Procedure code overutilization detection from healthcare claims using unsupervised deep learning methods.

BACKGROUND: Fraud, Waste, and Abuse (FWA) in medical claims have a negative impact on the quality and cost of healthcare. A major component of FWA in claims is procedure code overutilization, where one or more prescribed procedures may not be relevant to a given diagnosis and patient profile, resulting in unnecessary and unwarranted treatments and medical payments. This study aims to identify such unwarranted procedures from millions of healthcare claims. In the absence of labeled examples of unwarranted procedures, the study focused on the application of unsupervised machine learning techniques. METHODS: Experiments were conducted with deep autoencoders to find claims containing anomalous procedure codes indicative of FWA, and were compared against a baseline density-based clustering model. Diagnoses, procedures, and demographic data associated with healthcare claims were used as features for the models. A dataset of one hundred thousand claims sampled from a larger claims database is used to initially train and tune the models, followed by experimentations on a dataset with thirty-three million claims. Experimental results show that the autoencoder model, when trained with a novel feature-weighted loss function, outperforms the density-based clustering approach in finding potential outlier procedure codes. RESULTS: Given the unsupervised nature of our experiments, model performance was evaluated using a synthetic outlier test dataset, and a manually annotated outlier test dataset. Precision, recall and F1-scores on the synthetic outlier test dataset for the autoencoder model trained on one hundred thousand claims were 0.87, 1.0 and 0.93, respectively, while the results for these metrics on the manually annotated outlier test dataset were 0.36, 0.86 and 0.51, respectively. The model performance on the manually annotated outlier test dataset improved further when trained on the larger thirty-three million claims dataset with precision, recall and F1-scores of 0.48, 0.90 and 0.63, respectively. CONCLUSIONS: This study demonstrates the feasibility of leveraging unsupervised, deep-learning methods to identify potential procedure overutilization from healthcare claims.

A Major Step in Opening the Black Box of High-Molecular-Weight Dissolved Organic Nitrogen by Isotopic Labeling of Synechococcus and Multibond Two-Dimensional NMR.

Dissolved organic nitrogen (DON) comprises the largest pool of fixed N in the surface ocean, yet its composition has remained poorly constrained. Knowledge of the chemical composition of this nitrogen pool is crucial for understanding its biogeochemical function and reactivity in the environment. Previous work has suggested that high-molecular-weight (high-MW) DON exists only in two closely related forms, the secondary amides of peptides and of N-acetylated hexose sugars. Here, we demonstrate that the chemical structures of high-MW DON may be much more diverse than previously thought. We couple isotopic labeling of cyanobacterially derived dissolved organic matter with advanced two-dimensional NMR spectroscopy to open the "black box" of uncharacterized high-MW DON. Using multibond NMR correlations, we have identified novel N-methyl-containing amines and amides, primary amides, and novel N-acetylated sugars, which together account for nearly 50% of cyanobacterially derived high-MW DON. This study reveals unprecedented compositional details of the previously uncharacterized DON pool and outlines the means to further advance our understanding of this biogeochemically and globally important reservoir of organic nitrogen.

Enhanced detection of the algal toxin PbTx-2 in marine waters by atmospheric pressure chemical ionization mass spectrometry.

RATIONALE: Karenia brevis, a marine dinoflagellate, biosynthesizes a unique class of polyether toxins called brevetoxins that produce significant health, environmental and economic impacts in and along coastal waters. Previous application of liquid chromatography/mass spectrometry for detection of the most common brevetoxin, PbTx-2, has relied almost exclusively upon electrospray ionization (ESI). A different ionization source is proposed in this study with improved sensitivity ultimately leading to lower limit of detection compared to (+) ESI. METHODS: Brevetoxin standards and samples (PbTx-2) were analyzed by liquid chromatography/mass spectrometry using both (+) atmospheric pressure chemical ionization and (+) electrospray ionization sources. RESULTS: LC/MS with (+) APCI exhibited an order of magnitude improvement in the limit of detection (7.7 × 10(-4) pg mass on-column) compared to the same method using (+) ESI (7.5 × 10(-3) pg mass on-column). The calibration sensitivity of (+) APCI (1.3 × 10(3)) was also five times higher than positive mode (+) ESI (0.26 × 10(3)). CONCLUSIONS: Positive mode APCI represents a significant improvement in detection and quantification of PbTx-2 by LC/MS allowing for smaller sample sizes compared to previous studies using (+) ESI. This in turn leads to higher throughput of samples during and after bloom events giving stakeholders detailed information on the fate of this potent marine toxin.

Leveraging deep survival models to predict quality of care risk in diverse hospital readmissions.

Hospital readmissions rate is reportedly high and has caused huge financial burden on health care systems in many countries. It is viewed as an important indicator of health care providers' quality of care. We examine the use of machine learning-based survival analysis to assess quality of care risk in hospital readmissions. This study applies various survival models to explore the risk of hospital readmissions given patient demographics and their respective hospital discharges extracted from a health care claims dataset. We explore advanced feature representation techniques such as BioBERT and Node2Vec to encode high-dimensional diagnosis code features. To our knowledge, this study is the first to apply deep-learning based survival-analysis models for predicting hospital readmission risk agnostic of specific medical conditions and a fixed window for readmission. We found that modeling the time from discharge date to readmission date as a Weibull distribution as in the SparseDeepWeiSurv model yields the best discriminative power and calibration. In addition, embedding representations of the diagnosis codes do not contribute to improvement in model performance. We find dependency of each model's performance on the time point at which it is evaluated. This time dependency of the models' performance on the health care claims data may necessitate a different choice of model in quality of care issue detection at different points in time. We show the effectiveness of deep-learning based survival-analysis models in estimating the quality of care risk in hospital readmissions.

Surface waters as a sink and source of atmospheric gas phase ethanol.

This study reports the first ethanol concentrations in fresh and estuarine waters and greatly expands the current data set for coastal ocean waters. Concentrations for 153 individual measurements of 11 freshwater sites ranged from 5 to 598 nM. Concentrations obtained for one estuarine transect ranged from 56 to 77 nM and levels in five coastal ocean depth profiles ranged from 81 to 334 nM. Variability in ethanol concentrations was high and appears to be driven primarily by photochemical and biological processes. 47 gas phase concentrations of ethanol were also obtained during this study to determine the surface water degree of saturation with respect to the atmosphere. Generally fresh and estuarine waters were undersaturated indicating they are not a source and may be a net sink for atmospheric ethanol in this region. Aqueous phase ethanol is likely converted rapidly to acetaldehyde in these aquatic ecosystems creating the undersaturated conditions resulting in this previously unrecognized sink for atmospheric ethanol. Coastal ocean waters may act as either a sink or source of atmospheric ethanol depending on the partial pressure of ethanol in the overlying air mass. Results from this study are significant because they suggest that surface waters may act as an important vector for the uptake of ethanol emitted into the atmosphere including ethanol from biofuel production and usage.

Photodegradation of the antihistamine cetirizine in natural waters.

The photodegradation rate of the anti-histamine cetirizine (Zyrtec®) was investigated in various water matrices. The average observed first-order photodegradation rate coefficient (kobs ), obtained by linear regression of the logarithmic-transformed cetirizine concentrations versus irradiation time in simulated sunlight, was 0.024 h(-1) (n = 6; standard deviation ± 0.004) in deionized water corresponding to a half-life of approximately 30 h. There was no statistical difference in the kobs of cetirizine photodegradation in coastal seawater compared with deionized water or deionized water amended with dissolved chromophoric organic matter. The quantum yield of cetirizine photodegradation decreased dramatically with increasing wavelength and decreasing energy of incoming radiation, with the average value ranging from 5.28 × 10(-4) to 6.40 × 10(-3) in the ultraviolet wavelength range (280-366 nm). The activation energy of cetirizine photodegradation was 10.3 kJ mol(-1) with an observed increase in cetirizine photodegradation as temperature increased. This is a significant environmental factor influencing half-life and an important consideration, given that cetirizine has been detected in wastewater and receiving waters from different locations globally.