Prevalence of Tendon Rupture and Tendinopathies Among Patients with Atherosclerotic Cardiovascular Disease Derived From United States Administrative Claims Data.

INTRODUCTION: The prevalence of tendon rupture and tendinopathies (TRT) has not been determined in a large population of patients with atherosclerotic cardiovascular disease (ASCVD). We investigated TRT prevalence among patients with ASCVD and in the general population, using data from the Symphony Health Integrated Dataverse, a large US medical and pharmacy claims database. METHODS: This retrospective, observational study included patients aged ≥ 19 years from the claims database during the identification period (January 2019 to December 2020) and 12 months of continuous enrollment. The primary outcome was evidence of TRT in the 12 months following the index date (first ASCVD diagnosis in the ASCVD cohort; first claim in the claims database in the overall population). Diagnostic codes (ICD-10 and/or CPT) were used to define ASCVD and TRT diagnosis. RESULTS: The ASCVD cohort and overall population included 5,589,273 and 61,715,843 patients, respectively. In the ASCVD cohort, use of medications with a potential or known association with TRT was identified in 67.9% (statins), 17.7% (corticosteroids), and 16.7% (fluoroquinolones) of patients. Bempedoic acid use was reported in 1556 (< 0.1%) patients. TRT prevalence during 12-month follow-up was 3.4% (ASCVD cohort) and 1.9% (overall population). Among patients with ASCVD, 83.5% experienced TRT in only one region of the body. Factors most associated with TRT in the ASCVD cohort were increasing age, most notably in those aged 45-|64 years (odds ratio [OR] 2.19; 95% confidence interval [CI] 2.07-2.32), obesity (OR 1.51; 95% CI 1.50-1.53), and rheumatoid arthritis (OR 1.47; 95% CI 1.45-1.79). Use of statins or bempedoic acid was not associated with increased TRT risk. CONCLUSION: Patients with ASCVD may have greater risk of TRT than the general population, which may be driven by an increased prevalence of comorbidities and use of medications with a potential or known association with TRT.

Patients with atherosclerosis, the main cause of heart attacks, strokes, and peripheral vascular disease, typically require several drugs to control the disease. Some of the drugs used to treat atherosclerosis have been linked to a higher occurrence of tendon tears (or ruptures) or swelling/inflammation of the tendons (tendinopathies). However, there may be other factors present in these patients that increase the risk of tendon injuries that are not related to these drugs. This study used the medical records of over 5.5 million patients with atherosclerosis and over 63 million patients reflecting the general population in the United States to determine the prevalence of tendon injury. Additionally, the researchers looked at other factors that might be related to a higher risk of tendon injury in each group. Over a 12-month period, tendon injuries occurred in 3.4% of patients with atherosclerosis and 1.8% of patients in the general population. In patients with atherosclerosis, factors such as being obese, older (45­64 years), or having rheumatoid arthritis were also linked to an increased risk of tendon injuries. There was no association seen between statin or bempedoic acid use and tendon injuries. These results may help healthcare providers to determine the underlying risk of tendon injuries and guide treatment of this patient population.

Burden of illness in patients with pulmonary hypertension due to interstitial lung disease: a real-world analysis.

BACKGROUND: Pulmonary hypertension due to interstitial lung disease (PH-ILD) is associated with high rates of respiratory failure and death. Healthcare resource utilization (HCRU) and cost data are needed to characterize PH-ILD disease burden. METHODS: A retrospective cohort analysis of the Truven Health MarketScan® Commercial Claims and Encounters Database and Medicare Supplemental Database between June 2015 to June 2019 was conducted. Patients with ILD were identified and indexed based on their first claim with a PH diagnosis. Patients were required to be 18 years of age on the index date and continuously enrolled for 12-months pre- and post-index. Patients were excluded for having a PH diagnosis prior to ILD diagnosis or the presence of other non-ILD, PH-associated conditions. Treatment patterns, HCRU, and healthcare costs were compared between the 12 months pre- versus 12 months post-index date. RESULTS: In total, 122 patients with PH-ILD were included (mean [SD] age, 63.7 [16.6] years; female, 64.8%). The same medication classes were most frequently used both pre- and post-index (corticosteroids: pre-index 43.4%, post-index 53.5%; calcium channel blockers: 25.4%, 36.9%; oxygen: 12.3%, 25.4%). All-cause hospitalizations increased 2-fold, with 29.5% of patients hospitalized pre-index vs. 59.0% post-index (P < 0.0001). Intensive care unit (ICU) utilization increased from 6.6 to 17.2% (P = 0.0433). Mean inpatient visits increased from 0.5 (SD, 0.9) to 1.1 (1.3) (P < 0.0001); length of stay (days) increased from 5.4 (5.9) to 7.5 (11.6) (P < 0.0001); bed days from 2.5 (6.6) to 8.0 (16.3) (P < 0.0001); ICU days from 3.8 (2.3) to 7.0 (13.2) (P = 0.0362); and outpatient visits from 24.5 (16.8) to 32.9 (21.8) (P < 0.0001). Mean (SD) total all-cause healthcare costs increased from $43,201 ($98,604) pre-index to $108,387 ($190,673) post-index (P < 0.0001); this was largely driven by hospitalizations (which increased from a mean [SD] of $13,133 [$28,752] to $63,218 [$75,639] [P < 0.0001]) and outpatient costs ($16,150 [$75,639] to $25,604 [$93,964] [P < 0.0001]). CONCLUSION: PH-ILD contributes to a high HCRU and cost burden. Timely identification, management, and treatment are needed to mitigate the clinical and economic consequences of PH-ILD development and progression.

Spinal fusion surgery use among adults with low back pain enrolled in a digital musculoskeletal program: an observational study.

OBJECTIVES: To compare 12-month spinal fusion surgery rates in the setting of low back pain among digital musculoskeletal (MSK) program participants versus a comparison cohort who only received usual care. STUDY DESIGN: Retrospective cohort study with propensity score matched comparison cohort using commercial medical claims data representing over 100 million commercially insured lives. METHODS: All study subjects experienced low back pain between January 2020 and December 2021. Digital MSK participants enrolled in the digital MSK low back program between January 2020 and December 2021. Non-participants had low back pain related physical therapy (PT) between January 2020 and December 2021. Digital MSK participants were matched to non-participants with similar demographics, comorbidities and baseline MSK-related medical care use. Spinal fusion surgery rates at 12 months post participation were compared. RESULTS: Compared to non-participants, digital MSK participants had lower rates of spinal fusion surgery in the post-period (0.7% versus 1.6%; p < 0.001). Additionally, in the augmented inverse probability weighting (AIPW) model, digital MSK participants were found to have decreased odds of undergoing spinal fusion surgery (adjusted odds ratio: 0.64, 95% CI: 0.51-0.81). CONCLUSIONS: This study provides evidence that participation in a digital MSK program is associated with a lower rate of spinal fusion surgery.

South Tyrol Healthcare Company: A Look at a Peculiar Model of Claims Management in Italy and Analysis of Its Last 11 Years.

BACKGROUND: Alternative Disputes Resolution (ADR) systems are becoming increasingly important tools in recent years for the management and resolution of health responsibility cases, but their dissemination and efficiency are still poorly described. The purpose of this paper is to present an ADR system in the autonomous province of Bolzano: the Conciliation Commission. METHODS: systematic collection of all claims of the South Tyrol Sanitary Service that were dealt with in the Conciliation Commission from 1 January 2012 to 31 December 2022. RESULTS: closing times of the applications received turn out to be less than a year in 63.8% of the cases, with a number of cases managed rather stably in the time, even if minimal if compared to the number of complaints received to the South Tirol Health Service. Only 5.3% of the application continued the legal process before a civil court. CONCLUSIONS: the Conciliation Commission of South Tirol appears to be an excellent instrument for the resolution of disputes in the healthcare field, with rapid resolution times and little to zero costs for the healthcare company, a public health institution. Despite its effectiveness, it seems to be a tool that is still little-known in South Tyrol.

Mental illness and sleep disorders among women with gynecological problems.

This retrospective cohort study identifies differences between rates of selected mental illnesses and sleep disorders according to eight gynecological problems. Analyses utilize medical claims data for adult employees of a large corporation during 2017-2021. Women with a gynecological problem (most notably pain, endometriosis, pelvic inflammation and bleeding) are significantly more likely to experience mental illness. Several gynecological problems are also significantly associated with sleep disorders. Women with a gynecological problem (vs. none) are 50% more likely to have a mental health problem and 44% more likely to have a sleep disorder after adjusting for age, marital status, dependent children and year. The largest differences between higher (%) mental illness and sleep disorders appear for hyperplasia (6% vs. 45%), cancer (11% vs. 68%), pelvic inflammation (46% vs. 79%) and pain (79% vs. 43%), respectively. On the other hand, the rate of having one or more gynecological problems ranges from 7.1% for women with no mental illness or sleep disorder to 20.6% for women with schizophrenia. Understanding the association between gynecological problems, mental illness and sleep disorders can help clinicians more effectively identify and treat patients.

Does reversible postpartum contraception reduce the risk of pregnancy condition recurrence? A longitudinal claims-based study from Maine.

PURPOSE: To estimate the effect of reversible postpartum contraception use on the risk of recurrent pregnancy condition in the subsequent pregnancy and if this effect was mediated through lengthening the interpregnancy interval (IPI). METHODS: We used data from the Maine Health Data Organization's Maine All Payer Claims dataset. Our study population was Maine women with a livebirth index pregnancy between 2007 and 2019 that was followed by a subsequent pregnancy starting within 60 months of index pregnancy delivery. We examined recurrence of three pregnancy conditions, separately, in groups that were not mutually exclusive: prenatal depression, hypertensive disorders of pregnancy (HDP), and gestational diabetes (GDM). Effective reversible postpartum contraception use was defined as any intrauterine device, implant, or moderately effective method (pills, patch, ring, injectable) initiated within 60 days of delivery. Short IPI was defined as ≤ 12 months. We used log-binomial regression models to estimate risk ratios and 95 % confidence intervals, adjusting for potential confounders. RESULTS: Approximately 41 % (11,448/28,056) of women initiated reversible contraception within 60 days of delivery, the prevalence of short IPI was 26 %, and the risk of pregnancy condition recurrence ranged from 38 % for HDP to 55 % for prenatal depression. Reversible contraception initiation within 60 days of delivery was not associated with recurrence of the pregnancy condition in the subsequent pregnancy (aRR ranged from 0.97 to 1.00); however, it was associated with lower risk of short IPI (aRR ranged from 0.67 to 0.74). CONCLUSION(S): Although initiation of postpartum reversible contraception within 60 days of delivery lengthens the IPI, our findings suggest that it does not reduce the risk of prenatal depression, HDP, or GDM recurrence. This indicates a missed opportunity for providing evidence-based healthcare and health interventions in the intrapartum period to reduce the risk of recurrence.

Healthcare utilization patterns of individuals with depression after national policy to increase the mental health workforce in primary care: a data linkage study.

BACKGROUND: The deployment of the mental health nurse, an additional healthcare provider for individuals in need of mental healthcare in Dutch general practices, was expected to substitute treatments from general practitioners and providers in basic and specialized mental healthcare (psychologists, psychotherapists, psychiatrists, etc.). The goal of this study was to investigate the extent to which the degree of mental health nurse deployment in general practices is associated with healthcare utilization patterns of individuals with depression. METHODS: We combined national health insurers' claims data with electronic health records from general practices. Healthcare utilization patterns of individuals with depression between 2014 and 2019 (N = 31,873) were analysed. The changes in the proportion of individuals treated after depression onset were assessed in association with the degree of mental health nurse deployment in general practices. RESULTS: The proportion of individuals with depression treated by the GP, in basic and specialized mental healthcare was lower in individuals in practices with high mental health nurse deployment. While the association between mental health nurse deployment and consultation in basic mental healthcare was smaller for individuals who depleted their deductibles, the association was still significant. Treatment volume of general practitioners was also lower in practices with higher levels of mental health nurse deployment. CONCLUSION: Individuals receiving care at a general practice with a higher degree of mental health nurse deployment have lower odds of being treated by mental healthcare providers in other healthcare settings. More research is needed to evaluate to what extent substitution of care from specialized mental healthcare towards general practices might be associated with waiting times for specialized mental healthcare.

Drivers of Geographic Patterns in Outpatient Antibiotic Prescribing in the United States.

In a retrospective, ecological analysis of US medical claims, visit rates explained more of the geographic variation in outpatient antibiotic prescribing rates than per-visit prescribing. Efforts to reduce antibiotic use may benefit from addressing the factors that drive higher rates of outpatient visits, in addition to continued focus on stewardship.

Spatiotemporal Trends in Group A Streptococcal Pharyngitis in the United States.

BACKGROUND: Group A Streptococcus (GAS) causes an estimated 5.2 million outpatient visits for pharyngitis annually in the United States, with incidence peaking in winter, but the annual spatiotemporal pattern of GAS pharyngitis across the United States is poorly characterized. METHODS: We used outpatient claims data from individuals with private medical insurance between 2010 and 2018 to quantify GAS pharyngitis visit rates across U.S. census regions, subregions, and states. We evaluated seasonal and age-based patterns of geographic spread and the association between school start dates and the summertime upward inflection in GAS visits. RESULTS: The South had the most visits per person (yearly average, 39.11 visits per 1000 people; 95% confidence interval, 36.21-42.01) and the West had the fewest (yearly average, 17.63 visits per 1000 people; 95% confidence interval, 16.76-18.49). Visits increased earliest in the South and in school-age children. Differences in visits between the South and other regions were most pronounced in the late summer through early winter. Visits peaked earliest in central southern states, in December to January, and latest on the coasts, in March. The onset of the rise in GAS pharyngitis visits correlated with, but preceded, average school start times. CONCLUSIONS: The burden and timing of GAS pharyngitis varied across the continental United States, with the South experiencing the highest overall rates and earliest onset and peak in outpatient visits. Understanding the drivers of these regional differences in GAS pharyngitis will help in identifying and targeting prevention measures.

Evaluation of deep learning-based depression detection using medical claims data.

Human accuracy in diagnosing psychiatric disorders is still low. Even though digitizing health care leads to more and more data, the successful adoption of AI-based digital decision support (DDSS) is rare. One reason is that AI algorithms are often not evaluated based on large, real-world data. This research shows the potential of using deep learning on the medical claims data of 812,853 people between 2018 and 2022, with 26,973,943 ICD-10-coded diseases, to predict depression (F32 and F33 ICD-10 codes). The dataset used represents almost the entire adult population of Estonia. Based on these data, to show the critical importance of the underlying temporal properties of the data for the detection of depression, we evaluate the performance of non-sequential models (LR, FNN), sequential models (LSTM, CNN-LSTM) and the sequential model with a decay factor (GRU-Δt, GRU-decay). Furthermore, since explainability is necessary for the medical domain, we combine a self-attention model with the GRU decay and evaluate its performance. We named this combination Att-GRU-decay. After extensive empirical experimentation, our model (Att-GRU-decay), with an AUC score of 0.990, an AUPRC score of 0.974, a specificity of 0.999 and a sensitivity of 0.944, proved to be the most accurate. The results of our novel Att-GRU-decay model outperform the current state of the art, demonstrating the potential usefulness of deep learning algorithms for DDSS development. We further expand this by describing a possible application scenario of the proposed algorithm for depression screening in a general practitioner (GP) setting-not only to decrease healthcare costs, but also to improve the quality of care and ultimately decrease people's suffering.

Comparing the Strength of Associations Between Male Genital Problems and Mental Illnesses and Sleep Disorders.

This study compares the rate of selected types of mental illnesses (stress, anxiety, depression) and sleep disorders (insomnia, sleep apnea) according to the status of eight male genital problems. Analyses utilize medical claims data for male employees aged 18 to 64 years of a large corporation, 2017 to 2021. Approximately 1,076 (7.3%) men per year have one or more genital problems. The most common being benign prostatic hyperplasia (BPH; 3.8%) and then erectile dysfunction (ED; 1.7%). For BPH patients, the rate experiencing stress, anxiety, depression, or a combination of these is 0.96%, 6.2%, 5.3%, and 5.1%, respectively. Corresponding rates for ED are 1.5%, 7.2%, 5.9%, and 7.5%. For BPH patients, the rate experiencing insomnia, sleep apnea, or both is 3.1%, 22.7%, and 2.0%, respectively. Corresponding rates for ED are 1.2%, 20.6%, and 2.2%. Male genital problems positively associate with having one or more mental illnesses (stress, anxiety, depression), except for hydrocele, with ED and penis disorder having the strongest associations. Male genital problems also positively associate with having insomnia and/or sleep apnea, except for infertility and orchitis, with BPH and ED having the strongest associations. The positive associations involving BPH and ED with mental illnesses are each more pronounced in the younger age group (18-49 vs. 50-64). Similar results are seen in the models involving sleep disorders. Thus, comorbid male genital problems, mental illnesses, and sleep disorders exist, with the strength of associations unique to the male genital problem and sometimes modified by age.

Characteristics of pediatric patients claimed with acute upper respiratory infection during otorhinolaryngology consultations: A descriptive study of a large Japanese medical claims database.

This study aimed to clarify other diseases claimed simultaneously with acute upper respiratory infection (URI), antibiotic prescriptions, and examinations associated with infectious diseases in pediatric patients with acute URI insurance claims at otorhinolaryngology outpatient visits. Pediatric patients who visited an otolaryngology department between 2019 and 2021 and were definitively diagnosed with URI were selected using a large Japanese medical claims database. Patient backgrounds, antibiotic use, and examinations were descriptively evaluated. In total, 8010 patients were included in the analysis. The median number (interquartile range) of diseases claimed in the same month as acute URI was 4 (3-6). Only 519 (6.5 %) patients were claimed as acute URI alone. Regardless of the prescription of antibiotics, the most commonly redundantly claimed disease in these patients was allergic rhinitis, followed by acute bronchitis, acute sinusitis, and earwax impaction. The frequently prescribed antibiotics were third-generation cephalosporins, macrolides, and penicillins with extended-spectrum, including amoxicillin which was recommended by the Japanese manual; the proportion of patients with examinations was low (2.9-21.7 %). Among patients with acute URI, diagnoses requiring antibiotics were also claimed; therefore, when evaluating acute URI using the Japanese medical claims database, care must be taken in patient selection. Moreover, the implementation rate of examinations necessary for diagnosis was low, so there is an urgent need to develop an environment where examinations can be conducted in outpatient settings.

Early Prediction of COVID-19 Associated Hospitalization at the Time of CDC Contact Tracing using Machine Learning: Towards Pandemic Preparedness.

Objective: To develop and validate machine learning models for predicting COVID-19 related hospitalization as early as CDC contact tracing using integrated CDC contact tracing and South Carolina medical claims data. Methods: Using the dataset (n=82,073, 1/1/2018 - 3/1/2020), we identified 3,305 patients with COVID-19 and were captured by contact tracing. We developed and validated machine learning models (i.e., support vector machine, random forest, XGboost), followed by multi-level validations and pilot statewide implementation. Results: Using 10-cross validation, random forest outperformed other models (F1=0.872 for general hospitalization and 0.763 for COVID-19 related hospitalization), followed by XGBoost (F1=0.845 and 0.682) and support vector machine (F1=0.845 and 0.644). We identified new self-reported symptoms from contact tracing (e.g., fatigue, congestion, headache, loss of taste) that are highly predictive of hospitalization. Conclusions: Our study demonstrated the feasibility of identifying individuals at risk of hospitalization at the time of contact tracing for early intervention and prevention. Policy implications: Our findings demonstrate existing promise for leveraging CDC contact tracing for establishing a cost-effective statewide surveillance and generalizability for nationwide adoption for enhancing pandemic preparedness in the US.

Risk of Cancer After Diagnosis of Cardiovascular Disease.

Background: Cardiovascular disease (CVD) and cancer share several risk factors. Although preclinical models show that various types of CVD can accelerate cancer progression, clinical studies have not determined the impact of atherosclerosis on cancer risk. Objectives: The objective of this study was to determine whether CVD, especially atherosclerotic CVD, is independently associated with incident cancer. Methods: Using IBM MarketScan claims data from over 130 million individuals, 27 million cancer-free subjects with a minimum of 36 months of follow-up data were identified. Individuals were stratified by presence or absence of CVD, time-varying analysis with multivariable adjustment for cardiovascular risk factors was performed, and cumulative risk of cancer was calculated. Additional analyses were performed according to CVD type (atherosclerotic vs nonatherosclerotic) and cancer subtype. Results: Among 27,195,088 individuals, those with CVD were 13% more likely to develop cancer than those without CVD (HR: 1.13; 95% CI: 1.12-1.13). Results were more pronounced for individuals with atherosclerotic CVD (aCVD), who had a higher risk of cancer than those without CVD (HR: 1.20; 95% CI: 1.19-1.21). aCVD also conferred a higher risk of cancer compared with those with nonatherosclerotic CVD (HR: 1.11; 95% CI: 1.11-1.12). Cancer subtype analyses showed specific associations of aCVD with several malignancies, including lung, bladder, liver, colon, and other hematologic cancers. Conclusions: Individuals with CVD have an increased risk of developing cancer compared with those without CVD. This association may be driven in part by the relationship of atherosclerosis with specific cancer subtypes, which persists after controlling for conventional risk factors.

Pneumoconiosis incidence and prevalence among US Medicare beneficiaries, 1999-2019.

BACKGROUND: Pneumoconiosis is a group of occupational lung diseases caused by dust and fiber exposure. This study analyzes Medicare claims to estimate the burden of pneumoconiosis among fee-for-service (FFS; Medicare Parts A and B) Medicare beneficiaries during 1999-2019 in the United States. METHODS: Claim and enrollment information from 81 million continuously enrolled FFS Medicare beneficiaries were analyzed. Beneficiaries with any pneumoconiosis and cause-specific pneumoconiosis (e.g., asbestosis, silicosis) were identified using three case definitions (broad, intermediate, and narrow) with varying diagnostic criteria based on claim International Classification of Diseases, Clinical Modification (ICD-CM) diagnosis codes and Healthcare Common Procedure Coding System codes. Results are presented as ranges of values for the three case definitions. RESULTS: The 21-year prevalence range for any pneumoconiosis was 345,383-677,361 (412-833 per 100,000 beneficiaries) using the three case definitions. The highest prevalence was among those ≥75 years of age, males, Whites, and North American Natives. Most claims (70.0%-72.5%) included an ICD-CM diagnosis code for asbestosis. The broad pneumoconiosis prevalence rate increased significantly (p < 0.001) during 2002-2009 by 3%-10% annually and declined significantly by 3%-5% annually starting in 2009. The average annual broad incidence rate declined significantly by 7% annually during 2009-2019. CONCLUSIONS: Despite the decline in rate for any pneumoconiosis among Medicare beneficiaries, which is primarily attributed to a decline in asbestosis, pneumoconiosis is prevalent among FFS Medicare beneficiaries.

Prediction of Future Medical Costs by Modifiable Measures of Health.

Introduction: Strategies to mitigate rising health-care costs are a priority for patients, employers, and health insurers. Yet gaps currently exist in whether health risk assessment can forecast medical claims costs. This study examined the ability of a health quotient (HQ) based on modifiable risk factors, age, sex, and chronic conditions to predict future medical claims spending. Methods: The study included 18,695 employees and adult dependents who participated in health assessments and were enrolled in an employer-sponsored health plan. Linear mixed effect models stratified by chronic conditions and adjusted for age and sex were utilized to evaluate the relationship between the health quotient (score of 0-100) and future medical claims spending. Results: Lower baseline health quotient was associated with higher medical claims cost over 2 years of follow up. For participants with chronic condition(s), costs were $3628 higher for those with a low health quotient (<73; N = 2673) compared to those with high health quotient (>85; N = 1045), after adjustment for age and sex (P value = 0.004). Each one-unit increase in health quotient was associated with a decrease of $154 (95% CI: 87.4, 220.3) in average yearly medical claims costs during follow up. Discussion: This study used a large employee population with 2 years of follow-up data, which provides insights that are applicable to other large employers. Results of this analysis contribute to our ability to predict health-care costs using modifiable aspects of health, objective laboratory testing and chronic condition status.

Susceptibility to Colorectal Cancer Based on HSD17B4 rs721673 and rs721675 Polymorphisms and Alcohol Intake among Taiwan Biobank Participants: A Retrospective Case Control Study Using the Nationwide Claims Data.

Colorectal cancer (CRC) is a major public health issue, and there are limited studies on the association between 17ß-hydroxysteroid dehydrogenase type 4 (HSD17B4) polymorphism and CRC. We used two national databases from Taiwan to examine whether HSD17B4 rs721673, rs721675, and alcohol intake were independently and interactively correlated with CRC development. We linked the Taiwan Biobank (TWB) participants' health and lifestyle information and genotypic data from 2012 to 2018 to the National Health Insurance Database (NHIRD) to confirm their medical records. We performed a genome-wide association study (GWAS) using data from 145 new incident CRC cases and matched 1316 healthy, non-CRC individuals. We calculated the odds ratios (OR) and 95% confidence intervals (CI) for CRC based on multiple logistic regression analyses. HSD17B4 rs721673 and rs721675 on chromosome 5 were significantly and positively correlated with CRC (rs721673 A > G, aOR = 2.62, p = 2.90 × 10-8; rs721675 A > T, aOR = 2.61, p = 1.01 × 10-6). Within the high-risk genotypes, significantly higher ORs were observed among the alcohol intake group. Our results demonstrated that the rs721673 and rs721675 risk genotypes of HSD17B4 might increase the risk of CRC development in Taiwanese adults, especially those with alcohol consumption habits.

Comparison of Machine Learning Algorithms for Predicting Hospital Readmissions and Worsening Heart Failure Events in Patients With Heart Failure With Reduced Ejection Fraction: Modeling Study.

BACKGROUND: Heart failure (HF) is highly prevalent in the United States. Approximately one-third to one-half of HF cases are categorized as HF with reduced ejection fraction (HFrEF). Patients with HFrEF are at risk of worsening HF, have a high risk of adverse outcomes, and experience higher health care use and costs. Therefore, it is crucial to identify patients with HFrEF who are at high risk of subsequent events after HF hospitalization. OBJECTIVE: Machine learning (ML) has been used to predict HF-related outcomes. The objective of this study was to compare different ML prediction models and feature construction methods to predict 30-, 90-, and 365-day hospital readmissions and worsening HF events (WHFEs). METHODS: We used the Veradigm PINNACLE outpatient registry linked to Symphony Health's Integrated Dataverse data from July 1, 2013, to September 30, 2017. Adults with a confirmed diagnosis of HFrEF and HF-related hospitalization were included. WHFEs were defined as HF-related hospitalizations or outpatient intravenous diuretic use within 1 year of the first HF hospitalization. We used different approaches to construct ML features from clinical codes, including frequencies of clinical classification software (CCS) categories, Bidirectional Encoder Representations From Transformers (BERT) trained with CCS sequences (BERT + CCS), BERT trained on raw clinical codes (BERT + raw), and prespecified features based on clinical knowledge. A multilayer perceptron neural network, extreme gradient boosting (XGBoost), random forest, and logistic regression prediction models were applied and compared. RESULTS: A total of 30,687 adult patients with HFrEF were included in the analysis; 11.41% (3184/27,917) of adults experienced a hospital readmission within 30 days of their first HF hospitalization, and nearly half (9231/21,562, 42.81%) of the patients experienced at least 1 WHFE within 1 year after HF hospitalization. The prediction models and feature combinations with the best area under the receiver operating characteristic curve (AUC) for each outcome were XGBoost with CCS frequency (AUC=0.595) for 30-day readmission, random forest with CCS frequency (AUC=0.630) for 90-day readmission, XGBoost with CCS frequency (AUC=0.649) for 365-day readmission, and XGBoost with CCS frequency (AUC=0.640) for WHFEs. Our ML models could discriminate between readmission and WHFE among patients with HFrEF. Our model performance was mediocre, especially for the 30-day readmission events, most likely owing to limitations of the data, including an imbalance between positive and negative cases and high missing rates of many clinical variables and outcome definitions. CONCLUSIONS: We predicted readmissions and WHFEs after HF hospitalizations in patients with HFrEF. Features identified by data-driven approaches may be comparable with those identified by clinical domain knowledge. Future work may be warranted to validate and improve the models using more longitudinal electronic health records that are complete, are comprehensive, and have a longer follow-up time.

Long Short-Term Memory-Based Deep Learning Models for Screening Parkinson's Disease Using Sequential Diagnostic Codes.

BACKGROUND AND PURPOSE: It is challenging to detect Parkinson's disease (PD) in its early stages, which has prompted researchers to develop techniques based on machine learning methods for detecting PD. However, previous studies did not fully incorporate the slow progression of PD over a long period of time nor consider that its symptoms occur in a time-sequential manner. Contributing to the literature on PD, which has relied heavily on cross-sectional data, this study aimed to develop a method for detecting PD early that can process time-series information using the long short-term memory (LSTM) algorithm. METHODS: We sampled 926 patients with PD and 9,260 subjects without PD using medical-claims data. The LSTM algorithm was tested using diagnostic histories, which contained the diagnostic codes and their respective time information. We compared the prediction power of the 12-month diagnostic codes under two different settings over the 4 years prior to the first PD diagnosis. RESULTS: The model that was trained using the most-recent 12-month diagnostic codes had the best performance, with an accuracy of 94.25%, a sensitivity of 82.91%, and a specificity of 95.26%. The other three models (12-month codes from 2, 3, and 4 years prior) were found to have comparable performances, with accuracies of 92.27%, 91.86%, and 91.81%, respectively. The areas under the curve from our data settings ranged from 0.839 to 0.923. CONCLUSIONS: We explored the possibility that PD specialists could benefit from our proposed machine learning method as an early detection method for PD.