OBJECTIVE: Serious illness conversations are complex clinical narratives that remain poorly understood. Natural Language Processing (NLP) offers new approaches for identifying hidden patterns within the lexicon of stories that may reveal insights about the taxonomy of serious illness conversations.
METHODS: We analyzed verbatim transcripts from 354 consultations involving 231 patients and 45 palliative care clinicians from the Palliative Care Communication Research Initiative. We stratified each conversation into deciles of "narrative time" based on word counts. We used standard NLP analyses to examine the frequency and distribution of words and phrases indicating temporal reference, illness terminology, sentiment and modal verbs (indicating possibility/desirability).
RESULTS: Temporal references shifted steadily from talking about the past to talking about the future over deciles of narrative time. Conversations progressed incrementally from "sadder" to "happier" lexicon; reduction in illness terminology accounted substantially for this pattern. We observed the following sequence in peak frequency over narrative time: symptom terms, treatment terms, prognosis terms and modal verbs indicating possibility.
CONCLUSIONS: NLP methods can identify narrative arcs in serious illness conversations.
PRACTICE IMPLICATIONS: Fully automating NLP methods will allow for efficient, large scale and real time measurement of serious illness conversations for research, education and system re-design.
OBJECTIVE: Clinicians frequently overestimate survival time among seriously ill patients, and this can result in medical treatment at end of life that does not reflect the patient's preferences. Little is known, however, about the sources of clinicians' optimistic bias in survival estimation. Related work in social networks and experimental psychology demonstrates that psychological states-such as optimism-can transfer from one person to another.
METHODS: We directly observed and audio recorded 189 initial inpatient palliative care consultations among hospitalized patients with advanced cancer. Patients self-reported their level of trait optimism and expectations for survival prognosis prior to the palliative care consultation, and the palliative care clinicians rated their expectations for the patient's survival time following the initial conversation with the patient. We followed patient mortality for 6 months.
RESULTS: Patient optimism was associated with clinician overestimation of their survival in a dose-response relationship. Clinicians were approximately three times as likely to overestimate the survival of patients endorsing both high trait optimism and optimistic ratings of their survival time compared with neither (OR: 2.95; 95% CI: 1.24-7.02). This association was not attenuated by adjustment for age, gender, race, ethnicity, education, income, cancer type, functional status, quality of life, or white blood cell count (ORadj : 3.45; 95% CI: 1.24-9.66).
CONCLUSION: Patients' optimism may have some influence over their clinicians' prognostic judgments.
OBJECTIVE: We examined whether conversations involving Black or Latino patients with advanced cancer differ in the presence or characteristics of prognosis communication.
METHODS: We audio-recorded initial consultations between 54 palliative care clinicians and 231 hospitalized people with advanced cancer. We coded for the presence and characteristics of prognosis communication. We examined whether the presence or characteristics of prognosis communication differed by patients' self-reported race/ethnicity.
RESULTS: In 231 consultations, 75.7% contained prognosis communication. Prognosis communication was less than half as likely to occur during conversations with Black or Latino patients (N = 48) compared to others. Among consultations in which prognosis was addressed, those involving Black or Latino patients were more than 8 times less likely to contain optimistically cued prognoses compared to others.
CONCLUSION: Prognosis communication occurred less frequently for Black and Latino patients and included fewer optimistic cues than conversations with other patients. More work is needed to better understand these observed patterns of prognosis communication that vary by race and ethnicity.
PRACTICE IMPLICATIONS: Growing evidence supports prognosis communication being important for end-of-life decision-making and disproportionately rare among non-White populations. Therefore, our findings identify a potentially salient target for clinical interventions that are focused on ameliorating disparities in end-of-life care.
CONTEXT: Prognosis communication is one hypothesized mechanism by which effective palliative care (PC) promotes preference-concordant treatment near end of life (EOL), but little is known about this relationship.
METHODS: This is a multisite cohort study of 231 hospitalized patients with advanced cancer who consulted with PC. We audio-recorded the initial consultation with the PC team and coded conversations for all statements regarding expectations for how long the patient will live. We refer to these statements as length-of-life talk. We followed patients for up to six months to determine EOL treatment utilization, including hospice enrollment. Patients completed a brief interviewer-facilitated questionnaire at study enrollment.
RESULTS: Forty-four percent (101/231) of observed conversations contained at least one statement about expectations for length of life, and 60% of patients (139/231) enrolled in hospice during the six months following these conversations. The association between length-of-life talk and hospice enrollment was strong among those (155/231) who endorsed treatment preferences favoring comfort over longevity in the last weeks to months of life (odds ratio [OR]adj = 2.98; 95% confidence interval [CI] = 1.34–6.65) and weak/absent among others (69/231; ORadj = 0.70; 95% CI = 0.16–3.04).
CONCLUSIONS: Talking about expectations for remaining length of life during PC consultations is associated with six-month hospice enrollment among people with advanced cancer who endorse preferences for EOL treatment that favor comfort over longevity.
Context: Clinicians frequently overestimate survival time in serious illness.
Objective: To understand the frequency of overestimation in palliative care (PC) and the relation with end-of-life (EOL) treatment.
Methods: This is a multi-site cohort study of 230 hospitalized patients with advanced cancer who consulted with PC between 2013 and 2016. We asked the consulting PC clinician to make their "best guess" about the patients' "most likely survival time, assuming that their illnesses are allowed to take their natural course." [<24 hours; 24 hours to < 2 weeks; 2 weeks to < 3 months; 3 months to < 6 months; 6 months or longer]. We followed patients for up to 6-months for mortality and EOL treatment utilization. Patients completed a brief interviewer-facilitated questionnaire at study enrollment.
Results: Median survival was 37 days (Interquartile Range: 12 days, 97 days) and 186/230 (81%) died during the follow-up period. Forty-one percent of clinicians' predictions were accurate. Among inaccurate prognoses, 85% were overestimates. Among those who died, overestimates were substantially associated with less hospice use (ORadj: 0.40; 95% CI: 0.16, 0.99) and later hospice enrollment (within 72 hours of death ORadj: 0.33; 95% CI: 0.15, ,0.74). PC clinicians were substantially more likely to overestimate survival for patients who identified as Black or Latino compared to others (ORadj: 3.89; 95% CI: 1.64, 9.22). EOL treatment preferences did not explain either of these findings.
Conclusions: Overestimation is common in PC, associated with lower hospice use and a potentially mutable source of racial/ethnic disparity in EOL care.
BACKGROUND: Systematic measurement of conversational features in the natural clinical setting is essential to better understand, disseminate, and incentivize high quality serious illness communication. Advances in machine-learning (ML) classification of human speech offer exceptional opportunity to complement human coding (HC) methods for measurement in large scale studies.
OBJECTIVES: To test the reliability, efficiency, and sensitivity of a tandem ML-HC method for identifying one feature of clinical importance in serious illness conversations: Connectional Silence.
DESIGN: This was a cross-sectional analysis of 354 audio-recorded inpatient palliative care consultations from the Palliative Care Communication Research Initiative multisite cohort study.
SETTING/SUBJECTS: Hospitalized people with advanced cancer.
MEASUREMENTS: We created 1000 brief audio "clips" of randomly selected moments predicted by a screening ML algorithm to be two-second or longer pauses in conversation. Each clip included 10 seconds of speaking before and 5 seconds after each pause. Two HCs independently evaluated each clip for Connectional Silence as operationalized from conceptual taxonomies of silence in serious illness conversations. HCs also evaluated 100 minutes from 10 additional conversations having unique speakers to identify how frequently the ML screening algorithm missed episodes of Connectional Silence.
RESULTS: Connectional Silences were rare (5.5%) among all two-second or longer pauses in palliative care conversations. Tandem ML-HC demonstrated strong reliability (kappa 0.62; 95% confidence interval: 0.47-0.76). HC alone required 61% more time than the Tandem ML-HC method. No Connectional Silences were missed by the ML screening algorithm.
CONCLUSIONS: Tandem ML-HC methods are reliable, efficient, and sensitive for identifying Connectional Silence in serious illness conversations.
OBJECTIVE: Automating conversation analysis in the natural clinical setting is essential to scale serious illness communication research to samples that are large enough for traditional epidemiological studies. Our objective is to automate the identification of pauses in conversations because these are important linguistic targets for evaluating dynamics of speaker involvement and turn-taking, listening and human connection, or distraction and disengagement.
DESIGN: We used 354 audio recordings of serious illness conversations from the multisite Palliative Care Communication Research Initiative cohort study.
SETTING/SUBJECTS: Hospitalized people with advanced cancer seen by the palliative care team.
MEASUREMENTS: We developed a Random Forest machine learning (ML) algorithm to detect Conversational Pauses of two seconds or longer. We triple-coded 261 minutes of audio with human coders to establish a gold standard for evaluating ML performance characteristics.
RESULTS: ML automatically identified Conversational Pauses with a sensitivity of 90.5 and a specificity of 94.5.
CONCLUSIONS: ML is a valid method for automatically identifying Conversational Pauses in the natural acoustic setting of inpatient serious illness conversations.
CONTEXT: Maximizing value in palliative care requires continued development and standardization of communication quality indicators.
OBJECTIVES: To describe the basic epidemiology of a newly-adopted patient-centered communication quality indicator for hospitalized palliative care patients with advanced cancer.
METHODS: Cross-sectional analysis of 207 advanced cancer patients who received palliative care consultation at two medical centers in the United States. Participants completed the Heard & Understood quality indicator immediately before and the day following the initial palliative care consultation: "Over the past two days ["24 hours" for the post-consultation version], how much have you felt heard and understood by the doctors, nurses and hospital staff? Completely/Quite a Bit/Moderately/Slightly/Not at All". We categorized "Completely" as indicating ideal quality.
RESULTS: Approximately one-third indicated ideal Heard & Understood quality before palliative care consultation. Age, financial security, emotional distress, preferences for comfort-longevity tradeoffs at end-of-life, and prognosis expectations were associated with pre-consultation quality. Among those with less-than-ideal quality at baseline, 56% rated feeling more Heard & Understood the day following palliative care consultation. The greatest pre-post improvement was among people who had unformed end-of-life treatment preferences or who reported having "no idea" about their prognosis at baseline.
CONCLUSION: Most patients felt incompletely heard and understood at the time of referral to palliative care consultation and more than half improved following consultation. Feeling heard and understood is an important quality indicator sensitive to interventions to improve care and key variations in the patient experience.
The Measuring What Matters (MWM) initiative identified 10 indicators of high-quality palliative and hospice care. Members of the AAHPM Research Committee, through a special series of articles, examined applications of the MWM quality indicators in research and practice settings. Many themes were present in these articles, including the important role of electronic health records in quality measurement, challenges and strategies for implementing and tracking measures over time, and the importance of identifying new measures. This article is the final commentary of the series and includes recommendations for next steps in quality measurement.
Etude qualitative qui décrit les perspectives des patients qui ont une insuffisance cardiaque et de leurs proches, en ce qui concerne les obstacles aux soins palliatifs et des services de soins palliatifs dans le cas d'insuffisance cardiaque à un stade avancé.