Digital Health Economics

Why this matters in health economics

Health systems worldwide are being offered a flood of digital tools: symptom-checker apps, video consultations, wearables that stream vital signs, remote monitoring for long-term conditions, and platforms that promise to move care out of hospitals and into homes. The promise is seductive because the marginal cost of serving one more user of a piece of software is close to zero, so the story goes that digital scales where clinics cannot. That economic peculiarity — high fixed cost, near-zero marginal cost — is real, but it is only half the ledger, and a director who acts on the promise without the proof is spending public money on faith.

The stakes are concrete. A telehealth service that substitutes for face-to-face care can free clinician time and cut travel, or it can add a video call on top of the same in-person visit and raise total cost while looking like progress. A remote monitoring programme can catch deterioration early and avoid admissions, or it can generate a torrent of alerts that swamp staff and change nothing. The difference between these outcomes is not the technology; it is whether the tool was evaluated honestly, procured against a standard, and designed for the whole population rather than the confident, connected minority who adopt first.

Two features make the economics distinctive enough to warrant its own chapter. First, digital prevention shares the long-lag problem that breaks standard return-on-investment tools: the benefits of monitoring blood pressure or supporting behaviour change may not show up as avoided events for years, so a tool that is genuinely cost-effective over a decade looks like pure cost inside a one-year budget (the evaluation machinery for this belongs to Chapter 2.1 — Economic Evaluation). Second, digital tools redistribute access — often away from older, poorer, rural, and less digitally literate people — so a tool that is efficient on average can widen inequality, turning an efficiency gain into an equity loss the average figure never shows.

Core concepts

Digital health is the umbrella term for the use of information and communication technologies to support health and care. Underneath it sit overlapping labels worth keeping straight. eHealth is the older, broad term for electronic health services and records; mHealth narrows to mobile devices — smartphones, tablets, and the apps and messaging that run on them. Telehealth covers the delivery of care and health information at a distance, with telemedicine usually reserved for remote clinical services specifically, such as a video consultation with a doctor. Remote patient monitoring uses devices — blood-pressure cuffs, glucose sensors, pulse oximeters, wearables — to collect physiological data outside a clinical setting and route it to a care team. Digital therapeutics are software products that deliver an evidence-based intervention to treat or manage a condition, positioning software itself as the treatment rather than a channel for it.

The economic point that unites these is the cost structure of software: most of the money is spent once, building and validating the product, after which each additional user costs very little to serve. This is why digital tools are pitched as answers to scarcity — and why the pitch can mislead. Near-zero marginal cost is a property of the software, not of the care around it; the clinician who reads the monitoring data, the helpdesk that resets the password, and the home visit for the patient whose device failed are all marginal costs that do not vanish.

Evidence standards for digital health exist because ordinary health technology assessment (see Chapter 2.4 — Pharmacoeconomics and Chapter 2.1 — Economic Evaluation) was built for drugs and devices that are fixed at launch, whereas software updates continuously and often reaches market with thin evidence. Several bodies have written frameworks to grade what proof a digital tool must show for its claimed function and financial risk. England's National Institute for Health and Care Excellence (NICE) publishes an Evidence Standards Framework for digital health technologies, tiering the evidence expected by the tool's potential for harm and the nature of its claim. The National Health Service (NHS) complements it with the Digital Technology Assessment Criteria (DTAC), covering clinical safety, data protection, technical security, interoperability, and usability. The World Health Organization (WHO) has issued a classification of digital health interventions and guideline recommendations on which interventions the evidence supports, giving lower- and middle-income systems a reference point that is not tied to any one country's payer.

Data as an economic asset is the second distinctive idea. The data a digital tool generates has value beyond the immediate care episode — for population health, research, service planning, and, commercially, for the vendor. That value is real but hard to price, and it raises questions of ownership, consent, and benefit-sharing that ordinary procurement ignores. Data governance — the rules for who may hold, use, and profit from health data — is therefore an economic question, not merely a legal one, because the terms on which you let a vendor keep your population's data can be worth more than the price on the invoice. Data also drives network effects: a platform that gets better as more people use it can entrench a single vendor, creating lock-in that raises long-run cost even when the initial price was low.

Digital exclusion is the equity counterpart. The digital divide — unequal access to devices, connectivity, and the digital literacy to use them — maps closely onto the social gradient in health, so the people least able to use digital tools are often those with the greatest need. When a system shifts a service to digital-by-default and treats the excluded as a rounding error, it converts a private disadvantage into a public one, and the opportunity cost falls on those already worst served (the equity machinery is developed in Chapter 3.4 — Equity).

Best practices

1. State what the tool replaces before you count what it costs. A digital tool is substitutive, additive, or supplementary, and the economics turn entirely on which. A video consultation that replaces a clinic visit can save money; one that is followed by the same in-person visit adds cost while feeling like progress. Define the counterfactual — what would happen to this patient without the tool — and build the evaluation around the difference, not the digital activity in isolation.

2. Match the evidence demanded to the tool's function and risk, not to its novelty. A wellness app that offers general information warrants far less proof than a digital therapeutic that guides treatment or a monitor that triggers clinical action. Use a tiered framework — NICE's Evidence Standards Framework for digital health technologies, the NHS Digital Technology Assessment Criteria, or the WHO classification and guideline recommendations as international reference points — to set the bar by potential harm and by the strength of the claim. Demanding a randomized trial of a low-risk information tool wastes money; accepting testimonials for a tool that changes treatment risks lives.

3. Cost the whole pathway, including the human work the software creates. Near-zero marginal cost is a property of the code, not of the care. Remote monitoring generates alerts that someone must triage; apps generate messages a clinician must answer; every deployment needs onboarding, support, and the fallback for when devices fail. Count these downstream costs, because a tool that shifts work rather than removing it can raise total system cost while cutting the line item you were watching.

4. Model the long lag explicitly, and do not let a one-year budget judge a ten-year benefit. Digital prevention and monitoring often pay back slowly: the avoided stroke, the prevented admission, the slowed progression may be years away. Standard return-on-investment tools built for quick paybacks will score a genuinely cost-effective tool as pure cost, exactly as they mis-score dementia and cardiovascular prevention. Use a time horizon long enough to capture the benefit, discount appropriately, and be explicit that the payer bearing the cost may not be the one collecting the saving (the methods live in Chapter 2.1 — Economic Evaluation).

5. Evaluate against the population you will actually serve, not the early adopters. Pilots recruit the willing, the connected, and the digitally confident, so pilot results flatter the tool and understate the cost of reaching everyone else. Stratify results by age, income, language, disability, and connectivity, and plan for the non-digital pathway from the start. If the tool only works for the people who were doing well anyway, it is an efficiency gain that widens inequality.

6. Treat data rights as part of the price. When a vendor keeps, reuses, or resells the data your population generates, they are extracting value that never appears on the invoice. Negotiate ownership, permitted uses, consent, and benefit-sharing as commercial terms, and price the deal accordingly. Data governance is an economic lever: giving away a population's longitudinal data can cost far more, over time, than the licence fee you saved.

7. Guard against lock-in and demand interoperability. A platform that becomes more valuable as more clinicians and patients join can entrench one vendor, and switching costs then let that vendor raise prices or let quality slide. Require open standards and data portability so tools can be replaced, and weigh the long-run cost of dependence against the short-run convenience of a single integrated supplier.

8. Budget for digital inclusion as a line item, not a hope. Reaching excluded groups costs money — devices, data allowances, assisted digital support, translated content, staff time — and if it is not funded it will not happen. Decide whether the digital route supplements or replaces the traditional one, and if it replaces it, fund the inclusion measures that keep the service universal. An unfunded inclusion plan is a decision to exclude.

9. Beware the app graveyard: plan for evidence decay and abandonment. Software changes, engagement falls off, and evidence generated on one version may not hold for the next. Build in re-evaluation triggered by major updates, monitor real-world use rather than trusting launch trials, and set explicit criteria for decommissioning tools that stop delivering. A tool that worked in the trial and is ignored in practice is spending money for nothing.

10. Keep the boundary with artificial intelligence clean. Many digital tools now embed machine learning, and the economics of continuously-learning, opaque algorithms — validation that must repeat, algorithmic bias as an equity cost, paying for a model that changes after purchase — differ enough to warrant separate treatment (see Chapter 5.3 — AI Health Economics). Assess the general digital tool on the practices here; where a claim rests on an adaptive algorithm, apply the additional scrutiny that chapter sets out.

11. Distinguish access gains from health gains. Telehealth reliably improves convenience and access, and that has genuine value, but access is not the same as outcome. A programme can raise the number of consultations and patient satisfaction while leaving clinical results unchanged. Be clear which you are buying, value convenience honestly rather than smuggling it in as a health effect, and hold outcome claims to outcome evidence.

Questions to discuss with your team

1. What exactly does this tool replace, and are we sure the old cost actually goes away? This question exposes the most common digital business case error: counting the new tool's benefits while quietly keeping the old pathway running. Trace a real patient through both the current and the proposed journey and ask, at each step, whether the digital step removes work or merely adds to it — does the video consultation end the episode, or does it precede the same face-to-face visit? Look hard at whether clinic slots, ward beds, or staff posts are genuinely released, or whether the saving is notional because the freed capacity is immediately filled by other demand. The honest answer distinguishes cash-releasing savings from capacity that is simply redeployed, and it names who actually banks any saving, since the budget bearing the cost of the tool is often not the one that collects the benefit. If the team cannot point to a cost that concretely disappears, the case rests on access and convenience, which are worth having but should be argued on their own terms.

2. Who in our population cannot or will not use this, and what happens to them? Digital tools redistribute access, usually along the same gradient as ill health, so this question decides whether a proposal is an efficiency gain or an equity loss dressed as one. Map your population by age, income, first language, disability, digital literacy, and connectivity, and estimate how many fall on the wrong side of each line — then ask concretely what their pathway becomes if the service goes digital-by-default. The tension is real and financial: serving the excluded properly costs money that erodes the efficiency saving the digital route promised, so there is always pressure to treat them as a small exception. An honest answer resists that, funds the non-digital route or the inclusion support explicitly, and states plainly whether the digital tool supplements the old service or replaces it, because replacement without inclusion is a decision to widen inequality. It also names the monitoring that will tell you, after launch, whether the gap is narrowing or growing.

3. When will the benefit actually arrive, and can our budget process wait that long? Many digital tools — especially in prevention and long-term-condition monitoring — pay back over years, and standard budgeting rewards what shows up this year, so a genuinely worthwhile tool can fail on a timescale mismatch rather than on merit. Ask over what horizon the claimed benefits materialize, how confident the evidence is at that horizon, and what the tool looks like on a one-year, three-year, and ten-year view. Name the split between cost and benefit across payers and time: the community budget that funds the monitor may not be the acute budget that avoids the admission three years later. The honest answer chooses a time horizon long enough to be fair to a slow-acting tool, acknowledges the discounting and the uncertainty that grows with distance, and is candid that a short-horizon return-on-investment tool will mis-score it. It also decides, in advance, what interim signals would justify continuing to fund something whose ultimate payoff is not yet visible.

4. What evidence does this tool actually need to earn its claim, and has the vendor met that bar? Digital procurement fails at both extremes: demanding a randomized trial of a harmless information app wastes money and delays a useful tool, while accepting glossy engagement statistics for a product that changes treatment risks patients. Locate the tool on a tiered framework — NICE's Evidence Standards Framework, the NHS Digital Technology Assessment Criteria, or the WHO classification — by what it claims to do and the harm it could cause if wrong, then ask whether the evidence on offer matches that tier or a lower one. Interrogate what the evidence is really made of: is it outcomes or only usage, drawn from the population you will serve or from a self-selected pilot, generated on the current version or an earlier one? The honest answer names the gap between the tier the tool sits in and the tier its evidence supports, and decides whether to reject, to commission the missing study, or to buy under a managed evaluation with agreed success criteria. It also resists the vendor's framing that novelty or momentum should substitute for proof, and holds a treatment-grade claim to treatment-grade evidence.

5. Who will own and control the data this tool generates, and what will it cost us to leave? The invoice price is often the least important number in a digital deal, because the vendor may be acquiring something more valuable — a longitudinal record of your population's health that it can reuse, resell, or use to entrench itself. Ask explicitly who owns the data, what secondary uses are permitted, whether consent and benefit-sharing are addressed, and whether you could export your records and switch supplier without losing them. Probe the network effects and proprietary formats that create lock-in: a platform that grows more valuable and more entangled as adoption rises can quietly convert a cheap entry price into an expensive dependence, letting the vendor raise fees or let quality drift once you cannot leave. The honest answer treats data rights, interoperability, and portability as priced commercial terms negotiated up front, not legal boilerplate discovered at renewal, and it is candid that giving away a population's data can cost far more over time than the licence fee it saved. It also names who inside your organization is accountable for defending those terms against a determined counterparty.

6. What is our plan for when the software changes, engagement fades, or the tool must be retired? Unlike a drug fixed at licensing, software updates continuously, its users drift away, and evidence generated on last year's version may not hold for this year's — so a tool that shone in the trial can quietly stop delivering while still drawing budget. Ask what will trigger re-evaluation, how real-world use and outcomes will be monitored rather than assumed, and what explicit criteria would tell you a tool has entered the app graveyard and should be decommissioned. Name the sunk-cost trap honestly: after a costly rollout and staff retraining, the pressure is to keep paying for a tool nobody uses rather than admit the investment is stranded. The honest answer builds re-evaluation on major updates and a decommissioning rule into the contract from the start, funds the monitoring that would reveal decay, and treats retirement as a normal, planned event rather than a failure to be hidden. It also asks who is watching, because a tool with no owner after go-live will decay unobserved.

In practice: a health economics example

The provincial health office of Nusa Timur — a fictional island province in a middle-income country, with a mix of a crowded coastal capital and scattered rural islands reachable only by ferry — is deciding whether to fund a remote monitoring and telehealth programme for hypertension. Roughly a third of adults have raised blood pressure, most undiagnosed or poorly controlled, and stroke and heart disease are the leading causes of premature death. The proposal, from a regional health-technology company, pairs a low-cost connected blood-pressure cuff with a smartphone app and a central nursing team who review readings and adjust medication remotely. The vendor's pitch quotes a striking return on investment and near-zero cost per additional patient.

The office's economist, Dr Halim, starts by fixing the counterfactual and the perspective. Current care requires a clinic visit that many islanders skip because of ferry costs and lost work; the realistic alternative to the programme is continued poor control, not perfect clinic attendance. Taking a health-system perspective over a ten-year horizon — chosen deliberately, because hypertension control prevents events that are years away — she rebuilds the case. The vendor's one-year return-on-investment figure collapses almost to nothing, exactly as the long-lag problem predicts: within a single budget year the programme is almost pure cost, and only over a decade of avoided strokes and admissions does it become cost-effective. She flags this to the finance director explicitly, because the community budget that would fund the cuffs is not the hospital budget that would bank the avoided admissions.

Next she costs the whole pathway rather than the software. The cuff and app are cheap, but the central nursing team, the helpdesk, the medication logistics, and the home support for patients whose devices fail are not — and the near-zero marginal cost the vendor quoted turns out to describe only the code. She then applies an evidence lens drawn from the NICE Evidence Standards Framework and the WHO guideline recommendations: because the tool triggers medication changes, it sits in a high-risk tier that demands real outcome evidence, and the vendor has offered engagement statistics and a small uncontrolled pilot instead. The pilot, moreover, recruited smartphone-owning residents of the capital — precisely the group least representative of the rural islanders the programme is meant to reach.

That last point reframes the decision. Stratifying the population, Dr Halim finds that the islanders with the worst control are the least likely to own a suitable phone, have reliable connectivity, or read the app's language comfortably. A digital-by-default rollout would improve control among the already-advantaged capital population and leave the rural gradient untouched or wider — an efficiency gain that worsens equity (see Chapter 3.4 — Equity). She also reads the data-rights clause: the vendor would retain and could commercialize the province's longitudinal blood-pressure data, value the invoice never mentions.

The office does not reject the programme; it re-specifies it. It funds a controlled evaluation with outcome measures over a realistic horizon; requires interoperability and provincial ownership of the data as procurement conditions; funds an assisted-digital pathway — community health workers taking readings for those without phones — as an explicit line item rather than a hope; and agrees interim signals (measured blood-pressure control at twelve and twenty-four months) that would justify continued funding before the avoided strokes are visible. The result is slower and less flattering than the vendor's slide, but it is a decision the province can defend when the benefit is still years away.

Four sector lenses

Startup

A digital health start-up lives or dies on evidence it can rarely afford to generate: investors reward user growth and engagement, which are cheap to show, while payers demand clinical outcomes, which are slow and expensive. The temptation is to sell access and satisfaction as though they were health gains, and to launch on a thin pilot of enthusiastic early adopters. The disciplined move is to identify the lowest evidence tier that still supports the intended claim — using a framework such as NICE's — and to build proportionate proof into the product from the start, treating a credible evaluation as a market-access asset rather than a compliance cost. A start-up that also offers fair data terms and interoperability, rather than betting on lock-in, is easier for a cautious public payer to buy.

Small business

An established small provider — a general-practice partnership, a single clinic, a care home, or a niche device supplier — buys digital tools rather than builds them, and its constraint is thin margins and no in-house evaluation capacity rather than the race for growth that drives a start-up. It cannot commission its own trials, so it must lean on recognized frameworks and shared assessments — the NHS Digital Technology Assessment Criteria, a NICE evidence rating, a buying-group appraisal — to avoid paying for tools whose downstream work it will absorb through its own staff's time. Its real exposure is the pathway cost the vendor externalizes: the reception team fielding app queries, the nurse triaging alerts, the fallback when a patient's device fails all land on a small payroll that has no slack. The prudent small business insists on interoperability with the systems it already runs, avoids proprietary lock-in it cannot afford to unwind, and chooses steady, supported tools over novelty, because a failed rollout it has already paid for is a loss it cannot easily absorb.

Enterprise

A large provider or insurer buys digital tools as part of a portfolio and carries the integration, procurement, and inclusion costs the start-up externalizes. Its risks are lock-in — a platform whose network effects and data hold make it expensive to leave — and stranded investment when a tool is abandoned after a costly rollout. The enterprise advantage is the ability to demand real-world evidence at scale, to insist on open standards and portability, and to negotiate data-governance terms from a position of strength. Its accountability is to demonstrate that a fleet of digital tools improves outcomes across the whole covered population, not just among the digitally confident members who adopt first.

Government

A ministry or national payer is answerable for equity and public money across everyone, so its central concern is that digital-by-default not become exclusion-by-default. Government sets the evidence standards other actors work to — NICE's framework, the NHS Digital Technology Assessment Criteria, and the WHO classification are all instruments of this kind — and it can mandate interoperability, data sovereignty, and inclusion funding as conditions of national adoption. It also carries the long-lag problem most acutely, because it funds prevention whose benefits arrive beyond the electoral and budgetary cycle. Its task is to build appraisal and budgeting processes that can say yes to a slow-acting, population-wide tool while protecting the excluded and the public purse.

Common failure modes

• Counting digital activity as health benefit. More consultations, logins, or app downloads are inputs, not outcomes. Fix: define and measure the clinical or economic outcome, and hold access and convenience to their own honest value.
• Additive tools sold as substitutive. A video call added on top of, not instead of, the clinic visit raises cost while looking efficient. Fix: fix the counterfactual and verify that the replaced cost actually disappears.
• One-year ROI on a ten-year benefit. Short-horizon tools mis-score prevention and monitoring as pure cost. Fix: use a horizon long enough to capture the benefit and name the cross-payer, cross-time split (see Chapter 2.1 — Economic Evaluation).
• Piloting on the willing. Results from digitally confident early adopters overstate benefit and hide the cost of universal reach. Fix: stratify by need and connectivity, and fund the non-digital pathway.
• Giving away the data. Accepting a low licence fee while a vendor keeps and monetizes population data. Fix: treat data rights and governance as priced commercial terms.
• Buying lock-in cheaply. A low entry price with proprietary formats becomes expensive when you cannot leave. Fix: require open standards and data portability up front.
• Evidence decay ignored. Trusting launch-era evidence for a product that has since changed. Fix: re-evaluate on major updates and monitor real-world use, with decommissioning criteria.

Maturity model

Checklist

• The counterfactual is defined and the replaced cost is shown to actually disappear.
• Evidence demanded is matched to the tool's function and risk using a recognized framework.
• The whole pathway is costed, including the human work the tool creates.
• The time horizon is long enough to capture a slow-arriving benefit, with discounting stated.
• Results are stratified by age, income, language, disability, and connectivity.
• A non-digital pathway or assisted-digital support is funded as an explicit line item.
• Data ownership, permitted uses, consent, and benefit-sharing are negotiated and priced.
• Interoperability and data portability are procurement conditions, not afterthoughts.
• Re-evaluation and decommissioning triggers are set for future versions.
• Access and convenience gains are valued honestly and not counted as health outcomes.
• Any reliance on an adaptive algorithm is flagged for the additional scrutiny of Chapter 5.3 — AI Health Economics.

Key sources

• NICE Evidence Standards Framework for digital health technologies — the leading tiered framework setting evidence expectations by function and financial risk.
• NHS Digital Technology Assessment Criteria (DTAC) — clinical safety, data protection, security, interoperability, and usability criteria for digital health tools in England.
• World Health Organization — Classification of Digital Health Interventions and the WHO guideline recommendations on digital interventions for health system strengthening — an internationally applicable reference for lower- and middle-income systems.
• GOV.UK — Health economics: a guide for public health teams — practitioner tools illustrating the long-lag problem in prevention return-on-investment analysis.
• Wikipedia — Health economics and the digital health entries listed in References — for concept overviews.

References

1. Digital health — Wikipedia — https://en.wikipedia.org/wiki/Digital_health
2. eHealth — Wikipedia — https://en.wikipedia.org/wiki/EHealth
3. mHealth — Wikipedia — https://en.wikipedia.org/wiki/MHealth
4. Telehealth — Wikipedia — https://en.wikipedia.org/wiki/Telehealth
5. Remote patient monitoring — Wikipedia — https://en.wikipedia.org/wiki/Remote_patient_monitoring
6. Digital therapeutics — Wikipedia — https://en.wikipedia.org/wiki/Digital_therapeutics
7. Wearable technology — Wikipedia — https://en.wikipedia.org/wiki/Wearable_technology
8. Data governance — Wikipedia — https://en.wikipedia.org/wiki/Data_governance
9. Network effect — Wikipedia — https://en.wikipedia.org/wiki/Network_effect
10. Digital divide — Wikipedia — https://en.wikipedia.org/wiki/Digital_divide
11. Digital literacy — Wikipedia — https://en.wikipedia.org/wiki/Digital_literacy
12. Opportunity cost — Wikipedia — https://en.wikipedia.org/wiki/Opportunity_cost
13. Evidence Standards Framework for digital health technologies — National Institute for Health and Care Excellence (NICE) — https://www.nice.org.uk/corporate/ecd7
14. Digital Technology Assessment Criteria (DTAC) — NHS England — https://transform.england.nhs.uk/key-tools-and-info/digital-technology-assessment-criteria-dtac/
15. Classification of Digital Health Interventions v1.0 — World Health Organization — https://www.who.int/publications/i/item/9789241550505
16. WHO guideline: recommendations on digital interventions for health system strengthening — World Health Organization — https://www.who.int/publications/i/item/9789240020924
17. Health economics: a guide for public health teams — GOV.UK — https://www.gov.uk/guidance/health-economics-a-guide-for-public-health-teams