Exploring the Value of Using Patient-Oriented MRI Reports in Clinical Practice

Nathan Perlis (  nathan.perlis@uhn.ca ) University Health Network https://orcid.org/0000-0003-1258-9847 Antonio Finelli University Health Network Mike Lovas University Health Network Alexis Lund University Health Network Amelia Di Meo University Health Network Katherine Lajkosz University Health Network Alejandro Berlin University Health Network Janet Papadakos University Health Network Sangeet Ghai University Health Network Dominik Deniffel University Health Network Eric Meng University Health Network David Wiljer University Health Network Shabbir Alibhai University Health Network Vasiliki Bakas University Health Network Adam Badzynski University Health Network Odelia Lee University Health Network Joseph Cafazzo University Health Network Masoom Haider University Health Network

MpMRI can direct biopsies towards lesions that are occult on ultrasound yielding more clinically signi cant PCa and less clinically insigni cant PCa than systematic biopsy alone [2,3]. MpMRI is also used to monitor patients on active surveillance, limiting repeat biopsies [4].
To be transparent, many organizations are connecting patients to their radiology reports via online patient portals. This enhanced access to medical records can make patients feel more in control [5,6]. However, in general, the quality of online records is variable [7].
Radiology reports contain vital information that is relied on for patients and families to discuss diagnosis, treatment, and prognosis with healthcare providers [8]. While medical records have become more accessible for patients, reports are not designed for patients. "Standard" radiology reports (SRR) are intended for communication amongst doctors. Despite standardization by the Prostate Imaging -Reporting and Data System v2.1 (PI-RADS v2.1), prostate mpMRI is a case study in reports that are di cult for patients to understand [9]. This is because they contain very technical medical language, esoteric scoring, and countless abbreviations.
In fact, sharing records that require high health literacy does not promote understanding and could harm patients [10]. Without knowledge to interpret reports or doctor-guidance, patients are left to speculate the implications of reports. Thus, a different approach is needed. To improve clinical encounters, several groups have proposed a work ow where patients have immediate access to augmented, patient-centred reports, in addition to the standard ones, to help transmit an appropriate amount of information with language and design intended for patients [11,12].
To address this gap, in collaboration with patients and expert groups, our team designed a patient-centered radiology report for prostate MRI (PACERR) [13]. PACERRs are individualized radiology reports with an annotated image of a prostate and explanation of any identi ed lesions. The aim of the current study is to test the usefulness of PACERR in a clinical setting. We hypothesize that, in comparison to SRRs alone, adding PACERRs improves patient understanding of key ndings on MRI, allows for a more patient-oriented clinic visit, and improves doctor-patient communication.

Patients and trial work ow
A prospective clinical trial in accordance with research ethics guidelines and Ontario Personal Health Information Protection Act (PHIPA) was performed with institutional research ethics approval (REB #18-6240) between June and October 2019. Patients with upcoming prostate MRIs were approached for study participation. Eligible patients were 18+, English speaking and pending prostate MRI for active surveillance, prior negative biopsy but rising PSA, or biopsy naïve with elevated PSA. After informed consent was provided, patients were alternately assigned to receive their prostate MRI report as either an SRR alone or an SRR and PACERR. Group assignment was strati ed by MRI status (normal MRI vs abnormal MRI result). Basic clinicopathologic data (e.g., age, PSA, clinical state) were collected.
Individualized PACERRs were generated with Adobe Acrobat by investigators using information from the SRRs (see Figure I for example PACERR). Scoring rules were designated by consensus during pilot testing and used to determine various MRI results (e.g., "enlarged prostate") and tumour characteristics (e.g. "small spot" -Appendix I).
There were no validated questionnaires available to assess understanding and experience with medical reports. Thus, study investigators designed a questionnaire based on patient input during prototype development. Questionnaires included 16 multiplechoice that target 4 domains hypothesized to improve with patient-oriented reports (understanding, usefulness, next steps, emotional experience) and 11 short answer questions, testing knowledge regarding MRI results.
Prostate MRI reports (either SRR alone or SRR + PACERR) were emailed to patients once available. There was typically a one-week delay between report production and the clinic visit where patients would be reviewing results with their doctor. Patients completed the questionnaire twice. The pre-visit questionnaire (Appendix II) was completed within 7 days of receiving the MRI report(s), while the post-visit questionnaire (Appendix III) was administered within 30 days of the clinical visit. During the clinical visit, there was also an observational component exploring doctor and patient behavior. The observational component was scored by a clinical research assistant observing the clinical interaction, tabulating various ways reports were being used (e.g. sketching on report) and calculating visit time (see scoring table -Appendix IV).

Sample size calculation and statistical analysis
We anticipated that patients in the PACERR group could score 10% higher on knowledge questions. With 20 patients in each group there is 80% power to detect a 10% difference between the two groups at a signi cance level of 5%. We also felt that 20 patients would be su cient to perform factor analysis to assess if questions were performing as expected.
Patient demographics, strati ed by arm, were summarized using descriptive statistics and compared using Fisher's Exact and Mann-Whitey U tests. Factor analysis and thematic synthesis were used to organize questionnaire questions into categories.
Oblimin rotation and a loading cutoff of 0.4 were utilized for the factor analysis. Cronbach's alpha was used to evaluate the internal consistency of each group. To analyze differences in responses to the multiple-choice questionnaire, responses were coded as ordinal outcomes (1 = strongly disagree, 2 = somewhat disagree, 3 = neutral, 4 = somewhat agree, 5 = strongly agree).
The group score was calculated by taking the mean of its components. Differences in response distribution between arms were visualized using density plots. Patients' understanding was quanti ed by assessing the ability to correctly identify spots, likelihood of cancer and accuracy of additional ndings. Multiple choice and observational data were summarized using descriptive statistics, and differences between arms were evaluated using the Mann-Whitney U test. Subgroup analyses strati ed by arm and MRI positivity status were explored, and differences between subgroups were assessed using Kruskal-Wallis tests. All statistical tests were 2-tailed and p<0.05 was considered statistically signi cant. Statistical analysis was performed using R version 3.6.1 (R Foundation for Statistical Computing, Vienna, Austria).

Patient and Clinical Characteristics
We contacted 90 participants for enrolment in this study. Forty-two patients consented (enrolment rate 47%) and 2 patients withdrew (retention rate 95%). Thirty-seven participants completed the pre-visit questionnaire and 40 participants completed the post-visit questionnaire. Patient and clinical characteristics were well-balanced between groups except for age (Table I)  In the subgroup analysis, presented in Figure II and Appendix VII, results were further strati ed by MRI status (i.e., whether patients had a positive or negative MRI result). In general, the differences between SRR and PACERR groups were similar to the main analysis regardless of MRI result. However, some of the questions in the emotional experience domain demonstrated differences between the groups that were not apparent in main analysis. For example, despite similar distribution of PIRADs scored between arms, patients in the SRR arm with an abnormal MRI were signi cantly more likely to feel that they received very bad news compared to those in the PACERR arm with an abnormal MRI (36% vs. 8%, p=0.036). Furthermore, patients with an abnormal PACERR were signi cantly more likely to state that receiving their MRI report ahead of time lowered their anxiety compared with SRR patients who received an abnormal report (75% vs. 45%, p=0.029).

Pre-visit Questionnaire results -short-answer section
In the short-answer knowledge section, overall, patients in the PACERR arm scored better than patients in the SRR arm (87% vs. 56%, p=0.004). 80% of participants with PACERR correctly acknowledged the presence of potentially malignant abnormality in their prostate on MRI report as compared to 38% of SRR participants (p=0.051). 84% of patients in the PACERR arm correctly identi ed the likelihood of cancer in a particular spot on their MRI compared to 50% of participants who received the SRR alone (p=0.041). The proportion of patients in the PACERR and SRR arms that correctly identi ed abnormalities on their MRI was 88% and 67%, respectively, though the difference was not statistically signi cant (p=0.22).

Doctor and Patient Behaviors
Identi ed behaviors from the clinical encounter strati ed by group assignment are shown in Table III. There was no difference in the clinical encounter time between the two arms (PACERR median 7 minutes, range 2-27 minutes; SRR median 8 minutes, range 2-37 minutes). Results strati ed by arm and MRI status are presented in Appendix VIII. The PACERR abnormal report was verbally referred to by doctors signi cantly more than the SRR abnormal report (mean = 2.79 times and 0.9 times, respectively (p=0.027)).

Post-visit Questionnaire
There were no statistically signi cant differences in questionnaire responses between arms or signi cant interactions for the postvisit questionnaire.

Discussion
PACERR prostate MRI was designed to improve patients' understanding of their MRI report, improve their experience and enable them to interact more productively with the healthcare system. In this trial, patients undergoing prostate MRI for common clinical indications were randomly assigned to usual care with standard radiology reports versus adding an individualized patient-centred report.
Patients who received both an SRR and PACERR reported that their MRI reports were easy to understand and useful for discussing treatment plans with their medical team.
Patients in the PACERR group were also better able to determine whether there were possible sites of malignant abnormality in their reports. Interestingly, even though there was an additional report to review, the PACERR did not increase the length of clinical encounters. While most observed behaviors were similar between interactions with and without a PACERR, it was noted that doctors more regularly referred to results during clinical encounters with PACERRs than SRRs. Thus, although the main aim of PACERR is to improve patient understanding and experience, doctors also found them useful for communication.
Interestingly, the differences that were seen between groups, when the questionnaire was administered prior to the clinical encounter (i.e., pre-visit), were not present when the questionnaire was re-administered after the clinical encounter (i.e., post-visit).
We hypothesize that this is, at least in part, because by the time the post-visit questionnaire was administered, patients already had the opportunity to review results with their doctors, contemplate the implications of the report and educate themselves in other ways. Thus, it appears that the PACERR is most bene cial in the period between the publication of the SRR and the clinical encounter.
We believed that providing patients with comprehensible information would allow them to approach the clinical visit informed and prepared for productive interactions. Indeed, patients in the PACERR arm had higher scores when asked how con dent they were to take part in the decision-making process. Interestingly, the bene ts of having a patient-centred report are not uniform and sometimes can depend on the actual test result. For example, we found that patients with a PACERR and a lesion on MRI were less worried and anxious about the result compared to patients who had a standard report with a lesion. It is possible that the design, language and targeted information present in the PACERR better prepares patients to receive di cult results on the report.
The patient centered report improved the overall patient experience and created a more patient-oriented visit by encouraging doctors to interactively use the report to refer to results and support their decision-making process. The PACERR also encouraged patients to use their report as an interactive take home tool, which was evidenced by the fact that the PACERR was drawn and written on more frequently than the SRR. We also believe that, despite having an extra document to review, the clinical encounter time was not prolonged with PACERR because patients had better understanding of the results. This study has several limitations. We limited inclusion to English speaking patients so it is unknown whether non-English speakers, a group that might bene t most from a patient-centred report, bene t from PACERR. Household income and education level are likely associated with comprehension and communication with doctors. This study contained a wealthier and highly educated population which may not be representative of the general population. Furthermore, with a small sample size we were unable to demonstrate statistically signi cant differences in several domains which were hypothesized to be improved with PACERR. For example, this was apparent in the questions about lowering anxiety which showed potentially important differences in score distribution but did not reach statistically signi cance. Finally, non-validated questionnaires were used. We use a nonvalidated questionnaire because there were no validated ones available that targeted patients' experiences with medical records.
Statistical assessment of the in-house developed questionnaires demonstrated good consistency and grouping. However, there are some unique strengths including using a PACERR that was developed with patient-input at every design step [13]. Also, doctors who were in the trial had no exposure to the PACERRs prior to the clinical trial, so it mimics a more real-world roll-out of such technology.

Conclusion
Through plain language and simple design, our team has created a patient centered radiology report that increases patients' selfreported understanding and their ability to accurately identify lesions and their approximate likelihood of cancer. Participants were empowered to actively involve themselves in discussions with their doctors, communicate their results to family and friends, and think realistically about the future. Future work will focus on encompassing patient demographics not adequately represented in this study and expanding to other disease sites and pathology reports in addition to automating the creation of individualized patient-centred reports.

Dr. Perlis reports from Canadian Urological Association Scholarship Foundation (CUASF) -Early Investigator Research
Scholarship, during the conduct of the study.

Con icts of interest/Competing interests
The authors have no con icts of interest to declare that are relevant to the content of this article.  Tables   Table I: Patient and clinical characteristics of entire cohort and strati ed by group assignment. PACERR = patient-centred radiology report, SRR= standard radiology report, SD= standard deviation, AS= Active surveillance, BN= Biopsy naïve, PSA= prostate speci c antigen