Dataset Information

Development of a keyword library for capturing PRO-CTCAE-focused "symptom talk" in oncology conversations.

ABSTRACT:

Objectives

As computational methods for detecting symptoms can help us better attend to patient suffering, the objectives of this study were to develop and evaluate the performance of a natural language processing keyword library for detecting symptom talk, and to describe symptom communication within our dataset to generate insights for future model building.

Materials and methods

This was a secondary analysis of 121 transcribed outpatient oncology conversations from the Communication in Oncologist-Patient Encounters trial. Through an iterative process of identifying symptom expressions via inductive and deductive techniques, we generated a library of keywords relevant to the Patient-Reported Outcome version of the Common Terminology Criteria for Adverse Events (PRO-CTCAE) framework from 90 conversations, and tested the library on 31 additional transcripts. To contextualize symptom expressions and the nature of misclassifications, we qualitatively analyzed 450 mislabeled and properly labeled symptom-positive turns.

Results

The final library, comprising 1320 terms, identified symptom talk among conversation turns with an F1 of 0.82 against a PRO-CTCAE-focused gold standard, and an F1 of 0.61 against a broad gold standard. Qualitative observations suggest that physical symptoms are more easily detected than psychological symptoms (eg, anxiety), and ambiguity persists throughout symptom communication.

Discussion

This rudimentary keyword library captures most PRO-CTCAE-focused symptom talk, but the ambiguity of symptom speech limits the utility of rule-based methods alone, and limits to generalizability must be considered.

Conclusion

Our findings highlight opportunities for more advanced computational models to detect symptom expressions from transcribed clinical conversations. Future improvements in speech-to-text could enable real-time detection at scale.

SUBMITTER: Durieux BN

PROVIDER: S-EPMC9912707 | biostudies-literature | 2023 Apr

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Development of a keyword library for capturing PRO-CTCAE-focused "symptom talk" in oncology conversations.

Durieux Brigitte N BN Zverev Samuel R SR Tarbi Elise C EC Kwok Anne A Sciacca Kate K Pollak Kathryn I KI Tulsky James A JA Lindvall Charlotta C

JAMIA open 20230209 1

<h4>Objectives</h4>As computational methods for detecting symptoms can help us better attend to patient suffering, the objectives of this study were to develop and evaluate the performance of a natural language processing keyword library for detecting symptom talk, and to describe symptom communication within our dataset to generate insights for future model building.<h4>Materials and methods</h4>This was a secondary analysis of 121 transcribed outpatient oncology conversations from the Communic ...[more]

PMID: 36789287

Similar Datasets

Project description:BackgroundIn the postoperative period, individual patient experiences vary widely and are based on a diverse set of input variables influenced by all stakeholders in and throughout the surgical process. Although clinical research has primarily focused on clinical and administrative datasets to characterize the postoperative recovery experience, there is increasing interest in patient-reported outcome measures (PROMs). The growth of online communities in which patients themselves participate provides a venue to study PROMs directly. One such forum-based community is HysterSisters, dedicated to helping individuals through the experience of hysterectomy, a major surgery which removes the uterus. The surgery can be performed by a variety of methods such as minimally invasive approaches or the traditional abdominal approach using a larger incision. The community offers support for "medical and emotional issues [...] from diagnosis, to treatment, to recovery." Users can specify when and what type of hysterectomy they underwent. They can discuss their shared experience of hysterectomy and provide, among other interactions, feedback, reassurance, sympathy, or advice, thus providing a unique view into conversations surrounding the hysterectomy experience.ObjectiveWe aimed to characterize conversations about hysterectomy recovery as experienced by users of the HysterSisters online community.MethodsA retrospective keyword frequency analysis of the HysterSisters Hysterectomy Recovery forum was performed.ResultsWithin the Hysterectomy Recovery forum, 33,311 unique users declared their hysterectomy date and type and posted during the first 12 weeks postsurgery. A taxonomy of 8 primary symptom groups was created using a seed list of keywords generated from a term frequency analysis of these threads. Pain and bleeding were the two most mentioned symptom groups and account for almost half of all symptom mentions (19,965/40,127). For symptoms categories such as pain and hormones and emotions, there was no difference in the proportion of users mentioning related keywords, regardless of the type of hysterectomy, whereas bleeding-related or intimacy-related keywords were mentioned more frequently by users undergoing certain minimally invasive approaches when compared with those undergoing abdominal hysterectomy. Temporal patterns in symptom mentions were noted as well. The majority of all posting activity occurred in the first 3 weeks. Across all keyword groups, individuals reporting minimally invasive procedures ceased forum use of these keywords significantly earlier than those reporting abdominal hysterectomy. Peaks in conversation volume surrounding particular symptom categories were also identified at 1, 3, and 6 weeks postoperatively.ConclusionsThe HysterSisters Hysterectomy Recovery forum and other such forums centered on users' health care experience can provide novel actionable insights that can improve patient-centered care during the postoperative period. This study adds another dimension to the utility of social media analytics by demonstrating that measurement of post volumes and distribution of symptom mentions over time reveal key opportunities for beneficial symptom-specific patient engagement.

Project description:AimsThe US National Cancer Institute recently developed the PRO-CTCAE (Patient-Reported Outcomes version of the Common Terminology Criteria for Adverse Events). PRO-CTCAE is a library of questions for clinical trial participants to self-report symptomatic adverse events (e.g. nausea). The objective of this study is to inform evidence-based selection of a recall period when PRO-CTCAE is included in a trial. We evaluated differences between 1-, 2-, 3-, and 4-week recall periods, using daily reporting as the reference.MethodsEnglish-speaking patients with cancer receiving chemotherapy and/or radiotherapy were enrolled at four US cancer centers and affiliated community clinics. Participants completed 27 PRO-CTCAE items electronically daily for 28 days, and then weekly over 4 weeks, using 1-, 2-, 3-, and 4-week recall periods. For each recall period, mean differences, effect sizes, and intraclass correlation coefficients were calculated to evaluate agreement between the maximum of daily ratings and the corresponding ratings obtained using longer recall periods (e.g. maximum of daily scores over 7 days vs 1-week recall). Analyses were repeated using the average of daily scores within each recall period rather than the maximum of daily scores.ResultsA total of 127 subjects completed questionnaires (57% male; median age: 57). The median of the 27 mean differences in scores on the PRO-CTCAE 5-point response scale comparing the maximum daily versus the longer recall period (and corresponding effect size) was -0.20 (-0.20) for 1-week recall, -0.36 (-0.31) for 2-week recall, -0.45 (-0.39) for 3-week recall, and -0.47 (-0.40) for 4-week recall. The median intraclass correlation across 27 items between the maximum of daily ratings and the corresponding longer recall ratings for 1-week recall was 0.70 (range: 0.54-0.82), for 2-week recall was 0.74 (range: 0.58-0.83), for 3-week recall was 0.72 (range: 0.61-0.84), and for 4-week recall was 0.72 (range: 0.64-0.86). Similar results were observed for all analyses using the average of daily scores rather than the maximum of daily scores.ConclusionA 1-week recall corresponds best to daily reporting. Although intraclass correlations remain stable over time, there are small but progressively larger differences between daily and longer recall periods at 2, 3, and 4 weeks, respectively. The preferred recall period for the PRO-CTCAE is the past 7 days, although investigators may opt for recall periods of 2, 3, or 4 weeks with an understanding that there may be some information loss.

Dataset Information

Development of a keyword library for capturing PRO-CTCAE-focused "symptom talk" in oncology conversations.

Objectives

Materials and methods

Results

Discussion

Conclusion

Publications

Development of a keyword library for capturing PRO-CTCAE-focused "symptom talk" in oncology conversations.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets