The Hidden Cost of Convenience Is What We Fail to Measure

What if the most expensive feature of modern health care is not a treatment, a device, or a prescription, but convenience?

That sounds backward. Convenience is usually presented as an unambiguous good: fewer clicks, shorter waits, no commute, care available from a phone. Yet evidence from acute respiratory care suggests that only 12 percent of direct to consumer telehealth visits replaced visits that would otherwise have happened elsewhere. The remaining 88 percent represented new utilization. Access improved, but annual spending for each telehealth user increased by $45.

This is not necessarily an argument against telehealth. It is an argument against evaluating innovation with instruments too crude to distinguish substitution from addition.

The same problem appears in a less obvious place: the design of questionnaires. A questionnaire is often treated as a static form, a collection of prompts that a participant fills out. But when its questions, branching logic, definitions, and response structures are represented in an open, machine readable format, it becomes something more powerful: a portable measurement system that different tools can execute and different researchers can inspect.

These two ideas belong together. Both ask whether an intervention can be understood outside the container that produced it. Both expose a central challenge of contemporary health care: we keep making access easier without making consequences easier to see.

Convenience expands behavior. Only good measurement tells us whether that expansion is beneficial, wasteful, or both.

The question beneath the technology

The usual question about a new health service is: does it increase access?

That question matters, but it is incomplete. Access is an input, not an outcome. More people using a service may mean that previously unmet needs are finally being addressed. It may also mean that people are seeking care for problems they would have managed at home, consulting multiple providers for the same episode, or receiving low value care because the barrier to entry has fallen.

The important question is therefore more precise:

When access becomes easier, what behavior does the system create, and what behavior does it replace?

This is a distinction between marginal access and marginal utilization. Marginal access asks who can now obtain care that was previously unavailable. Marginal utilization asks what additional care occurs because the new channel exists. The two can rise together, and a system can improve access while increasing unnecessary spending.

Imagine a town with one crowded clinic. A virtual service opens overnight. Some residents who could not take time off work now receive appropriate treatment. That is genuine access. But other residents use the virtual service at midnight, then visit their primary care clinician the next morning because the diagnosis felt uncertain. Still others seek care for symptoms they would previously have watched for two days. The same technology has produced substitution, supplementation, and entirely new demand.

A simple visit count cannot distinguish these pathways. Neither can a satisfaction score. A person may be delighted with a convenient visit that added cost without adding clinical value. A health plan may report faster access while missing the fact that the new service has become an extra layer in the care journey.

This is where the architecture of measurement becomes part of the architecture of care.

A questionnaire is not just a form

Most organizations treat questionnaires as documents. They are drafted, placed inside a platform, sent to participants, and eventually exported as data. That workflow seems harmless until the same instrument must be revised, translated, reused in another study, or connected to data from a different system.

At that point, the questionnaire reveals itself as a small software application. It contains variables, conditional logic, validation rules, display instructions, and pathways that change according to a participant’s answers. A person who answers no to a screening question may skip five later questions. Someone who reports a particular symptom may see a more detailed branch. The order and wording of questions shape the data that can ultimately be analyzed.

If this logic exists only inside a proprietary platform, the instrument becomes difficult to inspect and difficult to move. The questionnaire may technically belong to its creators, yet its usable form is trapped inside a vendor’s ecosystem. Reusing it may require rebuilding it by hand. Comparing versions may be arduous. Combining responses across studies may require reconstructing the meaning of fields after the fact.

An open markup format offers a different model. Questions and logic can be represented in plain text that humans can read and applications can render. The same specification can potentially be used by different interfaces, stored in repositories, versioned like code, and mapped to shared data standards.

The important shift is conceptual: the questionnaire becomes a description of an instrument rather than a possession of a platform.

That shift has a direct parallel in telehealth. A visit should not be understood merely as an event recorded by the company that delivered it. It is one observation in a larger care pathway. To know whether it substituted for another visit, we need interoperable records that connect the virtual encounter with subsequent claims, diagnoses, prescriptions, follow up visits, and patient reported outcomes.

Without that connection, the telehealth platform can count its own activity but cannot reliably describe the system level consequence of that activity. It knows what happened inside its boundary. It does not know what happened because of it.

The substitution problem is a measurement problem

The 12 percent and 88 percent distinction is striking because it changes the meaning of the headline. If telehealth visits are counted as evidence of successful adoption, then growth looks like progress. If visits are classified according to whether they replaced existing care or added new utilization, the interpretation becomes more complicated.

This is a general problem in innovation metrics. Organizations count what their systems can easily observe. A platform can count logins, consultations, messages, and prescriptions. A hospital can count admissions and procedures. A questionnaire tool can count completed forms. But the most consequential variables often sit between systems:

• Did the new encounter replace another encounter?
• Did it prevent a later escalation?
• Did it duplicate an existing service?
• Did it help someone who would otherwise receive no care?
• Did it produce information that another clinician could use?

These are relational questions. They cannot be answered by looking at one transaction in isolation.

Consider a simple accounting identity:

Total utilization after an innovation equals displaced utilization plus additional utilization.

That identity is not a prediction. It is a reminder to separate two forces that are often combined in a single adoption number. If displaced utilization is large and additional utilization is small, the innovation may improve convenience without greatly increasing cost. If displacement is small and additional utilization is large, access may improve at a higher financial price. If the additional utilization prevents hospitalizations or complications, the initial increase may still be worthwhile. The point is not that new utilization is automatically bad. The point is that it must be traced.

Tracing requires a shared language for events and outcomes. A questionnaire can ask whether a patient sought care elsewhere after a virtual visit, but the response is more useful when the question has a stable definition, a documented answer structure, and a format that can be reused across studies. Claims data can show a subsequent visit, but that observation becomes more interpretable when it can be linked to the patient’s reported symptoms, perceived urgency, and reason for choosing telehealth.

This is the deeper role of interoperability. It is not merely a technical convenience. It allows a system to distinguish what a service did from what occurred around the service.

From data collection to causal visibility

There is a useful mental model here: think of every health intervention as casting a measurement shadow.

The intervention itself is easy to see. A visit occurred. A questionnaire was completed. A prescription was issued. The shadow consists of the consequences that are harder to observe: avoided care, repeated care, delayed care, reassurance, anxiety, downstream testing, improved adherence, or unnecessary treatment.

Poorly designed data systems illuminate only the object and leave the shadow dark. Better systems illuminate both by capturing context before the intervention and consequences afterward.

A reusable questionnaire can support this process if it is designed as a modular instrument. One module might record symptoms and duration. Another might capture the patient’s alternatives, such as inability to get an appointment or preference for convenience. A third might ask what happened after the encounter. Because each module is formally described, researchers can deploy the same questions in different settings and compare results without treating every new study as a fresh invention.

The same approach can improve telehealth evaluation. Instead of asking only whether a patient used virtual care, an evaluation could record:

1. The patient’s reason for choosing the channel.
2. Whether an in person appointment was available.
3. Whether another clinician was consulted for the same episode.
4. Whether the patient received testing, medication, or follow up care.
5. Whether the episode resolved, persisted, or escalated.
6. Whether the patient believed the visit replaced care or supplemented it.

Some of these variables belong in administrative data, some in clinical records, and some in patient reported questionnaires. Their value increases when their definitions are explicit and their formats are portable.

This suggests a principle that health systems often neglect: measurement should be designed for movement. Data should move across researchers, platforms, institutions, and time. If the meaning of a response depends on the proprietary interface where it was collected, it is not truly reusable. If the meaning of a telehealth claim cannot be connected to the rest of a patient’s episode, it is not enough to evaluate the service.

The platform is part of the policy

There is a temptation to think of software architecture as separate from public policy. On this view, policymakers decide what services are covered, clinicians decide how to provide them, and technologists decide how to store the data.

In practice, the boundaries are porous. A platform that makes a service easy to access also influences how often the service is used. A platform that makes questionnaire logic opaque also influences what can be learned about that use. Technical design therefore affects both behavior and accountability.

Proprietary systems are not inherently useless. They may provide polished interfaces, sophisticated branching, and reliable operations. The problem emerges when the convenience of the front end is purchased with the invisibility of the underlying structure. A health system may gain a smooth workflow while losing the ability to compare instruments, audit logic, or create a shared data commons.

The alternative is not to reject platforms. It is to make the underlying specification portable. A questionnaire should be able to travel from one renderer to another. A definition of a telehealth episode should be understandable across vendors. A question about substitution should preserve its meaning whether asked through a research portal, a patient application, or a clinical intake process.

This is the difference between interoperability as plumbing and interoperability as institutional memory. Plumbing lets data flow. Institutional memory lets a field retain what it has learned when software, vendors, and studies change.

Key Takeaways

• Measure replacement, not just adoption. For any new care channel, separate visits that substitute for existing care from visits that add utilization.
• Treat questionnaires as executable specifications. Document questions, branching logic, definitions, and response formats in a human readable, machine readable form.
• Track the whole episode of care. Link the initial encounter to prior alternatives, subsequent visits, prescriptions, testing, outcomes, and patient experience.
• Design data for reuse from the beginning. Use stable concepts, explicit metadata, open formats, and versioned instruments instead of relying on one platform’s internal structure.
• Ask why the service was used. Convenience, lack of access, reassurance, urgency, and clinical necessity can produce identical visit records but imply very different value.

The new definition of access

Access is often imagined as a door. The door is either open or closed. Telehealth opens another door, and that can be transformative for people who face distance, disability, inflexible work, or scarce appointments.

But health care is not a building with a finite number of doors. It is a network of pathways. Opening one pathway changes traffic across the others. Some congestion falls. Some traffic is newly generated. Some travelers take several routes for the same destination.

The central challenge is therefore not simply to make care easier to reach. It is to make the consequences of reaching care visible, comparable, and learnable.

That requires a change in the unit of thought. Stop treating the visit as the basic object. The basic object is the care episode, including the alternatives that preceded it, the services that followed it, and the patient’s account of what changed. Stop treating the questionnaire as a disposable form. Treat it as a portable instrument whose logic can participate in a shared evidence system.

The future of convenient care will not be judged by how many people can click into a consultation. It will be judged by whether we can tell who benefited, what was replaced, what was added, and what became possible because the system was easier to use.

The paradox is simple: the more frictionless care becomes, the more deliberate measurement must become. Convenience accelerates behavior. Interoperability gives that behavior a memory. Without both, innovation may move faster than understanding.