From Interface Friction to Avoidance: The Feedback Loop
For older adults in digital primary care, resistance can begin with an ordinary task: signing in, reading a result, or explaining several symptoms through a new channel. This chapter traces how that task-level difficulty can become lower confidence, anxiety, perceived clinical or privacy risk, and eventual avoidance. It also separates what the evidence directly reports from what the review’s conceptual model proposes, so that the feedback loop becomes a useful way to investigate exclusion rather than an untested law about every older user.
The first break is usually a care task
The resistance review identifies usage barriers in 16 of its 17 included primary-care studies. That count tells us how consistently the category appeared in the reviewed literature; it is not a prevalence estimate for older adults as a whole. The category includes interface complexity, poor navigation, repetitive authentication, difficult terminology, inaccessible text or audio, sensory and cognitive demands, trouble articulating symptoms, and the learning curve involved in a new system.
Each item matters because it changes the work of the encounter. Consider a patient portal that requires a person to remember a password, locate the correct service, choose a category, and upload a document before they can ask a question. None of those steps is necessarily impossible. Together, however, they create a sequence in which a small mistake can interrupt access to care. A video consultation introduces a different demand: the person must explain symptoms in real time, perhaps without the nonverbal cues or physical examination they associate with a clinical visit. A form may ask for a concise description when the person has several concerns that feel connected.
The same design can be easy for one task and difficult for another. A person may use a phone to call family but struggle with a portal’s terminology. They may read a message but not hear a clinician clearly over a poor connection. They may understand a single appointment request but become overwhelmed when a service requires a series of linked actions. “Can this person use a smartphone?” is therefore too broad a test. The relevant question is whether the person can complete the particular health-care sequence with enough clarity and control.
This is where usage becomes more than a skills category. A difficulty may reflect the user’s prior experience, but it may also reflect the service’s assumptions about memory, vision, hearing, dexterity, language, and tolerance for repetition. When the system makes a health task cognitively expensive, the user has fewer spare resources for interpreting instructions, checking information, or asking for help. The immediate result may be hesitation. The deeper consequence is that the next attempt begins with an expectation of difficulty.
Confidence is a mechanism, not a demographic label
The review repeatedly links the learning curve to a self-efficacy deficit. Self-efficacy means confidence in one’s ability to complete a particular task successfully. It is not a fixed measure of intelligence, age, or general motivation. It is shaped by what happens during interaction with the system.
A first failed attempt can therefore do more than delay one appointment. If the person cannot tell whether the error came from their action, the platform, or the connection, they have little basis for correcting it. If the interface offers no clear recovery route, the person may infer that future attempts will be just as uncertain. The task becomes a test of personal competence rather than a supported interaction with a service. The resulting loss of confidence can make the next unfamiliar step feel threatening before it is tried.
Technology anxiety is the emotional expression of this uncertainty. The reviewed studies describe fear or discomfort about making mistakes, mishandling a system, or causing harm through improper use. Anxiety can narrow attention: the person concentrates on not pressing the wrong button and has less capacity to describe symptoms, read a result, or evaluate what the clinician is saying. A task that was already demanding now feels even less manageable.
The causal language needs care. The review’s evidence comes from heterogeneous qualitative, cross-sectional, and mixed-methods studies. It does not experimentally show that a particular interface causes anxiety in a fixed sequence. Its conceptual integration proposes a recurring pathway in which usability problems contribute to a learning difficulty, the learning difficulty erodes self-efficacy, and lower self-efficacy increases technology anxiety. That pathway is plausible because it connects reported experiences, but plausibility is not the same as longitudinal proof.
Even with that limitation, the mechanism changes the practical question. If low confidence is treated as a trait, the response is likely to be generic instruction directed at the individual. If it is treated as a mediator between the task and the decision to engage, the response can also examine whether the service makes practice safe, whether errors are recoverable, and whether someone can provide timely support. The goal is not to persuade a person to feel confident by assertion. It is to create interactions that give confidence a reasonable basis.
When uncertainty becomes a risk judgment
Anxiety becomes especially consequential in health care because the user may connect a technical mistake to a clinical outcome. The review identifies diagnostic uncertainty, fear of missed diagnoses, technology-misuse anxiety, privacy and security concerns, fraud or scams, and concerns about autonomy. These are not interchangeable fears, but they can accumulate around the same encounter.
A person may worry that a video visit cannot convey a physical sign, that a clinician will misunderstand a description, or that an important issue will be overlooked. They may be unsure whether a message or uploaded result reached the care team. They may fear entering information incorrectly, deleting something important, or exposing personal health data. When the system does not make its boundaries and safeguards clear, the user must supply the missing certainty themselves. Avoiding the channel can then appear safer than attempting a task whose consequences are unclear.
The distinction between perceived and actual risk is essential. The evidence documents what older adults report fearing; it does not quantify the clinical or cybersecurity risk of every digital-health modality. Nor does it establish that digital care always produces more errors. Yet perceived risk still affects behavior. A service can be objectively secure and still lose trust if users cannot understand how information is handled. A digital encounter can be clinically appropriate for one purpose and seem inadequate for another because the user cannot tell what information can be assessed remotely.
Risk also has a communication component. In a familiar in-person encounter, a person may expect to clarify a misunderstanding immediately, observe a clinician’s reaction, or receive guidance about what happens next. In a digital interaction, uncertainty about whether the message was received or whether the symptom was understood can make the person feel less in control. This is why the review connects risk to both technology misuse and loss of autonomy: the concern is not merely that the software might fail, but that the person may be unable to detect or repair the failure.
A risk response therefore cannot be reduced to reassurance. It must distinguish tasks that require different kinds of information, make errors and handoffs visible, and preserve a credible way to obtain appropriate human judgment. The resistance evidence supports the need to address those concerns; it does not specify one universal technical safeguard or prove that any single feature will restore trust.
Why avoidance stabilizes the problem
Once a person avoids the digital channel, the immediate anxiety may fall. But avoidance also removes opportunities to practice and learn. The next invitation arrives without the experience that could have made the process more familiar. The person’s low confidence is not corrected, and the unfamiliar service continues to look risky. In this sense, avoidance can stabilize the very conditions that made it seem necessary.
The review describes this as a self-reinforcing cycle: usability problems contribute to low self-efficacy; low self-efficacy increases anxiety; anxiety amplifies concerns about privacy, misuse, diagnosis, and care quality; distrust leads to withdrawal; and withdrawal limits exposure and skill acquisition. The model also includes a second pathway. A strong preference for face-to-face care and known providers can reinforce the perception that digital encounters are impersonal, less legitimate, or designed for younger users. That legitimacy concern makes the next difficult interaction easier to reject.
Different starting points, similar endpoint
The loop does not require every barrier to appear in a fixed order. One older adult may begin with a usage problem: the portal’s navigation is confusing, so a request remains unfinished. Another may have adequate technical skill but begin with risk: a privacy warning or fear of a missed diagnosis makes the digital channel unacceptable before any practice occurs. A third may begin with tradition or image, preferring a known clinician and viewing a remote encounter as less legitimate. If all three eventually stop using the service, the visible outcome—nonuse—looks the same even though the mechanism differs.
That difference matters because a response aimed at the wrong entry point can shift effort without changing the judgment. More instruction may help the first person if the service also makes recovery possible. It may do little for the second person if the system leaves data handling or clinical limits unexplained. It may be irrelevant to the third person if the service removes continuity with a trusted care team. The point is not that each barrier requires a separate program; it is that the service must identify which uncertainty or loss is driving the decision before choosing a remedy.
The loop can also contain reinforcing combinations. A person who already doubts the legitimacy of video care may interpret a minor audio problem as evidence that the channel is inherently inferior. A person with low confidence may interpret a routine security step as proof that the system is too risky to use. Conversely, a clear explanation from a known clinician may make a new task feel worth attempting, giving the person an opportunity to build skill. These examples describe plausible interactions from the review’s synthesis, not universal psychological rules. They show why resistance should be examined at the level of the encounter and its surrounding relationship.
The practical endpoint is not necessarily permanent rejection. Avoidance can mean postponing a task, limiting use to familiar functions, asking someone else to act, or returning to a phone or in-person route. Those behaviors can protect the person from an immediate uncertainty while still leaving the underlying loop intact for the next digital request. A service team that counts only completed logins may miss this partial disengagement. Understanding the sequence requires asking what the person tried, where control was lost, what consequence they feared, and whether another channel remained available.
The arrows should not be read as a claim that every user moves through the same states in the same order. A person might begin with a privacy concern, a prior negative experience, a preference for direct examination, or an initial usage difficulty. Several barriers may coexist, and an improvement at one point may leave another untouched. The diagram is useful because it shows why a single intervention can have a limited effect: fewer clicks may reduce task friction without changing concerns about diagnosis or legitimacy; a privacy notice may clarify data handling without making symptom communication easier.
The model also clarifies the difference between a trigger, a mediator, and an outcome. Interface complexity or a sensory barrier can trigger difficulty. Self-efficacy and anxiety can mediate how that difficulty is interpreted. Perceived risk and distrust can shape the decision. Avoidance is a behavioral outcome that then feeds back by reducing practice. These terms describe a proposed relationship in the review’s synthesis, not a set of variables whose timing has been fully demonstrated.
What the evidence can—and cannot—tell us
The feedback loop is grounded in recurring findings across the primary-care studies, including participant accounts of limiting use to familiar functions, becoming nervous outside a comfort zone, preferring phone calls, or wanting a clinician to examine them directly. The review classified links as explicit when a study directly stated them, inferential when the connection followed from a study’s narrative, and integrative when similar patterns recurred across studies. That hierarchy matters. A repeated pattern is more informative than an isolated impression, but it is still not the same as a prospective test of the sequence.
The evidence also does not establish that older adults are passive recipients of the loop. Avoidance may be a rational response to a service that is inaccessible, poorly supported, unsuitable for the clinical problem, or inconsistent with the person’s values. Nor does the model imply that all in-person care should remain unchanged or that every digital alternative is illegitimate. It asks a narrower and more useful question: what conditions are causing this person to judge the digital channel as too difficult, too risky, or not worth changing for?
That question points toward testing changes at more than one link: task difficulty, risk communication, and continuity may need to be considered together rather than treated as unrelated fixes. The review does not establish a clinically validated intervention bundle. The next task is to examine which kinds of support have actually improved skills, confidence, perceptions, or technology use—and how much confidence those results deserve.