Evidence and guidelines do not support use of systemic steroids for acute respiratory tract infections (ARTIs), but such practice appears common. We aim to quantify such use and determine its ...predictors.
We conducted a cohort study based on a large United States national commercial claims database, the IBM MarketScan, to identify patients aged 18-64 years with an ARTI diagnosis (acute bronchitis, sinusitis, pharyngitis, otitis media, allergic rhinitis, influenza, pneumonia, and unspecified upper respiratory infections) recorded in ambulatory visits from 2007 to 2016. We excluded those with systemic steroid use in the prior year and an extensive list of steroid-indicated conditions, including asthma, chronic obstructive pulmonary disease, and various autoimmune diseases. We calculated the proportion receiving systemic steroids within 7 days of the ARTI diagnosis and determined its significant predictors. We identified 9,763,710 patients with an eligible ARTI encounter (mean age 39.6, female 56.0%) and found 11.8% were prescribed systemic steroids (46.1% parenteral, 47.3% oral, 6.6% both). All ARTI diagnoses but influenza predicted receiving systemic steroids. There was high geographical variability: the adjusted odds ratio (aOR) of receiving parenteral steroids was 14.48 (95% confidence interval CI 14.23-14.72, p < 0.001) comparing southern versus northeastern US. The corresponding aOR was 1.68 (95% CI 1.66-1.69, p < 0.001) for oral steroids. Other positive predictors for prescribing included emergency department (ED) or urgent care settings (versus regular office), otolaryngologist/ED doctors (versus primary care), fewer comorbidities, and older patient age. There was an increasing trend from 2007 to 2016 (aOR 1.93 95% CI 1.91-1.95 comparing 2016 to 2007, p < 0.001). Our findings are based on patients between 18 and 64 years old with commercial medical insurance and may not be generalizable to older or uninsured populations.
In this study, we found that systemic steroid use in ARTI is common with a great geographical variability. These findings call for an effective education program about this practice, which does not have a clear clinical net benefit.
Several sources of bias can affect the performance of machine learning systems used in medicine and potentially impact clinical care. Here, we discuss solutions to mitigate bias across the different ...development steps of machine learning-based systems for medical applications.
To determine the impact of electronic health record (EHR)-discontinuity on the performance of prediction models.
The study population consisted of patients with a history of cardiovascular (CV) ...comorbidities identified using US Medicare claims data from 2007 to 2017, linked to EHR from two networks (used as model training and validation set, respectively). We built models predicting one-year risk of mortality, major CV events, and major bleeding events, stratified by high vs. low algorithm-predicted EHR-continuity. The best-performing models for each outcome were chosen among 5 commonly used machine-learning models. We compared model performance by Area under the ROC curve (AUROC) and Area under the precision-recall curve (AUPRC).
Based on 180,950 in the training and 103,061 in the validation set, we found EHR captured only 21.0-28.1% of all the non-fatal outcomes in the low EHR-continuity cohort but 55.4-66.1% of that in the high EHR-continuity cohort. In the validation set, the best-performing model developed among high EHR-continuity patients had consistently higher AUROC than that based on low-continuity patients: AUROC was 0.849 vs. 0.743 when predicting mortality; AUROC was 0.802 vs. 0.659 predicting the CV events; AUROC was 0.635 vs. 0.567 predicting major bleeding. We observed a similar pattern when using AUPRC as the outcome metric.
Among patients with CV comorbidities, when predicting mortality, major CV events, and bleeding outcomes, the prediction models developed in datasets with low EHR-continuity consistently had worse performance compared to models developed with high EHR-continuity.
AbstractObjectiveTo determine the extent to which late stage development of new drugs relies on support from public funding.DesignCohort study.SettingAll new drugs containing one or more new ...molecular entities approved by the US Food and Drug Administration (FDA) between January 2008 and December 2017 via the new drug application pathway.Main outcome measuresPatents or drug development histories documenting late stage research contributions by a public sector research institution or a spin-off company, as well as each drug’s regulatory approval pathway and first-in-class designation.ResultsOver the 10 year study period, the FDA approved 248 drugs containing one or more new molecular entities. Of these drugs, 48 (19%) had origins in publicly supported research and development and 14 (6%) originated in companies spun off from a publicly supported research program. Drugs in these groups were more likely to receive expedited FDA approval (68% v 47%, P=0.005) or be designated first in class (45% v 26%, P=0.007), indicating therapeutic importance.ConclusionsA review of the patents associated with new drugs approved over the past decade indicates that publicly supported research had a major role in the late stage development of at least one in four new drugs, either through direct funding of late stage research or through spin-off companies created from public sector research institutions. These findings could have implications for policy makers in determining fair prices and revenue flows for these products.
The US government supports drug innovation. It is therefore crucial that it distinguish between high-value and low-value innovation in purchasing expensive prescription drugs and medical devices and ...ensure the continued discovery of transformative drugs and that patient and taxpayer funds are not wasted.