International Classification Of Disease Codes Research Articles

Atherosclerotic cardiovascular disease following a diagnosis of idiopathic inflammatory myopathy: analysis from a retrospective cohort in the TriNetX registry.

Idiopathic inflammatory myopathies (IIM) confer an increased risk of morbidity from atherosclerotic cardiovascular disease (ASCVD). While ASCVD risk has been studied in other countries, these results may not be applicable to patients with dermatomyositis (DM) and polymyositis (PM) in the United States. This retrospective analysis of a cohort of patients identified by ICD code from TriNetX investigated the incidence of ASCVD after International Classification of Disease (ICD) codes of DM, PM, dermatopolymyositis (DPM) or juvenile dermatomyositis (JDM). Patients were identified by entry of two ICD codes separated by at least 6months, according to their first diagnosis code; ASCVD was defined as first ICD code for myocardial infarction, ischemic stroke, transient ischemic attack, or peripheral arterial disease. Cox proportional hazards regression modeled time from first IIM ICD code to ASCVD event. A total of 35,554 patients were identified with the mean age at first IIM code of 54 and 26.1% were male. The most common comorbidity for all groups except JDM was hyperlipidemia (39.9%) though 79.2% of patients were on no cholesterol lowering medication. ASCVD occurred in 30.4% of patients with PM, 24.3% of patients with DM and 0.9% of patients with JDM. Patients with PM had a median time to event of 9.7years (95% Confidence interval (CI) 9.1, 10.7) and 14.3years (95% CI 12.6, 14.8) for DM. This study demonstrates that ASCVD is a comorbidity occurring after a median of 12.5years (95% CI 11.9, 13.6) in patients with IIM. ASCVD appears to be a long-term complication for IIM patients occurring in nearly a quarter of US patients without prior ASCVD with at least two ICD codes for IIM, with a median time to event of 12.5years. There appears to be a practice gap in the recognition and treatment of hyperlipidemia in these patients. Key Points • Hyperlipidemia was a common comorbidity identified in patients with IIM though most patients were not on cholesterol lowering medication. • Development of ASCVD appears to be a long-term complication for patients with IIM in the United States.

Prescription patterns of home rescue benzodiazepines for febrile seizures

ObjectiveTo describe the prescription patterns of home, non-intravenous rescue benzodiazepines (non-IV-rBZDs) for febrile seizures and the factors associated with their prescription. MethodsRetrospective descriptive study using the MarketScan Commercial Database, a large database of employer-sponsored privately insured patients in the United States. We used data from January 1st 2006 to December 31st 2022. We studied patients with febrile seizures as the main code for the healthcare encounter (identified with International Classification of Diseases codes) with age from 6 months to 5 years of age and with at least 1 month of follow-up. ResultsThere were a total of 82,835 patients [median (p25-p75) age 1.0 (1.0–2.0) years, 56.7 % males] with at least one febrile seizure, of whom 9,737 (11.8 %) filled at least one non-IV-rBZD prescription. Among the 9,737 patients who filled at least one prescription, the median (p25-p75) time from first febrile seizure to non-IV-rBZD prescription was 27 (2–186) days. Among the factors known at the time of the first febrile seizure, complex febrile seizure (OR: 3.51, 95 % CI: 3.24–3.79), and an initial inpatient hospitalization for febrile seizure (OR: 3.53, 95 % CI: 3.29–3.79) were the factors most strongly associated with filling a non-IV-rBZD prescription. In contrast, sex, rural patient's residence, and salary employment (versus other employment class) were not independently associated with filling a non-IV-rBZD prescription. Among the factors known at the end of follow-up, complex febrile seizures, type of initial encounter, and an eventual diagnosis of epilepsy were major independent factors associated with filling a non-IV-rBZD prescription. ConclusionOnly approximately 12 % of children with a febrile seizure filled a prescription for a home non-IV-rBZD. The major factors independently associated with prescription were complex febrile seizure, hospital admission, recurrent febrile seizures, and an eventual diagnosis of epilepsy.

The Right ICD Code, Right Now: A Call to Action for Pragmatic Language Disorders After Right Hemisphere Stroke.

Diagnosis of language impairments after stroke is important to optimizing stroke outcomes. After right hemisphere brain damage (RHD), apragmatism can impact the comprehension and production of pragmatic language. However, despite decades of empirical evidence, there is no International Classification of Diseases (ICD) code for RHD pragmatic language impairments. The absence of an ICD code has far reaching ramifications that impact patient outcomes, including reduced clinical and public awareness, limited curricular content, and underdiagnosis. This viewpoint justifies the need to appropriately classify the pragmatic language symptomology after RHD with an ICD code. An ICD code can positively influence health care practitioner knowledge, education, and practice while informing public health considerations vital to epidemiological analyses.

Enhanced ICD-10 code assignment of clinical texts: A summarization-based approach

BackgroundAssigning International Classification of Diseases (ICD) codes to clinical texts is a common and crucial practice in patient classification, hospital management, and further statistics analysis. Current auto-coding methods mainly transfer this task to a multi-label classification problem. Such solutions are suffering from high-dimensional mapping space and excessive redundant information in long clinical texts. To alleviate such a situation, we introduce text summarization methods to the ICD coding regime and apply text matching to select ICD codes. MethodWe focus on the tenth revision of the ICD (ICD-10) coding and design a novel summarization-based approach (SuM) with an end-to-end strategy to efficiently assign ICD-10 code to clinical texts. In this approach, a knowledge-guided pointer network is purposed to distill and summarize key information in clinical texts precisely. Then a matching model with matching-aggregation architecture follows to align the summary result with code, tuning the one-vs-all scenario to one-vs-one matching so that the large-label-space obstacle laid in classification approaches would be avoided. ResultThe 12,788 ICD-10 coded discharge summaries from a Chinese hospital were collected to evaluate the proposed approach. Compared with existing methods, the purposed model achieves the greatest coding results with Micro AUC of 0.9548, MRR@10 of 0.7977, Precision@10 of 0.0944, and Recall@10 of 0.9439 for the TOP-50 Dataset. Results on the FULL-Dataset remain consistent. Also, the proposed knowledge encoder and applied end-to-end strategy are proven to facilitate the whole model to gain efficacy in selecting the most suitable code. ConclusionThe proposed automatic ICD-10 code assignment approach via text summarization can effectively capture critical messages in long clinical texts and improve the performance of ICD-10 coding of clinical texts.

Bidirectional association between type 2 diabetes and irritable bowel syndrome: A large-scale prospective cohort study.

To examine the bidirectional association between type 2 diabetes (T2D) and irritable bowel syndrome (IBS) in a large prospective population cohort. Participants free of IBS at baseline in the UK Biobank were included in the analysis of T2D and incident IBS (cohort 1), with 11 140 T2D patients and 413 979 non-T2D patients. Similarly, those free of T2D at baseline were included in the analysis of IBS and incident T2D (cohort 2), with 21 944 IBS patients and 413 979 non-IBS patients. Diagnoses of T2D and IBS were based on International Classification of Disease-10 codes. The Cox proportional hazards model was used to estimate adjusted hazard ratios (HRs). In cohort 1, 8984 IBS cases were identified during a median 14.5-year follow-up. Compared with non-T2D, T2D patients had a 39.0% increased risk of incident IBS (HR = 1.39, 95% confidence interval [CI]: 1.23-1.56, P < .001), with a higher IBS risk in those with higher fasting blood glucose levels (HR = 1.43, 95% CI: 1.19-1.72, P < .001) or longer T2D duration (HR = 1.47, 95% CI: 1.23-1.74, P < .001). In cohort 2, 29 563 incident T2D cases were identified. IBS patients had an 18.0% higher risk of developing T2D versus non-IBS patients (HR = 1.18, 95% CI: 1.12-1.24, P < .001). A similar excess T2D risk was observed in IBS patients with a duration of either less than 10 years, or of 10 years or longer. Further sensitivity analysis and subgroup analysis indicated consistent findings. T2D and IBS exhibit a bidirectional association, with an increased risk of co-morbidity. Awareness of this association may improve the prevention and management of both diseases.

Association between primary hypothyroidism and metabolic dysfunction-associated steatotic liver disease: an updated meta-analysis

ObjectiveEpidemiological studies have reported an association between primary hypothyroidism and metabolic dysfunction-associated steatotic liver disease (MASLD). However, the magnitude of the risk and whether this risk changes with the severity...

Improved Proteinuria May Attenuate the Risk of Atrial Fibrillation: A Nationwide Population-Based Cohort Study.

Background/Objectives: Proteinuria is documented as a risk factor for atrial fibrillation (AF) and can manifest in either reversible or continued forms. Our objective was to examine the relationship between the change in status for proteinuria and the risk of AF in a longitudinal cohort study on the general population nationwide. Methods: We included participants (n = 1,708,103) who underwent repetitive health examinations. The presence of proteinuria was determined by dipstick urinalysis results. The outcome was the occurrence of AF (International Classification of Diseases-10 code: I48). Results: All included participants, 1,666,111 (97.5%), 17,659 (1.0%), 19,696 (1.2%), and 4637 (0.3%), were categorized into groups of proteinuria-free, improved, progressed, and persistent, respectively. During a median follow-up of 14.5 years, 41,190 (2.4%) cases of AF occurred. In the multivariable analysis, the risk of AF was increased as the initial severity was more severe in the proteinuria-improved and proteinuria-persistent groups (p for trend < 0.001). In a further pairwise comparison, the proteinuria-improved group had a relatively lower risk of AF compared to the proteinuria-persistent group (HR: 0.751, 95% CI: 0.652-0.865, p < 0.001). Conclusions: Our study showed that the risk of AF can change according to alterations in proteinuria status. Notably, recovering from proteinuria can also be considered a modifiable risk factor for AF.

ICDXML: enhancing ICD coding with probabilistic label trees and dynamic semantic representations

Accurately assigning standardized diagnosis and procedure codes from clinical text is crucial for healthcare applications. However, this remains challenging due to the complexity of medical language. This paper proposes a novel model that incorporates extreme multi-label classification tasks to enhance International Classification of Diseases (ICD) coding. The model utilizes deformable convolutional neural networks to fuse representations from hidden layer outputs of pre-trained language models and external medical knowledge embeddings fused using a multimodal approach to provide rich semantic encodings for each code. A probabilistic label tree is constructed based on the hierarchical structure existing in ICD labels to incorporate ontological relationships between ICD codes and enable structured output prediction. Experiments on medical code prediction on the MIMIC-III database demonstrate competitive performance, highlighting the benefits of this technique for robust clinical code assignment.

Social Determinants of Health Phenotypes and Cardiometabolic Condition Prevalence Among Patients in a Large Academic Health System: Latent Class Analysis.

Adverse social determinants of health (SDoH) have been associated with cardiometabolic disease; however, disparities in cardiometabolic outcomes are rarely the result of a single risk factor. This study aimed to identify and characterize SDoH phenotypes based on patient-reported and neighborhood-level data from the institutional electronic medical record and evaluate the prevalence of diabetes, obesity, and other cardiometabolic diseases by phenotype status. Patient-reported SDoH were collected (January to December 2020) and neighborhood-level social vulnerability, neighborhood socioeconomic status, and rurality were linked via census tract to geocoded patient addresses. Diabetes status was coded in the electronic medical record using International Classification of Diseases codes; obesity was defined using measured BMI ≥30 kg/m2. Latent class analysis was used to identify clusters of SDoH (eg, phenotypes); we then examined differences in the prevalence of cardiometabolic conditions based on phenotype status using prevalence ratios (PRs). Complete data were available for analysis for 2380 patients (mean age 53, SD 16 years; n=1405, 59% female; n=1198, 50% non-White). Roughly 8% (n=179) reported housing insecurity, 30% (n=710) reported resource needs (food, health care, or utilities), and 49% (n=1158) lived in a high-vulnerability census tract. We identified 3 patient SDoH phenotypes: (1) high social risk, defined largely by self-reported SDoH (n=217, 9%); (2) adverse neighborhood SDoH (n=1353, 56%), defined largely by adverse neighborhood-level measures; and (3) low social risk (n=810, 34%), defined as low individual- and neighborhood-level risks. Patients with an adverse neighborhood SDoH phenotype had higher prevalence of diagnosed type 2 diabetes (PR 1.19, 95% CI 1.06-1.33), hypertension (PR 1.14, 95% CI 1.02-1.27), peripheral vascular disease (PR 1.46, 95% CI 1.09-1.97), and heart failure (PR 1.46, 95% CI 1.20-1.79). Patients with the adverse neighborhood SDoH phenotype had higher prevalence of poor cardiometabolic conditions compared to phenotypes determined by individual-level characteristics, suggesting that neighborhood environment plays a role, even if individual measures of socioeconomic status are not suboptimal.

Cannabis and Craniotomy for Glioblastoma: Impact on Complications and Health Care Utilization

Despite advances in treatment of glioblastomas (GBMs), the median survival remains 14-16 months. In the United States, 52.5 million people ≥12 years of age used cannabis in 2021. We aim to elucidate differences in complications after craniotomy for resection of GBM between users and nonusers of cannabis. Merative MarketScan Research Data (2008-2019) (includes >265 million patients) were used to extract adults (≥18 years of age) with GBM diagnosis (International Classification of Diseases-9 code 191.x and International Classification of Diseases-10 code C71.x) who had a craniotomy (Current Procedure Terminology code 61510) from inpatient admission data. The inverse probability treatment weighted analysis balanced the groups of cannabis abuse disorder (CAD) and no CAD in terms of age, gender, insurance coverage, comorbidities, and prior 12-month opioid dependence. Individuals with CAD were younger (median, 37 vs. 51 years; P < 0.0001). There was a lower percentage of women (19% vs. 45%; P < 0.0001). In the CAD group, opioid abuse pattern for ≥12 months was higher (16% vs. 5%; P=0.001) and the rate of complications was higher (32% vs. 15%; P= 0.001) during index hospital stay. At 6 months postdischarge, neurologic complications were higher among the CAD group (27% vs. 8%; P < 0.001). At 1 year postdischarge, patients with CAD sought feweroutpatient services (P= 0.012). More neurologic complications were seen in the CAD group (31% vs. 12%; P< 0.001). This retrospective population-based study sounds a higher rate of neurologic complications among patients using cannabis who also had a newly diagnosed GBM. This suggests the lack of a protective effect from use of cannabis in patients with primary malignant brain tumors.

Prognosis of Patients With Ischemic Stroke With Prior Anticoagulant Therapy: Direct Oral Anticoagulants Versus Warfarin.

Direct oral anticoagulants (DOACs) have been the drug of choice for preventing ischemic stroke in patients with atrial fibrillation since 2014. In previous studies, the stroke risk while taking warfarin was 2 per 100 patient-years and 1.5% per year while taking DOACs. We hypothesized that even if ischemic stroke occurred during anticoagulation therapy with DOACs, the prognosis was likely to be better than that with warfarin. Data from 2002 to 2019, sourced from a nationwide claims database, were used to identify atrial fibrillation patients using International Classification of Diseases codes. Patients who experienced an ischemic stroke during anticoagulation were categorized by the drugs used (warfarin, dabigatran, apixaban, rivaroxaban, and edoxaban). The primary outcome was mortality within 3 months and 1 year after the ischemic stroke. Among the 9578 patients with ischemic stroke during anticoagulation, 3343 received warfarin, and 6235 received DOACs (965 dabigatran, 2320 apixaban, 1702 rivaroxaban, 1248 edoxaban). The DOACs group demonstrated lower risks of 3-month (adjusted hazard ratio [HR], 0.550, [95% CI, 0.473-0.639]; P<0.0001) and 1-year mortality (adjusted HR, 0.596 [95% CI, 0.536-0.663]; P<0.0001) than the warfarin group. Apixaban and edoxaban within the DOAC group exhibited particularly reduced 1-year mortality risk compared with other DOACs (P<0.0001). Our study confirmed that DOACs have a better prognosis than warfarin after ischemic stroke. The apixaban and edoxaban groups had a lower risk of death after ischemic stroke than the other DOAC groups.

The risk and risk factors of chikungunya virus infection and rheumatological sequelae in a cohort of U.S. Military Health System beneficiaries: Implications for the vaccine era.

Understanding the risk of chikungunya virus (CHIKV) infection and rheumatic sequelae across populations, including travelers and the military, is critical. We leveraged healthcare delivery data of over 9 million U.S. Military Health System (MHS) beneficiaries to identify cases, and sampled controls, to estimate the risk of post-CHIKV rheumatic sequelae. MHS beneficiary CHIKV infections diagnosed 2014-2018 were identified from the Disease Reporting System internet, TRICARE Encounter Data Non-Institutional, and Comprehensive Ambulatory/Professional Encounter Record systems. Non-CHIKV controls were matched (1:4) by age, gender, beneficiary status, and encounter date. The frequency of comorbidities and incident rheumatic diagnoses through December 2018 were derived from International Classification of Diseases codes and compared between cases and controls. Poisson regression models estimated the association of CHIKV infection with rheumatic sequelae. We further performed a nested case-control study to estimate risk factors for post-CHIKV sequelae in those with prior CHIKV. 195 CHIKV cases were diagnosed between July 2014 and December 2018. The median age was 42 years, and 43.6% were active duty. 63/195 (32.3%) of CHIKV cases had an incident rheumatic diagnosis, including arthralgia, polyarthritis, polymyalgia rheumatica, and/or rheumatoid arthritis, compared to 156/780 (20.0%) of controls (p < 0.001). CHIKV infection remained associated with rheumatic sequelae (aRR = 1.579, p = 0.008) after adjusting for prior rheumatic disease and demography. Those with rheumatic CHIKV sequelae had a median 7 healthcare encounters (IQR 3-15). Among CHIKV infections, we found no association between post-CHIKV rheumatic sequelae and demography, service characteristics, or comorbidities. CHIKV infection is uncommon but associated with rheumatic sequelae among MHS beneficiaries, with substantial healthcare requirements in a proportion of cases with such sequelae. No demographic, clinical, or occupational variables were associated with post-CHIKV rheumatic sequelae, suggesting that prediction of these complications is challenging in MHS beneficiaries. These findings are important context for future CHIKV vaccine decision making in this and other populations.

Mild traumatic brain injury caused by workplace violence in a US workers’ compensation system

ObjectivesApproximately 81% of traumatic brain injury cases are considered to be mild (mTBI), but few studies have reviewed mTBI caused by workplace violence (WPV). This study aimed to (1) determine...

A New Start with HAART: Evaluating Breast Reconstruction in the Era of Highly Active Antiretroviral Therapy.

As HIV-positive individuals utilizing highly active antiretroviral therapy live longer, the burden of breast cancer increases in the population. Breast reconstruction is an integral aspect of surgical treatment for many patients after a breast cancer diagnosis, prompting this examination of the characteristics and outcomes of breast reconstruction in this growing patient population. Using Merative MarketScan Research Databases, a large multipayer database, HIV-positive adult patients who underwent autologous or implant-based breast reconstruction between 2007 and 2021 were identified using International Classification of Disease codes and Common Procedural Terminology codes. In both HIV-positive and -negative cohorts, patient demographics, procedure-related complications, and postoperative revisions were recorded. Shapiro-Wilk, chi-square, Wilcoxon-Mann-Whitney, and multivariable logistic regression tests were used for statistical analysis. Of 173,421 patients who underwent breast reconstruction, 1816 had an HIV diagnosis. HIV-positive patients were younger (P < 0.001), underwent surgery more recently (P < 0.001), more often underwent immediate breast reconstruction (P < 0.001), and had higher comorbidity levels (P < 0.001). There was a regional variation in which the patient cohorts underwent breast reconstruction. There was no significant difference in overall complication rates between patient groups, but HIV-negative patients more often underwent revision procedures (P = 0.009). When compared to their HIV-negative counterparts, breast reconstruction can be considered safe and efficacious in patients living with HIV. HIV-positive patients are a growing demographic who seek breast reconstruction, and surgeons must continue to further understand the unique implications of breast reconstruction in this population.

The Prevalence of Persistent Toe Walking in Children With and Without Autism Spectrum Disorder and the Odds of Subsequent Surgery

Persistent toe walking is associated with autism spectrum disorder. The true prevalence of persistent toe walking and odds of progression to surgery in children with and without autism remains unclear. This retrospective descriptive study identified patients ages 3 to 17 years who were enrolled in our healthcare system over a 2-year period. Using international classification of disease codes, we identified all children with autism and persistent toe walking, and excluded children with conditions that may independently cause toe walking. Data on Achilles lengthening surgeries, sex, race and body mass index was gathered. The toe walking prevalence amongst children with and without autism was calculated. Multivariable logistic regression analysis controlling for sex, race and body mass index was used to determine independent risk factors for persistent toe walking and surgery. Of the children who met inclusion criteria (N = 284,925), 4622 (1.6%) had persistent toe walking. Prevalence of persistent toe walking was higher amongst children with autism (6.3% vs 1.5%, p < .01), as were odds of persistent toe walking (OR 4.13, 95% CI 3.74 to 4.56, p < .01). Males and White patients had higher odds of persistent toe walking compared to females and patients of any other race, respectively (p < .01 for all). Although children with autism and toe walking had higher rates of surgery than their counterparts without autism (4.3% vs 2.6%, p = .04), this difference was not significant after controlling for sex, race and BMI (OR 1.59, 95% CI 0.95 to 2.69, p > .05).

National 22-year epilepsy surgery landscape shows increasing open and minimally invasive pediatric epilepsy surgery.

A surgical "treatment gap" in pediatric epilepsy persists despite the demonstrated safety and effectiveness of surgery. For this reason, the national surgical landscape should be investigated such that an updated assessment may more appropriately guide health care efforts. In our retrospective cross-sectional observational study, the National Inpatient Sample (NIS) database was queried for individuals 0 to <18 years of age who had an International Classification of Diseases (ICD) code for drug-resistant epilepsy (DRE). This cohort was then split into a medical group and a surgical group. The former was defined by ICD codes for -DRE without an accompanying surgical code, and the latter was defined by DRE and one of the following epilepsy surgeries: any open surgery; laser interstitial thermal therapy (LITT); vagus nerve stimulation; or responsive neurostimulation (RNS) from 1998 to 2020. Demographic variables of age, gender, race, insurance type, hospital charge, and hospital characteristics were analyzed between surgical options. Continuous variables were analyzed with weight-adjusted quantile regression analysis, and categorical variables were analyzed by weight-adjusted counts with percentages and compared with weight-adjusted chi-square test results. These data indicate an increase in epilepsy surgeries over a 22-year period, primarily due to a statistically significant increase in open surgery and a non-significant increase in minimally invasive techniques, such as LITT and RNS. There are significant differences in age, race, gender, insurance type, median household income, Elixhauser index, hospital setting, and size between the medical and surgical groups, as well as the procedure performed. An increase in open surgery and minimally invasive surgeries (LITT and RNS) account for the overall rise in pediatric epilepsy surgery over the last 22 years. A positive inflection point in open surgery is seen in 2005. Socioeconomic disparities exist between medical and surgical groups. Patient and hospital sociodemographics show significant differences between the procedure performed. Further efforts are required to close the surgical "treatment gap."

Stigmatizing and affirming provider language in medical records on hospitalized patients with opioid use disorder.

Stigma within the healthcare environment limits access to treatment for opioid use disorder (OUD), even as OUD results in significant morbidity and mortality. Language in clinical documentation affects patient experience and future care through the transmission of stigma or positive regard. With the passage of the 21st Century Cures Act, patients have full access to their medical records online. The objective of our study was to understand providers' use of stigmatizing and affirming language in the electronic health record (EHR) for OUD patients with long hospital stays. We selected patients with a first-time referral to the Duke University Hospital OUD consult service who met diagnostic criteria for OUD with a hospital stay ≥28 days from July 2019 to February 2022. Two reviewers independently evaluated each admission and discharge note for stigmatizing or affirming language and the group met weekly to validate coding reliability. Forty-eight patients (96 notes) met our inclusion criteria. We identified 434 occurrences of stigmatizing and 47 occurrences of affirming language. One-third (34%) of stigmatizing language appeared in system-generated fields (drop-down categories and diagnosis codes) and the rest was authored by providers. Stigmatizing language was present in both provider- and system-generated language and was nine times more frequent than affirming language in the medical records of hospitalized patients with OUD. While provider education may reduce stigmatizing language, institutional level changes to the EHR and International Classification of Disease codes are necessary to decrease stigmatizing language within medical records.

Association Between Greater Social Vulnerability and Delayed Glaucoma Surgery

PurposeTiming of surgical intervention in glaucoma is crucial to preserving sight. While ocular characteristics that increase surgical risk are known, the impact of neighborhood-level social risk factors such as the Social Vulnerability Index (SVI) and Area Deprivation Index (ADI) on time to glaucoma surgery is unknown. The objective of this study was to evaluate the association between SVI or ADI scores and the timing of glaucoma surgical intervention. DesignRetrospective cohort study. MethodsAdult subjects with open-angle glaucoma were identified from the Bascom Palmer Glaucoma Repository using International Classification of Disease-10 codes. Subject demographics, ocular characteristics, and standard automated perimetry (SAP) data were extracted. Geocoded data were obtained using subject residences and American Community Survey data. Univariable and multivariable time-to-event survival analyses using accelerated failure time (AFT) models were completed to evaluate whether geocoded SVI and ADI scores accelerated or delayed time to glaucoma surgery from initial glaucoma diagnosis in the electronic health record. ResultsA total of 10,553 eyes from 6,934 subjects were evaluated, of which 637 eyes (6.0%) from 568 subjects (8.2%) underwent glaucoma surgery. Mean age was 68.3±13.5 years, with 57.9% female, 21.5% Black, and 34.5% Hispanic subjects. Mean follow-up time was 5.0±2.1 years, with time to surgery of 3.2±1.9 years. Multivariable AFT models demonstrated that higher mean intraocular pressure (time ratio (TR) 0.27 per 5 mmHg higher; 95% CI: 0.23-0.31, p<0.001), faster SAP rate of progression (TR 0.74 per 0.5 dB/year faster; 95% CI: 0.69-0.78, p<0.001), moderate (TR 0.69; 95% CI: 0.56-0.85, p<0.001) or severe baseline severity (TR 0.39; 95% CI: 0.32-0.47, p<0.001), and thinner central corneal thickness (TR 0.85 per 50µm thinner; 95% CI: 0.77-0.95, p=0.003) all accelerated time to surgery. In contrast, overall SVI delayed surgery (TR 1.11 per 25% increase; 95% CI: 1.03-1.20, p=0.006). Specifically, SVI Themes 1 (TR 1.08; 95% CI: 1.01-1.17, p=0.037) and 4 (TR 1.11; 95% CI: 1.03-1.19, p=0.006) were significant. Patients from the most deprived neighborhoods (highest national ADI quartile) had a 68% increase in time to surgery compared to the least deprived quartile (TR 1.68; 95% CI: 1.20-2.36, p=0.002). ConclusionsResidence in areas with higher SVI or ADI scores was associated with delayed glaucoma surgery after controlling for demographic and ocular parameters. Awareness of such disparities can guide initiatives aimed at achieving parity in health outcomes.

Unsupervised Feature Selection to Identify Important ICD-10 and ATC Codes for Machine Learning on a Cohort of Patients With Coronary Heart Disease: Retrospective Study.

The application of machine learning in health care often necessitates the use of hierarchical codes such as the International Classification of Diseases (ICD) and Anatomical Therapeutic Chemical (ATC) systems. These codes classify diseases and medications, respectively, thereby forming extensive data dimensions. Unsupervised feature selection tackles the "curse of dimensionality" and helps to improve the accuracy and performance of supervised learning models by reducing the number of irrelevant or redundant features and avoiding overfitting. Techniques for unsupervised feature selection, such as filter, wrapper, and embedded methods, are implemented to select the most important features with the most intrinsic information. However, they face challenges due to the sheer volume of ICD and ATC codes and the hierarchical structures of these systems. The objective of this study was to compare several unsupervised feature selection methods for ICD and ATC code databases of patients with coronary artery disease in different aspects of performance and complexity and select the best set of features representing these patients. We compared several unsupervised feature selection methods for 2 ICD and 1 ATC code databases of 51,506 patients with coronary artery disease in Alberta, Canada. Specifically, we used the Laplacian score, unsupervised feature selection for multicluster data, autoencoder-inspired unsupervised feature selection, principal feature analysis, and concrete autoencoders with and without ICD or ATC tree weight adjustment to select the 100 best features from over 9000 ICD and 2000 ATC codes. We assessed the selected features based on their ability to reconstruct the initial feature space and predict 90-day mortality following discharge. We also compared the complexity of the selected features by mean code level in the ICD or ATC tree and the interpretability of the features in the mortality prediction task using Shapley analysis. In feature space reconstruction and mortality prediction, the concrete autoencoder-based methods outperformed other techniques. Particularly, a weight-adjusted concrete autoencoder variant demonstrated improved reconstruction accuracy and significant predictive performance enhancement, confirmed by DeLong and McNemar tests (P<.05). Concrete autoencoders preferred more general codes, and they consistently reconstructed all features accurately. Additionally, features selected by weight-adjusted concrete autoencoders yielded higher Shapley values in mortality prediction than most alternatives. This study scrutinized 5 feature selection methods in ICD and ATC code data sets in an unsupervised context. Our findings underscore the superiority of the concrete autoencoder method in selecting salient features that represent the entire data set, offering a potential asset for subsequent machine learning research. We also present a novel weight adjustment approach for the concrete autoencoders specifically tailored for ICD and ATC code data sets to enhance the generalizability and interpretability of the selected features.

Effect of liver abnormalities on mortality in Fontan patients: a systematic review and meta-analysis

BackgroundFontan-associated liver disease (FALD) is one of the most common complications following Fontan procedure, but the impact of FALD on survival outcomes remains controversial. The aim of this systematic review and meta-analysis was to examine and quantify the influence of liver disease on the survival of Fontan patients.MethodsThe Preferred Reporting Items for Systematic reviews and Meta-Analyses guidelines were followed, and relevant human studies published from inception up to 12 August 2022 were searched. Stata (version 17.0) was applied to perform the meta-analysis, using random effects (Mantel-Haenszel) models. The I2 statistic was used to assess the heterogeneity. Subgroup analysis and meta-regression were employed to explore the potential sources of heterogeneity and sensitivity analysis was performed to determine the potential influence of each study on the overall pooled results.ResultsA total of 312 records were initially identified and 8 studies involving 2,466 patients were selected for inclusion. Results revealed a significant association between the severity of liver disease following Fontan procedure and mortality, which was confirmed by sensitivity analysis and subgroup analysis assessing post-HT mortality. Meta-regression showed that diagnostic methods for liver disease may be a source of heterogeneity. After removal of the FALD patients identified by international classification of disease codes, heterogeneity was markedly reduced, and the positive association between all-cause mortality and the severity of liver disease became significant.ConclusionsThis meta-analysis showed the severity of liver disease following the Fontan procedure has a significant association with mortality. Lifelong follow-up is necessary and imaging examinations are recommended for routine surveillance of liver disease. Among patients with failing Fontan and advanced liver disease, combined heart-liver transplantation may provide additional survival benefits.