Data-driven identification of heart failure disease states and progression pathways using electronic health records

Evolution of economic burden of heart failure by ejection fraction in newly diagnosed patients in Spain

OBJECTIVE

To describe healthcare resource utilization (HCRU) and costs, in patients with newly diagnosed heart failure (HF) according to ejection fraction (EF) in Spain.

METHODS

Retrospective cohort study that analyzed anonymized, integrated and computerised medical records in Spain. Patients with ≥ 1 new HF diagnosis between January 2013 and September 2019 were included and followed-up during a 4-year period. Rates per 100 person-years of HCRU and costs were estimated.

RESULTS

Nineteen thousand nine hundred sixty-one patients were included, of whom 43.5%, 26.3%, 5.1% and 25.1% had HF with reduced, preserved, mildly reduced and unknown EF, respectively. From year 1 to 4, HF rates of outpatient visits decreased from 1149.5 (95% CI 1140.8-1159.3) to 765.5 (95% CI 745.9-784.5) and hospitalizations from 61.7 (95% CI 60.9-62.7) to 15.7(14.7-16.7) per 100 person-years. The majority of HF-related healthcare resource costs per patient were due to hospitalizations (year 1-4: 63.3-38.2%), followed by indirect costs (year 1-4: 12.2-29.0%), pharmacy (year 1-4: 11.9-19.9%), and outpatient care (year 1-4: 12.6-12.9%). Mean (SD) per patient HF-related costs decreased from 2509.6 (3518.5) to 1234.6 (1534.1) Euros (50% cost reduction). At baseline, 70.1% were taking beta-blockers, 56.3% renin-angiotensin system inhibitors, 11.8% mineralocorticoid receptor antagonists and 8.9% SGLT2 inhibitors. At 12 months, these numbers were 72.3%, 65.4%, 18.9% and 9.8%, respectively.

CONCLUSIONS

Although the economic burden of HF decreased over time since diagnosis, it is still substantial. This reduction could be partially related to a survival bias (sick patients died early), but also to a better HF management. Despite that, there is still much room for improvement.

Is the patient speaking or the nurse? Automatic speaker type identification in patient-nurse audio recordings

OBJECTIVES

Patient-clinician communication provides valuable explicit and implicit information that may indicate adverse medical conditions and outcomes. However, practical and analytical approaches for audio-recording and analyzing this data stream remain underexplored. This study aimed to 1) analyze patients' and nurses' speech in audio-recorded verbal communication, and 2) develop machine learning (ML) classifiers to effectively differentiate between patient and nurse language.

MATERIALS AND METHODS

Pilot studies were conducted at VNS Health, the largest not-for-profit home healthcare agency in the United States, to optimize audio-recording patient-nurse interactions. We recorded and transcribed 46 interactions, resulting in 3494 "utterances" that were annotated to identify the speaker. We employed natural language processing techniques to generate linguistic features and built various ML classifiers to distinguish between patient and nurse language at both individual and encounter levels.

RESULTS

A support vector machine classifier trained on selected linguistic features from term frequency-inverse document frequency, Linguistic Inquiry and Word Count, Word2Vec, and Medical Concepts in the Unified Medical Language System achieved the highest performance with an AUC-ROC = 99.01 ± 1.97 and an F1-score = 96.82 ± 4.1. The analysis revealed patients' tendency to use informal language and keywords related to "religion," "home," and "money," while nurses utilized more complex sentences focusing on health-related matters and medical issues and were more likely to ask questions.

CONCLUSION

The methods and analytical approach we developed to differentiate patient and nurse language is an important precursor for downstream tasks that aim to analyze patient speech to identify patients at risk of disease and negative health outcomes.