Threats to the Validity of the Clinical Interview: Can Anything Be Done?

Levy and colleagues¹ report the findings of 2 online surveys asking large convenience samples (N = 4510) whether they had ever withheld any of 7 different types of medically relevant information from their health care clinicians and, if so, their reasons. Most respondents acknowledged withholding information at least once. Their chief reasons for this choice were disagreement with the clinician’s recommendations and failure to understand the clinician’s instructions. The investigators infer that nondisclosure is motivated by the wish to avoid embarrassment or concern about being disapproved of by their clinicians. The study in effect explores some threats to the validity of the clinical interview.

Perhaps the oldest question in the history of health care is, “What brings you [the patient] to see me [the healer] today?” Before the modern era, there were patients with illnesses and people to consult for help. The diagnostic tools for most of history were the patient’s account, enshrined today in the chief concern and the medical history, and a rudimentary examination for signs and symptoms. Modern health care began with the understanding that much was going on that was not accessible with those methods. Diagnostic technology has proliferated. With the rise of modern psychology, the trustworthiness of the patient’s account has also been increasingly challenged. We now understand that a patient’s account of the ailment is only 1 possible account, and not necessarily the true state of affairs. A report of an episode is not a photographic record of experience; it is structured in some way. A patient might not remember features that the clinician would regard as very important, resulting in gaps in the story, or might choose not to disclose elements in stored memory that are readily accessible. The degree to which restructuring is driven by affective factors, such as self-interest or the wish to create a favorable impression, or occurs essentially as a consequence of automatic processing inherent in the machinery, is a long-standing issue separating psychodynamic theories of memory² from contemporary neuroscience. But the upshot is that, on the one hand, not all experiences can be retrieved on demand and, on the other, patients are sometimes reluctant to tell all they know. Both are threats to the validity of the clinical interview. It is not surprising, therefore, that health care clinicians increasingly rely on diagnostic technologies to illuminate the patient’s history.

These considerations underlie the study being discussed. The study begins with the commonly accepted view that patients purposefully withhold information from their clinicians, at least on some occasions. The goal is to identify the most important reasons why patients might make this decision. Since patients have a choice about what to disclose, under different conditions they might have disclosed what they decided not to reveal.

The major findings of interest in my view are that failure to disclose seems motivated by the belief that the information would reflect poorly on the patient (eg, not exercising, not taking a prescribed medication faithfully) and that an association with race and the deception rate was found in only 1 sample. The first conclusion is intuitively obvious—the second is not. One would like to know if nondisclosure is related to discordance between the races of patient and clinicians. Are patients generally more forthcoming with clinicians who they perceive to be similar to them in some relevant dimension? Were the respondents asked to identify the race of their primary clinician?

The investigators note that both online samples are “not fully representative.” The white population is overrepresented, while Hispanic and African American populations are underrepresented. Both samples also overrepresent patients with a college education, probably because college-educated white individuals are more likely to enroll in online surveys. Despite their size, these are not stratified random samples. Because this method of data collection is inexpensive and convenient, the day of face-to-face or telephone surveys of carefully stratified samples may soon be over. We have a new type of convenience sample: people who actively use the internet and wish to participate in surveys. The errors and biases introduced by this method will become clear as it is more widely used. Optimistically, they may be self-limiting as internet access and use become more widely diffused.

The study depends on memory and self-report; both are 2-edged swords. On the one hand, it is unlikely that anyone recalls accurately every instance of not telling the whole truth about a sensitive matter; on the other, the investigators assume that some patients recall more than they choose to tell; the results support that assumption. It is ironic that one self-report identifies reasons why another self-report might be inaccurate. Why would patients disclose on a survey form that they had not disclosed some relevant information to their clinicians? A plausible reply is that the online survey and the anonymous investigators behind it are perceived to be neutral and nonjudgmental, in contrast to the respondents’ concern about the response of their clinicians. It seems to me that nondisclosure is likely to be more prevalent among the underrepresented population because the social gap between them and their clinicians would tend to increase concern about being disapproved or disrespected. The magnitude of this socioeconomic bias awaits further study.

A skeptical view of the magnitude of deception would point out that the respondents were asked whether they had ever deceived their clinicians, so that 1 omission counts as much as a consistent pattern. On this view, the analysis may overstate the magnitude of the problem. Yet it seems equally likely that most people simply do not recall every incident of nondisclosure and that at least some respondents chose not to disclose an instance they recalled. So the findings may, as the investigators believe, underestimate the magnitude of the problem. In my opinion, the results may be as close to the truth as we are likely to get with ethically acceptable methods, given the sampling problems already discussed.

What can be done to ameliorate this problem? The authors recognize the need to increase patients’ comfort with disclosing information that may be embarrassing. If clinicians could offer convincingly a nonthreatening, nonjudgmental environment, the frequency of nondisclosure might be reduced. In my opinion, any effort to change human behavior is costly and inefficient compared with implementing a technical fix, were one available. In this case, the authors have implicitly suggested a remedy worth trying: to shift questions that are the most difficult to answer candidly in a face-to-face clinical interview to an impersonal survey instrument. The success of this study in eliciting data on the prevalence of nondisclosure suggests that this technology might be useful in solving the problem. A trial of this strategy would almost surely be less expensive than a sustained effort to change the clinical culture and the interviewing style of a generation of clinicians.

Published: November 30, 2018. doi:10.1001/jamanetworkopen.2018.5298

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2018 Elstein AS. JAMA Network Open.

Corresponding Author: Arthur S. Elstein, PhD, 3800 N Lake Shore Dr, Apt 13E, Chicago, IL 60613 (aelstein30@gmail.com).

Conflict of Interest Disclosures: None reported.