Enhancing Suicide Risk Prediction Models with Temporal Clinical Note Features.

OBJECTIVE: The objective of this study was to investigate the impact of enhancing a structured-data-based suicide attempt risk prediction model with temporal Concept Unique Identifiers (CUIs) derived from clinical notes. We aimed to examine how different temporal schemes, model types, and prediction ranges influenced the model's predictive performance. This research sought to improve our understanding of how the integration of temporal information and clinical variable transformation could enhance model predictions. MATERIALS AND METHODS: We identified modeling targets using diagnostic codes for suicide attempts within 30, 90, or 365 days following a temporally grouped visit cluster. Structured data included medications, diagnoses, procedures, and demographics, while unstructured data consisted of terms extracted with regular expressions from clinical notes. We compared models trained only on structured data (controls) to hybrid models trained on both structured and unstructured data. We used two temporalization schemes for clinical notes: fixed 90-day windows and flexible epochs. We trained and assessed random forests and hybrid LSTM neural networks using AUPRC and AUROC, with additional evaluation of sensitivity and PPV at 95% specificity. RESULTS: The training set included 2,364,183 visit clusters with 2,009 30-day suicide attempts, and the testing set contained 471,936 visit clusters with 480 suicide attempts. Models trained with temporal CUIs outperformed those trained with only structured data. The window-temporalized LSTM model achieved the highest AUPRC (0.056 ± 0.013) for the 30-day prediction range. Hybrid models generally showed better performance compared to controls across most metrics. DISCUSSION AND CONCLUSION: This study demonstrated that incorporating EHR-derived clinical note features enhanced suicide attempt risk prediction models, particularly with window-temporalized LSTM models. Our results underscored the critical value of unstructured data in suicidality prediction, aligning with previous findings. Future research should focus on integrating more sophisticated methods to continue improving prediction accuracy, which will enhance the effectiveness of future intervention.

The Path to Diagnosis of Severe Asthma-A Qualitative Exploration.

Purpose: Severe asthma poses a significant health burden in those with the disease, therefore a timely diagnosis can ensure patients receive specialist care and appropriate medication management. This study qualitatively explored the patient experience of adult Australians with severe asthma regarding specialist referral, to identify potential opportunities to streamline the process of severe asthma diagnosis and treatment and optimise referral pathways. Patients and Methods: Adults currently being treated with medication for severe asthma were invited to participate in this study. Participants were interviewed and asked to describe initial diagnosis of their asthma or severe asthma, and how they came to be referred to secondary care. Interviews were transcribed verbatim, coded by two members of the research team and thematically analysed. Results: Thirty-two people completed the study; 72% were female. Mean interview length was 33 minutes. The major themes generated were patient-related factors contributing to seeking a severe asthma diagnosis; perceptions of health care provision; diagnosis of severe asthma and the referral journey. Key findings were that both patient and healthcare provider attitudes contributed to participants' willingness to seek or receive a referral, and referral to respiratory specialists was often delayed. Contributing factors included a mismatch between patient expectations and general practice, lack of continuity of primary care, and a lack of patient understanding of the role of the respiratory specialist. Conclusion: Timely severe asthma diagnosis in Australia appears to be hampered by an absence of a clear referral process, lack of general practitioner (GP) knowledge of additional treatment options, underutilisation of pharmacists, and multiple specialists treating patient comorbidities. Directions for future research might include interviewing healthcare providers regarding how well the referral process works for severe asthma patients, and researching the time between referral and when a patient sees the respiratory specialist.

Enhancing Postmarketing Surveillance of Medical Products With Large Language Models.

Importance: The Sentinel System is a key component of the US Food and Drug Administration (FDA) postmarketing safety surveillance commitment and uses clinical health care data to conduct analyses to inform drug labeling and safety communications, FDA advisory committee meetings, and other regulatory decisions. However, observational data are frequently deemed insufficient for reliable evaluation of safety concerns owing to limitations in underlying data or methodology. Advances in large language models (LLMs) provide new opportunities to address some of these limitations. However, careful consideration is necessary for how and where LLMs can be effectively deployed for these purposes. Observations: LLMs may provide new avenues to support signal-identification activities to identify novel adverse event signals from narrative text of electronic health records. These algorithms may be used to support epidemiologic investigations examining the causal relationship between exposure to a medical product and an adverse event through development of probabilistic phenotyping of health outcomes of interest and extraction of information related to important confounding factors. LLMs may perform like traditional natural language processing tools by annotating text with controlled vocabularies with additional tailored training activities. LLMs offer opportunities for enhancing information extraction from adverse event reports, medical literature, and other biomedical knowledge sources. There are several challenges that must be considered when leveraging LLMs for postmarket surveillance. Prompt engineering is needed to ensure that LLM-extracted associations are accurate and specific. LLMs require extensive infrastructure to use, which many health care systems lack, and this can impact diversity, equity, and inclusion, and result in obscuring significant adverse event patterns in some populations. LLMs are known to generate nonfactual statements, which could lead to false positive signals and downstream evaluation activities by the FDA and other entities, incurring substantial cost. Conclusions and Relevance: LLMs represent a novel paradigm that may facilitate generation of information to support medical product postmarket surveillance activities that have not been possible. However, additional work is required to ensure LLMs can be used in a fair and equitable manner, minimize false positive findings, and support the necessary rigor of signal detection needed for regulatory activities.

Characterization and Association of the Missing Ventral Tubercle(s) from the Sixth Cervical Vertebra and Transpositions on the Ventral Surface of the Seventh Cervical Vertebra in Modern Equus ferus caballus.

In recent years, equine complex vertebral malformation (ECVM) has been of concern in the equine community, with studies identifying numerous associative morphological variations. Here, we examine the morphological association between C6 and C7 for dependency in ECVM cases, where the partially absent ventral process of C6 transposes on the ventral surface of C7. A C6 ventral process presents two tubercles, one cranial (CrVT) and one caudal (CVT). In this study, the C6 osseous specimens (n = 85) demonstrated a partial or completely absent CVT (aCVT) graded 1-4 that often extended cranially creating a partially absent cranial ventral tubercle (aCrVT) graded 1-3. In the 85 C6 osseous specimens examined, the corresponding C7s demonstrated either a complete or incomplete transposition of the ventral process from C6 in 44/85, with 30/44 replicating a transverse foramen. A strong statistical dependency existed between C6 grade 4 aCVTs and grades 1-3 aCrVTs and C7 transpositions with replicated transverse foramen. Sidedness was also demonstrated, where a left sided absent C6 associated with transposition on the left ventral surface of C7. This likewise applied to right sidedness and most bilateral cases. These findings might benefit practitioners when radiographing the extent of the ECVM configuration in patients presenting caudal cervical pain.

Still Separate, Still Not Equal: An Ecological Examination of Redlining and Racial Segregation with COVID-19 Vaccination Administration in Washington D.C.

Racial residential segregation has been deemed a fundamental cause of health inequities. It is a result of historical and contemporary policies such as redlining that have created a geographic separation of races and corresponds with an inequitable distribution of health-promoting resources. Redlining and racial residential segregation may have contributed to racial inequities in COVID-19 vaccine administration in the early stages of public accessibility. We use data from the National Archives (historical redlining), Home Mortgage Disclosure Act (contemporary redlining), American Community Survey from 1940 (historical racial residential segregation) and 2015-2019 (contemporary racial residential segregation), and Washington D.C. government (COVID-19 vaccination administration) to assess the relationships between redlining, racial residential segregation, and COVID-19 vaccine administration during the early stages of vaccine distribution when a tiered system was in place due to limited supply. Pearson correlation was used to assess whether redlining and racial segregation, measured both historically and contemporarily, were correlated with each other in Washington D.C. Subsequently, linear regression was used to assess whether each of these measures associate with COVID-19 vaccine administration. In both historical and contemporary analyses, there was a positive correlation between redlining and racial residential segregation. Further, redlining and racial residential segregation were each positively associated with administration of the novel COVID-19 vaccine. This study highlights the ongoing ways in which redlining and segregation contribute to racial health inequities. Eliminating racial health inequities in American society requires addressing the root causes that affect access to health-promoting resources.

Prevalence and burden of difficult-to-treat and severe asthma in Australia: A national population survey.

BACKGROUND AND OBJECTIVE: Most evidence about difficult-to-treat and severe asthma (DTTA) comes from clinical trials and registries. We aimed to identify people with DTTA from a large nationally representative asthma population and describe their characteristics and healthcare utilization compared with people whose asthma was not 'difficult-to-treat'. METHODS: We conducted a cross-sectional survey of Australians aged ≥18 years with current asthma from large web-based survey panels. Enrolment was stratified by gender, age-group and state/territory based on national population data for people with asthma. Difficult-to-treat or severe asthma was defined by poor symptom control, exacerbations and/or oral corticosteroid/biologic use despite medium/high-dose inhaled therapy. Outcomes included exacerbations, healthcare utilization, multimorbidity, quality of life and coronavirus disease of 2019 (COVID-19)-related behaviour. Weighted data were analysed using SAS version 9.4. RESULTS: The survey was conducted in February-March 2021. The weighted sample comprised 6048 adults with current asthma (average age 47.3 ± SD 18.1 years, 59.9% female), with 1313 (21.7%) satisfying ≥1 DTTA criteria. Of these, 50.4% had very poorly controlled symptoms (Asthma Control Test ≤15), 36.2% were current smokers, and 85.4% had ≥1 additional chronic condition, most commonly anxiety/depression. More than twice as many participants with DTTA versus non-DTTA had ≥1 urgent general practitioner (GP) visit (61.4% vs. 27.5%, OR 4.8 [4.2-5.5, p < 0.0001]), or ≥1 emergency room visit (41.9% vs. 17.9%, OR 3.8 [3.3-4.4, p < 0.0001]) in the previous 12 months. CONCLUSION: Our findings emphasize the burden of uncontrolled symptoms, current smoking, multimorbidity and healthcare utilization in people with DTTA in the community, who may be under-represented in registries or clinical trials.

Mobile app activity engagement by cancer patients and their caregivers informs remote monitoring.

Mobile phone applications ("apps") are potentially an effective, low-burden method to collect patient-reported outcomes outside the clinical setting. Using such apps consistently and in a timely way is critical for complete and accurate data capture, but no studies of concurrent reporting by cancer patient-caregiver dyads have been published in the peer-reviewed literature. This study assessed app engagement, defined as adherence, timing, and attrition with two smartphone applications, one for adult cancer patients and one for their informal caregivers. This was a single-arm, pilot study in which adult cancer patients undergoing IV chemotherapy or immunotherapy used the DigiBioMarC app, and their caregivers used the TOGETHERCare app, for approximately one month to report weekly on the patients' symptoms and wellbeing. Using app timestamp metadata, we assessed user adherence, overall and by participant characteristics. Fifty patient-caregiver dyads completed the study. Within the one-month study period, both adult cancer patients and their informal caregivers were highly adherent, with app activity completion at 86% for cancer patients and 84% for caregivers. Caregivers completed 86% of symptom reports, while cancer patients completed 89% of symptom reports. Cancer patients and their caregivers completed most activities within 48 h of availability on the app. These results suggest that the DigiBioMarC and TOGETHERCare apps can be used to collect patient- and caregiver-reported outcomes data during intensive treatment. From our research, we conclude that metadata from mobile apps can be used to inform clinical teams about study participants' engagement and wellbeing outside the clinical setting.

Designing mHealth Apps to Incorporate Evidence-Based Techniques for Prolonging User Engagement.

Maintaining user engagement with mobile health (mHealth) apps can be a challenge. Previously, we developed a conceptual model to optimize patient engagement in mHealth apps by incorporating multiple evidence-based methods, including increasing health literacy, enhancing technical competence, and improving feelings about participation in clinical trials. This viewpoint aims to report on a series of exploratory mini-experiments demonstrating the feasibility of testing our previously published engagement conceptual model. We collected data from 6 participants using an app that showed a series of educational videos and obtained additional data via questionnaires to illustrate and pilot the approach. The videos addressed 3 elements shown to relate to engagement in health care app use: increasing health literacy, enhancing technical competence, and improving positive feelings about participation in clinical trials. We measured changes in participants' knowledge and feelings, collected feedback on the videos and content, made revisions based on this feedback, and conducted participant reassessments. The findings support the feasibility of an iterative approach to creating and refining engagement enhancements in mHealth apps. Systematically identifying the key evidence-based elements intended to be included in an app's design and then systematically testing the implantation of each element separately until a satisfactory level of positive impact is achieved is feasible and should be incorporated into standard app design. While mHealth apps have shown promise, participants are more likely to drop out than to be retained. This viewpoint highlights the potential for mHealth researchers to test and refine mHealth apps using approaches to better engage users.

Sustainable deployment of clinical prediction tools-a 360° approach to model maintenance.

BACKGROUND: As the enthusiasm for integrating artificial intelligence (AI) into clinical care grows, so has our understanding of the challenges associated with deploying impactful and sustainable clinical AI models. Complex dataset shifts resulting from evolving clinical environments strain the longevity of AI models as predictive accuracy and associated utility deteriorate over time. OBJECTIVE: Responsible practice thus necessitates the lifecycle of AI models be extended to include ongoing monitoring and maintenance strategies within health system algorithmovigilance programs. We describe a framework encompassing a 360° continuum of preventive, preemptive, responsive, and reactive approaches to address model monitoring and maintenance from critically different angles. DISCUSSION: We describe the complementary advantages and limitations of these four approaches and highlight the importance of such a coordinated strategy to help ensure the promise of clinical AI is not short-lived.

Correction: Correlation Between Remote Symptom Reporting by Caregivers and Adverse Clinical Outcomes: Mixed Methods Study.

[This corrects the article DOI: 10.2196/49100.].