Lessons and tips for designing a machine learning study using EHR data

Efficient management of pulmonary embolism diagnosis using a two-step interconnected machine learning model based on electronic health records data

Pulmonary Embolism (PE) is a life-threatening clinical disease with no specific clinical symptoms and Computed Tomography Angiography (CTA) is used for diagnosis. Clinical decision support scoring systems like Wells and rGeneva based on PE risk factors have been developed to estimate the pre-test probability but are underused, leading to continuous overuse of CTA imaging. This diagnostic study aimed to propose a novel approach for efficient management of PE diagnosis using a two-step interconnected machine learning framework directly by analyzing patients' Electronic Health Records data. First, we performed feature importance analysis according to the result of LightGBM superiority for PE prediction, then four state-of-the-art machine learning methods were applied for PE prediction based on the feature importance results, enabling swift and accurate pre-test diagnosis. Throughout the study patients' data from different departments were collected from Sina educational hospital, affiliated with the Tehran University of medical sciences in Iran. Generally, the Ridge classification method obtained the best performance with an F1 score of 0.96. Extensive experimental findings showed the effectiveness and simplicity of this diagnostic process of PE in comparison with the existing scoring systems. The main strength of this approach centered on PE disease management procedures, which would reduce avoidable invasive CTA imaging and be applied as a primary prognosis of PE, hence assisting the healthcare system, clinicians, and patients by reducing costs and promoting treatment quality and patient satisfaction.

Clinical and socioeconomic predictors of hospital use and emergency department visits among children with medical complexity: A machine learning approach using administrative data

OBJECTIVES

The primary objective of this study was to identify clinical and socioeconomic predictors of hospital and ED use among children with medical complexity within 1 and 5 years of an initial discharge between 2010 and 2013. A secondary objective was to estimate marginal associations between important predictors and resource use.

METHODS

This retrospective, population-cohort study of children with medical complexity in Alberta linked administrative health data with Canadian census data and used tree-based, gradient-boosted regression models to identify clinical and socioeconomic predictors of resource use. Separate analyses of cumulative numbers of hospital days and ED visits modeled the probability of any resource use and, when present, the amount of resource use. We used relative importance in each analysis to identify important predictors.

RESULTS

The analytic sample included 11 105 children with medical complexity. The best short- and long-term predictors of having a hospital stay and number of hospital days were initial length of stay and clinical classification. Initial length of stay, residence rurality, and other socioeconomic factors were top predictors of short-term ED use. The top predictors of ED use in the long term were almost exclusively socioeconomic, with rurality a top predictor of number of ED visits. Estimates of marginal associations between initial length of stay and resource use showed that average number of hospital days increases as initial length of stay increases up to approximately 90 days. Children with medical complexity living in rural areas had more ED visits on average than those living in urban or metropolitan areas.

CONCLUSIONS

Clinical factors are generally better predictors of hospital use whereas socioeconomic factors are more predictive of ED use among children with medical complexity in Alberta. The results confirm existing literature on the importance of socioeconomic factors with respect to health care use by children with medical complexity.

Oncologic Applications of Artificial Intelligence and Deep Learning Methods in CT Spine Imaging-A Systematic Review

In spinal oncology, integrating deep learning with computed tomography (CT) imaging has shown promise in enhancing diagnostic accuracy, treatment planning, and patient outcomes. This systematic review synthesizes evidence on artificial intelligence (AI) applications in CT imaging for spinal tumors. A PRISMA-guided search identified 33 studies: 12 (36.4%) focused on detecting spinal malignancies, 11 (33.3%) on classification, 6 (18.2%) on prognostication, 3 (9.1%) on treatment planning, and 1 (3.0%) on both detection and classification. Of the classification studies, 7 (21.2%) used machine learning to distinguish between benign and malignant lesions, 3 (9.1%) evaluated tumor stage or grade, and 2 (6.1%) employed radiomics for biomarker classification. Prognostic studies included three (9.1%) that predicted complications such as pathological fractures and three (9.1%) that predicted treatment outcomes. AI's potential for improving workflow efficiency, aiding decision-making, and reducing complications is discussed, along with its limitations in generalizability, interpretability, and clinical integration. Future directions for AI in spinal oncology are also explored. In conclusion, while AI technologies in CT imaging are promising, further research is necessary to validate their clinical effectiveness and optimize their integration into routine practice.

The Utility of Artificial Intelligence and Machine Learning in the Diagnosis of Takotsubo Cardiomyopathy: A Systematic Review

Introduction:

Takotsubo cardiomyopathy (TTC) is a cardiovascular disease caused by physical/psychological stressors with significant morbidity if left untreated. Because TTC often mimics acute myocardial infarction in the absence of obstructive coronary disease, the condition is often underdiagnosed in the population. Our aim was to discuss the role of artificial intelligence (AI) and machine learning (ML) in diagnosing TTC.

Methods:

We systematically searched electronic databases from inception until April 8, 2023, for studies on the utility of AI- or ML-based algorithms in diagnosing TTC compared with other cardiovascular diseases or healthy controls. We summarized major findings in a narrative fashion and tabulated relevant numerical parameters.

Results:

Five studies with a total of 920 patients were included. Four hundred and forty-seven were diagnosed with TTC via International Classification of Diseases codes or the Mayo Clinic diagnostic criteria, while there were 473 patients in the comparator group (29 of healthy controls, 429 of myocardial infarction, and 14 of acute myocarditis). Hypertension and smoking were the most common comorbidities in both cohorts, but there were no statistical differences between TTC and comparators. Two studies utilized deep-learning algorithms on transthoracic echocardiographic images, while the rest incorporated supervised ML on cardiac magnetic resonance imaging, 12-lead electrocardiographs, and brain magnetic resonance imaging. All studies found that AI-based algorithms can increase the diagnostic rate of TTC when compared to healthy controls or myocardial infarction patients. In three of these studies, AI-based algorithms had higher sensitivity and specificity compared to human readers.

Conclusion:

AI and ML algorithms can improve the diagnostic capacity of TTC and additionally reduce erroneous human error in differentiating from MI and healthy individuals.

Biomarkers vs Machines: The Race to Predict Acute Kidney Injury

BACKGROUND

Acute kidney injury (AKI) is a serious complication affecting up to 15% of hospitalized patients. Early diagnosis is critical to prevent irreversible kidney damage that could otherwise lead to significant morbidity and mortality. However, AKI is a clinically silent syndrome, and current detection primarily relies on measuring a rise in serum creatinine, an imperfect marker that can be slow to react to developing AKI. Over the past decade, new innovations have emerged in the form of biomarkers and artificial intelligence tools to aid in the early diagnosis and prediction of imminent AKI.

CONTENT

This review summarizes and critically evaluates the latest developments in AKI detection and prediction by emerging biomarkers and artificial intelligence. Main guidelines and studies discussed herein include those evaluating clinical utilitiy of alternate filtration markers such as cystatin C and structural injury markers such as neutrophil gelatinase-associated lipocalin and tissue inhibitor of metalloprotease 2 with insulin-like growth factor binding protein 7 and machine learning algorithms for the detection and prediction of AKI in adult and pediatric populations. Recommendations for clinical practices considering the adoption of these new tools are also provided.

SUMMARY

The race to detect AKI is heating up. Regulatory approval of select biomarkers for clinical use and the emergence of machine learning algorithms that can predict imminent AKI with high accuracy are all promising developments. But the race is far from being won. Future research focusing on clinical outcome studies that demonstrate the utility and validity of implementing these new tools into clinical practice is needed.

Machine Learning for prediction of violent behaviors in schizophrenia spectrum disorders: a systematic review

Background

Schizophrenia spectrum disorders (SSD) can be associated with an increased risk of violent behavior (VB), which can harm patients, others, and properties. Prediction of VB could help reduce the SSD burden on patients and healthcare systems. Some recent studies have used machine learning (ML) algorithms to identify SSD patients at risk of VB. In this article, we aimed to review studies that used ML to predict VB in SSD patients and discuss the most successful ML methods and predictors of VB.

Methods

We performed a systematic search in PubMed, Web of Sciences, Embase, and PsycINFO on September 30, 2023, to identify studies on the application of ML in predicting VB in SSD patients.

Results

We included 18 studies with data from 11,733 patients diagnosed with SSD. Different ML models demonstrated mixed performance with an area under the receiver operating characteristic curve of 0.56-0.95 and an accuracy of 50.27-90.67% in predicting violence among SSD patients. Our comparative analysis demonstrated a superior performance for the gradient boosting model, compared to other ML models in predicting VB among SSD patients. Various sociodemographic, clinical, metabolic, and neuroimaging features were associated with VB, with age and olanzapine equivalent dose at the time of discharge being the most frequently identified factors.

Conclusion

ML models demonstrated varied VB prediction performance in SSD patients, with gradient boosting outperforming. Further research is warranted for clinical applications of ML methods in this field.

Identifying diagnostic indicators for type 2 diabetes mellitus from physical examination using interpretable machine learning approach

Background

Identification of patients at risk for type 2 diabetes mellitus (T2DM) can not only prevent complications and reduce suffering but also ease the health care burden. While routine physical examination can provide useful information for diagnosis, manual exploration of routine physical examination records is not feasible due to the high prevalence of T2DM.

Objectives

We aim to build interpretable machine learning models for T2DM diagnosis and uncover important diagnostic indicators from physical examination, including age- and sex-related indicators.

Methods

In this study, we present three weighted diversity density (WDD)-based algorithms for T2DM screening that use physical examination indicators, the algorithms are highly transparent and interpretable, two of which are missing value tolerant algorithms.

Patients

Regarding the dataset, we collected 43 physical examination indicator data from 11,071 cases of T2DM patients and 126,622 healthy controls at the Affiliated Hospital of Southwest Medical University. After data processing, we used a data matrix containing 16004 EHRs and 43 clinical indicators for modelling.

Results

The indicators were ranked according to their model weights, and the top 25% of indicators were found to be directly or indirectly related to T2DM. We further investigated the clinical characteristics of different age and sex groups, and found that the algorithms can detect relevant indicators specific to these groups. The algorithms performed well in T2DM screening, with the highest area under the receiver operating characteristic curve (AUC) reaching 0.9185.

Conclusion

This work utilized the interpretable WDD-based algorithms to construct T2DM diagnostic models based on physical examination indicators. By modeling data grouped by age and sex, we identified several predictive markers related to age and sex, uncovering characteristic differences among various groups of T2DM patients.

Risk factors associated with post-acute sequelae of SARS-CoV-2: an N3C and NIH RECOVER study

BACKGROUND

More than one-third of individuals experience post-acute sequelae of SARS-CoV-2 infection (PASC, which includes long-COVID). The objective is to identify risk factors associated with PASC/long-COVID diagnosis.

METHODS

This was a retrospective case-control study including 31 health systems in the United States from the National COVID Cohort Collaborative (N3C). 8,325 individuals with PASC (defined by the presence of the International Classification of Diseases, version 10 code U09.9 or a long-COVID clinic visit) matched to 41,625 controls within the same health system and COVID index date within ± 45 days of the corresponding case's earliest COVID index date. Measurements of risk factors included demographics, comorbidities, treatment and acute characteristics related to COVID-19. Multivariable logistic regression, random forest, and XGBoost were used to determine the associations between risk factors and PASC.

RESULTS

Among 8,325 individuals with PASC, the majority were > 50 years of age (56.6%), female (62.8%), and non-Hispanic White (68.6%). In logistic regression, middle-age categories (40 to 69 years; OR ranging from 2.32 to 2.58), female sex (OR 1.4, 95% CI 1.33-1.48), hospitalization associated with COVID-19 (OR 3.8, 95% CI 3.05-4.73), long (8-30 days, OR 1.69, 95% CI 1.31-2.17) or extended hospital stay (30 + days, OR 3.38, 95% CI 2.45-4.67), receipt of mechanical ventilation (OR 1.44, 95% CI 1.18-1.74), and several comorbidities including depression (OR 1.50, 95% CI 1.40-1.60), chronic lung disease (OR 1.63, 95% CI 1.53-1.74), and obesity (OR 1.23, 95% CI 1.16-1.3) were associated with increased likelihood of PASC diagnosis or care at a long-COVID clinic. Characteristics associated with a lower likelihood of PASC diagnosis or care at a long-COVID clinic included younger age (18 to 29 years), male sex, non-Hispanic Black race, and comorbidities such as substance abuse, cardiomyopathy, psychosis, and dementia. More doctors per capita in the county of residence was associated with an increased likelihood of PASC diagnosis or care at a long-COVID clinic. Our findings were consistent in sensitivity analyses using a variety of analytic techniques and approaches to select controls.

CONCLUSIONS

This national study identified important risk factors for PASC diagnosis such as middle age, severe COVID-19 disease, and specific comorbidities. Further clinical and epidemiological research is needed to better understand underlying mechanisms and the potential role of vaccines and therapeutics in altering PASC course.

Behavioral and neurocognitive factors distinguishing post-traumatic stress comorbidity in substance use disorders

Significant trauma histories and post-traumatic stress disorder (PTSD) are common in persons with substance use disorders (SUD) and often associate with increased SUD severity and poorer response to SUD treatment. As such, this sub-population has been associated with unique risk factors and treatment needs. Understanding the distinct etiological profile of persons with co-occurring SUD and PTSD is therefore crucial for advancing our knowledge of underlying mechanisms and the development of precision treatments. To this end, we employed supervised machine learning algorithms to interrogate the responses of 160 participants with SUD on the multidimensional NIDA Phenotyping Assessment Battery. Significant PTSD symptomatology was correctly predicted in 75% of participants (sensitivity: 80%; specificity: 72.22%) using a classification-based model based on anxiety and depressive symptoms, perseverative thinking styles, and interoceptive awareness. A regression-based machine learning model also utilized similar predictors, but failed to accurately predict severity of PTSD symptoms. These data indicate that even in a population already characterized by elevated negative affect (individuals with SUD), especially severe negative affect was predictive of PTSD symptomatology. In a follow-up analysis of a subset of 102 participants who also completed neurocognitive tasks, comorbidity status was correctly predicted in 86.67% of participants (sensitivity: 91.67%; specificity: 66.67%) based on depressive symptoms and fear-related attentional bias. However, a regression-based analysis did not identify fear-related attentional bias as a splitting factor, but instead split and categorized the sample based on indices of aggression, metacognition, distress tolerance, and interoceptive awareness. These data indicate that within a population of individuals with SUD, aberrations in tolerating and regulating aversive internal experiences may also characterize those with significant trauma histories, akin to findings in persons with anxiety without SUD. The results also highlight the need for further research on PTSD-SUD comorbidity that includes additional comparison groups (i.e., persons with only PTSD), captures additional comorbid diagnoses that may influence the PTSD-SUD relationship, examines additional types of SUDs (e.g., alcohol use disorder), and differentiates between subtypes of PTSD.

Electronic health records identify timely trends in childhood mental health conditions

BACKGROUND

Electronic health records (EHRs) data provide an opportunity to collect patient information rapidly, efficiently and at scale. National collaborative research networks, such as PEDSnet, aggregate EHRs data across institutions, enabling rapid identification of pediatric disease cohorts and generating new knowledge for medical conditions. To date, aggregation of EHR data has had limited applications in advancing our understanding of mental health (MH) conditions, in part due to the limited research in clinical informatics, necessary for the translation of EHR data to child mental health research.

METHODS

In this cohort study, a comprehensive EHR-based typology was developed by an interdisciplinary team, with expertise in informatics and child and adolescent psychiatry, to query aggregated, standardized EHR data for the full spectrum of MH conditions (disorders/symptoms and exposure to adverse childhood experiences (ACEs), across 13 years (2010-2023), from 9 PEDSnet centers. Patients with and without MH disorders/symptoms (without ACEs), were compared by age, gender, race/ethnicity, insurance, and chronic physical conditions. Patients with ACEs alone were compared with those that also had MH disorders/symptoms. Prevalence estimates for patients with 1+ disorder/symptoms and for specific disorders/symptoms and exposure to ACEs were calculated, as well as risk for developing MH disorder/symptoms.

RESULTS

The EHR study data set included 7,852,081 patients < 21 years of age, of which 52.1% were male. Of this group, 1,552,726 (19.8%), without exposure to ACEs, had a lifetime MH disorders/symptoms, 56.5% being male. Annual prevalence estimates of MH disorders/symptoms (without exposure to ACEs) rose from 10.6% to 2010 to 15.1% in 2023, a 44% relative increase, peaking to 15.4% in 2019, prior to the Covid-19 pandemic. MH categories with the largest increases between 2010 and 2023 were exposure to ACEs (1.7, 95% CI 1.6-1.8), anxiety disorders (2.8, 95% CI 2.8-2.9), eating/feeding disorders (2.1, 95% CI 2.1-2.2), gender dysphoria/sexual dysfunction (43.6, 95% CI 35.8-53.0), and intentional self-harm/suicidality (3.3, 95% CI 3.2-3.5). White youths had the highest rates in most categories, except for disruptive behavior disorders, elimination disorders, psychotic disorders, and standalone symptoms which Black youths had higher rates. Median age of detection was 8.1 years (IQR 3.5-13.5) with all standalone symptoms recorded earlier than the corresponding MH disorder categories.

CONCLUSIONS

These results support EHRs' capability in capturing the full spectrum of MH disorders/symptoms and exposure to ACEs, identifying the proportion of patients and groups at risk, and detecting trends throughout a 13-year period that included the Covid-19 pandemic. Standardized EHR data, which capture MH conditions is critical for health systems to examine past and current trends for future surveillance. Our publicly available EHR-mental health typology codes can be used in other studies to further advance research in this area.

Seasonality of acute kidney injury phenotypes in England: an unsupervised machine learning classification study of electronic health records

BACKGROUND

Acute Kidney Injury (AKI) is a multifactorial condition which presents a substantial burden to healthcare systems. There is limited evidence on whether it is seasonal. We sought to investigate the seasonality of AKI hospitalisations in England and use unsupervised machine learning to explore clustering of underlying comorbidities, to gain insights for future intervention.

METHODS

We used Hospital Episodes Statistics linked to the Clinical Practice Research Datalink to describe the overall incidence of AKI admissions between 2015 and 2019 weekly by demographic and admission characteristics. We carried out dimension reduction on 850 diagnosis codes using multiple correspondence analysis and applied k-means clustering to classify patients. We phenotype each group based on the dominant characteristics and describe the seasonality of AKI admissions by these different phenotypes.

RESULTS

Between 2015 and 2019, weekly AKI admissions peaked in winter, with additional summer peaks related to periods of extreme heat. Winter seasonality was more evident in those diagnosed with AKI on admission. From the cluster classification we describe six phenotypes of people admitted to hospital with AKI. Among these, seasonality of AKI admissions was observed among people who we described as having a multimorbid phenotype, established risk factor phenotype, and general AKI phenotype.

CONCLUSION

We demonstrate winter seasonality of AKI admissions in England, particularly among those with AKI diagnosed on admission, suggestive of community triggers. Differences in seasonality between phenotypes suggests some groups may be more likely to develop AKI as a result of these factors. This may be driven by underlying comorbidity profiles or reflect differences in uptake of seasonal interventions such as vaccines.

Bottom-up and top-down paradigms of artificial intelligence research approaches to healthcare data science using growing real-world big data

OBJECTIVES

As the real-world electronic health record (EHR) data continue to grow exponentially, novel methodologies involving artificial intelligence (AI) are becoming increasingly applied to enable efficient data-driven learning and, ultimately, to advance healthcare. Our objective is to provide readers with an understanding of evolving computational methods and help in deciding on methods to pursue.

TARGET AUDIENCE

The sheer diversity of existing methods presents a challenge for health scientists who are beginning to apply computational methods to their research. Therefore, this tutorial is aimed at scientists working with EHR data who are early entrants into the field of applying AI methodologies.

SCOPE

This manuscript describes the diverse and growing AI research approaches in healthcare data science and categorizes them into 2 distinct paradigms, the bottom-up and top-down paradigms to provide health scientists venturing into artificial intelligent research with an understanding of the evolving computational methods and help in deciding on methods to pursue through the lens of real-world healthcare data.

Bias in artificial intelligence algorithms and recommendations for mitigation

The adoption of artificial intelligence (AI) algorithms is rapidly increasing in healthcare. Such algorithms may be shaped by various factors such as social determinants of health that can influence health outcomes. While AI algorithms have been proposed as a tool to expand the reach of quality healthcare to underserved communities and improve health equity, recent literature has raised concerns about the propagation of biases and healthcare disparities through implementation of these algorithms. Thus, it is critical to understand the sources of bias inherent in AI-based algorithms. This review aims to highlight the potential sources of bias within each step of developing AI algorithms in healthcare, starting from framing the problem, data collection, preprocessing, development, and validation, as well as their full implementation. For each of these steps, we also discuss strategies to mitigate the bias and disparities. A checklist was developed with recommendations for reducing bias during the development and implementation stages. It is important for developers and users of AI-based algorithms to keep these important considerations in mind to advance health equity for all populations.

Artificial intelligence and machine learning overview in pathology & laboratory medicine: A general review of data preprocessing and basic supervised concepts

Machine learning (ML) is becoming an integral aspect of several domains in medicine. Yet, most pathologists and laboratory professionals remain unfamiliar with such tools and are unprepared for their inevitable integration. To bridge this knowledge gap, we present an overview of key elements within this emerging data science discipline. First, we will cover general, well-established concepts within ML, such as data type concepts, data preprocessing methods, and ML study design. We will describe common supervised and unsupervised learning algorithms and their associated common machine learning terms (provided within a comprehensive glossary of terms that are discussed within this review). Overall, this review will offer a broad overview of the key concepts and algorithms in machine learning, with a focus on pathology and laboratory medicine. The objective is to provide an updated useful reference for those new to this field or those who require a refresher.

Development and internal validation of predictive models to assess risk of post-acute care facility discharge in adults undergoing multi-level instrumented fusions for lumbar degenerative pathology and spinal deformity

PURPOSE

To develop a model for factors predictive of Post-Acute Care Facility (PACF) discharge in adult patients undergoing elective multi-level (≥ 3 segments) lumbar/thoracolumbar spinal instrumented fusions.

METHODS

The State Inpatient Databases acquired from the Healthcare Cost and Utilization Project from 2005 to 2013 were queried for adult patients who underwent elective multi-level thoracolumbar fusions for spinal deformity. Outcome variables were classified as discharge to home or PACF. Predictive variables included demographic, pre-operative, and operative factors. Univariate and multivariate logistic regression analyses informed development of a logistic regression-based predictive model using seven selected variables. Performance metrics included area under the curve (AUC), sensitivity, and specificity.

RESULTS

Included for analysis were 8866 patients. The logistic model including significant variables from multivariate analysis yielded an AUC of 0.75. Stepwise logistic regression was used to simplify the model and assess number of variables needed to reach peak AUC, which included seven selected predictors (insurance, interspaces fused, gender, age, surgical region, CCI, and revision surgery) and had an AUC of 0.74. Model cut-off for predictive PACF discharge was 0.41, yielding a sensitivity of 75% and specificity of 59%.

CONCLUSIONS

The seven variables associated significantly with PACF discharge (age > 60, female gender, non-private insurance, primary operations, instrumented fusion involving 8+ interspaces, thoracolumbar region, and higher CCI scores) may aid in identification of adults at risk for discharge to a PACF following elective multi-level lumbar/thoracolumbar spinal fusions for spinal deformity. This may in turn inform discharge planning and expectation management.

Describing a complex primary health care population to support future decision support initiatives

Introduction

Developing decision support tools using data from a health care organization, to support care within that organization, is a promising paradigm to improve care delivery and population health. Descriptive epidemiology may be a valuable supplement to stakeholder input towards selection of potential initiatives and to inform methodological decisions throughout tool development. We additionally propose that to properly characterize complex populations in large-scale descriptive studies, both simple statistical and machine learning techniques can be useful.

Objective

To describe sociodemographic, clinical, and health care use characteristics of primary care clients served by the Alliance for Healthier Communities, which provides team-based primary health care through Community Health Centres (CHCs) across Ontario, Canada.

Methods

We used electronic health record data from adult ongoing primary care clients served by CHCs in 2009-2019. We performed traditional table-based summaries for each characteristic; and applied three unsupervised learning techniques to explore patterns of common condition co-occurrence, care provider teams, and care frequency.

Results

There were 221,047 eligible clients. Sociodemographics: We described 13 characteristics, stratified by CHC type and client multimorbidity status. Clinical characteristics: Eleven-year prevalence of 24 investigated conditions ranged from 1% (Hepatitis C) to 63% (chronic musculoskeletal problem) with non-uniform risk across the care history; multimorbidity was common (81%) with variable co-occurrence patterns. Health care use characteristics: Most care was provided by physician and nursing providers, with heterogeneous combinations of other provider types. A subset of clients had many issues addressed within single-visits and there was within- and between-client variability in care frequency. In addition to substantive findings, we discuss methodological considerations for future decision support initiatives.

Conclusions

We demonstrated the use of methods from statistics and machine learning, applied with an epidemiological lens, to provide an overview of a complex primary care population and lay a foundation for stakeholder engagement and decision support tool development.

Is primary health care ready for artificial intelligence? What do primary health care stakeholders say?

Background

Effective deployment of AI tools in primary health care requires the engagement of practitioners in the development and testing of these tools, and a match between the resulting AI tools and clinical/system needs in primary health care. To set the stage for these developments, we must gain a more in-depth understanding of the views of practitioners and decision-makers about the use of AI in primary health care. The objective of this study was to identify key issues regarding the use of AI tools in primary health care by exploring the views of primary health care and digital health stakeholders.

Methods

This study utilized a descriptive qualitative approach, including thematic data analysis. Fourteen in-depth interviews were conducted with primary health care and digital health stakeholders in Ontario. NVivo software was utilized in the coding of the interviews.

Results

Five main interconnected themes emerged: (1) Mismatch Between Envisioned Uses and Current Reality—denoting the importance of potential applications of AI in primary health care practice, with a recognition of the current reality characterized by a lack of available tools; (2) Mechanics of AI Don’t Matter: Just Another Tool in the Toolbox– reflecting an interest in what value AI tools could bring to practice, rather than concern with the mechanics of the AI tools themselves; (3) AI in Practice: A Double-Edged Sword—the possible benefits of AI use in primary health care contrasted with fundamental concern about the possible threats posed by AI in terms of clinical skills and capacity, mistakes, and loss of control; (4) The Non-Starters: A Guarded Stance Regarding AI Adoption in Primary Health Care—broader concerns centred on the ethical, legal, and social implications of AI use in primary health care; and (5) Necessary Elements: Facilitators of AI in Primary Health Care—elements required to support the uptake of AI tools, including co-creation, availability and use of high quality data, and the need for evaluation.

Conclusion

The use of AI in primary health care may have a positive impact, but many factors need to be considered regarding its implementation. This study may help to inform the development and deployment of AI tools in primary health care.

Risk Factors Associated with Post-Acute Sequelae of SARS-CoV-2 in an EHR Cohort: A National COVID Cohort Collaborative (N3C) Analysis as part of the NIH RECOVER program

Background

More than one-third of individuals experience post-acute sequelae of SARS-CoV-2 infection (PASC, which includes long-COVID).

Objective

To identify risk factors associated with PASC/long-COVID.

Design

Retrospective case-control study.

Setting

31 health systems in the United States from the National COVID Cohort Collaborative (N3C).

Patients

8,325 individuals with PASC (defined by the presence of the International Classification of Diseases, version 10 code U09.9 or a long-COVID clinic visit) matched to 41,625 controls within the same health system.

Measurements

Risk factors included demographics, comorbidities, and treatment and acute characteristics related to COVID-19. Multivariable logistic regression, random forest, and XGBoost were used to determine the associations between risk factors and PASC.

Results

Among 8,325 individuals with PASC, the majority were >50 years of age (56.6%), female (62.8%), and non-Hispanic White (68.6%). In logistic regression, middle-age categories (40 to 69 years; OR ranging from 2.32 to 2.58), female sex (OR 1.4, 95% CI 1.33-1.48), hospitalization associated with COVID-19 (OR 3.8, 95% CI 3.05-4.73), long (8-30 days, OR 1.69, 95% CI 1.31-2.17) or extended hospital stay (30+ days, OR 3.38, 95% CI 2.45-4.67), receipt of mechanical ventilation (OR 1.44, 95% CI 1.18-1.74), and several comorbidities including depression (OR 1.50, 95% CI 1.40-1.60), chronic lung disease (OR 1.63, 95% CI 1.53-1.74), and obesity (OR 1.23, 95% CI 1.16-1.3) were associated with increased likelihood of PASC diagnosis or care at a long-COVID clinic. Characteristics associated with a lower likelihood of PASC diagnosis or care at a long-COVID clinic included younger age (18 to 29 years), male sex, non-Hispanic Black race, and comorbidities such as substance abuse, cardiomyopathy, psychosis, and dementia. More doctors per capita in the county of residence was associated with an increased likelihood of PASC diagnosis or care at a long-COVID clinic. Our findings were consistent in sensitivity analyses using a variety of analytic techniques and approaches to select controls.

Conclusions

This national study identified important risk factors for PASC such as middle age, severe COVID-19 disease, and specific comorbidities. Further clinical and epidemiological research is needed to better understand underlying mechanisms and the potential role of vaccines and therapeutics in altering PASC course.

Psychotropic Medication Use Is Associated With Greater 1-Year Incidence of Dementia After COVID-19 Hospitalization

Background

COVID-19 has been associated with an increased risk of incident dementia (post-COVID dementia). Establishing additional risk markers may help identify at-risk individuals and guide clinical decision-making.

Methods

We investigated pre-COVID psychotropic medication use (exposure) and 1-year incidence of dementia (outcome) in 1,755 patients (≥65 years) hospitalized with COVID-19. Logistic regression models were used to examine the association, adjusting for demographic and clinical variables. For further confirmation, we applied the Least Absolute Shrinkage and Selection Operator (LASSO) regression and a machine learning (Random Forest) algorithm.

Results

One-year incidence rate of post-COVID dementia was 12.7% (N = 223). Pre-COVID psychotropic medications (OR = 2.7, 95% CI: 1.8-4.0, P < 0.001) and delirium (OR = 3.0, 95% CI: 1.9-4.6, P < 0.001) were significantly associated with greater 1-year incidence of post-COVID dementia. The association between psychotropic medications and incident dementia remained robust when the analysis was restricted to the 423 patients with at least one documented neurological or psychiatric diagnosis at the time of COVID-19 admission (OR = 3.09, 95% CI: 1.5-6.6, P = 0.002). Across different drug classes, antipsychotics (OR = 2.8, 95% CI: 1.7-4.4, P < 0.001) and mood stabilizers/anticonvulsants (OR = 2.4, 95% CI: 1.39-4.02, P = 0.001) displayed the greatest association with post-COVID dementia. The association of psychotropic medication with dementia was further confirmed with Random Forest and LASSO analysis.

Conclusion

Confirming prior studies we observed a high dementia incidence in older patients after COVID-19 hospitalization. Pre-COVID psychotropic medications were associated with higher risk of incident dementia. Psychotropic medications may be risk markers that signify neuropsychiatric symptoms during prodromal dementia, and not mutually exclusive, contribute to post-COVID dementia.

The Critical Role of Health Information Technology in the Safe Integration of Behavioral Health and Primary Care to Improve Patient Care

Integrated behavioral health and primary care is a set of related approaches to enhance care delivery, safety, and quality through closer collaboration and coordination among clinicians and care organizations in these two fields. Technology (e.g., electronic health records (EHR), telehealth, clinical decision support, and standards-based interoperability) can enable integration and improve delivery of care. Reinforcing these opportunities, virtual care and telehealth have dramatically changed care availability and delivery, especially during the COVID-19 pandemic. There are different levels of integration, ranging from minimal to full collaboration. The Partnership for Health IT Patient Safety and the HIMSS Electronic Health Record Association (EHRA) formed a workgroup to examine using information technology to facilitate integration. Taking a three-pronged approach focused on (1) screening for behavioral health issues, (2) clinician documentation, and (3) sharing data among clinicians, patients, and authorized parties, the workgroup developed action-oriented recommendations and strategies for safe use of health IT for stakeholders and policymakers seeking to advance efforts to integrate behavioral health with primary care.