Patients’ Question-Asking Behavior During Primary Care Visits: A Report From the AAFP National Research Network

Kevin Fiscella is absolutely correct in his suggestion of this hypothesis – the 10 intervention practices may have differentially implemented the AM3 program – some more rigorously than others. We have some data obtained from practices via a post-study survey given to intervention practice staff (lead physicians and others who were familiar with the AM3 program). A cursory glance at these data suggests what one might expect – variability across practices and also within practices on the AM3 implementation. This observation of variability across (and within) busy primary care practices in “how things get accomplished” even operating under a standard protocol remains a central and defining characteristic of primary care.

Dr. Fiscella also raises the issue that the intervention (AM3 program) may not have had an observed effect because it was not potent enough. The authors are in full agreement. Expecting a low-burden and low-intensity intervention to show an effect after the patient’s first exposure to the intervention is highly unlikely. We would love to be in the position to conduct a longitudinal study with a cohort of study patients to see how/if continued exposure to the intervention would affect the communication patterns between patients and their clinicians.

We also agree with Kevin that the context of the visit may matter – the AM3 intervention may have a differential effect depending on the type of visit (acute care, chronic care, wellness, etc). As our time allows, we will return to these data and address this question.

Competing interests:   None declared

Paul Kinnersley contends that the AM3 questions may not have been the “right” ones for these patients to ask – indeed they were asking other several questions (on average) all along. Thus, there was no reason to ask one or more of these specific AM3 questions. And, as he indicates, practices and clinicians should encourage their patients to ask their own questions – the ones that ostensibly make sense to patients.

Dr. Kinnersley also suggests that encouraging patients immediately before their visit to ask specific questions (ones that may have little meaning to them or may actually be distracting/confusing) is not likely to be effective. This is a very good point and again suggests the need for a longitudinal study in which patients might be exposed to the same intervention across time to observe any possible effect in question asking and/or content of the conversations between clinicians and patients. We might also focus on patients who don’t ask many questions and/or those with relatively low health literacy skills.

One issue that we did not raise that Paul does is the nature of clinician behavior during their visits with patients. He suggests that a clinicians’ behavior in encouraging question-asking and also in responding to their patients’ questions is something to be studied along with patient question-asking.

However, his assertion that “Thirdly the [study] physicians were told not to change their approach to the patients…” is only true of control practice physicians – not the AM3 intervention practice physicians (see p. 152, last paragraph third sentence).

Competing interests:   None declared

Ian Bennett contends that one shortcoming of the design is that “the selection scheme did not result in a perfectly random selection by a number of potentially critical variables like ethnicity and educational attainment.” Just to be clear, he is referring to the desired property that each demographic category (females, males, Hispanics, high school graduates, low health literacy, etc) has approximately equal proportions in both the intervention and control groups. One can have a “perfect” random sample (i.e., no shortcomings or mistakes in the actual sampling process), yet the two groups may differ simply due to random sampling fluctuations. Each random sample or group based on random assignment (of patients or practices) will differ due to sampling error – in some cases, the differences may be statistically significant at some level (e.g., p<.05). This simply means that randomization is not a guarantee of statistical comparability between study groups.

Dr. Bennett observes that a better design would be to measure patient question-asking at baseline (before the intervention) and again afterwards with both the control and intervention patients to determine if the AM3 had an impact and to control statistically for any observed differences between the two groups at baseline. He is certainly correct here, and this would have been a much more powerful design, but also more time-consuming, costly, and burdensome to the study practices. But as he suggests, measuring baseline rates of patient-question asking before implementing the intervention would inform the investigators of the variability in questions asked – we then could concentrate on those practices whose patients do not approach some level of question-asking of their clinicians during their visits. This is the group for whom a program such as AM3 might be more beneficial – as he mentions.

We have performed numerous sub-analyses with these data. We did not systematically address the statistical bivariate associations between each of the control (demographic items) and question-asking in either or both groups (intervention and controls). Dr. Bennett also suggests creating a group of “low question askers” (some range of questions asked – e.g., first quartile or below the median) to see if exposure to AM3 resulted in more questions asked compared to study patients not exposed to AM3. Even if not statistically significant due to lack of ample power, this would suggest that with a larger sample of low question askers, the AM3 resulted in more questions asked – whether of the specific AM3 type or not.

Competing interests:   None declared

While congratulating the authors on a generally well-conducted study, perhaps some additional observations might be helpful.

First, these might not have been the right questions. It is not clear the degree to which patients were consulted about the importance of the questions, and whether alternatives were presented. I cannot help but notice that the questions proposed were about enhancing patient acquiescence to the physician’s view of the problem and its treatment. Imagine if the questions suggested to patients were something like, “Can you give me some idea about how much of a difference the treatment will really make?” In oncology, these questions are rarely asked (1) and likely similarly uncommon in the context of otitis media, diabetes and dysthymia.

Second, studies in oncology (1) suggest that patient question-asking is quickly suppressed if physicians do not acknowledge the question, seem interested and ask if there are additional concerns. How did physicians respond to patient questions in this study? Is there a need for a brief physician intervention to help them be more responsive?

Considering the costs associated with poor communication, further refining of interventions to improve communication seems to have sufficient justification.

Reference List

(1) Clayton JM, Butow PN, Tattersall MH et al. Randomized controlled trial of a prompt list to help advanced cancer patients and their caregivers to ask questions about prognosis and end-of-life care. J Clin Oncol 2007;25:715-723.

Competing interests:   None declared

This study highlights the types of practical questions that can be addressed using a randomized controlled design implemented in a practice based research network. The investigators asked a simple question: does low- intensity promotion of the AM3 increase patient question asking during family medicine office visits?

The study shows that it does not. Previous comments have suggested this might be due to ceiling effects because 92% of patients in both groups asked at least one question. This is not surprising - few patients would be expected to ask no questions during an entire office visit. What is notable is that there was no ceiling in asking AM3 questions. Only 30% of control patient and 26% of the intervention patients asked at least one of these questions. Clearly, the intervention had no effect on AM3 questions.

So why did the intervention fail to change patient behavior? Clues to an answer will require the investigators to explore issues surrounding implementation with their data. Did participating practices always ensure that waiting rooms were stocked with brochures? How often did the receptionist forget to hand a brochure to a patient? How often and by what means did medical assistants encourage patients to ask these questions?

If the intervention was effectively implemented it may be that it was simply not potent enough. Previous studies show that formal coaching of patients improve rates of questions during visits. Simply handing out brochures and suggesting patients try out the questions may not be sufficient to change their behavior during the visit. Qualitative exit interview data from participating patients about why they didn't use these questions might shed light on this key issue.

Last, context may matter. The investigators did not restrict enrollment to patients with certain types of visits or problems. Clues about visit context might emerge from further analysis of the audiotapes. Were patients more likely to ask AM3 questions during certain types of visits? Intuitively, one would expect patients to find AM3 questions more relevant for a new presenting complaint or discussion of a new problem such as an abnormal finding.

Determining the reasons for the negative study findings will maximize the value of the study. Discovering whether the findings reflect suboptimal implementation of the intervention, wrong intervention or wrong context will inform next steps in research related to question asking during office visits.

Competing interests:   None declared

This is a large, apparently well conducted randomised trial. For those of us who support methods of trying to help patients ask more questions in their consultations the results may appear disappointing but I think there are several possible explanations. Firstly this study constrained patients to ask three particular questions rather than encouraging them to think of their own questions. This approach may not have suited some patients and in fact Table 5 shows that more of the control group asked the AM3 questions than the intervention patients! Secondly the intervention was delivered to the patients immediately before their consultations which means it has little time to work or for patients to really consider changing their approach to the consultation. Thirdly the physicians were told not to change their approach to the patients so even if the intervention patients were feeling empowered to ask more questions they may have found this difficult if the physicians were not open to this. Finally as the authors point out, both groups of patients asked a relatively high number of questions and there may have been a ceiling effect preventing further improvement.

Given the lack of differences between the groups on question asking it is unsurprising that there were no differences in the adherence outcomes. But it is interesting to note that only 69-74% recalled a new prescription and 59-68% recalled a lifestyle recommendation. This demonstrates that there is a need for improvement in the communication in these consultations.

With colleagues I have conducted a systematic review [1] of 33 similar interventions in a range of settings (but mainly family practice and Oncology). Meta-analysis showed that the interventions do increase question asking and patient satisfaction without leading to an increase in consultation length. I would support the authors conclusion that research is needed into interventions which are used over several visits as I think this would increase their power and provide us with more interesting data.

Reference

1. Kinnersley P, Edwards A, Hood K, Cadbury N, Ryan R, Prout H, Owen D, MacBeth F, Butow P, Butler C. Interventions before consultations for helping patients address their information needs. BMJ 2008;337:485- doi:10.1136/bmj.a485

Competing interests:   None declared