Detection of Gastrointestinal Bleeding With Large Language Models to Aid Quality Improvement and Appropriate Reimbursement.

BACKGROUND & AIMS: Early identification and accurate characterization of overt gastrointestinal bleeding (GIB) enables opportunities to optimize patient management and ensures appropriately risk-adjusted coding for claims-based quality measures and reimbursement. Recent advancements in generative artificial intelligence, particularly large language models (LLMs), create opportunities to support accurate identification of clinical conditions. In this study, we present the first LLM-based pipeline for identification of overt GIB in the electronic health record (EHR). We demonstrate 2 clinically relevant applications: the automated detection of recurrent bleeding and appropriate reimbursement coding for patients with GIB. METHODS: Development of the LLM-based pipeline was performed on 17,712 nursing notes from 1108 patients who were hospitalized with acute GIB and underwent endoscopy in the hospital from 2014 to 2023. The pipeline was used to train an EHR-based machine learning model for detection of recurrent bleeding on 546 patients presenting to 2 hospitals and externally validated on 562 patients presenting to 4 different hospitals. The pipeline was used to develop an algorithm for appropriate reimbursement coding on 7956 patients who underwent endoscopy in the hospital from 2019 to 2023. RESULTS: The LLM-based pipeline accurately detected melena (positive predictive value, 0.972; sensitivity, 0.900), hematochezia (positive predictive value, 0.900; sensitivity, 0.908), and hematemesis (positive predictive value, 0.859; sensitivity, 0.932). The EHR-based machine learning model identified recurrent bleeding with area under the curve of 0.986, sensitivity of 98.4%, and specificity of 97.5%. The reimbursement coding algorithm resulted in an average per-patient reimbursement increase of $1299 to $3247 with a total difference of $697,460 to $1,743,649. CONCLUSIONS: An LLM-based pipeline can robustly detect overt GIB in the EHR with clinically relevant applications in detection of recurrent bleeding and appropriate reimbursement coding.

Results of inaugural international Down Syndrome Societal Services and Supports survey.

PURPOSE: We previously designed the Down Syndrome Societal Services and Supports Survey (DS-4S) to measure country-specific supports for people with Down syndrome (DS) across multiple life domains (healthcare, education, policy, independence, and community inclusion). We now report and analyze the results. METHODS: We partnered with international DS consortia, who distributed the DS-4S to 154 cumulative members representing over 100 countries. Organizations were included if they had a holistic focus on the lives of people with DS and if at least 50% of their members either have DS or are family members of people with DS. Factor analysis was used to analyze the results. RESULTS: We received survey responses from 55 different organizations in 50 countries who met inclusion criteria. Each country had complete data for at least 4 of the 5 domains. The lowest 5 scores were from countries in Africa and Asia; the highest 5 scores were in Europe and North America. CONCLUSION: The responses to the DS-4S stratified countries within each surveyed domain. The DS-4S can now be used to track countries' progress over time and to determine which countries have best practices that might be replicated. We will publish the results and update them biennially at www.DownSyndromeQualityOfLife.com.

Orbital optimisation in xTC transcorrelated methods.

We present a combination of the bi-orthogonal orbital optimisation framework with the recently introduced xTC version of transcorrelation. This allows us to implement non-iterative perturbation based methods on top of the transcorrelated Hamiltonian. Additionally, the orbital optimisation influences results of other truncated methods, such as the distinguishable cluster with singles and doubles. The accuracy of these methods in comparison to standard xTC methods is demonstrated, and the advantages and disadvantages of the orbital optimisation are discussed.

Two determinant distinguishable cluster.

A two reference determinant version of the distinguishable cluster with singles and doubles (DCSD) has been developed. We have implemented the two determinant distinguishable cluster (2D-DCSD) and the corresponding traditional 2D-CCSD method in a new open-source package written in Julia called ElemCo.jl. The methods were benchmarked on singlet and triplet excited states of valence and Rydberg character, as well as for singlet-triplet gaps of diradicals. It is demonstrated that the distinguishable cluster approximation improves the accuracy of 2D-CCSD.

xTC: An efficient treatment of three-body interactions in transcorrelated methods.

An efficient implementation for approximate inclusion of the three-body operator arising in transcorrelated methods via exclusion of explicit three-body components (xTC) is presented and tested against results in the "HEAT" benchmark set [Tajti et al., J. Chem. Phys. 121, 011599 (2004)]. Using relatively modest basis sets and computationally simple methods, total, atomization, and formation energies within near-chemical accuracy from HEAT results were obtained. The xTC ansatz reduces the nominal scaling of the three-body part of transcorrelation by two orders of magnitude to O(N5) and can readily be used with almost any quantum chemical correlation method.

Transcorrelated coupled cluster methods. II. Molecular systems.

We demonstrate the accuracy of ground-state energies of the transcorrelated Hamiltonian, employing sophisticated Jastrow factors obtained from variational Monte Carlo, together with the coupled cluster and distinguishable cluster methods at the level of singles and doubles excitations. Our results show that already with the cc-pVTZ basis, the transcorrelated distinguishable cluster method gets close to the complete basis limit and near full configuration interaction quality values for relative energies of over thirty atoms and molecules. To gauge the performance in different correlation regimes, we also investigate the breaking of the nitrogen molecule with transcorrelated coupled cluster methods. Numerical evidence is presented to further justify an efficient way to incorporate the major effects coming from the three-body integrals without explicitly introducing them into the amplitude equations.

Designing an international survey for organisations serving people with Down syndrome.

BACKGROUND: Down syndrome is the most common liveborn genetic condition. However, there are no surveys measuring societal services and supports for people with Down syndrome. We developed a questionnaire so that initiatives could be targeted towards countries most in need of assistance. METHOD: We formed a geographically diverse group of physicians, family members of people with Down syndrome, and members of Down syndrome not-for-profit organisations to create a survey of societal services and supports. We used a modified Delphi method and disseminated the survey to Down syndrome non-profit organisations worldwide. RESULTS: Our survey consists of 61 items categorised within five domains: Education, Community Inclusion, Independence, Healthcare, and Social and Policy Issues. CONCLUSIONS: We developed a survey to measure societal services and supports available to people with Down syndrome as perceived by organisational leaders. Our methods might serve as a blueprint for other populations of people with intellectual and developmental disabilities.

Per Oral Endoscopic Myotomy for Zenker's Diverticulum: A Novel and Superior Technique Compared With Septotomy?

INTRODUCTION: Endoscopic management of Zenker diverticuli (ZD) has traditionally been via septotomy technique. The recent development of tunneling technique has shown to be both efficacious and safe. The aim of this study is to evaluate the tunneling technique using per oral endoscopic myotomy (Z-POEM) versus septotomy. METHODS: Patients who underwent endoscopic management of ZD either by Z-POEM or septotomy from March 2017 until November 2020 from 9 international academic centers were included. Demographics, clinical data preprocedure and postprocedure, procedure time, adverse events, and hospital length of stay were analyzed. RESULTS: A total of 101 patients (mean age 74.9 y old, 55.4% male) were included: septotomy (n=49), Z-POEM (n=52). Preprocedure Functional Oral Intake Scale score and Eckardt score was 5.3 and 5.4 for the septotomy group and 5.9 and 5.15 for the Z-POEM group. Technical success was achieved in 98% of the Z-POEM group and 100% of the septotomy group. Clinical success was achieved in 84% and 92% in the septotomy versus Z-POEM groups. Adverse events occurred in 30.6% (n=15) in septotomy group versus 9.6% (n=5) in the Z-POEM group (P=0.017). Reintervention for ongoing symptoms occurred in 7 patients in the septotomy group and 3 patients in the Z-POEM group. Mean hospital length of stay was shorter for the Z-POEM group, at 1.5 versus 1.9 days. CONCLUSIONS: A tunneling technique via the Z-POEM procedure is an efficacious and safe endoscopic treatment for ZD. Z-POEM is a safer procedure with a statistically significant reduction in adverse events compared with traditional septotomy technique.

Spin Purification in Full-CI Quantum Monte Carlo via a First-Order Penalty Approach.

In this article, we demonstrate that a first-order spin penalty scheme can be efficiently applied to the Slater determinant based Full-CI Quantum Monte Carlo (FCIQMC) algorithm, as a practical route toward spin purification. Two crucial applications are presented to demonstrate the validity and robustness of this scheme: the 1Δg ← 3Σg vertical excitation in O2 and key spin gaps in a [Mn3(IV)O4] cluster. In the absence of a robust spin adaptation/purification technique, both applications would be unattainable by Slater determinant based ground state methods, with any starting wave function collapsing into the higher-spin ground state during the optimization. This strategy can be coupled to other algorithms that use the Slater determinant based FCIQMC algorithm as configuration interaction eigensolver, including the Stochastic Generalized Active Space, the similarity-transformed FCIQMC, the tailored-CC, and second-order perturbation theory approaches. Moreover, in contrast to the GUGA-FCIQMC technique, this strategy features both spin projection and total spin adaptation, making it appealing when solving anisotropic Hamiltonians. It also provides spin-resolved reduced density matrices, important for the investigation of spin-dependent properties in polynuclear transition metal clusters, such as the hyperfine-coupling constants.

Full configuration interaction quantum Monte Carlo treatment of fragments embedded in a periodic mean field.

We present an embedded fragment approach for high-level quantum chemical calculations on local features in periodic systems. The fragment is defined as a set of localized orbitals (occupied and virtual) corresponding to a converged periodic Hartree-Fock solution. These orbitals serve as the basis for the in-fragment post-Hartree-Fock treatment. The embedding field for the fragment, consisting of the Coulomb and exchange potential from the rest of the crystal, is included in the fragment's one-electron Hamiltonian. As an application of the embedded fragment approach, we investigate the performance of full configuration interaction quantum Monte Carlo (FCIQMC) with the adaptive shift. As the orbital choice, we use the natural orbitals from the distinguishable cluster method with singles and doubles. FCIQMC is a stochastic approximation to the full CI method and can be routinely applied to much larger active spaces than the latter. This makes this method especially attractive in the context of open shell defects in crystals, where fragments of adequate size can be rather large. As a test case, we consider dissociation of a fluorine atom from a fluorographane surface. This process poses a challenge for high-level electronic structure models as both the static and dynamic correlations are essential here. Furthermore, the active space for an adequate fragment (32 electrons in 173 orbitals) is already quite large even for FCIQMC. Despite this, FCIQMC delivers accurate dissociation and total energies.

Endoscopic Ultrasound Guided Gastroenterostomy: What Is the Learning Curve?

INTRODUCTION: Endoscopic ultrasound guided gastroenterostomy (EUS-GE) is a minimally invasive option for gastric outlet obstruction. It requires skills in endoscopic ultrasound, fluoroscopy, and lumen-apposing metal stent deployment. The aim of this study was to determine the learning curve for EUS-GE. METHODS: Consecutive patients undergoing EUS-GE by a single operator were included from a prospective registry over 3 years. Demographics, procedure info, postprocedure follow-up data, and adverse events were collected. Nonlinear regression and cumulative sum analyses were conducted for the learning curve. Clinical success was defined as tolerating a diet postprocedure. RESULTS: Twenty-three patients were included (39% male, mean age 65.8 y). Technical success was achieved in 22 (96%) patients. Clinical success was achieved in 21/22 (95%) patients. Average follow-up time 10.8 months (9.1 SD). Five patients had minor postprocedure complications; 1 patient had a periprocedural esophageal tear treated with clips. Four patients required repeat intervention for stent revision or removal if no longer needed.Median procedure time was 88 minutes (range: 45 to 140 min). Cumulative sum chart shows 88-minute procedure time was achieved at the seventh procedure indicating efficiency. Even with bridging of a misdeployed lumen-apposing metal stent, the procedure duration further reduced with consequent procedures indicating continued improvement with experience (nonlinear regression P<0.0001). CONCLUSIONS: Endoscopists experienced in EUS-GE achieve a reduction in procedure time over successive cases, with efficiency reached at 88 minutes and a learning rate of 7 cases. Misdeployed stents that require bridging add to the procedure time even after competency is achieved but do not affect the overall learning curve trend.

Assessing Different Approaches to Leveraging Historical Smoking Exposure Data to Better Select Lung Cancer Screening Candidates: A Retrospective Validation Study.

INTRODUCTION: There is mounting interest in the use of risk prediction models to guide lung cancer screening. Electronic health records (EHRs) could facilitate such an approach, but smoking exposure documentation is notoriously inaccurate. While the negative impact of inaccurate EHR data on screening practices reliant on dichotomized age and smoking exposure-based criteria has been demonstrated, less is known regarding its impact on the performance of model-based screening. AIMS AND METHODS: Data were collected from a cohort of 37 422 ever-smokers between the ages of 55 and 74, seen at an academic safety-net healthcare system between 1999 and 2018. The National Lung Cancer Screening Trial (NLST) criteria, PLCOM2012 and LCRAT lung cancer risk prediction models were validated against time to lung cancer diagnosis. Discrimination (area under the receiver operator curve [AUC]) and calibration were assessed. The effect of substituting the last documented smoking variables with differentially retrieved "history conscious" measures was also determined. RESULTS: The PLCOM2012 and LCRAT models had AUCs of 0.71 (95% CI, 0.69 to 0.73) and 0.72 (95% CI, 0.70 to 0.74), respectively. Compared with the NLST criteria, PLCOM2012 had a significantly greater time-dependent sensitivity (69.9% vs. 64.5%, p < .01) and specificity (58.3% vs. 56.4%, p < .001). Unlike the NLST criteria, the performances of the PLCOM2012 and LCRAT models were not prone to historical variability in smoking exposure documentation. CONCLUSIONS: Despite the inaccuracies of EHR-documented smoking histories, leveraging model-based lung cancer risk estimation may be a reasonable strategy for screening, and is of greater value compared with using NLST criteria in the same setting. IMPLICATIONS: EHRs are potentially well suited to aid in the risk-based selection of lung cancer screening candidates, but healthcare providers and systems may elect not to leverage EHR data due to prior work that has shown limitations in structured smoking exposure data quality. Our findings suggest that despite potential inaccuracies in the underlying EHR data, screening approaches that use multivariable models may perform significantly better than approaches that rely on simpler age and exposure-based criteria. These results should encourage providers to consider using pre-existing smoking exposure data with a model-based approach to guide lung cancer screening practices.

Accuracy of the distinguishable cluster approximation for triple excitations for open-shell molecules and excited states.

The distinguishable cluster approximation for triple excitations has been applied to calculate thermochemical properties and excited states involving closed-shell and open-shell species, such as small molecules, 3d transition metal atoms, ozone, and an iron-porphyrin model. Excitation energies have been computed using the ΔCC approach by directly optimizing the excited states. A fixed-reference technique has been introduced to target selected spin-states for open-shell molecular systems. The distinguishable cluster approximation consistently improves coupled cluster with singles doubles and triples results for absolute and relative energies. For excited states dominated by a single configuration state function, the fixed-reference approach combined with high-level coupled-cluster methods has a comparable accuracy to the corresponding equation-of-motion coupled-cluster methods with a negligible amount of spin contamination.

Transcorrelated coupled cluster methods.

Transcorrelated coupled cluster and distinguishable cluster methods are presented. The Hamiltonian is similarity transformed with a Jastrow factor in the first quantization, which results in up to three-body integrals. The coupled cluster with singles and doubles equations on this transformed Hamiltonian are formulated and implemented. It is demonstrated that the resulting methods have a superior basis set convergence and accuracy to the corresponding conventional and explicitly correlated methods. Additionally, approximations for three-body integrals are suggested and tested.

The Molpro quantum chemistry package.

Molpro is a general purpose quantum chemistry software package with a long development history. It was originally focused on accurate wavefunction calculations for small molecules but now has many additional distinctive capabilities that include, inter alia, local correlation approximations combined with explicit correlation, highly efficient implementations of single-reference correlation methods, robust and efficient multireference methods for large molecules, projection embedding, and anharmonic vibrational spectra. In addition to conventional input-file specification of calculations, Molpro calculations can now be specified and analyzed via a new graphical user interface and through a Python framework.

Toward fast and accurate ab initio calculation of magnetic exchange in polynuclear lanthanide complexes.

Ab initio calculations of the magnetic exchange in polynuclear lanthanide complexes are very challenging and often not feasible, due to large active spaces, the large number of required states or the necessity to include dynamical correlation into the calculations. We present an approach which allows for the computationally efficient calculation of exchange splittings in polynuclear lanthanide complexes including dynamical correlation. This is achieved by extending the local-density-fitted configuration-averaged Hartree-Fock (LDF-CAHF) method to systems with more than one group of open-shell orbitals (e.g. at different metal atoms) and combining it with linear-scaling many-state pair-natural-orbital complete active space perturbation theory of second order (PNO-CASPT2). In order to assess the performance of the method, we apply it to the asymmetric dinuclear complex [hqH2][Yb2(hq)4(NO3)3]·MeOH (hqH = 8-hydroxyquinoline).

On the distinguishable cluster approximation for triple excitations.

The distinguishable cluster approximation applied to coupled cluster doubles equations greatly improves absolute and relative energies. We apply the same approximation to the triples equations and demonstrate that it can also improve the results of the coupled cluster method with singles, doubles, and triples. The resulting method has a nominal computational scaling of O(N7) in the real-space representation, and is orbital invariant, size extensive, and exact for three electrons.

Multi-state local complete active space second-order perturbation theory using pair natural orbitals (PNO-MS-CASPT2).

A multistate complete active space second-order perturbation theory (CASPT2) method is presented, which utilizes domains of pair natural orbitals and projected atomic orbitals for the virtual space to achieve linear scaling of the computational effort with the number of inactive orbitals. The method is applied to compute excitation energies of medium size aromatic molecules, and it is shown that the impact of the local approximations on the computed excitation energies is negligible. The applicability and efficiency of the method are demonstrated for two large molecular systems with up to 400 correlated electrons, nearly 3000 basis functions, and 45 electronic states. Furthermore, some approximations in the CASPT2 zeroth-order Hamiltonian, which decouple different configuration spaces, are proposed and tested. These approximations allow us to reuse many integrals and amplitudes from the ground state in the excited states, thereby significantly reducing the computational effort for calculations with many states. Using appropriate correction terms, the impact of these approximations is shown to be small.