Predictive analytics for cardiovascular patient readmission and mortality: An explainable approach

BACKGROUND

Cardiovascular patients experience high rates of adverse outcomes following discharge from hospital, which may be preventable through early identification and targeted action. This study aimed to investigate the effectiveness and explainability of machine learning algorithms in predicting unplanned readmission and death in cardiovascular patients at 30 days and 180 days from discharge.

METHODS

Gradient boosting machines were trained and evaluated using data from hospital electronic medical records linked to hospital administrative and mortality data for 39,255 patients admitted to four hospitals in New South Wales, Australia between 2017 and 2021. Sociodemographic variables, admission history, and clinical information were used as potential predictors. The performance was compared to LASSO regression, as well as the HOSPITAL and LACE risk score indices. Important risk factors identified by the gradient-boosting machine model were explored using Shapley values.

RESULTS

The models performed well, especially for the mortality outcomes. Area under the receiver operating characteristic curve values were 0.70 for readmission and 0.87-0.90 for mortality using the full gradient boosting machine algorithms. Among the top predictors for 30-day and 180-day readmission were increased red cell distribution width, old age (especially above 80 years), high measured troponin and urea levels, not being married or in a relationship, and low albumin levels. For mortality, these included increased red cell distribution width, old age (especially older than 70 years), high measured troponin and urea levels, high neutrophil and monocyte counts, and low eosinophil and lymphocyte counts. The Shapley values gave clear insight into the dynamics of decision-tree-based models.

CONCLUSIONS

We demonstrated an explainable predictive algorithm to identify cardiovascular patients who are at high risk of readmission or death at discharge from the hospital and identified key risk factors.

Web-Based Application Based on Human-in-the-Loop Deep Learning for Deidentifying Free-Text Data in Electronic Medical Records: Development and Usability Study

BACKGROUND

The narrative free-text data in electronic medical records (EMRs) contain valuable clinical information for analysis and research to inform better patient care. However, the release of free text for secondary use is hindered by concerns surrounding personally identifiable information (PII), as protecting individuals' privacy is paramount. Therefore, it is necessary to deidentify free text to remove PII. Manual deidentification is a time-consuming and labor-intensive process. Numerous automated deidentification approaches and systems have been attempted to overcome this challenge over the past decade.

OBJECTIVE

We sought to develop an accurate, web-based system deidentifying free text (DEFT), which can be readily and easily adopted in real-world settings for deidentification of free text in EMRs. The system has several key features including a simple and task-focused web user interface, customized PII types, use of a state-of-the-art deep learning model for tagging PII from free text, preannotation by an interactive learning loop, rapid manual annotation with autosave, support for project management and team collaboration, user access control, and central data storage.

METHODS

DEFT comprises frontend and backend modules and communicates with central data storage through a filesystem path access. The frontend web user interface provides end users with a user-friendly workspace for managing and annotating free text. The backend module processes the requests from the frontend and performs relevant persistence operations. DEFT manages the deidentification workflow as a project, which can contain one or more data sets. Customized PII types and user access control can also be configured. The deep learning model is based on a Bidirectional Long Short-Term Memory-Conditional Random Field (BiLSTM-CRF) with RoBERTa as the word embedding layer. The interactive learning loop is further integrated into DEFT to speed up the deidentification process and increase its performance over time.

RESULTS

DEFT has many advantages over existing deidentification systems in terms of its support for project management, user access control, data management, and an interactive learning process. Experimental results from DEFT on the 2014 i2b2 data set obtained the highest performance compared to 5 benchmark models in terms of microaverage strict entity-level recall and F1-scores of 0.9563 and 0.9627, respectively. In a real-world use case of deidentifying clinical notes, extracted from 1 referral hospital in Sydney, New South Wales, Australia, DEFT achieved a high microaverage strict entity-level F1-score of 0.9507 on a corpus of 600 annotated clinical notes. Moreover, the manual annotation process with preannotation demonstrated a 43% increase in work efficiency compared to the process without preannotation.

CONCLUSIONS

DEFT is designed for health domain researchers and data custodians to easily deidentify free text in EMRs. DEFT supports an interactive learning loop and end users with minimal technical knowledge can perform the deidentification work with only a shallow learning curve.

Formation of a statewide community pharmacy practice-based research network: Pharmacist opinions on research participation and engagement

OBJECTIVE

To gather the insights and opinions of pharmacist stakeholders to inform the creation of a community pharmacy practice-based research network (PBRN) in Pennsylvania.

DESIGN

A stakeholder advisory board of pharmacists, patients, and researchers was established to guide this research. This was a qualitative study using a semistructured interview guide.

SETTING AND PARTICIPANTS

Community pharmacists from the Pennsylvania Pharmacist Care Network.

OUTCOME MEASURES

Themes were identified that describe pharmacist insights and opinions on research participation and preferences for engagement in the PBRN.

RESULTS

A total of 16 pharmacists participated in the study. The pharmacists believed that participating in research would help demonstrate their value and commitment to improving patients' health. Enhancing patient-pharmacist relationships and driving innovation were additional benefits that were reported. The pharmacists believed that they could effectively leverage their relationships with patients to engage them in research opportunities. The pharmacists reported that they would like to share research ideas and successful research practices with other members of the PBRN.

CONCLUSION

Gathering pharmacists' opinions on participating in research was an important step in developing a community pharmacy PBRN that meets stakeholder needs. The results of this study can help others who seek to form community pharmacy PBRNs that facilitate stakeholder-driven research.

Positive evidence for neonatal imitation: A general response, adaptive engagement

The study employed four gestural models using frame-by-frame microanalytic methods, and followed how the behaviours unfolded over time. Forty-two human newborns (0-3 days) were examined for their imitation of tongue protrusion, 'head tilt with looking up', three-finger and two-finger gestures. The results showed that all three gesture groups were imitated. Results of the temporal analyses revealed an early and a later, second stage of responses. Later responses were characterized by a suppression of similar, but non-matching movements. Perinatal imitation is not a phenomenon served by a single underlying mechanism; it has at least two different stages. An early phase is followed by voluntary matching behaviour by the neonatal infant.

Designing Online Interventions in Consideration of Young People's Concepts of Well-Being: Exploratory Qualitative Study

BACKGROUND

A key challenge in developing online well-being interventions for young people is to ensure that they are based on theory and reflect adolescent concepts of well-being.

OBJECTIVE

This exploratory qualitative study aimed to understand young people's concepts of well-being in Australia.

METHODS

Data were collected via workshops at five sites across rural and metropolitan sites with 37 young people from 15 to 21 years of age, inclusive. Inductive, data-driven coding was then used to analyze transcripts and artifacts (ie, written or image data).

RESULTS

Young adults' conceptions of well-being were diverse, personally contextualized, and shaped by ongoing individual experiences related to physical and mental health, along with ecological accounts acknowledging the role of family, community, and social factors. Key emerging themes were (1) positive emotions and enjoyable activities, (2) physical wellness, (3) relationships and social connectedness, (4) autonomy and control, (5) goals and purpose, (6) being engaged and challenged, and (7) self-esteem and confidence. Participants had no difficulty describing actions that led to positive well-being; however, they only considered their own well-being at times of stress.

CONCLUSIONS

In this study, young people appeared to think mostly about their well-being at times of stress. The challenge for online interventions is to encourage young people to monitor well-being prior to it becoming compromised. A more proactive focus that links the overall concept of well-being to everyday, concrete actions and activities young people engage in, and that encourages the creation of routine good habits, may lead to better outcomes from online well-being interventions.

An evaluation of behaviour change techniques in health and lifestyle mobile applications

Despite the current popularity and potential use of mobile applications (apps) in the area of behaviour change, health promotion, and well-being for young people, it is unclear whether their design is underpinned by theory-based behaviour change techniques. Understanding the design of these apps may improve the way they can be used to support young people's well-being.The objectives of this study were to investigate what behaviour change techniques are included in the content of health and lifestyle apps, and determine which of these are prominent in app design. Thirty of the top-listed health and lifestyle apps across three categories (physical activity, diet, and sleep) were freely downloaded from the two most popular app stores (GooglePlay™ and AppStore™). Selected apps were used by trained researchers and the features identified coded against the Behaviour Change Techniques Taxonomy 1, a systematic classification of techniques used in behaviour change interventions. It was found that 9 of the 93 behaviour change techniques listed in the Behaviour Change Techniques Taxonomy 1 were common across the chosen health and lifestyle apps. The app found to include the most behaviour change techniques had 20 (21%), while the app found to include the least had 1 behaviour change technique (1%). The most frequently used behaviour change techniques were related to goal setting and feedback. Entire categories in the Behaviour Change Techniques Taxonomy 1 were absent in the design of the selected apps.

Facilitating mental health help-seeking by young adults with a dedicated online program: a feasibility study of Link

OBJECTIVE

To explore the feasibility of a dedicated online youth mental health help-seeking intervention and to evaluate using a randomised controlled trial (RCT) study design in order to identify any modifications needed before commencement of the full-scale RCT.

DESIGN

A pilot RCT with 1:1 randomisation to either the intervention or comparison arm.

SETTING

An online study conducted Australia-wide.

PARTICIPANTS

18-25 year olds living in Australia were recruited via social media.

INTERVENTION

Link is a dedicated online mental health help-seeking navigation tool that matches user's mental health issues, severity and service-type preferences (online, phone and face-to-face) with appropriate youth-friendly services. The comparison arm was usual help-seeking strategies with a link to Google.com.

MAIN OUTCOME MEASURES

The primary outcome was the number of acceptability and feasibility criteria successfully met. Intervention and study design acceptability and feasibility were assessed by nine criteria. Secondary outcomes, via online surveys (at baseline, 1 week and 1 month) measured service use, help-seeking intentions, psychological distress, barriers to help-seeking, attitudes towards mental health help-seeking, mental health literacy, satisfaction and trust.

RESULTS

Fifty-one participants were randomised (intervention: n=24; comparison: n=27). Three out of four of the intervention and two out of five of the study design criteria were met. Unmet criteria could be addressed by modifications to the study design. Qualitative analysis demonstrated that Link was useful to participants and may have increased their positive experiences towards help-seeking. There were no observable differences between arms in any outcome measures and no harms were detected.

CONCLUSION

Generally, the Link intervention and study design were acceptable and feasible with modifications suggested for the four out of nine unmet criteria. The main trial will hence have shorter surveys and a simpler recruitment process, use positive affect as the primary outcome and will not link to Google.com for the comparison arm.

TRIAL REGISTRATION NUMBER

Australian New Zealand Clinical Trials Registry, ACTRN12614000386639.

Efficacy of a Web-Based Guided Recommendation Service for a Curated List of Readily Available Mental Health and Well-Being Mobile Apps for Young People: Randomized Controlled Trial

Background

Mental disorders are highly prevalent for the people who are aged between 16 and 25 years and can permanently disrupt the development of these individuals. Easily available mobile health (mHealth) apps for mobile phones have great potential for the prevention and early intervention of mental disorders in young adults, but interventions are required that can help individuals to both identify high-quality mobile apps and use them to change health and lifestyle behavior.

Objectives

The study aimed to assess the efficacy of a Web-based self-guided app recommendation service (“The Toolbox”) in improving the well-being of young Australians aged between 16 and 25 years. The intervention was developed in collaboration with young adults and consists of a curated list of 46 readily available health and well-being apps, assessed and rated by professionals and young people. Participants are guided by an interactive quiz and subsequently receive recommendations for particular apps to download and use based on their personal goals.

Methods

The study was a waitlist, parallel-arm, randomized controlled trial. Our primary outcome measure was change in well-being as measured by the Mental Health Continuum-Short Form (MHC-SF). We also employed ecological momentary assessments (EMAs) to track mood, energy, rest, and sleep. Participants were recruited from the general Australian population, via several Web-based and community strategies. The study was conducted through a Web-based platform consisting of a landing Web page and capabilities to administer study measures at different time points. Web-based measurements were self-assessed at baseline and 4 weeks, and EMAs were collected repeatedly at regular weekly intervals or ad hoc when participants interacted with the study platform. Primary outcomes were analyzed using linear mixed-models and intention-to-treat (ITT) analysis.

Results

A total of 387 participants completed baseline scores and were randomized into the trial. Results demonstrated no significant effect of “The Toolbox” intervention on participant well-being at 4 weeks compared with the control group (P=.66). There were also no significant differences between the intervention and control groups at 4 weeks on any of the subscales of the MHC-SF (psychological: P=.95, social: P=.42, emotional: P=.95). Repeat engagement with the study platform resulted in a significant difference in mood, energy, rest, and sleep trajectories between intervention and control groups as measured by EMAs (P<.01).

Conclusions

This was the first study to assess the effectiveness of a Web-based well-being intervention in a sample of young adults. The design of the intervention utilized expert rating of existing apps and end-user codesign approaches resulting in an app recommendation service. Our finding suggests that recommended readily available mental health and well-being apps may not lead to improvements in the well-being of a nonclinical sample of young people, but might halt a decline in mood, energy, rest, and sleep.

Trial Registration

Australian New Zealand Clinical Trials Registry (ANZCTR): ACTRN12614000710628; https://www.anzctr.org.au/Trial/Registration/TrialReview.aspx?id=366145 (Archived by WebCite at http://www.webcitation.org/ 6pWDsnKme)

Development of an online well-being intervention for young people: an evaluation protocol

BACKGROUND

Research has shown that improving well-being using positive mental health interventions can be useful for predicting and preventing mental illness. Implementing online interventions may be an effective way to reach young people, given their familiarity with technology.

OBJECTIVE

This study will assess the effectiveness of a website called the "Online Wellbeing Centre (OWC)," designed for the support and improvement of mental health and well-being in young Australians aged between 16 and 25 years. As the active component of the study, the OWC will introduce a self-guided app recommendation service called "The Toolbox: The best apps for your brain and body" developed by ReachOut.com. The Toolbox is a responsive website that serves as a personalized, ongoing recommendation service for technology-based tools and apps to improve well-being. It allows users to personalize their experience according to their individual needs.

METHODS

This study will be a two-arm, randomized controlled trial following a wait-list control design. The primary outcome will be changes in psychological well-being measured by the Mental Health Continuum Short Form. The secondary outcomes will be drawn from a subsample of participants and will include depression scores measured by the Center for Epidemiologic Studies Depression Scale, and quality of life measured by the Assessment of Quality of Life-four dimensions (AQOL-4D) index. Cost-effectiveness analysis will be conducted based on a primary outcome of cost per unique visit to the OWC. Utility-based outcomes will also be incorporated into the analysis allowing a secondary outcome to be cost per quality-adjusted life year gained (based on the AQOL-4D values). Resource use associated with both the intervention and control groups will be collected using a customized questionnaire. Online- and community-based recruitment strategies will be implemented, and the effectiveness of each approach will be analyzed. Participants will be recruited from the general Australian population and randomized online. The trial will last for 4 weeks.

RESULTS

Small but clinically significant increases in well-being symptoms are expected to be detected in the intervention group compared with the control group.

CONCLUSIONS

If this intervention proves to be effective, it will have an impact on the future design and implementation of online-based well-being interventions as a valid and cost-effective way to support mental health clinical treatment. Findings regarding recruitment effectiveness will also contribute to developing better ways to engage this population in research.

CLINICALTRIAL

This study is registered in the Australian New Zealand Clinical Trials Registry (ANZCTR): ACTRN12614000710628.

A multicenter phase II study of pemetrexed (P), cisplatin (C), and bevacizumab (B) in patients (pts) with advanced malignant mesothelioma (MM)

7578

Background: Vascular endothelial growth factor (VEGF) is a key growth factor for MM. In pre-clinical models, anti-VEGF antibodies inhibit MM proliferation. Standard chemotherapy for MM, pemetrexed + cisplatin, yields a response rate of 41%, progression-free survival (PFS) of 5.7 months (mo) and median overall survival (OS) of 12.1 mo. We added the anti-VEGF antibody B to PC in an ongoing phase II multi-center study in MM pts. Methods: Eligible pts have unresectable, histologically-confirmed MM, no prior chemotherapy, and PS 0–1. Pts receive C 75 mg/m2, P 500 mg/m2 and B 15 mg/kg Q21 days for 6 cycles, then B Q21 days until progression. CT scans are obtained Q 2 cycles. Primary endpoint: Progression-fere survival (PFS). Correlative studies include IHC for the VEGF/KDR complex and PCR for simian virus 40 (sv40) T antigen. Results: 43 pts enrolled at 4 centers from 3/06 to 12/08; currently, 34 are evaluable for PFS and 36 for toxicity. Pt characteristics: male 88%; median age 66 (range 24–81); histology: epithelial 62%, sarcomatoid 15%, biphasic 20%, unknown 3%; site of origin: pleural 85%, peritoneal 12%, tunica vaginalis 3%; PS 0 32%, PS 1 68%; thrombocytosis (>400) 32%. Cycles administered 229 (median 4, range 1–18). Grade 3/4 toxicity (%pts): neutropenia 8%; anemia 0%; thrombocytopenia 3%; thrombosis 11%; hypertension 8%; vomiting 8%; mucositis 6%; CVA 3%; proteinuria 0%; perforation 0%. Partial response: 41%, stable disease: 35%. Median PFS: 5.6 mo (95% CI: 4.1, 7.2). Median OS 11.5 mo (95% CI: 9.4, 24). Conclusions: These data suggest that the addition of B to PC does not improve PFS when compared with historical controls of PC in advanced MM pts. Correlative studies evaluating the VEGF/KDR complex and sv40 T antigen are pending.

[Table: see text]