We present motivation and new Stata commands for modeling count data. While the focus of this article is on modeling data with underdispersion, the new command for fitting generalized Poisson regression models is also suitable as an alternative to negative binomial regression for overdispersed data.

We present new Stata commands for carrying out exact Wilcoxon one-sample and two-sample comparisons of the median. Nonparametric tests are often used in clinical trials, in which it is not uncommon to have small samples. In such situations, researchers are accustomed to making inferences by using exact statistics. The ranksum and signrank commands in Stata provide only asymptotic results, which assume normality. Because large-sample results are unacceptable in many clinical trials studies, these researchers must use other software packages. To address this, we have developed new commands for Stata that provide exact statistics in small samples. Additionally, when samples are large, we provide results based on the Student's t distribution that outperform those based on the normal distribution.

We present motivation and new commands for modeling count data. While our focus is to present new commands for estimating count data, we also discuss generalized binomial regression and present the zero-inflated versions of each model.

Oral factor Xa (FXa) inhibitors significantly reduce incidence of stroke and thromboembolic events in patients with atrial fibrillation or venous thromboembolism. Due to various factors and the lack of a randomized controlled trial comparing andexanet alfa to usual care, non-specific replacement agents including 4 F-PCC are still used off-label for FXa inhibitor bleed management. Clinical and mortality data were extracted from the inpatient medical data and Veteran Affairs (VA) vital status files over the time of March 2014 through December 2020. Propensity score-weighted models were used for this retrospective cohort study using data from the Veterans Affairs Informatics and Computing Infrastructure (VINCI). The study included 255 patients (85-andexanet alfa and 170-4 F-PCC) exposed to an oral factor Xa inhibitor and hospitalized with an acute major, gastrointestinal (GI), intracranial (ICH) or other bleed. In-hospital mortality was significantly lower in the andexanet alfa cohort compared to the 4 F-PCC cohort (10.6% vs. 25.3%, p = 0.01). Propensity score–weighted Cox models reveal a 69% lower hazard of in-hospital mortality for those treated with andexanet alfa (HR 0.31, 95% CI 0.14–0.71) compared to those treated with 4 F-PCC. Additionally, those treated with andexanet alfa had a lower 30-day mortality rate and lower 30-day hazard of mortality in the weighted Cox model (20.0% vs. 32.4%, p = 0.039; HR 0.54, 95% CI 0.30–0.98) compared to those treated with 4 F-PCC. Among 255 US veterans with major bleeding in the presence of an oral factor Xa inhibitor, treatment with andexanet alfa was associated with lower in-hospital and 30-day mortality than treatment with 4 F-PCC.

Background: Obesity is a risk factor for the development of influenza by leading to a chronic inflammatory state and T-cell dysfunction. Based upon preclinical research, metformin has influenza activity by restoring T-cell function and improving the immune response. Objective: We aimed to evaluate the potential drug repurposing of metformin for the management of influenza among patients with obesity utilizing national claims data in an electronic health record database. Methods: The VA Informatics and Computing Infrastructure (VINCI) was utilized to obtain individual-level information on demographics, administrative claims, and pharmacy dispensation. A cohort was created among individuals with laboratory confirmed diagnosis of influenza with a diagnosis of fever, cough, influenza, or acute upper respiratory infection in an outpatient setting. The study outcome was death after diagnosis of influenza. Cohorts were formed using diabetes status and metformin exposure prior to a positive influenza diagnosis. Hazard ratios for mortality were estimated using a cox proportional hazards model adjusting for baseline covariates and a sub-analysis was conducted utilizing propensity score matching. A greedy nearest neighbor algorithm was utilized to match 1 to 1 non-metformin diabetic controls and non-diabetic controls to diabetic patients receiving metformin. Results: A total of 3551 patients met the inclusion criteria and were evaluated in our study. The cohorts consisted of 1461 patients in the non-diabetic cohort, 1597 patients in the diabetic / metformin cohort, and 493 patients in the diabetic no metformin cohort. Compared to non-diabetic patients, diabetic patients with metformin had a lower rate of death (aHR 0.78, 95% CI 0.609–0.999). There was not a statistical difference between the non-diabetic patients and the diabetic patients without metformin (aHR 1.046, 95% CI 0.781–1.400). The propensity score matched cohorts revealed consistent results with the primary analysis. Conclusion: Our results demonstrated patients with obesity and a history of metformin treatment have lower influenza mortality.