offline Android app, but it covers only registration and clinical forms, with orders, medications and diagnoses not able to be recorded until the device reconnects to the facility server. This makes it challenging to cover primary care clinics cost-effectively at scale with currently available open source software solutions. At the same time, the software itself also requires significant training, which has cost implications as well as affecting long-term adoption rates and the quality of the data captured on the system.
Keeping open source based systems current in today's world of rapidly advancing software capability is also a challenge. Each national instance carries its own customized forms, reports and modules, so every upgrade has to be retested and adapted before it is installed facility by facility, often on local servers with limited connectivity and on-site support. Health centers surveyed in Rwanda for a study published in 2022 were still running OpenMRS 1.6 and 1.11, and community figures from March 2025 show only around 3,000 of the 8,100+ reported OpenMRS sites on the current version, OpenMRS 3.
The technological strengths of proprietary solutions are often overshadowed by the question of ceding technical control to a third party
Proprietary platforms such as our own ElephantOS address many of the technological gaps left by open source solutions. ElephantOS works fully offline on each device, so a facility needs only tablets and SIM cards in place of local servers. Clinical decision support is built into the clinical workflow and the interface is designed for speed of documentation, using a "tapping not typing" design, with health workers onboarded in less than two hours. Under a software-as-a-service (SaaS) model, vendors like Elephant spread their engineering costs across customers, and the solution continuously evolves and upgrades as part of the ongoing service. This makes ElephantOS more capable and more cost-effective than open source solutions, priced at around $1 per patient per year including hardware, hosting, support and training, against $2.28 to $41.40 for KenyaEMR.
These cost differences make a substantial difference to the cost of deploying an EHR when scaled up across a primary care network, as seen in the example below.
| Unit | Cost | OpenMRS / OpenClinic | ElephantOS |
|---|---|---|---|
| Per primary health clinic | Local server | $1,500 | Not needed |
| Power backup for the server | $1,500 | Not needed | |
| Local network | $800 | Not needed | |
| Server running costs, per year | $300 | Not needed | |
| IT support, per year | $800 | Included in support contract | |
| Per 1,000 primary health clinics over 5 years | Upfront server hardware | $3.8m | Not needed |
| Estimated server and IT support running costs | $5.5m | Included in support contract | |
| National software license | No license fee | $0.75m to $2m | |
| Total | $9.3m | $0.75m to $2m |
In addition to lower total costs of ownership, Elephant has addressed some of the commercial concerns of governments around vendor lock-in and control of pricing by offering multi-year licenses or a patient-pay model where patients fund the ongoing costs rather than the government.
However, in many cases, lower costs and stronger technology do not alleviate governments' concerns about ceding technical control over a critical piece of their national health infrastructure to a third party, who may not honor its service-level agreements, or may cease trading.
Neither model, open source or proprietary SaaS, is giving ministries what they ask us for: a system they control, that is continuously improving, at an affordable long-term cost.
How high-income health systems handle the "control" trade-off
Large hospital systems in the US and Europe face the same tension. To address it, they typically license their EHR from a vendor such as Epic or Oracle Health on perpetual terms, and pay an annual maintenance fee for new releases and support. The health system then employs its own team of analysts to configure the EHR to its workflows, forms and reports. The vendor keeps the core software current and the health system tailors it to its needs. Control over the system sits with the people running it, while the cost of building and maintaining the software is shared across the vendor's customers. This model shows that ownership and continuous technological improvement can coexist.
What the ideal model for LMIC governments needs
A third option for these governments would keep what works in the high-income model and adjust the parts that don't fit. From our conversations with ministries over the years, we believe the ideal model needs six things.
- 1
Ownership that lasts. A one-off payment for a perpetual national license, which fits how donors and ministries fund capital investment. The system keeps running if payments stop, patient data stays in-country on servers the ministry chooses, and source code is held in escrow in case the supplier ceases to trade.
- 2
Control without an engineering team. Ministry staff can change forms, registers, reports and programs themselves. Each change has safeguards built in, ensuring that it is tested before it goes live, approved by a named person, logged and reversible.
- 3
One shared core. Every country runs the same release, and local changes are stored as configuration so upgrades don't break them. Engineering costs are shared across countries, so each government benefits from improvements paid for elsewhere.
- 4
Fit for primary care. Offline use on each device with no facility server, clinical decision support built in, and open standards such as HL7 FHIR, ICD-11 and SNOMED so data can move between systems.
- 5
Small, predictable running costs. An optional annual fee for new releases and support, with the system kept running whether or not the optional support fee is paid.
- 6
AI integration. The ability to plug the growing range of proven third-party AI tools into clinical workflows.
What we are building
At Elephant Healthcare we are building this third choice. We call it Elephant SovereignOS. Elephant SovereignOS gives a ministry a perpetual national license to the core OS, with patient data held in-country and an optional annual fee for new releases. Open source solutions and high-income vendors such as Epic rely on an in-house engineering team to adapt the EHR. SovereignOS replaces that team with a built-in AI customization tool that lets ministry staff describe what they need in plain language, such as a new register, screening program or patient messaging campaign, much as tools like Lovable let non-developers build software. Each change is tested on practice patients and goes live only once a named ministry lead approves it.
Governments benefit from the continuously improving technology behind ElephantOS, while owning their instance of the EHR and adapting it to their country's needs without depending on an engineering team, internal or external. The result is a national EHR with a lower total cost of ownership than open source and digital public infrastructure that responds faster to the changing needs of government.
Sources
- Africa CDC, Outcome Statement of the Primary Health Care Digitalization Expert Committee, 27 November 2025
- TechCabal, Africa's digital dependency crisis: how USAID cuts exposed the fragility of Africa's digital infrastructure
- The Standard, Patients on their own as US takes custody of critical healthcare data, 27 March 2025
- Kevany S, Shade S, Waters C, Puttkammer N. Implementation of Kenya Electronic Medical Records (KenyaEMR): costs and efficiencies. Science Journal of Public Health 2023;11(5)
- OpenMRS Talk, Retiring offline in O3, 22 September 2026
- Fraser HSF et al. User perceptions and use of an enhanced electronic health record in Rwanda with and without clinical alerts: cross-sectional survey. JMIR Medical Informatics 2022;10(5):e32305
- OpenMRS, OpenMRS Around the World (updated March 2025)
- Cerner Corporation, Form 10-K for fiscal year 2020
- What contracting with Epic for an EHR license really involves
