Citations

Abstract Objectives To comprehensively evaluate the validity of International Classification of Diseases, Tenth Revision, Clinical Modification (ICD-10-CM) codes for both prevalent diagnoses and less common diseases, and to assess the performance of a large language model (LLM)-based system in validating these codes. Materials and Methods This retrospective study analyzed hospital admissions from the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. We developed a validated LLM-based system using GPT-4o, refined through iterative prompt engineering, to assess ICD-10-CM code validity. We measured the positive predictive value (PPV) of ICD-10-CM codes, PPV of principal and secondary diagnoses, and the performance of an LLM-based system in code validation. Results Among 865 079 assigned codes, the PPV was 84.6% (95% CI, 84.5%-84.6%). Principal diagnoses had a PPV of 93.9% (95% CI, 93.7%-94.1%), while secondary diagnoses had a PPV of 83.8% (95% CI, 83.7%-83.9%). The LLM system demonstrated high performance in validating ICD codes, achieving 93.6% accuracy, 95.4% sensitivity, and 85.2% specificity. Among correctly assigned secondary diagnoses, the majority (67.9%) represented historical or baseline conditions, while 32.1% reflected active conditions that deviated from baseline status; 22.3% of these emerged after hospital admission. PPV decreases with later diagnosis positions, with the largest decline occurring between principal and secondary diagnoses. Discussion and Conclusion In this large-scale evaluation, ICD-10-CM codes exhibited generally high accuracy, though variability existed by position and condition type. A validated LLM system performed comparably to physician review and offers a scalable means to improve coding accuracy. These findings support the potential for integrating LLM-based auditing into routine workflows to strengthen the quality of administrative and research data.

ABSTRACT Objective To comprehensively evaluate the validity of ICD-10-CM codes for both prevalent diagnoses and less common diseases, and to assess the performance of a large language model (LLM)-based system in validating these codes. Materials and Methods This retrospective study analyzed hospital admissions from the Medical Information Mart for Intensive Care (MIMIC-IV) database. We developed a validated LLM-based system using GPT-4o, refined through iterative prompt engineering, to assess ICD-10-CM code validity. We measured the PPV of ICD-10-CM codes, PPV of principal and secondary diagnoses, and the performance of an LLM-based system in code validation. Results Among 865,079 assigned codes, the PPV was 84.6% (95% CI, 84.5%-84.6%). Principal diagnoses had a PPV of 93.9% (95% CI, 93.7%-94.1%), while secondary diagnoses had a PPV of 83.8% (95% CI, 83.7%-83.9%). The LLM system demonstrated high performance in validating ICD codes, achieving 93.6% accuracy, 95.4% sensitivity and 85.2% specificity. Among correctly assigned secondary diagnoses, the majority (67.9%) represented historical or baseline conditions, while 32.1% reflected active conditions that deviated from baseline status; 22.3% of these emerged after hospital admission. PPV decreases with later diagnosis positions, with the largest decline occurring between principal and secondary diagnoses. Discussion and Conclusion In this large-scale evaluation, ICD-10-CM codes exhibited generally high accuracy, though variability existed by position and condition type. A validated LLM system performed comparably to physician review and offers a scalable means to improve coding accuracy. These findings support the potential for integrating LLM-based auditing into routine workflows to strengthen the quality of administrative and research data.

Background/Objectives: Coronavirus disease 2019 (COVID-19) is associated with gastrointestinal (GI) symptoms. Small intestinal bacterial overgrowth (SIBO) is emerging as a significant GI sequela post-COVID-19 infection. This study aimed to evaluate the prevalence and incidence of SIBO post-COVID-19 infection across different age groups and to identify associated risk factors in a global cohort. Methods: A retrospective study utilized the TriNetX database and included adult patients (≥18 years) diagnosed with SIBO following COVID-19 infection (1 January 2022–30 May 2024). A propensity score matching (1:1) was used to adjust for demographics and SIBO risk factors. Kaplan–Meier survival analysis assessed the SIBO incidence within 12 months. Results: Among 1,660,092 COVID-19 patients and 42,322,017 controls, SIBO was diagnosed in 353 COVID-19 patients without hydrogen breath tests (BT) and 78 with BT, compared to 3368 controls without BT and 871 with BT. Age-specific analysis demonstrated a clear, progressive increase in the SIBO incidence, becoming distinctly significant by 6 months and highest at 12 months post-infection. The highest risks were noted in ages 60–69 (0.011% vs. 0.004%, OR 2.6, p = 0.0003) and 70–79 (0.011% vs. 0.005%, OR 2.0, p = 0.0004), with younger age groups (30–49 years) also showing significantly increased risks. The medication analysis revealed strong associations with chronic opioid, proton pump inhibitor, and antidiarrheal medication. Conclusions: COVID-19 significantly increased the risk of SIBO, particularly within the first 12 months post-infection, across various age groups and, notably, in association with certain chronic medications. Clinical vigilance and targeted management strategies are recommended to mitigate long-term GI consequences.

Accurate measures of insulin regulation of adipose tissue lipolysis (free fatty acid release rates) in humans can be accomplished using a two-step, hyperinsulinemic clamp experimental design, but careful scrutiny of the data is needed. The FFA palmitate IC 50 measure is reproducible in weight-stable adults and responds to treatment interventions in a quantitative manner.

Background National guidelines recommend that waist circumference (WC) be measured in patients with a body mass index (BMI) 27–35 kg/m2. Unfortunately, perhaps due to logistical reasons, WC is seldom measured in clinical settings. Herein, we describe the performance of a self-operated waist measurement device (SOWMD) as a potential means to overcome barriers to measuring WC. Materials and methods Ten volunteers underwent WC measures by professionals and SOWMD on 5 separate days to assess the reproducibility and accuracy. We then compared SOWMD measures with CT-derived fat content by recruiting 81 patients scheduled for a diagnostic abdominal CT scan. Results There was no difference between professionally measured and SOWMD-measured WC; the intraindividual coefficient of variation over the 5 days was between 0.4% and 2.2%. The WC measured manually, by SOWMD and CT scan were highly correlated (r=0.90–0.92, all p<0.001). The minimal WC measured by SOWMD was a better predictor (r=0.81 for all patients, r=0.87 for men, both p<0.001) of CT-measured visceral adipose tissue volume than other approaches. The minimal WC measured by SOWMD was correlated with fasting plasma glucose (r=0.40, p<0.05), triglyceride (r=0.41, p<0.01) and high-density cholesterol (r=−0.49, p<0.001) concentrations. Conclusion SOWMD is a reproducible, accurate and convenient way to measure WC that can provide added value for health care providers when combined with BMI information.

Mendelian randomization has become a popular tool to assess causal relationships using existing observational data. While randomized controlled trials are considered the gold standard for establishing causality between exposures and outcomes, it is not always feasible to conduct a trial. Mendelian randomization is a causal inference method that uses observational data to infer causal relationships by using genetic variation as a surrogate for the exposure of interest. Publications using the approach have increased dramatically in recent years, including in the field of hepatology. In this concise review, we describe the concepts, assumptions, and interpretation of Mendelian randomization as related to studies in hepatology. We focus on the strengths and weaknesses of the approach for a non-statistical audience, using an illustrative example to assess the causal relationship between body mass index and NAFLD.

Abstract Context The impact of insulin, particularly exogenous hyperinsulinemia, on insulin secretion in humans is debated. Objective We assessed the effects of exogenous hyperinsulinemia on insulin secretion and whether the response is altered in insulin resistance associated with obesity. Methods Insulin secretion rates (ISRs) during euglycemic hyperinsulinemic clamp studies (52 volunteers) were calculated using a model that employs plasma C-peptide concentrations. One study involved a 2-step insulin clamp and the other study was a single step insulin clamp. For both studies the goal was to achieve plasma glucose concentrations of 95 mg/dL during the clamp irrespective of fasting glucose concentrations. The percent change in ISR from fasting to the end of the insulin clamp interval was the main outcome. Linear regression and analysis of covariance were used to test for the effects of insulin on ISR and to test for group differences. Results ISR was greater in obese volunteers (P &lt; .001) under fasting and hyperinsulinemic clamp conditions. The change in plasma glucose from baseline to the end of the insulin clamp interval was highly correlated with the change in ISR (r = 0.61, P &lt; .001). From baseline to the end of the clamp we observed a 27% (SD 20) suppression of ISR. The participants who underwent a 2-step insulin clamp had greater suppression of ISR during the second step than the first step (P &lt; .001). The proportional suppression of ISR during euglycemic hyperinsulinemia was not different between nonobese and obese groups (P = .19). Conclusion Hyperinsulinemia suppresses endogenous insulin secretion and the relative change in insulin secretion produced by exogenous insulin did not differ between nonobese and obese people.

This experiment studied the effect of changes in overnight concentrations of free fatty acids (FFAs) and glucose on β-cell function and glucose metabolism. In response to elevation of these metabolites, the dynamic component of the β-cell response to glucose was impaired. This suggests that in health overnight hyperglycemia and FFA elevation can deplete preformed insulin granules in the β-cell.

AbstractInterest in analyzing X chromosome single nucleotide polymorphisms (SNPs) is growing and several approaches have been proposed. Prior studies have compared power of different approaches, but bias and interpretation of coefficients have received less attention. We performed simulations to demonstrate the impact of X chromosome model assumptions on effect estimates. We investigated the coefficient biases of SNP and sex effects with commonly used models for X chromosome SNPs, including models with and without assumptions of X chromosome inactivation (XCI), and with and without SNP–sex interaction terms. Sex and SNP coefficient biases were observed when assumptions made about XCI and sex differences in SNP effect in the analysis model were inconsistent with the data‐generating model. However, including a SNP–sex interaction term often eliminated these biases. To illustrate these findings, estimates under different genetic model assumptions are compared and interpreted in a real data example. Models to analyze X chromosome SNPs make assumptions beyond those made in autosomal variant analysis. Assumptions made about X chromosome SNP effects should be stated clearly when reporting and interpreting X chromosome associations. Fitting models with SNP × Sex interaction terms can avoid reliance on assumptions, eliminating coefficient bias even in the absence of sex differences in SNP effect.