Dataset Information

Detecting conversation topics in primary care office visits from transcripts of patient-provider interactions.

ABSTRACT: OBJECTIVE:Amid electronic health records, laboratory tests, and other technology, office-based patient and provider communication is still the heart of primary medical care. Patients typically present multiple complaints, requiring physicians to decide how to balance competing demands. How this time is allocated has implications for patient satisfaction, payments, and quality of care. We investigate the effectiveness of machine learning methods for automated annotation of medical topics in patient-provider dialog transcripts. MATERIALS AND METHODS:We used dialog transcripts from 279 primary care visits to predict talk-turn topic labels. Different machine learning models were trained to operate on single or multiple local talk-turns (logistic classifiers, support vector machines, gated recurrent units) as well as sequential models that integrate information across talk-turn sequences (conditional random fields, hidden Markov models, and hierarchical gated recurrent units). RESULTS:Evaluation was performed using cross-validation to measure 1) classification accuracy for talk-turns and 2) precision, recall, and F1 scores at the visit level. Experimental results showed that sequential models had higher classification accuracy at the talk-turn level and higher precision at the visit level. Independent models had higher recall scores at the visit level compared with sequential models. CONCLUSIONS:Incorporating sequential information across talk-turns improves the accuracy of topic prediction in patient-provider dialog by smoothing out noisy information from talk-turns. Although the results are promising, more advanced prediction techniques and larger labeled datasets will likely be required to achieve prediction performance appropriate for real-world clinical applications.

SUBMITTER: Park J

PROVIDER: S-EPMC6857514 | biostudies-literature | 2019 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Detecting conversation topics in primary care office visits from transcripts of patient-provider interactions.

Park Jihyun J Kotzias Dimitrios D Kuo Patty P Logan Iv Robert L RL Merced Kritzia K Singh Sameer S Tanana Michael M Karra Taniskidou Efi E Lafata Jennifer Elston JE Atkins David C DC Tai-Seale Ming M Imel Zac E ZE Smyth Padhraic P

Journal of the American Medical Informatics Association : JAMIA 20191201 12

<h4>Objective</h4>Amid electronic health records, laboratory tests, and other technology, office-based patient and provider communication is still the heart of primary medical care. Patients typically present multiple complaints, requiring physicians to decide how to balance competing demands. How this time is allocated has implications for patient satisfaction, payments, and quality of care. We investigate the effectiveness of machine learning methods for automated annotation of medical topics ...[more]

PMID: 31532490

Similar Datasets

Project description:BackgroundDirect-to-consumer telemedicine is an increasingly used modality to access primary care. Previous research on assessment using synchronous virtual visits showed mixed results regarding antibiotic prescription rates, and research on assessment using asynchronous chat-based eVisits is lacking.ObjectiveThe goal of the research was to investigate if eVisit management of sore throat, other respiratory symptoms, or dysuria leads to higher rates of antibiotic prescription compared with usual management using physical office visits.MethodsData from 3847 eVisits and 759 office visits for sore throat, dysuria, or respiratory symptoms were acquired from a large private health care provider in Sweden. Data were analyzed to compare antibiotic prescription rates within 3 days, antibiotic type, and diagnoses made. For a subset of sore throat visits (n=160 eVisits, n=125 office visits), Centor criteria data were manually extracted and validated.ResultsAntibiotic prescription rates were lower following eVisits compared with office visits for sore throat (169/798, 21.2%, vs 124/312, 39.7%; P<.001) and respiratory symptoms (27/1724, 1.6%, vs 50/251, 19.9%; P<.001), while no significant differences were noted comparing eVisits to office visits for dysuria (1016/1325, 76.7%, vs 143/196, 73.0%; P=.25). Guideline-recommended antibiotics were prescribed similarly following sore throat eVisits and office visits (163/169, 96.4%, vs 117/124, 94.4%; P=.39). eVisits for respiratory symptoms and dysuria were more often prescribed guideline-recommended antibiotics (26/27, 96.3%, vs 37/50, 74.0%; P=.02 and 1009/1016, 99.3%, vs 135/143, 94.4%; P<.001, respectively). Odds ratios of antibiotic prescription following office visits compared with eVisits after adjusting for age and differences in set diagnoses were 2.94 (95% CI 1.99-4.33), 11.57 (95% CI 5.50-24.32), 1.01 (95% CI 0.66-1.53), for sore throat, respiratory symptoms, and dysuria, respectively.ConclusionsThe use of asynchronous eVisits for the management of sore throat, dysuria, and respiratory symptoms is not associated with an inherent overprescription of antibiotics compared with office visits.Trial registrationClinicalTrials.gov NCT03474887; https://clinicaltrials.gov/ct2/show/NCT03474887.

Project description:BackgroundThe relationship between worklife factors, clinician outcomes, and time pressure during office visits is unclear.ObjectiveTo quantify associations between time pressure, workplace characteristics ,and clinician outcomes.DesignProspective analysis of data from the Healthy Work Place randomized trial.Participants168 physicians and advanced practice clinicians in 34 primary care practices in Upper Midwest and East Coast.Main measures and methodsTime pressure was present when clinicians needed more time than allotted to provide quality care. Other metrics included work control, work pace (calm to chaotic), organizational culture and clinician satisfaction, stress, burnout, and intent to leave the practice. Hierarchical analysis assessed relationships between time pressure, organizational characteristics, and clinician outcomes. Adjusted differences between clinicians with and without time pressure were expressed as effect sizes (ESs).Key resultsSixty-seven percent of clinicians needed more time for new patients and 53% needed additional time for follow-up appointments. Time pressure in new patient visits was more prevalent in general internists than in family physicians (74% vs 55%, p?<?0.05), women versus men (78% vs 55%, p?<?0.01), and clinicians with larger numbers of complex psychosocial (81% vs 59%, p?<?0.01) and Limited English Proficiency patients (95% vs 57%, p?<?0.001). Time pressure in new patient visits was associated with lack of control, clinician stress, and intent to leave (ESs small to moderate, p?<?0.05). Time pressure in follow-up visits was associated with chaotic workplaces and burnout (small to moderate ESs, p's?<?0.05). Time pressure improved over time in workplaces with values alignment and an emphasis on quality.ConclusionsTime pressure, more common in women and general internists, was related to chaos, control and culture, and stress, burnout, and intent to leave. Future studies should evaluate these findings in larger and more geographically diverse samples.

Project description:Background: Many medical centers are beginning to use OpenNotes (ON) to empower patients. However, there is a lack of literature reviewing the ON system in dermatology and any differences in attitudes between men and women. If so, it is uncertain what concerns are more important to female patients. Given the complex lexicon of notes in dermatology, the outpatient setting of dermatology practices, and the often-complex nature of treatment regimens, investigation was merited. Objective: This paper aimed to evaluate a survey of dermatologic patients on their attitudes toward the ON system. Methods: From July through October 2015, 333 dermatologic patients at the Beth Israel Deaconess Medical Center completed an anonymous, voluntary, 25-question survey of the ON system while in the waiting room. Approximately 60% of respondents were female and 40% were male. Respondents were older, with 27% age >65 years, 21% between 56 and 65 years, 16% between 46 and 55 years, 17% between 36 and 45 years, 14% between 26 and 35 years, and 4% between 18 and 25 years. Eighty-five percent of respondents were white, and 73% had, at minimum, graduated from college. Results: Patient response to ON was positive, with 93% agreeing ON is a good idea. Of the patients who accessed their own notes (69% of respondents), 99.6% desired continued access. In addition, 85.6% of patients felt ON allowed them to control their own health, and 70% reported increased confidence in their dermatologist. Nineteen percent of respondents thought ON presented a privacy concern. Conclusion: The results showed that female patients strongly desire access to their medical records, but concerns about privacy and security exist. Preliminary analysis by a statistician did not find any statistically significant variations between men and women within the results of the survey. Due to the wide agreement in responses, it is unlikely that there are significant differences in opinion on ON between men and women.

Project description:ImportanceLittle is known about the association between the coronavirus disease 2019 (COVID-19) pandemic and the level and content of primary care delivery in the US.ObjectiveTo quantify national changes in the volume, type, and content of primary care delivered during the COVID-19 pandemic, especially with regard to office-based vs telemedicine encounters.Design, setting, and participantsAnalysis of serial cross-sectional data from the IQVIA National Disease and Therapeutic Index, a 2-stage, stratified nationally representative audit of outpatient care in the US from the first calendar quarter (Q1) of 2018 to the second calendar quarter (Q2) of 2020.Main outcomes and measuresVisit type (office-based or telemedicine), overall and stratified by patient population and geographic region; assessment of blood pressure or cholesterol measurement; and initiation or continuation of prescription medications.ResultsIn the 8 calendar quarters between January 1, 2018, and December 31, 2019, between 122.4 million (95% CI, 117.3-127.5 million) and 130.3 million (95% CI, 124.7-135.9 million) quarterly primary care visits occurred in the US (mean, 125.8 million; 95% CI, 121.7-129.9 million), most of which were office-based (92.9%). In 2020, the total number of encounters decreased to 117.9 million (95% CI, 112.6-123.2 million) in Q1 and 99.3 million (95% CI, 94.9-103.8 million) in Q2, a decrease of 21.4% (27.0 million visits) from the average of Q2 levels during 2018 and 2019. Office-based visits decreased 50.2% (59.1 million visits) in Q2 of 2020 compared with Q2 2018-2019, while telemedicine visits increased from 1.1% of total Q2 2018-2019 visits (1.4 million quarterly visits) to 4.1% in Q1 of 2020 (4.8 million visits) and 35.3% in Q2 of 2020 (35.0 million visits). Decreases occurred in blood pressure level assessment (50.1% decrease, 44.4 million visits) and cholesterol level assessment (36.9% decrease, 10.2 million visits) in Q2 of 2020 compared with Q2 2018-2019 levels, and assessment was less common during telemedicine than during office-based visits (9.6% vs 69.7% for blood pressure; P < .001; 13.5% vs 21.6% for cholesterol; P < .001). New medication visits in Q2 of 2020 decreased by 26.0% (14.1 million visits) from Q2 2018-2019 levels. Telemedicine adoption occurred at similar rates among White individuals and Black individuals (19.3% vs 20.5% of patient visits, respectively, in Q1/Q2 of 2020), varied by region (low of 15.1% of visits [East North Central region], high of 26.8% of visits [Pacific region]), and was not correlated with regional COVID-19 burden.Conclusions and relevanceThe COVID-19 pandemic has been associated with changes in the structure of primary care delivery, with the content of telemedicine visits differing from that of office-based encounters.

Dataset Information

Detecting conversation topics in primary care office visits from transcripts of patient-provider interactions.

Publications

Detecting conversation topics in primary care office visits from transcripts of patient-provider interactions.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure