Bayesian multitask learning for medicine recommendation based on online patient reviews.

MOTIVATION: We propose a drug recommendation model that integrates information from both structured data (patient demographic information) and unstructured texts (patient reviews). It is based on multitask learning to predict review ratings of several satisfaction-related measures for a given medicine, where related tasks can learn from each other for prediction. The learned models can then be applied to new patients for drug recommendation. This is fundamentally different from most recommender systems in e-commerce, which do not work well for new customers (referred to as the cold-start problem). To extract information from review texts, we employ both topic modeling and sentiment analysis. We further incorporate variable selection into the model via Bayesian LASSO, which aims to filter out irrelevant features. To our best knowledge, this is the first Bayesian multitask learning method for ordinal responses. We are also the first to apply multitask learning to medicine recommendation. The sample code and data are made available at GitHub: https://github.com/thrushcyc-github/BMull. RESULTS: We evaluate the proposed method on two sets of drug reviews involving 17 depression/high blood pressure-related drugs. Overall, our method performs better than existing benchmark methods in terms of accuracy and AUC (area under the receiver operating characteristic curve). It is effective even with a small sample size and only a few available features, and more robust to possible noninformative covariates. Due to our model explainability, insights generated from our model may work as a useful reference for doctors. In practice, however, a final decision should be carefully made by combining the information from the proposed recommender with doctors' domain knowledge and past experience. AVAILABILITY AND IMPLEMENTATION: The sample code and data are publicly available at GitHub: https://github.com/thrushcyc-github/BMull.

Persistent TLR4 Activation Promotes Hepatocellular Carcinoma Growth through Positive Feedback Regulation by LIN28A/Let-7g miRNA.

Chronic inflammation caused by liver damage or infection plays an important role in the development and progression of hepatocellular carcinoma (HCC). The activation of Toll-like receptors 4 (TLR4) is involved in HCC tumorigenesis. Moreover, high TLR4 expression in HCC has been linked to poor prognosis. Although the expression of TLR4 in HCC is relatively low compared to hematopoietic cells, it is important to explore the molecular mechanism leading to the elevation of TLR4 in HCC. In this study, we aimed to investigate the positive regulating loop for TLR4 expression in HCC in response to chronic inflammation. Our results confirm that the mRNA expression of TLR4 and proinflammatory cytokines, including interleukin 6 (IL6) and C-C motif chemokine ligand 2 (CCL2), positively correlate in human HCC samples. High TLR4 expression in HCC is more susceptible to lipopolysaccharide (LPS); TLR4 activation in HCC provides growth and survival advantages and thus promotes tumorigenesis. It has been shown that the LIN28/let-7 microRNA (miRNA) axis is a downstream effector of the TLR4 signal pathway, and let-7 miRNA is a potential post-transcriptional regulator for TLR4. Thus, we investigated the correlation between TLR4 and LIN28A mRNA and let-7g miRNA in HCC clinical samples and found that the expression of TLR4 was positively correlated with LIN28A and negatively correlated with let-7g miRNA. Moreover, by culturing PLC/PRF5 (PLC5) HCC cells in low-dose LPS-containing medium to mimic chronic inflammation for persistent TLR4 activation, the mRNA and protein levels of TLR4 and LIN28A were elevated, and let-7g miRNA was decreased. Furthermore, the 3' untranslated region (3'UTR) of TLR4 mRNA was shown to be the target of let-7g miRNA, suggesting that inhibition of let-7g miRNA is able to increase TLR4 mRNA. While parental PLC5 cells have a low susceptibility to LPS-induced cell growth, long-term LPS exposure for PLC5 cells leads to increased proliferation, cytokine expression and stemness properties. In conclusion, our studies demonstrate positive feedback regulation for chronic TLR4 activation in the modulation of TLR4 expression level through the LIN28A/let-7g pathway in HCC and suggest a connection between chronic inflammation and TLR4 expression level in HCC for promoting tumorigenesis.

Transcriptome and Metabolome Profiling Reveal the Resistance Mechanisms of Rice against Brown Planthopper.

Brown planthopper (Nilaparvata lugens Stål, BPH) is one of the most destructive insects affecting rice production. To better understand the physiological mechanisms of how rice responds to BPH feeding, we analyzed BPH-induced transcriptomic and metabolic changes in leaf sheaths of both BPH-susceptible and -resistant rice varieties. Our results demonstrated that the resistant rice reduced the settling, feeding and growth of BPH. Metabolic analyses indicated that BPH infestation caused more drastic overall metabolic changes in the susceptible variety than the resistant rice. Differently accumulated metabolites (DAMs) belonging to flavonoids were downregulated in the susceptible rice but upregulated in resistant variety. Transcriptomic analyses revealed more differentially expressed genes (DEGs) in susceptible rice than resistant rice, and DEGs related to stimulus were significantly upregulated in resistant rice but downregulated in susceptible rice. Combined analyses of transcriptome and metabolome showed that many DEGs and DAMs were enriched in phenylpropane biosynthesis, flavonoid biosynthesis, and plant hormone signal transduction. We conducted correlation analyses of DEGs and DAMs in these pathways and found a high correlation between DEGs and DAMs. Then, we found that the contents of endogenous indole 3-acetic acid (IAA) in resistant rice was lower than that of susceptible rice after BPH feeding, while the salicylic acid (SA) content was the opposite. For functional analysis, an exogenous application of IAA decreased rice resistance to BPH, but the exogenous application of SA increased resistance. In addition, biochemical assessment and quantitative PCR analysis showed that the lignin content of resistant accession was constitutively higher than in susceptible accession. By adding epigallocatechin, the substrate of anthocyanidin reductase (ANR), to the artificial diet decreased the performance of BPH. We first combined a transcriptome-metabolome-wide association study (TMWAS) on rice resistance to BPH in this study. We demonstrated that rice promoted resistance to BPH by inducing epigallocatechin and decreasing IAA. These findings provided useful transcriptomic and metabolic information for understanding the rice-BPH interactions.

Posttranslational Modification of Waxy to Genetically Improve Starch Quality in Rice Grain.

The waxy (Wx) gene, encoding the granule-bound starch synthase (GBSS), is responsible for amylose biosynthesis and plays a crucial role in defining eating and cooking quality. The waxy locus controls both the non-waxy and waxy rice phenotypes. Rice starch can be altered into various forms by either reducing or increasing the amylose content, depending on consumer preference and region. Low-amylose rice is preferred by consumers because of its softness and sticky appearance. A better way of improving crops other than downregulation and overexpression of a gene or genes may be achieved through the posttranslational modification of sites or regulatory enzymes that regulate them because of their significance. The impact of posttranslational GBSSI modifications on extra-long unit chains (ELCs) remains largely unknown. Numerous studies have been reported on different crops, such as wheat, maize, and barley, but the rice starch granule proteome remains largely unknown. There is a need to improve the yield of low-amylose rice by employing posttranslational modification of Wx, since the market demand is increasing every day in order to meet the market demand for low-amylose rice in the regional area that prefers low-amylose rice, particularly in China. In this review, we have conducted an in-depth review of waxy rice, starch properties, starch biosynthesis, and posttranslational modification of waxy protein to genetically improve starch quality in rice grains.

Genome-Wide Identification of QTLs for Grain Protein Content Based on Genotyping-by-Resequencing and Verification of qGPC1-1 in Rice.

To clarify the genetic mechanism underlying grain protein content (GPC) and to improve rice grain qualities, the mapping and cloning of quantitative trait loci (QTLs) controlling the natural variation of GPC are very important. Based on genotyping-by-resequencing, a total of 14 QTLs were detected with the Huanghuazhan/Jizi1560 (HHZ/JZ1560) recombinant inbred line (RIL) population in 2016 and 2017. Seven of the fourteen QTLs were repeatedly identified across two years. Using three residual heterozygote-derived populations, a stably inherited QTL named as qGPC1-1 was validated and delimited to a ~862 kb marker interval JD1006-JD1075 on the short arm of chromosome 1. Comparing the GPC values of the RIL population determined by near infrared reflectance spectroscopy (NIRS) and Kjeldahl nitrogen determination (KND) methods, high correlation coefficients (0.966 and 0.983) were observed in 2016 and 2017. Furthermore, 12 of the 14 QTLs were identically identified with the GPC measured by the two methods. These results indicated that instead of the traditional KND method, the rapid and easy-to-operate NIRS was suitable for analyzing a massive number of samples in mapping and cloning QTLs for GPC. Using the gel-based low-density map consisted of 208 simple sequence repeat (SSR) and insert/deletion (InDel) markers, the same number of QTLs (fourteen) were identified in the same HHZ/JZ1560 RIL population, and three QTLs were repeatedly detected across two years. More stably expressed QTLs were identified based on the genome resequencing, which might be attributed to the high-density map, increasing the detection power of minor QTLs. Our results are helpful in dissecting the genetic basis of GPC and improving rice grain qualities through molecular assisted selection.

Identifying disease-associated copy number variations by a doubly penalized regression model.

Copy number variation (CNV) of DNA plays an important role in the development of many diseases. However, due to the irregularity and sparsity of the CNVs, studying the association between CNVs and a disease outcome or a trait can be challenging. Up to now, not many methods have been proposed in the literature for this problem. Most of the current researchers reply on an ad hoc two-stage procedure by first identifying CNVs in each individual genome and then performing an association test using these identified CNVs. This potentially leads to information loss and as a result a lower power to identify disease associated CNVs. In this article, we describe a new method that combines the two steps into a single coherent model to identify the common CNV across patients that are associated with certain diseases. We use a double penalty model to capture CNVs' association with both the intensities and the disease trait. We validate its performance in simulated datasets and a data example on platinum resistance and CNV in ovarian cancer genome.

Molecular signatures associated with ZIKV exposure in human cortical neural progenitors.

Zika virus (ZIKV) infection causes microcephaly and has been linked to other brain abnormalities. How ZIKV impairs brain development and function is unclear. Here we systematically profiled transcriptomes of human neural progenitor cells exposed to Asian ZIKVC, African ZIKVM, and dengue virus (DENV). In contrast to the robust global transcriptome changes induced by DENV, ZIKV has a more selective and larger impact on expression of genes involved in DNA replication and repair. While overall expression profiles are similar, ZIKVC, but not ZIKVM, induces upregulation of viral response genes and TP53. P53 inhibitors can block the apoptosis induced by both ZIKVC and ZIKVM in hNPCs, with higher potency against ZIKVC-induced apoptosis. Our analyses reveal virus- and strain-specific molecular signatures associated with ZIKV infection. These datasets will help to investigate ZIKV-host interactions and identify neurovirulence determinants of ZIKV.

Group association test using a hidden Markov model.

In the genomic era, group association tests are of great interest. Due to the overwhelming number of individual genomic features, the power of testing for association of a single genomic feature at a time is often very small, as are the effect sizes for most features. Many methods have been proposed to test association of a trait with a group of features within a functional unit as a whole, e.g. all SNPs in a gene, yet few of these methods account for the fact that generally a substantial proportion of the features are not associated with the trait. In this paper, we propose to model the association for each feature in the group as a mixture of features with no association and features with non-zero associations to explicitly account for the possibility that a fraction of features may not be associated with the trait while other features in the group are. The feature-level associations are first estimated by generalized linear models; the sequence of these estimated associations is then modeled by a hidden Markov chain. To test for global association, we develop a modified likelihood ratio test based on a log-likelihood function that ignores higher order dependency plus a penalty term. We derive the asymptotic distribution of the likelihood ratio test under the null hypothesis. Furthermore, we obtain the posterior probability of association for each feature, which provides evidence of feature-level association and is useful for potential follow-up studies. In simulations and data application, we show that our proposed method performs well when compared with existing group association tests especially when there are only few features associated with the outcome.

Cobalt salt-catalyzed carbocyclization reactions of α-bromo-N-phenylacetamide derivatives.

An efficient and low-cost method for the synthesis of 4-substituted quinolin-2-(1H)-ones has been developed. In the presence of TBHP, the cobalt(ii)-catalyzed carbocyclization reactions of arylethenyl substituted α-bromo-N-phenylacetamides, involving sequential 6-exo-trig radical cyclization, t-butylperoxy radical cross-coupling reaction and the base-promoted ionic Kornblum-DeLaMare reaction, produced 4-benzoylquinolin-2-(1H)-ones. A variety of useful functional groups such as methoxy, fluoro, chloro, bromo, methoxycarbonyl, and cyano groups, are compatible with the reaction conditions. This strategy was further applied to arylethynyl substituted α-bromo-N-phenylacetamides, and 4-benzylquinolin-2-(1H)-ones were formed effectively.

Copy number alterations detected by whole-exome and whole-genome sequencing of esophageal adenocarcinoma.

BACKGROUND: Esophageal adenocarcinoma (EA) is among the leading causes of cancer mortality, especially in developed countries. A high level of somatic copy number alterations (CNAs) accumulates over the decades in the progression from Barrett's esophagus, the precursor lesion, to EA. Accurate identification of somatic CNAs is essential to understand cancer development. Many studies have been conducted for the detection of CNA in EA using microarrays. Next-generation sequencing (NGS) technologies are believed to have advantages in sensitivity and accuracy to detect CNA, yet no NGS-based CNA detection in EA has been reported. RESULTS: In this study, we analyzed whole-exome (WES) and whole-genome sequencing (WGS) data for detecting CNA from a published large-scale genomic study of EA. Two specific comparisons were conducted. First, the recurrent CNAs based on WGS and WES data from 145 EA samples were compared to those found in five previous microarray-based studies. We found that the majority of the previously identified regions were also detected in this study. Interestingly, some novel amplifications and deletions were discovered using the NGS data. In particular, SKI and PRKCZ detected in a deletion region are involved in transforming growth factor-ß pathway, suggesting the potential utility of novel biomarkers for EA. Second, we compared CNAs detected in WGS and WES data from the same 15 EA samples. No large-scale CNA was identified statistically more frequently by WES or WGS, while more focal-scale CNAs were detected by WGS than by WES. CONCLUSIONS: Our results suggest that NGS can replace microarrays to detect CNA in EA. WGS is superior to WES in that it can offer finer resolution for the detection, though if the interest is on recurrent CNAs, WES can be preferable to WGS for its cost-effectiveness.

Aberrant expression of annexin A10 is closely related to gastric phenotype in serrated pathway to colorectal carcinoma.

Annexin A10 (ANXA10) is a member of the ANX family that is normally expressed in gastric mucosa. ANXA10 was recently observed to be upregulated in sessile serrated adenoma, a precursor to microsatellite-unstable colorectal cancer. We investigated the use of ANXA10 in diagnosing colorectal carcinoma. In an immunohistochemical analysis, the intensity and quantity of ANXA10, MUC5AC, MUC6 and CDX2 in 123 colorectal carcinomas were graded. We determined the molecular status of BRAF and KRAS mutations, as well as the microsatellite instability status and the CpG island methylator phenotype in all colorectal carcinomas, and subcategorized into four molecular subgroups according to the molecular derangements. Nuclear ANXA10 staining was present in 36 colorectal carcinomas, exhibiting a strong significant association with the BRAF mutation status (P<0.0001) and positive CpG island methylator phenotype (P<0.0001), and a borderline significant association with high levels of microsatellite instability (P=0.072). The ANXA10-positive colorectal carcinomas were frequently positive for MUC5AC and MUC6, and were associated with absent or reduced CDX2 expression (all P<0.0001). According to a classification and regression tree analysis, ANXA10 is a superior marker for the molecular subtyping of colorectal carcinomas and represents a specific marker for colorectal cancers of the serrated pathway. Our results indicated that ANXA10 expression is implicated in gastric programming in serrated-pathway-associated colorectal carcinoma. ANXA10-positive colorectal carcinoma is highly associated with the molecular features of the serrated neoplasia pathway.

Traditional serrated adenoma has two pathways of neoplastic progression that are distinct from the sessile serrated pathway of colorectal carcinogenesis.

Traditional serrated adenoma is one type of colorectal serrated neoplasm and a precursor of colorectal cancer. We evaluated the pathologic and molecular features of 60 traditional serrated adenomas with cytologic dysplasia and/or invasive carcinoma. On the basis of morphological features, 16 cases (27%) were categorized as traditional serrated adenoma with serrated dysplasia and 25 cases (42%) as traditional serrated adenoma with conventional adenomatous dysplasia. In addition, 19 cases (31%) showed an overall tubulovillous adenomatous structure but with focal serrated feature. Traditional serrated adenoma with serrated dysplasia had a significantly higher frequency of BRAF mutation than traditional serrated adenoma with conventional adenomatous dysplasia and tubulovillous adenoma with serrated feature (P=0.006), whereas traditional serrated adenoma with conventional adenomatous dysplasia and tubulovillous adenoma with serrated feature had higher frequencies of KRAS mutation than traditional serrated adenoma with serrated dysplasia (P<0.0001). Only traditional serrated adenoma with serrated dysplasia showed sessile serrated adenoma-like lesions at the periphery (n=3) and developed invasive carcinomas when the lesions were <15 mm in size. Abnormal nuclear accumulation of ß-catenin was detected in traditional serrated adenoma with conventional adenomatous dysplasia and tubulovillous adenoma with serrated feature but not in traditional serrated adenoma with serrated dysplasia. The frequency of the positive CpG island methylator phenotype was similar among the three dysplastic subtypes, and immunostaining of four mismatch repair proteins in the nucleus was retained in all traditional serrated adenomas and associated invasive malignancies. Traditional serrated adenoma-associated adenocarcinomas (n=28) displayed distinctive morphological features: oval cell nuclei, serrated glands, infiltrating borders, rare occurrences of necrosis and mucinous differentiation. Overexpression of p53 was detected only in high-grade dysplasia and invasive adenocarcinoma. Our findings indicate that traditional serrated adenoma is a heterogeneous neoplasm with two pathways of neoplastic progression, which are distinct from the sessile serrated pathway of colorectal carcinogenesis.

Dissecting the Genetic Basis of Yield Traits and Validation of a Novel Quantitative Trait Locus for Grain Width and Weight in Rice.

Grain yield in rice is a complex trait and it is controlled by a number of quantitative trait loci (QTL). To dissect the genetic basis of rice yield, QTL analysis for nine yield traits was performed using an F2 population containing 190 plants, which was developed from a cross between Youyidao (YYD) and Sanfenhe (SFH), and each plant in the population evaluated with respect to nine yield traits. In this study, the correlations among the nine yield traits were analyzed. The grain yield per plant positively correlated with six yield traits, except for grain length and grain width, and showed the highest correlation coefficient of 0.98 with the number of filled grains per plant. A genetic map containing 133 DNA markers was constructed and it spanned 1831.7 cM throughout 12 chromosomes. A total of 36 QTLs for the yield traits were detected on nine chromosomes, except for the remaining chromosomes 5, 8, and 9. The phenotypic variation was explained by a single QTL that ranged from 6.19% to 36.01%. Furthermore, a major QTL for grain width and weight, qGW2-1, was confirmed to be newly identified and was narrowed down to a relatively smaller interval of about ~2.94-Mb. Collectively, we detected a total of 36 QTLs for yield traits and a major QTL, qGW2-1, was confirmed to control grain weight and width, which laid the foundation for further map-based cloning and molecular design breeding in rice.

Endosymbiotic Fungal Diversity and Dynamics of the Brown Planthopper across Developmental Stages, Tissues, and Sexes Revealed Using Circular Consensus Sequencing.

Endosymbiotic fungi play an important role in the growth and development of insects. Understanding the endosymbiont communities hosted by the brown planthopper (BPH; Nilaparvata lugens Stål), the most destructive pest in rice, is a prerequisite for controlling BPH rice infestations. However, the endosymbiont diversity and dynamics of the BPH remain poorly studied. Here, we used circular consensus sequencing (CCS) to obtain 87,131 OTUs (operational taxonomic units), which annotated 730 species of endosymbiotic fungi in the various developmental stages and tissues. We found that three yeast-like symbionts (YLSs), Polycephalomyces prolificus, Ophiocordyceps heteropoda, and Hirsutella proturicola, were dominant in almost all samples, which was especially pronounced in instar nymphs 4-5, female adults, and the fat bodies of female and male adult BPH. Interestingly, honeydew as the only in vitro sample had a unique community structure. Various diversity indices might indicate the different activity of endosymbionts in these stages and tissues. The biomarkers analyzed using LEfSe suggested some special functions of samples at different developmental stages of growth and the active functions of specific tissues in different sexes. Finally, we found that the incidence of occurrence of three species of Malassezia and Fusarium sp. was higher in males than in females in all comparison groups. In summary, our study provides a comprehensive survey of symbiotic fungi in the BPH, which complements the previous research on YLSs. These results offer new theoretical insights and practical implications for novel pest management strategies to understand the BPH-microbe symbiosis and devise effective pest control strategies.

Determinants of outpatient substance use disorder treatment length-of-stay and completion: the case of a treatment program in the southeast U.S.

Successful outcomes of outpatient substance use disorder treatment result from many factors for clients-including intersections between individual characteristics, choices made, and social determinants. However, prioritizing which of these and in what combination, to address and provide support for remains an open and complex question. Therefore, we ask: What factors are associated with outpatient substance use disorder clients remaining in treatment for > 90 days and successfully completing treatment? To answer this question, we apply a virtual twins machine learning (ML) model to de-identified data for a census of clients who received outpatient substance use disorder treatment services from 2018 to 2021 from one treatment program in the Southeast U.S. We find that primary predictors of outcome success are: (1) attending self-help groups while in treatment, and (2) setting goals for treatment. Secondary predictors are: (1) being linked to a primary care provider (PCP) during treatment, (2) being linked to supplemental nutrition assistance program (SNAP), and (3) attending 6 or more self-help group sessions during treatment. These findings can help treatment programs guide client choice making and help set priorities for social determinant support. Further, the ML method applied can explain intersections between individual and social predictors, as well as outcome heterogeneity associated with subgroup differences.

Examining Physical Wellness as the Fundamental Element for Achieving Holistic Well-Being in Older Persons: Review of Literature and Practical Application in Daily Life.

This review examines the impact of physical activity, nutrition, and sleep evaluations on the physical wellness (PW) and overall well-being of older individuals. A comprehensive search was conducted in databases like PubMed, Google Scholar, and EBSCO Information Services. The search spanned from January 2000 to December 2022, resulting in 19,400 articles, out of which 98 review articles met the inclusion criteria. Through the analysis of these articles, key characteristics of the literature were summarized, and opportunities to enhance the practical application of physical activity (PA), nutrition, and sleep evaluations in the daily lives of older persons were identified. Regular physical activity is crucial for older persons to maintain their physical, mental, and emotional well-being and prevent age-related health issues. Older persons have specific nutritional needs, including increased protein, vitamin D, calcium, and vitamin B12 intake. Poor sleep quality in older persons is associated with negative health outcomes such as cognitive decline, physical disability, and mortality. This review emphasizes the significance of considering physical wellness as a fundamental element for achieving holistic well-being in older persons and highlights the importance of physical activity, nutrition, and sleep evaluations in improving their overall health and well-being. By understanding and implementing these findings, we can enhance the quality of life and promote healthy aging in older persons.

Effects of heat-killed Lactiplantibacillus plantarum TWK10 on exercise performance, fatigue, and muscle growth in healthy male adults.

Consumption of Lactiplantibacillus plantarum TWK10 (TWK10) has beneficial probiotic effects, improves exercise endurance performance, regulates body composition, and mitigates aging-related problems in mice and humans. Here, we investigated the effects of heat-killed TWK10 on exercise endurance performance, muscle weight and strength, fatigue, and body composition in a double-blind, placebo-controlled clinical trial. Thirty healthy males aged 20-40 years were assigned to the Control group or heat-killed TWK10 group (TWK10-HK) in a balanced order according to each individual's initial maximal oxygen uptake. After 6-week administration, the exercise endurance time in the TWK10-HK was significantly increased (p = 0.0028) compared with that in the Control group. The grip strength on the right and left hands of the subjects was significantly increased (p = 0.0002 and p = 0.0140, respectively) in the TWK10-HK compared with that in the Control group. Administration of heat-killed TWK10 resulted in a significant increase (p = 0.0275) in muscle weight. After 6-week administration, serum lactate, and ammonia levels were significantly lower in the TWK10-HK group than in the Control group during the exercise and recovery periods. These findings demonstrate that heat-killed TWK10 has significant potential to be used as a postbiotic for humans.

Consumer perceptions of telehealth for mental health or substance abuse: a Twitter-based topic modeling analysis.

Objective: The objective of this study is to understand the primary topics of consumer discussion on Twitter associated with telehealth for mental health or substance abuse for prepandemic versus during-pandemic time-periods, using a state-of-the-art machine learning (ML) natural language processing (NLP) method. Materials and Methods: The primary methodological phases of this project were: (1) collecting, cleaning, and filtering data (tweets) from January 2014 to June 2021, (2) describing the final corpus, (3) running and optimizing Bidirectional Encoder Representations from Transformers (BERT; using BERTopic in Python) models, and (4) human refinement of topic model results and thematic classification of topics. Results: The number of tweets in this context increased by 4 times during the pandemic (2017 tweets prepandemic vs 8672 tweets during the pandemic). During the pandemic topics were more frequently mental health related than substance abuse related. Top during-pandemic topics were therapy, suicide, pain (associated with burnout and drinking), and mental health diagnoses such as ADHD and autism. Anxiety was a key topic of discussion both pre- and during the pandemic. Discussion: Telehealth for mental health and substance abuse is being discussed more frequently online, which implies growing demand. Given the topics extracted as proxies for demand, the most demand is currently for telehealth for mental health primarily, especially for children, parents, and therapy for those with anxiety or depression, and substance abuse secondarily. Conclusions: Scarce telehealth resources can be allocated more efficiently if topics of consumer discussion are included in resource allocation decision- and policy-making processes.

Telehealth Adoption and Discontinuation by US Hospitals: Results From 2 Quasi-Natural Experiments.

BACKGROUND: Prior US hospital telehealth (video visit) studies have focused on describing factors that influence telehealth adoption or performance effects for specific patient segments, hospital systems, or geographic regions. To our knowledge, a larger-scale, national-level (US) study has yet to be conducted on the causal impacts of hospital telehealth adoption as well as discontinuation. OBJECTIVE: The aim of this study is to understand the causal impact of US hospital telehealth adoption or discontinuation on hospital performance from 2016 to 2018. METHODS: We analyzed impacts of telehealth adoption or discontinuation by US hospitals on emergency department visits, total ambulatory visits (minus emergency department visits), outpatient services revenue, total facility expenses, and total hospital revenue for the 2016-2018 period. We specifically focused on performance effects for hospitals that switched from not having telehealth to adopting telehealth, or vice versa, during the 2016-2018 period, thus exploiting 2 quasi-natural experiments. We applied a difference-in-differences research design to each of the 2 main analyses. We compared hospitals that have made a telehealth change to groups of hospitals with similar characteristics that did not make a telehealth change, which established a counterfactual. To appropriately match hospitals between treatment and control groups, we applied propensity score matching. Our primary data were from the American Hospital Association Annual Survey and the Healthcare Cost Report Information System data. Several control variables were obtained from additional sources, including the Area Health Resource File and the Federal Communications Commission. RESULTS: We found that telehealth adoption by US hospitals during the 2016-2018 period resulted in, on average, an increased number of total ambulatory visits (P=.008), increased total facility expenses (P<.001), and increased hospital revenue (P=.004) compared with the control group. We found that telehealth discontinuation during the same period resulted in, on average, decreased outpatient services revenue (P=.02) compared with the control group. CONCLUSIONS: Our findings suggest that telehealth adoption increases use but has mixed impacts on performance, given that cost and revenue increase. However, once telehealth is offered, removing it can have a negative impact on performance, implying that returning to prior performance levels, if telehealth is removed, may be challenging.

Use of machine learning to examine disparities in completion of substance use disorder treatment.

The objective of this work is to examine disparities in the completion of substance use disorder treatment in the U.S. Our data is from the Treatment Episode Dataset Discharge (TEDS-D) datasets from the U.S. Substance Abuse and Mental Health Services Administration (SAMHSA) for 2017-2019. We apply a two-stage virtual twins model (random forest + decision tree) where, in the first stage (random forest), we determine differences in treatment completion probability associated with race/ethnicity, income source, no co-occurrence of mental health disorders, gender (biological), no health insurance, veteran status, age, and primary substance (alcohol or opioid). In the second stage (decision tree), we identify subgroups associated with probability differences, where such subgroups are more or less likely to complete treatment. We find the subgroups most likely to complete substance use disorder treatment, when the subgroup represents more than 1% of the sample, are those with no mental health condition co-occurrence (4.8% more likely when discharged from an ambulatory outpatient treatment program, representing 62% of the sample; and 10% more likely for one of the more specifically defined subgroups representing 10% of the sample), an income source of job-related wages/salary (4.3% more likely when not having used in the 30 days primary to discharge and when primary substance is not alcohol only, representing 28% of the sample), and white non-Hispanics (2.7% more likely when discharged from residential long-term treatment, representing 9% of the sample). Important implications are that: 1) those without a co-occurring mental health condition are the most likely to complete treatment, 2) those with job related wages or income are more likely to complete treatment, and 3) racial/ethnicity disparities persist in favor of white non-Hispanic individuals seeking to complete treatment. Thus, additional resources may be needed to combat such disparities.