How do you calculate telehealth ROI accurately?

Direct answer

The direct answer is that telehealth ROI should be calculated as the value created by a clearly defined remote-care workflow, not as the revenue from every video visit. For payer and provider operations, the defensible formula is: ROI percentage equals net benefit divided by fully loaded program cost, multiplied by 100. Net benefit equals measurable financial benefit plus approved value credit minus all recurring and one-time costs. The result should be shown alongside payback period, break-even volume, and a confidence range because a single point estimate can hide weak assumptions.

Also worth reading: How do payers and providers accurately calculate value-based care software ROI in 2026? · How do you accurately calculate and track RCM automation ROI metrics for healthcare revenue cycle operations? · How do healthcare organizations accurately measure SaaS ROI for cost-containment and care-coordination platforms?

For a 12-month pilot, a practical baseline is annual net benefit of $300,000 against fully loaded cost of $200,000, producing $100,000 net benefit and 50% ROI. Payback is 8 months when monthly net cash benefit is $8,333. This example is intentionally simple: it does not assign dollar value to member satisfaction, faster access, or clinician convenience. Those outcomes may still matter, but they should be reported separately unless finance has approved a conversion method.

The answer becomes much stronger when benefits and costs are compared with a real counterfactual. Sending every appointment online is not automatically cheaper than an in-person visit, and a low-cost platform can still produce poor returns if staff time, no-show handling, transcription, devices, and integration work are omitted. The best calculation asks what would have happened to the same patients, volume, and operating model without the telehealth intervention.

Baseline and counterfactual

A defensible calculation starts with a baseline period of at least 90 days, although 12 months is preferable for programs with seasonal utilization. Compare the same cohort, service line, geography, and payer population before and after launch, while adjusting for visit mix and case complexity. The baseline should include total cost of care where available, not merely telehealth platform fees. For claims-based programs, use a 3- to 6-month lookback and distinguish paid claims from incurred but not reported claims.

The counterfactual is the operating choice that would occur without remote care. It may be an in-person primary-care visit, an emergency-department visit, a telephone encounter, a cancellation, or no visit at all. A payer may compare a telehealth behavioral-health pathway with usual in-person behavioral-health care, while a provider may compare virtual urgent care with walk-in demand. The comparison should include clinical appropriateness and patient selection, because treating lower-acuity cases remotely creates a different result from substituting remote visits for all appointments.

For a provider baseline, record total visits, in-person visits, no-shows, staff minutes, room utilization, equipment use, and cost per completed encounter. For a payer baseline, add allowed claims, utilization-management minutes, care-management hours, avoidable admissions, and avoidable emergency visits. A 20% reduction in no-shows is a useful planning scenario, but it is not an industry promise. The actual reduction should come from the pilot, matched comparison group, or a conservative assumption that is tested in sensitivity analysis.

Benefit model

The core benefit formula is annualized financial benefit equals avoided cost plus collected net revenue plus labor or capacity value, minus any benefit transfers that must be reported separately. Avoided cost should include only expenses that actually disappear or can be redeployed. Examples include an avoided emergency visit, fewer unnecessary transports, reduced facility use, lower patient travel support, and a measured decline in staff overtime. Do not count a lower-paid virtual visit as savings if the same staff, technology, and overhead remain in place.

Collected net revenue is calculated as eligible encounters multiplied by net allowed amount, minus contractual adjustments, refunds, denials, and collection costs. For a provider, a virtual visit can add revenue when it converts a no-show into a completed encounter or opens capacity that produces additional billable work. For a payer, the same encounter may be a cost shift rather than enterprise savings. The model should therefore show payer, provider, health-system, and patient perspectives separately.

Labor value is based on minutes saved multiplied by loaded hourly cost. A 15-minute reduction across 2,000 encounters equals 500 hours, but that is only an economic benefit if staffing is reduced, overtime falls, or the time is used to complete other billable or necessary work. Capacity value should use contribution margin per added encounter, not gross billing. Patient travel savings can be included for a social-value view when mileage, time, parking, and caregiver costs are measured consistently.

Cost model

Fully loaded telehealth cost includes the subscription or usage fee, implementation, integration, data migration, device procurement, clinical training, support, cybersecurity review, payment processing, transcription, and internal project time. A $25 per-user monthly license can become much more expensive when 5,000 users, 20,000 encounters, recording storage, and support seats are added. Usage pricing should be tested against both minimum monthly fees and per-encounter charges because volume changes can reverse the expected cost order.

One-time implementation cost should be spread over the expected useful life of the capability. A $60,000 integration and workflow project over three years is $20,000 per year, or about $1,667 per month. Recurring cost should include the platform, support, licenses, cloud storage, integration maintenance, training refreshers, and the internal owner's time. Exclude sunk costs from the decision model, but recognize that past spending can affect whether a continuation decision is economically rational.

A simple annual cost example is $180,000 in licenses and usage, $45,000 in implementation amortized over three years, $30,000 in integration and support, and $25,000 in training and project labor. The fully loaded annual cost is $280,000, or roughly $23,333 per month. If the program serves 12,000 encounters, the average fully loaded cost is $23.33 per encounter before counting any clinical staff time that is not already included.

ROI and payback calculations

The basic ROI calculation is annual net benefit divided by annual fully loaded cost. Using the example above, assume avoided emergency and inpatient cost of $220,000, added net revenue of $90,000, and verified labor or capacity value of $50,000. Gross measurable benefit is $360,000, net benefit is $80,000, and ROI is 28.6% against $280,000 of cost. The calculation is transparent because every input has a source, a period, and an owner.

Payback period is implementation and initial launch cost divided by average monthly net cash benefit. If the first-year cost is $280,000 and recurring annual benefit is $360,000, simple payback is 9.3 months. A discounted payback calculation is preferable for a multi-year investment because it discounts future cash flows at the organization's approved rate. It should not be confused with ROI, since payback measures timing while ROI measures the size of the return.

Break-even volume shows how much activity is needed to cover cost. With $280,000 annual cost, $20 benefit per encounter, and $5 variable cost per encounter, the contribution per encounter is $15. The program needs about 18,667 encounters per year, or 1,556 per month, to break even. If the actual volume is 12,000 encounters, the model should be revised rather than forcing a positive result.

Sensitivity and attribution

The first sensitivity test changes the volume assumption by 20% above and below the baseline. The second changes the avoided-cost estimate by 25%, because claims and utilization estimates are often less certain than platform invoices. The third changes the conversion rate from no-show to completed encounter, the net allowed amount, and the labor minutes saved. A result that remains positive across reasonable ranges is more useful than a base case built around one optimistic forecast.

Attribution should be based on a matched comparison group when possible. Compare members or patients with similar risk, diagnosis, geography, and prior utilization who did not receive the same remote-care pathway. Difference-in-differences can separate a general trend from the effect of the program, but the comparison group must be credible. Randomization is not always practical, and a before-and-after chart alone cannot prove that telehealth caused a decline in utilization.

Use confidence ranges rather than false precision. A base case may show 28.6% ROI, while a conservative case shows a loss and an upside case shows 45%. The range is not a defect; it communicates the quality of the evidence. For a pilot, require a predefined success threshold such as positive net benefit, a 10% reduction in avoidable utilization, or at least a 15% reduction in no-shows before scaling.

Comparison and alternatives

The choice is not simply telehealth versus no telehealth. Remote care can be video, audio-only, asynchronous messaging, remote patient monitoring, hybrid scheduling, or an in-person pathway. Each option has a different clinical fit, cost structure, access effect, and attribution problem. The right comparison is the lowest-cost appropriate pathway for the patient and the workflow.

An asynchronous workflow may be cheaper for a simple medication review, while remote patient monitoring may be justified for a high-risk group with a measurable reduction in exacerbations. Audio-only care can improve access for patients without reliable video, but it may not support every examination or consent process. A hybrid model often avoids the false choice between digital and physical care, but it requires clear triage rules so that patients are not routed to the wrong channel.

The alternative should also be compared with a no-change case. If the current problem is long wait times, adding video capacity may help, but adding scheduling automation or same-day in-person slots may be cheaper. If the problem is missed follow-up, a reminder and care-navigation workflow may produce more value than an expensive platform. The financial model should therefore price the workflow outcome, not merely the software category.

Common mistakes

The most common mistake is counting gross billing as savings. A $150 visit is not $150 of value when contractual adjustments, variable cost, and capacity limits are ignored. The model should use net collected revenue or contribution margin, depending on the decision being made. The same distinction matters when comparing payer and provider economics.

Another error is including every telehealth encounter as a substitution for an in-person visit. Some encounters would not have occurred without remote access, while others are add-ons that increase total utilization. The baseline must identify whether the program reduces, replaces, or creates encounters. Duplicate visits, recording fees, transcription, and support time should be checked for double counting.

Costs are often understated when internal labor is excluded. A clinician spending 10 extra minutes on documentation is a real cost even if payroll does not change immediately. Conversely, saved time is not savings unless it is redeployed to billable work, reduced overtime, or lower staffing. Benefits should also be separated from transfers between payer, provider, and patient budgets.

Attribution errors are equally important. A decline in emergency visits may reflect seasonality, a policy change, or a change in coding rather than telehealth. The report should disclose data lag, missing claims, patient mix, and the number of encounters included. A clean-looking ROI number is not credible if the inputs cannot be traced to finance, operations, or claims data.

When to act

Act when the workflow has a defined target population, reliable baseline data, and a plausible mechanism for savings or capacity creation. A pilot is appropriate when the expected benefit is large enough to cover the measurement cost and when the organization can observe utilization for at least 90 to 180 days. For a payer program, a 6- to 12-month claims window may be needed to capture avoided admissions and follow-up care. For a provider program, 3 months of encounter and no-show data may be enough for an initial operating decision, with a longer review before expansion.

Scale only after the model shows positive net benefit under conservative assumptions and the operational data is repeatable. A useful threshold is a 12-month ROI above 0% in the base case, no worse than 0% in a clearly labeled downside case, and a payback period below the approved investment limit. The threshold should be set before launch so that a favorable result is not defined after the fact.

Do not scale merely because adoption is high. A program can have 80% patient satisfaction and still fail to reduce total cost if it adds visits without changing utilization. Conversely, a modest financial return may be acceptable when the program protects access for a high-risk population, but that should be recorded as a strategic value case rather than disguised as direct ROI.

Cost and pricing framework

Pricing should be modeled from the total operating cost per completed encounter, not from the advertised license alone. The formula is annual platform and usage cost plus implementation amortization plus support, integration, devices, and internal labor, divided by eligible completed encounters. This figure can then be compared with the contribution margin or avoided cost per encounter in the target population.

For a small provider pilot, a practical planning range might be $5,000 to $25,000 in first-year software and implementation cost, with additional device, integration, and staff expenses. A payer or multi-site provider program can require substantially more because of identity management, claims integration, analytics, support, and compliance review. These figures are planning ranges, not quotes, and should be replaced with vendor proposals and internal labor rates.

Ask vendors for per-user, per-encounter, tiered, and minimum-volume pricing side by side. Test the model at 50%, 100%, and 150% of expected volume. Include cancellation, recording retention, interpreter, e-prescribing, payment, and support fees rather than assuming they are included. The final decision should compare the fully loaded cost with the measurable benefit per encounter and with the next-best operational option.

Practical implementation plan

Begin with a one-page operating charter that names the target population, clinical exclusions, counterfactual, owner, and success metric. Define whether the program is intended to reduce avoidable utilization, improve access, increase completed visits, or shorten time to follow-up. A single program can have several goals, but the financial model should identify which ones are expected to create cash value.

Build the baseline from encounter, claims, scheduling, staffing, and cost data before launch. Reconcile encounter counts with billing records and exclude test visits, duplicate records, and incomplete claims. Assign an owner to each input so that the finance team can challenge the assumptions. The data dictionary should state whether a metric is paid, incurred, collected, or estimated.

Run the pilot with a control or comparison group and review results monthly. Track utilization, no-shows, completion rate, staff minutes, net revenue, avoidable admissions, emergency visits, and patient access measures. At the end of the period, update the model with actuals and document what changed. The next decision should be based on observed performance, not on the original business case alone.

Final calculation template

Use the following template as an auditable worksheet. It is a calculation framework, not a promise that a particular percentage will be achieved.

The template keeps the answer useful without pretending that every healthcare cost can be measured with equal confidence. Claims data may lag, patient behavior may change, and a workflow that works in one service line may not work in another. The strongest result is therefore a documented range, a clear counterfactual, and a decision rule that can be reviewed after the pilot. That is the practical standard for a telehealth ROI calculation guide.

FAQ

Is telehealth ROI the same as revenue? No. Revenue is money collected for services, while ROI is net benefit after the full cost of the program. A telehealth service can produce revenue and still lose money when platform, staff, integration, and support costs exceed the contribution generated. Should avoided emergency visits be included? Yes, when the decrease can be attributed to the telehealth pathway and the cost is measured consistently. Use a comparison group or risk adjustment where possible, and report the assumption separately from confirmed savings. Do not count every emergency visit that does not occur after launch as telehealth savings. How many months of data are needed? At least 90 days is useful for operational measures such as no-shows and completion rates. Claims-based outcomes often need 6 to 12 months because admissions, procedures, and incurred-but-not-reported costs appear with a delay. The right period depends on the care pathway and the frequency of the outcome being measured. What is a good telehealth ROI percentage? There is no universal percentage. A positive result means the measured benefit exceeds the fully loaded cost, while a higher percentage generally indicates a stronger financial case. The acceptable threshold should be set by the organization's cost of capital, risk tolerance, and strategic goals. Can patient satisfaction be included in ROI? It can be included in a broader value case, but it should not be converted into dollars without a defensible method. Report satisfaction, access, and clinical outcomes alongside the financial calculation. This keeps the result useful without mixing subjective measures with verified cash savings.