Introduction: To identify predictors of hypoglycemia and five other clinical and economic outcomes among treated patients with type 2 diabetes (T2D) using machine learning and structured data from a large, geographically diverse administrative claims database. Methods: A retrospective cohort study design was applied to Optum Clinformatics claims data indexed on first antidiabetic prescription date. A hypothesis-free, Bayesian machine learning analytics platform (GNS Healthcare REFS TM : Reverse Engineering and Forward Simulation) was used to build ensembles of generalized linear models to predict six outcomes defined in patients' 1-year post-index claims history, including hypoglycemia, antidiabetic class persistence, glycated hemoglobin (HbA1c) target Enhanced Digital Features To view enhanced digital features for this article go to https://doi.org/10.6084/ m9.figshare.11527902.

No abstract

IMPORTANCE Colorectal cancer (CRC) screening reduces CRC incidence and mortality. It is important to examine screening patterns over time, including after the introduction of new screening modalities.OBJECTIVE To compare use of CRC screening tests before and after the availability of the multitarget stool DNA (mt-sDNA) test, given that endorsed options have changed.

This economic evaluation examines whether adult patients in the US who have commercial or Medicare insurance pay out-of-pocket costs associated with follow-up colonoscopy within 6 months of a noninvasive stool-based test.

ImportanceNoninvasive stool-based screening tests (SBTs) are effective alternatives to colonoscopy. However, a positive SBT result requires timely follow-up colonoscopy (FU-CY) to complete the colorectal cancer screening paradigm.ObjectivesTo evaluate FU-CY rates after a positive SBT result and to assess the association of the early COVID-19 pandemic with FU-CY rates.Design, Setting, and ParticipantsThis mixed-methods cohort study included retrospective analysis of deidentified administrative claims and electronic health records data between June 1, 2015, and June 30, 2021, from the Optum Labs Data Warehouse and qualitative, semistructured interviews with clinicians from 5 health care organizations (HCOs). The study population included data from average-risk primary care patients aged 50 to 75 years with a positive SBT result between January 1, 2017, and June 30, 2020, at 39 HCOs.Main Outcomes and MeasuresThe primary outcome was the FU-CY rate within 1 year of a positive SBT result according to patient age, sex, race, ethnicity, insurance type, Charlson Comorbidity Index (CCI), and prior SBT use.ResultsThis cohort study included 32 769 individuals (16 929 [51.7%] female; mean [SD] age, 63.1 [7.1] years; 2092 [6.4%] of Black and 28 832 [88.0%] of White race; and 825 [2.5%] of Hispanic ethnicity). The FU-CY rates were 43.3% within 90 days of the positive SBT result, 51.4% within 180 days, and 56.1% within 360 days (n = 32 769). In interviews, clinicians were uniformly surprised by the low FU-CY rates. Rates varied by race, ethnicity, insurance type, presence of comorbidities, and SBT used. In the Cox proportional hazards regression model, the strongest positive association was with multitarget stool DNA use (hazard ratio, 1.63 [95% CI, 1.57-1.68] relative to fecal immunochemical tests; P < .001), and the strongest negative association was with the presence of comorbidities (hazard ratio, 0.64 [95% CI, 0.59-0.71] for a CCI of >4 relative to 0; P < .001). The early COVID-19 pandemic was associated with lower FU-CY rates.Conclusions and RelevanceThis study found that FU-CY rates after a positive SBT result for colorectal cancer screening were low among an average-risk population, with the median HCO achieving a 53.4% FU-CY rate within 1 year. Socioeconomic factors and the COVID-19 pandemic were associated with lower FU-CY rates, presenting opportunities for targeted intervention by clinicians and health care systems.