Decision aid use during post‐biopsy consultations for localized prostate cancer

Abstract Background Decision Aids (DAs) effectively translate medical evidence for patients but are not routinely used in clinical practice. Little is known about how DAs are used during patient‐clinician encounters. Objective To characterize the content and communicative function of high‐quality DAs during diagnostic clinic visits for prostate cancer. Participants 252 men newly diagnosed with localized prostate cancer who had received a DA, 45 treating physicians at 4 US Veterans Administration urology clinics. Methods Qualitative analysis of transcribed audio recordings was used to inductively develop categories capturing content and function of all direct references to DAs (booklet talk). The presence or absence of any booklet talk per transcript was also calculated. Results Booklet talk occurred in 55% of transcripts. Content focused on surgical procedures (36%); treatment choice (22%); and clarifying risk classification (17%). The most common function of booklet talk was patient corroboration of physicians’ explanations (42%), followed by either physician or patient acknowledgement that the patient had the booklet. Codes reflected the absence of DA use for shared decision‐making. In regression analysis, predictors of booklet talk were fewer years of patient education (P = .027) and more time in the encounter (P = .027). Patient race, DA type, time reading the DA, physician informing quality and physician age did not predict booklet talk. Conclusions Results show that good decision aids, systematically provided to patients, appeared to function not to open up deliberations about how to balance benefits and harms of competing treatments, but rather to allow patients to ask narrow technical questions about recommended treatments.


| INTRODUCTION
Patient decision aids (DAs) describing treatment options and risk/benefit trade-offs among treatments have been successfully developed and tested over several decades, beginning, for early-stage prostate cancer, in 1988. 1,2 . While DAs are effective information translation tools, they are not routinely used in clinical practice. 3  While previous studies have shown that DAs have potential to positively impact both patient informing and patient-clinician interaction, little is known about the role that DAs play during the exchanges between patients and clinicians. The impact of DAs on the clinical encounter is assumed more often than examined. Of the 105 studies included in the most recent Cochrane review, 1 10 studied the effect on communication. Of those, the five studies that implemented the DA in preparation for the consultation all used self-report measures of decision-making. [11][12][13][14][15] To our knowledge, no previous study has used data from direct observation of patient-clinician communication (ie, data from transcripts or recordings of clinic visits) to identify how patients and clinicians actually use and discuss DAs during encounters.
Analysis of transcripts or recordings (rather than reports based on patient or clinician recollection) is not subject to hindsight bias 16 and is generally considered the most accurate method for assessing the content of communication during clinic visits. 17,18 In this study, we analysed visit transcripts to investigate the content and communicative function of direct references to DAs in patients' post-biopsy urology clinic encounters during which initial treatment decisions for clinically localized prostate cancer were made. We analysed transcripts for these visits to inductively develop categories capturing content and function of direct references to DAs ("booklet talk"). We also examined patient and clinician characteristics associated with the presence of a reference to a DA during the en-  20 The current version of the MCC DA can be found at www.
prostatecancerdecision.org. The NCCN DA was chosen because of its high-quality information and the high credibility of the sponsoring organizations. The current version of the NCCN DA can be found at https://www.nccn.org/patients/guidelines/prostate. Both decision aids used the terminology "watchful waiting" because the term, "active surveillance" was not a commonly used term when this study was conducted. Therefore, we use watchful waiting throughout. (More detailed quality analysis of the DAs can be found in the Appendix S1).
Block randomization was used to ensure that equal numbers of African American and low-literacy patients received each decision aid. Physicians were aware that patients received a DA, but not given any further instructions in DA use. In addition to transcripts of audio recordings, survey data describing patient characteristics and selfreported DA use were available for analysis from the parent trial.
Patients with clinically localized prostate cancer (Gleason score 6 or 7, PSA < 20 ng/mL) were asked to participate in audio recording of the first post-biopsy encounter, the one at which the patient first received his diagnosis and discussed initial treatment options. Surveys were administered at three time points: biopsy, immediately before the physician encounter and 7-10 days following the physician encounter.
Patients were called 2 days before the physician encounter and reminded to read the DA, but were not informed of their diagnosis. They learned their diagnosis from the physician, with the exception of one site that followed a practice of giving the diagnosis over the telephone.
Participants at that site were interviewed before the diagnosis phone call. Physician participants were urology residents and attending physicians. All provided demographic data at the time of recruitment. The study was approved by the VA Institutional Review Boards at each participating site; written informed consent was obtained from each patient and physician participant. The funding agencies had no role in conduct or reporting of the parent study or the analysis presented in this manuscript.

| Measures from the parent study
We obtained descriptive data from the parent study. Survey measures completed by patients before the clinical encounter included patient literacy 21 and numeracy 22 , preference for shared decision-making 23 , prostate cancer treatment knowledge related to survival benefit and side-effects associated with treatments, 24-26 treatment preference, use of and satisfaction with DA and demographics (patients' race, ethnicity, age, marital status and education).
A measure of the quality of physician informing was obtained through a transcript analysis. The Informed Decision Making (IDM) score, 27 is a standardized observational measure of the quality of physician informing behaviour, scored by analysing transcripts of audiorecorded patient encounters. 28 Patients' PSA level, Gleason Score and treatment received were obtained from electronic medical records.

| Audio recordings and transcripts for this analysis
A research associate sets up an audio recorder in the examination room at the start of each visit and then waited outside the examination room until the visit was complete. Recordings were later anonymized and transcribed verbatim. Of 256 transcripts, 252 were available for inclusion. Two transcripts were excluded because of recorder malfunction; two encounters were only to obtain a referral to radiation oncology. Time in the encounter was measured directly from the audio recordings. Time when the physician was out of the room was subtracted from total time to yield the net time the physician was in the room with the patient.

| Coding and qualitative analysis
In this analysis, we coded and analysed direct references to the DA and used a two-step coding process to identify the content categories and function categories to describe how the DA functioned in the exchange. In step one, two coders independently identified all direct DA references. In addition, a word search of the text was performed using the words "booklet", "pamphlet", "book" and "decision aid" to check for missing episodes. Booklet talk was classified into one of four transactional categories: (i) patient initiates and doctor responds, (ii) patient initiates and doctor fails to respond, (iii) doctor initiates and patient responds and (iv) doctor initiates and patient fails to respond.
Coding exchanges (ie, topic initiation and response) accounts for the interactional nature of clinic visits and is a common approach when coding patient-clinician communication. 29,30 Because we previously noted that communication tasks during these visits occurred in a predictable sequence, 31 we analysed a random sample of 28 transcripts to evaluate whether booklet talk also occurred in predictable portions of visits, for example. after diagnosis delivery, during treatment choice discussions, at the close of the encounter. To do this, we calculated the percentage of total words in each transcript before each episode of booklet talk and analysed the distribution of results in the 28 randomly sampled transcripts. The wide distribution of percentages of words before episodes of booklet talk (range = 1-99) and no clear clustering pattern, suggested there was no part of the clinical routine that triggered booklet talk. We therefore did not pursue a separate structural analysis of the encounters.
In step two, we inductively developed the set of content and function codes for each coded exchange by carefully analysing a ran-

| Regression analysis
To identify predictors of booklet talk during the consultation, we conducted two mixed effects logistic regression models, using patient, physician and encounter level variables from the parent study to predict the presence of booklet talk in the transcript. Among the variables available in the parent study, we prioritized those with a theoretical relationship to the likelihood of mentioning the DA in the encounter. Patient education and race have been previously associated with how much patients participate in encounters with physicians. [34][35][36][37] Time spent reading the DA before the encounter was included as a measure of interest in the content. Age was not included because of the narrow range of patient ages. Time in the encounter, measured in minutes from the recordings, was included because trials of DAs have been shown across studies to sometimes shorten and sometimes lengthen encounter times. 38,39 DA type (plain vs standard language) was also included as a variable, as randomization in the original study was based on DA type.

| RESULTS
Demographic characteristics for the 252 patients are shown in Table 3.
The mean age of the patient sample was 63.3 years (SD = 5.9); 33% were non-white; 40% had high school education or less. The mean age of 45 treating physicians was 33 (SD = 7.2); 20% were women, 34%  Expression of concern Doctor or patient expresses concern specifically from something seen in the booklet 4 T A B L E 2 Booklet talk function codes

| Content and function
DAs were referenced most frequently during discussion of treatment options (36%). The most common specific content code was details of surgery. Direct references to making a treatment decision constituted 22% of all content codes; clarification of technical information about risk classification, 18%. Frequencies for all content codes are shown in Table 1, and examples of each content code appear in Table 4. The functions were dominated by "learn more or validate" (41%) and "acknowledging the booklet" (28%). "Learn more" was usually a patient request to hear the physician's explanation for or interpretation of something the patient read in the DA. "Acknowledging the booklet" was usually a physician question about whether the patient received a DA or a comment that s/he saw the patient carrying a copy of the DA. The third most frequent category was consistent with the design of DAs, "using the booklet to ask a question" (12%). Examples of each category of function codes appear in Table 5.
Occasionally, patients referred to the booklet to explain how worried they were about their prostate cancer or specific treatments. The DA, in these instances, appeared to serve as either reassurance, or as a vehicle for expressing concern to the physician. (See Table 5.) Patient references to the DA to challenge the physician's recommendation were rare.

| Predictors of booklet talk
Results of the first mixed effects regression model revealed that only education and time in the clinical encounter predicted reference to PAT-Okay.
DOC-Okay, and um, you know, we're happy to kind of help you make whatever decision it is that you want to make, whether that's surgery or radiation.

Example 2
DOC-Okay um, and even with aggressive disease….. the chance it can affect your lifespan at five years is low. It interests me that some patients say, "Listen I really want this tumor out." And we get the tumor out, it's cancer, even despite the fact that I tell them that not all cancer is the same.

DOC-Okay
PAT-Well, like she had breast cancer and she immediately wanted it out. I basically said the same thing the other day. If I find out I have cancer I immediately want it out. But now, that I you know, read some of that and after talking with you I got a little, "Yeah it's, we'll do the wait and see approach for awhile." DOC-The only caveat about the wait and see approach again is that, you're a little different than the typical wait and see approach patient.

| DISCUSSION
While many DAs have been shown to be effective in translating medical evidence for patients, they are not routinely used in practice. 1,3 Our results contribute to better understanding of this implementation conundrum. We found that DAs appeared to function not to open up deliberations about how to balance benefits and harms of competing treatments, but to allow patients to ask narrow technical questions about recommended treatments. This was contrary to expectations, as we chose high-quality DAs, shown previously to be engaging to patients. 19 We found no evidence that DAs functioned to facilitate shared decision-making in the encounter.
Direct references to DAs occurred in over half of encounters.
Direct references to the DA were more frequently initiated by patients than physicians. This may in part, be attributable to the fact that physicians did not receive any training in DA use, while patients were asked to use the DAs to prepare for the encounter. However, the analysis PAT-This book is talking about a PSA number and then they're talking about a Gleason s….  PAT-If the high is ten, you're over half. DOC 2-Right, but the lowest grade that they call is six DOC 1-Right DOC 2-It's a scale of six to ten, not, not zero to ten PAT-Not according to this book DOC 2-Yeah. Well the pathologists don't call Gleason fives anymore, they used to, but they don't anymore.

Expression of concern
Example 1 PAT-Whenever I start getting upset or nervous about this I can take this out and start reading through it again. The way things are explained in here kind of calms you down.
DOC-Well yeah, that's good to know. Specifically, 41% of booklet talk functioned to validate or prompt additional discussion of a topic ("learn more"); the most common topic discussed when referencing the DA was details of specific treatment options. Patients only used the DA as a platform for asking the physician a question in 12% of transcripts (see Table 2 Regression analysis showed that less well-educated patients were more likely to mention the DA. This finding is not unexpected. In our prior research, we found that patients with less education gained more knowledge from a DA. 42 It is one of the unique characteristics of DAs that those who are less knowledgeable before reading a DA gain the most knowledge. DAs are also designed to provide an authoritative source to help patients ask questions. 1 Time in the encounter, but not the quality of physician informing (IDM score) predicted booklet talk.
This lends support to earlier findings that discussing patient questions raised by a DA may take a small amount of extra time in the encounter.
The Cochrane Review of DAs shows that the association of DA use with time in the encounter is highly variable, sometimes associated with shorter and sometimes with longer encounters (range −4 minutes to +23 minutes), with an average of 2.6 minutes longer 1 . It is important to note that in this study, which over-sampled African American patients, race was not a predictor of booklet talk. This suggests that minority and white patients were equally likely to use the DA in the encounter.
While there are studies of patient-clinician communication focused on measuring the presence of shared decision-making, 43 we know of none that investigates how DAs function in real time