A private clinic group signs off on its own telemedicine platform. The board deck shows a video screen with a doctor in a white coat, an app in the Apple and Google stores and a six-month timeline. The first vendor demos a working video call after two weeks, and the board is delighted. Five months later the app works, but clinicians cannot issue an e-prescription through it, the visit summary ends up as a PDF that nobody can attach to the practice management system, and payment only succeeds if the patient does not close the browser tab at the wrong moment.
None of this requires a bad vendor. It only requires a project that started with what is easiest to show rather than what is hardest to build. Video is impressive, but it is a commodity now: you buy it for fractions of a cent per minute. The hard and expensive parts are the visit workflow, the medical record, integration with national e-prescription infrastructure, remote prescribing rules and billing.
This article is for CEOs and executives of healthcare providers and health-tech companies who are weighing whether to build a telemedicine platform, and for anyone comparing telemedicine software development companies. It defends a thesis many vendors will dispute: a telemedicine platform is a records, prescribing and billing system with video as one channel, not a video app with extras. Three decisions follow from that: buy the video, build the clinical workflow, and launch the first version in the browser with e-prescribing working before you spend money on native mobile apps.
What is a telemedicine platform, and why is video the least important part?
A telemedicine platform lets a provider deliver a healthcare service remotely from start to finish: patient registration, identity and consent checks, the consultation itself, documentation, prescriptions, referrals or sick notes, then payment and reporting. Regulators in Europe and the United States treat a remote consultation as a real clinical encounter. Records must be kept to the same standard as an in-person visit, prescriptions are subject to the same or stricter limits, and patients keep the same rights.
That definition reveals a proportion that is easy to miss. The video call lasts fifteen or twenty minutes and covers one step of the visit. The record created during that visit may have to survive for decades, the prescription travels to a national or commercial network and on to a pharmacy, and the payment flows into accounting and possibly an insurer claim. Each of those has its own legal requirements, its own interfaces and its own ways to fail. Video has essentially one failure mode, poor call quality, and choosing a good provider solves most of it.
So why do projects start with video? Because it is the easiest thing to show a board and patients. A video demo can be assembled from an SDK in days. A demo of a correctly issued, signed e-prescription requires a test environment, certificates and an understanding of the document format. The first looks like progress and the second looks like delay, even though the second decides whether the platform can earn money at all.
We looked at the wider picture, including where telemedicine fits alongside AI and connected devices, in our article on digital transformation in healthcare. Here the focus is narrower: a platform that a provider or health-tech company wants to own.
Which modules make up a telemedicine platform?
Before the board asks about cost, it helps to know what it is buying. A telemedicine platform that generates revenue from clinical services usually consists of nine modules. The order below follows how patients and clinicians use them, not the order in which it is wise to build them.
Scheduling. Clinician rosters, visit types (video, phone, chat, in-person), booking, reminders, cancellations and, where public payers are involved, synchronisation with central booking or referral systems.
Identity and consent. Patient accounts, identity verification before care, guardians for minors, consent to data processing and to remote care, delegated access to records.
Pre-visit intake. Forms, photo and document upload, medication lists. This module saves clinician minutes, and it is also the place where a platform most easily crosses into medical device territory if it starts scoring the answers on its own.
Consultation channels. Video, voice, chat with attachments. Every channel must leave a trace in the record.
Clinician workspace. Patient view, visit history, notes, coded diagnoses, orders, templates. Clinicians spend their time here, and whether they want to work on the platform depends on this screen.
Prescriptions, referrals and certificates. E-prescribing, referrals, sick notes, reporting to national systems. Technically the hardest module.
Payments and billing. Pay-per-visit, bundles and subscriptions, invoices, refunds, employer and insurer programmes and, in the US, claims to payers.
Records repository. Storage, versioning, sharing with patients and other providers, long-term archiving, export when the platform changes hands.
Administration and audit. Role permissions, access logs, management reporting, handling of data subject or patient access requests.
If you scan this list asking “where is the video?”, you will find it in one of nine items. That is a good starting point for a conversation about budget.
When should you buy a telemedicine platform and when should you build one?
Before deciding to build, give buying a fair hearing. Practice management systems with a telehealth module and white-label telemedicine platforms exist in every major market. For many clinics that is the right answer: e-prescribing integration is already built and maintained by the vendor, and the provider pays a subscription.
Building your own platform makes sense in three situations. First, telemedicine is the product, not a channel. The company sells remote care, such as an employer subscription, a chronic care programme or a service for an insurer, and competes on patient experience or a care model that off-the-shelf systems do not support. Second, the platform has to connect to something off-the-shelf systems do not anticipate: remote monitoring devices, an in-house lab, a partner’s systems. Third, scale is large enough that subscription fees exceed the cost of your own team and dependence on one vendor becomes a business risk.
There is a fourth option, and it is often the best: a hybrid. Your own patient-facing front end and your own care workflow, built on purchased components: video from a communications platform as a service (CPaaS), payments from a processor, sometimes even e-prescribing through an established system with an API. Build what differentiates you, buy what is the same everywhere. We describe this reasoning in more detail in our build vs buy software decision framework.
The counter-argument is serious: a hybrid means more vendors, more data processing agreements and more points of failure. True. But each of those components has a vendor whose only job is keeping it running. A home-grown video server or payment gateway is not cheaper just because it has no invoice. It is more expensive, with the cost spread across salaries and incidents.
Should you build WebRTC video yourself or use a CPaaS video API?
Browser video consultations rely on WebRTC. The standard is free and built into browsers, but a working service needs signalling servers, TURN servers to traverse firewalls, media servers for calls with more than two participants, recording if you need it at all, quality monitoring and a team that keeps it running at night and on weekends.
The alternative is a CPaaS video API: the provider supplies the SDK and infrastructure, and the platform pays per minute. As one example, Vonage’s Video API documentation lists USD 0.0041 (EUR 0.00373) per participant minute for sessions with up to 25 active publishers, with HIPAA compliance and a business associate agreement offered as a separate subscription priced by sales. Do the arithmetic: a 20-minute consultation with two participants is 40 participant minutes, about USD 0.16. Ten thousand such visits a month come to roughly USD 1,640. That is less than the monthly cost of one engineer maintaining your own infrastructure.
Four questions matter when choosing a video provider. Where are streams routed and processed, and can you pin an EU or US region? Will the provider sign a GDPR Article 28 data processing agreement or a HIPAA business associate agreement, and who are its sub-processors? Does it record by default (it usually should not; recording needs its own legal basis and decision)? How does it behave on a weak connection, because many patients join from a phone on a mobile network, not from fibre?
Under HIPAA the BAA point is not optional. 45 CFR 164.308(b)(1) allows a covered entity to let a business associate handle electronic protected health information only after obtaining satisfactory assurances that it will safeguard it. A video vendor that will not sign a BAA is not a candidate for a US covered entity, however good its demo is.
Your own WebRTC infrastructure makes sense in narrow cases: volumes in the millions of minutes a month, a requirement that data never leave your own data centre, or features the market does not sell. For most platforms in their first two years it is an investment in the least differentiating part of the system.
One practical note: put video behind your own internal interface so it can be swapped. If the provider changes its pricing or terms, switching should take weeks, not a rewrite of the clinician workspace. A well-defined internal API between the visit workflow and the communication channel costs little at the start and a lot if you have to retrofit it. Our glossary entry on creating APIs covers the basics.
What does e-prescribing integration actually involve?
E-prescribing is where most telemedicine revenue is realised and where most projects underestimate the work. Every market has its own infrastructure, and none of it is a single API call.
In the EU, prescriptions run through national systems. Poland is a useful example because it is fully digital: e-prescriptions have been mandatory since January 2020 and e-referrals since January 2021, both routed through the P1 platform run by the national e-Health Centre (Centrum e-Zdrowia). Documents follow a national implementation of HL7 CDA and must be signed by an authorised clinician. Access requires TLS and WSS certificates issued by the P1 Certification Centre, valid for two years, testing on a dedicated integration environment and a separate application for production access. Sick notes go through a different system run by the social insurance institution, so a platform that issues them needs a second integration. Germany’s E-Rezept, mandatory since January 2024, runs through its own telematics infrastructure with its own access requirements. Other member states differ again.
In the US, prescriptions run through commercial networks and federal rules. Electronic prescriptions are routed through e-prescribing networks, and electronic prescriptions for controlled substances fall under DEA requirements for identity proofing, two-factor authentication and certified applications. Most platforms integrate through a certified e-prescribing vendor rather than building this themselves.
Three consequences matter for a board. First, e-prescribing integration is a family of tasks, not one task: each document type has its own validation rules, error messages and edge cases (cancellations, corrections, prescriptions for controlled substances, compounded medicines). Second, the infrastructure changes. National agencies and networks publish new versions of their specifications, and a platform that stops keeping up will one day stop issuing prescriptions. That is a recurring cost, not a one-off. Third, obligations keep growing. In Poland, for example, the national health fund announced in June 2026 that from 1 July 2026 it would pay for first-stage services (cardiology and cervical and breast cancer screening) only when the visit had been booked through central e-registration, which requires clinics to keep schedule data flowing continuously between their software and the central system.
This is why the thesis is phrased as it is. A first version of a telemedicine platform that cannot issue e-prescriptions is not a minimum viable product. It is a prototype you cannot earn money on, because most remote consultations end with a prescription, referral or certificate. If budget or time do not allow full integration in the first stage, consider the hybrid route: e-prescribing through an established certified system with an API, with your own care workflow on top. It is less elegant, but it lets you start delivering paid care.
Which telehealth and prescribing rules have to live in the code?
Regulation in telemedicine does not stop at data protection. Some of it governs how care may be delivered, and it has to be reflected in the platform’s logic, or the system will let a clinician do something that is not allowed.
US controlled substances. Under the Ryan Haight Act of 2008, a practitioner must generally perform an in-person medical evaluation before prescribing a controlled substance via telemedicine. Since the pandemic, the DEA and HHS have repeatedly extended temporary flexibilities that waive this requirement; the fourth extension, issued at the end of 2025, runs through December 31, 2026. A special registration framework proposed in January 2025 had not been finalised as of the law-firm analyses we reviewed. For a platform this means the rule set may change at the end of the year, so the logic that decides whether a remote prescription is allowed must be configurable, not hard-coded.
US licensure. Clinicians generally need to be licensed in the state where the patient is located at the time of the visit. The platform should capture the patient’s location at booking and match it against clinician licences, rather than trusting a profile address.
EU national rules. Member states set their own conditions. Poland again provides a concrete example. A 2023 regulation (Journal of Laws 2023, item 1368) requires prescribers of certain narcotic and psychotropic medicines to verify, through the national system or by interview, what the patient has already been prescribed; where a prescription is issued without an examination, no more than three months may have passed since the patient was last examined, and the verification must be described in the record. Separately, Polish primary care rules define situations in which a remote consultation is not appropriate and an in-person visit is required, such as for young children.
These rules hit online “prescription mill” business models directly. For a responsible platform the design lesson is that rules derived from regulation today may change tomorrow, so they belong in one configurable rules layer, not scattered across screen code. The platform should show the clinician the date of the last examination, block or warn when a prescription falls outside the permitted pathway, and require a verification note.
None of these rules is especially hard to program. What is expensive is having to know them before design starts. A vendor building its first healthcare system will learn about them from the first clinician who files a bug, or from the first audit. Our glossary entry on regulatory compliance outlines how to structure this work across GDPR, HIPAA and sector rules.
How do you design medical records that must be kept for decades?
Retention periods in healthcare outlast most software systems. HIPAA itself does not set a medical-record retention period; in the US that is a matter of state law, and periods differ by state and by patient age. In the EU it is national law. Poland is at the long end: under Article 29 of its Act on Patient Rights, records must be kept for 20 years from the end of the year of the last entry, 30 years where the patient died from injury or poisoning, and 22 years for children under two.
That leads to architectural decisions that have to be made at the start, not the end.
A format independent of the application. Records must be readable in twenty years, when today’s front end, and perhaps today’s vendor, are gone. That argues for standardised document formats and a regularly tested export, not just database backups.
Immutability and versioning. An entry cannot disappear or be overwritten without trace. A correction is a new version recording who changed what and when. The audit log is part of the record, not an add-on.
Retention by document type. If the law sets different periods for different documents, the system must know what type each document is, when the last entry was made and when the document may be deleted. A single retention setting for the whole database will either break the law or break GDPR’s data minimisation principle.
Continuity when the provider changes or closes. Records follow legal successors, and contracts with platform vendors should guarantee a complete handover, not just “access to the dashboard.”
Call recordings. By default, do not make them. A video recording of a consultation becomes health data that needs a legal basis, long-term storage and protection, and its clinical value is usually lower than a well-written note. If your care model requires recordings, that is a decision for the board and the data protection officer, not a default setting in the video provider’s console.
Which GDPR, HIPAA and security requirements change the architecture?
Health data is a special category of personal data under GDPR Article 9. Processing it is prohibited by default and allowed only under specific exceptions, in healthcare usually the provision of care by professionals bound by confidentiality. A telemedicine platform processing such data at scale should undergo a data protection impact assessment under Article 35 before launch, because the findings can change the design. In the US, the HIPAA Security Rule sets administrative, physical and technical safeguards for electronic PHI.
In practice both regimes translate into a handful of architectural properties that are hard to add later.
Encryption at rest and in transit for all patient data, including pre-visit uploads and backups, with keys managed outside the database.
Access control based on the treatment relationship. A clinician sees the records of patients they treat, not the whole database. Break-glass access exists, but it is logged and reviewed.
An access log that answers “who opened this patient’s record, when and why.” Without it you cannot handle a patient complaint or an incident.
Separation of identifying and clinical data where possible, and strict limits on what reaches analytics and marketing tools. The most common failure in health apps is not a breach but an analytics script sending page URLs that contain test names to a third-party service.
A register of processors and business associates. Video, cloud, SMS, payments, support tooling. Each needs an agreement and an assessment of where it processes data.
Cybersecurity regulation adds to this. The EU NIS2 directive covers the health sector, including healthcare providers, and large clients will push supplier security requirements into contracts. The platform should be penetration tested before launch and after every major change; our article on penetration test pricing explains how such tests are scoped and priced.
How should payments, subscriptions and insurer billing work?
Payments look simple until you have to handle a visit that did not happen. The clinician was late, the call dropped after three minutes, the patient did not show up, the visit was rescheduled. Each case needs a rule: full refund, partial refund, credit moved to a new slot, no-show fee. The business sets these rules, not the developer, and they should be written down before the first sprint.
A platform usually supports several models at once: pay-per-visit through a payment processor, visit bundles, monthly subscriptions with renewal, visits paid by an employer or insurer against an eligibility list and, in the US, claims to payers with their own coding and documentation requirements. Each model is a different flow, with different documents and different reports.
Three principles save the most trouble. First, payment must never be a precondition for saving the record. A visit that took place must be documented even if the payment failed; settlement is a separate process. Second, card data should not touch the platform’s servers at all. The processor’s hosted payment form and tokenisation reduce the platform’s PCI DSS scope to a minimum. Third, billing with insurers and employer programmes has its own formats and deadlines, worth understanding before signing partner contracts, because these partners, not card payments, often generate most of the revenue.
Where public payers are involved, reporting and, as the Polish central e-registration example shows, funding conditions increasingly depend on integration with central systems. That is another argument for a single, well-defined register of services delivered, from which both the record and the billing are produced.
When does a telemedicine platform become a medical device, and what does the EHDS change?
Scheduling, video and clinician notes are usually not a medical device. The line moves when software gets a medical intended purpose: it assesses symptoms and routes the patient to a pathway on its own, interprets readings from a blood pressure monitor or glucose meter, suggests a diagnosis or a dose. Then the EU MDR and its Rule 11 for software come into play, or the FDA’s device framework in the US, bringing a quality system, technical documentation and often a notified body. That is a separate topic; we flag it here only so the board knows that a triage or remote monitoring feature is a regulatory decision, not just a product one. If such a module is planned, separate it architecturally from the rest of the platform so that device requirements do not cover the whole system.
The second regulatory horizon in Europe is the European Health Data Space (EHDS), established by Regulation (EU) 2025/327, which entered into force in March 2025. It applies generally from 26 March 2027, and its requirements for electronic health record systems phase in. According to Osborne Clarke’s analysis, systems processing the priority categories of patient summaries, e-prescriptions and e-dispensations will be subject to mandatory CE marking and essential requirements, including the European interoperability and logging components, from March 2029. From March 2031 the same applies to systems processing imaging, test results and discharge reports, and to systems put into service within health institutions.
For a platform built in 2026 this is a practical argument, not a theoretical one. A system that still handles e-prescriptions and records in 2029 will likely fall within these requirements. Designing today around standardised data formats, export and a complete audit log is cheaper than a rebuild in three years.
How much does telemedicine platform development cost?
An honest answer depends on the option, and every figure below is an estimate, not a price. We base it on ARDURA’s published team costs: a typical three-person team (two developers and a QA engineer) costs about EUR 14,500-28,700 a month, roughly EUR 4,800-9,600 per person-month depending on seniority. Then there is duration: our service pages put simple MVPs at 2-3 months, mid-size applications at 4-8 months and enterprise systems at 8-18 months. The factors that move the result most are the number of e-prescribing and document types, the number of payment models, native mobile apps and partner integrations. A US or Western European team would cost more per person-month; the structure of the estimate stays the same.
| Option | What it covers | Team and duration (estimate) | Build cost (estimate) | Biggest risk |
|---|---|---|---|---|
| A. White-label platform | Configuration, branding, training, integration with your practice management system | A few weeks of client and vendor effort | Vendor subscription plus internal costs | Vendor lock-in, no differentiation |
| B. Own care workflow, purchased components | Scheduling, consent, clinician workspace, records, e-prescribing and referrals, CPaaS video, processor payments, browser app | 4-5 people for 5-7 months, i.e. 20-35 person-months | about EUR 100,000-340,000 | Underestimating e-prescribing integration and prescribing rules |
| C. Full product platform | Option B plus native iOS and Android apps, subscriptions and bundles, sick notes, lab and partner integrations, employer or insurer portal | 6-8 people for 9-12 months, i.e. 54-96 person-months | about EUR 260,000-920,000 | Scope spreading to everything at once |
Add the costs boards most often leave out. Maintenance and development: changes in national specifications and e-prescribing networks, library updates, security patches, mobile app compatibility with new OS versions. A rule of thumb, not a fact, is somewhere in the mid-teens to mid-twenties as a percentage of build cost per year, and the upper part of that range for systems tightly integrated with national platforms. Variable costs: video minutes, SMS, payment fees, cloud. Costs on the provider side: clinician time for testing and training, the data protection officer, the DPIA, penetration tests, patient support.
The table also shows why sequence matters. The gap between options B and C is largely native apps and additional integrations. If option B goes live and earns money, the decision on option C is made on patient data rather than slides. If the project starts with native apps and postpones e-prescribing, it pays for the most expensive part before testing whether the business model works. Our MVP development cost guide covers how to budget a first product version in more detail.
How do you choose a telemedicine software development company?
Proposals for a telemedicine platform can differ several-fold, often not because anyone is overcharging but because vendors priced different things. One priced video with a clinician screen, another priced the full workflow with e-prescribing and records. So the key questions come before the price comparison.
Does the vendor understand your e-prescribing infrastructure and can it show how it will integrate? Ask how it will get access to the test environment, who applies for certificates or network access, and which document types the first version covers. “We’ll integrate via the API” without detail is a warning sign.
Does it know the remote prescribing and telehealth rules? Ask how the system will handle in-person requirements, licensure and controlled substance rules in your markets. The vendor does not need to be a law firm, but it must know these rules exist and where it will put them in the system.
How will it design records for long retention? Format, versioning, retention by document type, export when you change vendors. If the answer is “we take backups,” the question was not understood.
What will it buy and what will it build? A good vendor tells you which components are not worth building. A vendor that wants to build everything, including the video server, is building itself a long contract.
Who will be on the team and for how long? Healthcare projects need continuity of knowledge. Ask about team composition, replacement terms and who stays after launch. We covered the competencies an e-health project team needs, and how to phase them, in our article on building the IT team for an e-health project.
Will it propose a discovery phase? A few weeks to pin down scope, rules and architecture, ending with a narrowed estimate, is the cheapest way to make proposals comparable. At ARDURA, discovery on software development projects takes one to two weeks and defines scope and architecture.
How should you plan the first 12 months?
Sequencing is the most important board decision in this project, more important than technology choices. We suggest an order that removes regulatory and integration risk first and invests in reach afterwards.
Month 1: discovery and regulatory decisions. Care model, service types, intended purpose (could any module become a medical device?), DPIA, list of integrations, choice of component vendors, architecture. It ends with a narrowed estimate and a decision on stage two.
Months 2-6: a revenue-generating browser version. Scheduling, consent, clinician workspace, records, e-prescribing and referrals, CPaaS video, pay-per-visit. In parallel, apply for production access to the e-prescribing infrastructure and test in its integration environment. Before launch, run penetration tests and tests with clinicians on real scenarios.
Months 6-8: limited launch. One specialty or one patient group, measuring visit duration, dropped-call rate, share of visits ending with a document, clinician satisfaction. These are the data on which the board decides further investment.
Months 8-12: expansion based on data. Native apps, if the data show patients need them (a well-built progressive web app with notifications is often enough), subscriptions and bundles, sick notes, partner integrations.
ARDURA works in two-week sprints and delivers working increments from the first sprint, so the board can see every two weeks what actually works rather than a progress report. At each stage, write down the criteria for moving on. If after six months the platform does not issue e-prescriptions in production, that is not the moment for native apps; it is the moment to discuss what is blocking the integration.
What are the key takeaways before deciding to build a telemedicine platform?
A telemedicine platform is a records, prescribing and billing system in which video is one channel. Starting with video means starting with the cheapest part and postponing the most expensive one, while the board judges progress by what is easiest to show.
Remember six things. First, give buying a white-label platform or a hybrid a fair hearing; building your own is justified by telemedicine as a product, unusual integrations or scale. Second, buy video from a CPaaS provider and hide it behind your own interface, because at around sixteen US cents a visit there is no case for your own infrastructure. Third, treat e-prescribing integration as part of the minimum viable product, not a later add-on, and budget its maintenance as a recurring cost. Fourth, encode telehealth and prescribing rules, from the DEA flexibilities to national rules in the EU, as configurable policies in one place. Fifth, design records for long retention and with the EHDS in mind, which from 2029 covers systems handling e-prescriptions. Sixth, sequence the work: discovery, a revenue-generating browser version, a limited launch, and only then native apps and expansion.
If you are preparing a decision on a telemedicine platform, see how we approach custom software development and platform engineering, from discovery to product growth, or how we can strengthen your team with healthcare IT specialists who join within two weeks. The simplest first step is a conversation about scope: after discovery you get a scope, an architecture and an estimate you can take to your board.