publications

2025

1. States With Substantial Increases In Buprenorphine Uptake Did So With Increased Medicaid Prescribing, 2018–24

   Stephen Crystal, Fangzhou Xie, Hillary Samples, and 6 more authors

   Health Affairs, Sep 2025

   Abs

   

   Multiple federal policy changes since 2018 have been intended to increase buprenorphine prescribing in response to a persistent treatment gap for opioid use disorder in the US. Anticipated national increases did not occur, but highly variable state-level trends provide important insights. We used IQVIA data to examine all-payer and per payer prescribing across states during the period 2018–24. All-payer prescriptions per 1,000 residents increased by more than 60 percent in ten states while decreasing by more than 10 percent in eight states. States expanding Medicaid during this period increased all-payer prescribing by 27.3 percent; those that had expanded earlier increased prescribing by 11.6 percent, whereas in nonexpansion states, prescribing declined by 2.1 percent. All-payer rates ranged thirty-eight-fold in 2024 (varying from 9.3 to 355.5 per 1,000; lowest quintile, 15.7 per 1,000; highest quintile, 118.6 per 1,000). Medicaid prescribing changes were key drivers of all-payer changes. In 2024, Medicaid prescribing ranged from 0.50 per 1,000 to 217.26 per 1,000, and from 3.6 percent to 66.8 percent of all-payer buprenorphine prescriptions. The self-pay proportion ranged from 0.7 percent to 30.8 percent, and this proportion was strongly associated with the Medicaid proportion. Highly disparate state-level changes suggest that federal policy impacts were mediated by state-specific factors. Medicaid’s key role in driving overall prescribing highlights the public health urgency of maintaining expansions and sustaining enrollment for the single adult population.

2. Antipsychotic Medication Use by Older Adults

   Mark Olfson, Fangzhou Xie, Greta Bushnell, and 3 more authors

   JAMA Psychiatry, Dec 2025

   Abs

   

   This cross-sectional study examines trends in antipsychotic medication use among adults aged 65 years or older in the US.

3. Buprenorphine Dispensation After X-Waiver Elimination by Clinician Specialty

   Elizabeth M. Stone, Fangzhou Xie, Jennifer Miles, and 3 more authors

   American Journal of Preventive Medicine, Nov 2025

   Abs

   

   Introduction Elimination of the X-waiver, which required clinicians to complete additional registration to prescribe buprenorphine for opioid use disorder, removed one barrier to treatment. This study examined the association of the X-waiver elimination with buprenorphine dispensations by clinician specialty. Methods Using IQVIA Longitudinal Prescription data, patients with 15 or more days of dispensed buprenorphine supply each month from May 2021 to December 2024 were identified. Interrupted time series analyses (conducted in 2025) examined changes in monthly counts of clinicians associated with dispensations and patients, overall and stratified by clinician specialty. Results During the study period, 189,771 clinicians dispensed buprenorphine to 2,699,441 patients. X-waiver elimination was associated with significant increases in the number of clinicians associated with dispensed buprenorphine prescriptions overall (change in level= 1,626 clinicians; 95% CI=577, 2,674; p<0.01; change in slope: 15 clinicians per month, 95% CI=13, 18, p<0.001) and across all specialties. X-waiver elimination was associated with a decrease in the number of patients with buprenorphine dispensations in January 2023 overall (change in level= -24,104 patients; 95% CI= -40,010, -8,198; p<0.01) and from all clinician groups except behavioral health physicians. Decreasing monthly rates of patients with buprenorphine dispensed by behavioral health physicians slowed after X-waiver elimination; monthly rates of buprenorphine patients with dispensations from primary care providers increased after (versus before) the policy change. Conclusions Although the number of clinicians associated with dispensed buprenorphine prescriptions after X-waiver elimination increased across all clinician types, patient-level gains associated with X-waiver elimination were limited.

4. Deriving the Gradients of Some Popular Optimal Transport Algorithms

   Fangzhou Xie

   Oct 2025

   Abs

   

   In this note, I review entropy-regularized Monge-Kantorovich problem in Optimal Transport, and derive the gradients of several popular algorithms popular in Computational Optimal Transport, including the Sinkhorn algorithms, Wasserstein Barycenter algorithms, and the Wasserstein Dictionary Learning algorithms.

2022

1. Rethnicity: An R Package for Predicting Ethnicity from Names

   Fangzhou Xie

   SoftwareX, Jan 2022

   Abs

   

   In this study, a new R package, rethnicity 11https://github.com/fangzhou-xie/rethnicity. It has also been published on [CRAN]. is provided for predicting ethnicity based on names. The Bidirectional Long Short-Term Memory (Bi-LSTM), a recurrent neural network architecture commonly used for natural language processing, was chosen as the model for our study. The Florida Voter Registration was used as the training and testing data. Special care was given for the accuracy of minority groups by adjusting the imbalance in the dataset. The models were trained and exported to C++ and then integrated with R using Rcpp. Additionally, the availability, accuracy, and performance of the package were compared with other solutions.

2020

1. Pruned Wasserstein Index Generation Model and Wigpy Package

   Fangzhou Xie

   arXiv:2004.00999 [cs, econ, q-fin], Jul 2020

   Abs

   

   Recent proposal of Wasserstein Index Generation model (WIG) has shown a new direction for automatically generating indices. However, it is challenging in practice to fit large datasets for two reasons. First, the Sinkhorn distance is notoriously expensive to compute and suffers from dimensionality severely. Second, it requires to compute a full \N\times N matrix to be fit into memory, where \N is the dimension of vocabulary. When the dimensionality is too large, it is even impossible to compute at all. I hereby propose a Lasso-based shrinkage method to reduce dimensionality for the vocabulary as a pre-processing step prior to fitting the WIG model. After we get the word embedding from Word2Vec model, we could cluster these high-dimensional vectors by \k\-means clustering, and pick most frequent tokens within each cluster to form the "base vocabulary". Non-base tokens are then regressed on the vectors of base token to get a transformation weight and we could thus represent the whole vocabulary by only the "base tokens". This variant, called pruned WIG (pWIG), will enable us to shrink vocabulary dimension at will but could still achieve high accuracy. I also provide a \textit{wigpy} module in Python to carry out computation in both flavor. Application to Economic Policy Uncertainty (EPU) index is showcased as comparison with existing methods of generating time-series sentiment indices.

2. Wasserstein Index Generation Model: Automatic Generation of Time-Series Index with Application to Economic Policy Uncertainty

   Fangzhou Xie

   Economics Letters, Jan 2020

   Abs

   

   I propose a novel method, the Wasserstein Index Generation model (WIG), to generate a public sentiment index automatically. To test the model’s effectiveness, an application to generate Economic Policy Uncertainty (EPU) index is showcased.