David Shilane’s research while affiliated with Columbia University and other places

This study investigates how effectively telehealth utilization (THU) can substitute for a usual source of care (USC) for taking prescription medication using data from the 2020-2022 National Health Interview Survey. We analyzed data for 69,581 patients. Of these, 5,994 patients (8.6%) lacked a USC. THU was 37.3% for patients with a USC and 15.8% for those without. For patients with no USC or THU, 25.4% had taken a prescription medication within 12 months, while patients with THU but no USC had a rate of 75.4%. In essentially all subgroups, telehealth was associated with substantially higher rates of taking prescription medications. Multivariate logistic regression showed that THU was associated with a 7.39-fold increase (95% CI: 6.19-8.84) in the odds of taking a prescription medication.

Older age is a potentially confounding variable in models of telehealth utilization. We compared unified and stratified logistic regression models using data from the 2021 National Health Interview Survey. A total of 27,626 patients were identified, of whom 38.9% had utilized telehealth. Unified and stratified modeling showed a number of important differences in their quantitative estimates, especially for gender, Hispanic ethnicity, heart disease, COPD, food allergies, high cholesterol, weak or failing kidneys, liver conditions, difficulty with self-care, the use of mobility equipment, health problems that limit the ability to work, problems paying bills, and filling a recent prescription. Telehealth utilization odds ratios differ meaningfully between younger and older patients in stratified modeling. Traditional statistical adjustments in logistic regression may not sufficiently account for the confounding influence of older age in models of telehealth utilization. Stratified modeling by age may be more effective in obtaining clinical inferences.

This study enumerates and compares the risks and rewards of different forms of statistical consulting contracts. We assess three different contract models: project‐based fees, hourly fees, and retainer agreements and three different planned durations: project‐based, time‐based, and evergreen contracts. The requirements of time and effort vary considerably for many aspects of consulting work. The risks of statistical consulting contracts include both the general risks of consulting projects along with the specialized risks of statistical investigations. We enumerate a number of general risks in the categories of unanticipated developments, revisions and collaboration, and changing scopes of projects. Meanwhile, the specialized statistical risks include issues of study design, data quality, statistical investigation, and communication of statistical issues. Because of these concerns, the specialized risks of statistical investigations add considerably to the general risks of consulting projects. Moreover, these issues can be exacerbated or mitigated by the form of the consulting agreement. With a greater understanding of the risks and benefits of each type of contract, statistical consultants and clients can negotiate more mutually beneficial contracts for either or both parties. Through this discussion, we hope to raise awareness of these issues and help to create working conditions with a greater likelihood of a successful project for both statistical consultants and their clients.

Educational curricula in data analysis are increasingly fundamental to statistics, data science, and a wide range of disciplines. The educational literature comparing coding syntaxes for instruction in data analysis recommends utilizing a simple syntax for introductory coursework. However, there is limited prior work to assess the pedagogical elements of coding syntaxes. The study investigates the paradigms of the dplyr, data.table, and DTwrappers packages for R programming from a pedagogical perspective. We enumerate the pedagogical elements of computer programming that are inherent to utilizing each package, including the functions, operators, general knowledge, and specialized knowledge. The merits of each package are also considered in concert with other pedagogical goals, such as computational efficiency and extensions to future coursework. The pedagogical considerations of this study can help instructors make informed choices about their curriculum and how best to teach their selected methods.

Introduction Telehealth is an important complement to in-person healthcare, with opportunities to overcome barriers to healthcare and improve health equity. Telehealth utilization increased sharply at the onset of the COVID-19 pandemic. This study assessed trends in telehealth utilization for the years 2020 through 2022, both overall and in subgroups. Methods We gathered data from the 2020–2022 National Health Interview Survey. The rates of telehealth utilization were calculated overall and within subgroups based on demographic factors, health conditions, healthcare utilization, challenges of ability, and social needs. Adjusted multivariable logistic regression models of telehealth utilization assessed the yearly trends. We also analyzed the ratios of subgroup utilization relative to the overall rates. Results A total of 69,581 patients were identified with complete information. The overall rates of telehealth utilization were 36.2% for 2020, 38.9% for 2021, and 31.3% for 2022. The reduction in telehealth utilization for 2022 was large and statistically significant (OR: 0.64 (95% CI: 0.62–0.67), p < 0.001). Subgroup analyses showed corresponding reductions in 2022 for essentially all patient subgroups. Telehealth was utilized at higher rates by patients with chronic conditions, challenges of ability, and other kinds of medical utilization. Ratio analyses showed evidence of widening disparities for patients of older age, in rural areas, and by geographic region, limited education, and of low income. Discussion The study demonstrates declining rates of telehealth utilization are occurring with widening gaps among patient subgroups. Addressing these disparities may be critical to improving equity in telehealth and healthcare overall.

Objective: This study assessed barriers and facilitators to telehealth utilization among patients living in New York City public housing with chronic conditions and a gap in clinical care. Methods: Community health workers performed outreach to eligible patients by telephone between January and March 2021. Consenting respondents answered questions about telehealth barriers, including internet and cell phone access, ownership of digital devices, comfort with using digital devices, comfort with telehealth, cost, awareness, and availability of written materials in patients' preferred language. We obtained demographic and medical information from patients' electronic health records. We used multivariable logistic regression to estimate the association of barriers with the odds of self-reported prior telehealth utilization. Results: A total of 304 consenting patients participated in the program. The average patient had 3.1 telehealth barriers; 76% reported at least one barrier. Regression analysis showed sizable reductions in prior telehealth utilization associated with the barriers of unlimited cell phone minutes (odds ratio [OR]: 0.21 [0.05-0.88], p = 0.033), technological comfort (OR: 0.33 [0.13-0.82], p = 0.016), conceptual comfort with telehealth (OR: 0.15 [0.04-0.54], p = 0.004), and materials in the patient's preferred language (OR: 0.23 [0.07-0.79], p = 0.02). Discussion: With a high prevalence of telehealth barriers, patients with limited income, a chronic condition, and a care gap may benefit from greater technological access and supportive programs for awareness, telehealth comfort, and navigation support. Addressing telehealth barriers could increase the quality of medical care and improve health outcomes for this population.