Scalable Incident Detection via Natural Language Processing and Probabilistic Language Models.

Post marketing safety surveillance depends in part on the ability to detect concerning clinical events at scale. Spontaneous reporting might be an effective component of safety surveillance, but it requires awareness and understanding among healthcare professionals to achieve its potential. Reliance on readily available structured data such as diagnostic codes risk under-coding and imprecision. Clinical textual data might bridge these gaps, and natural language processing (NLP) has been shown to aid in scalable phenotyping across healthcare records in multiple clinical domains. In this study, we developed and validated a novel incident phenotyping approach using unstructured clinical textual data agnostic to Electronic Health Record (EHR) and note type. It's based on a published, validated approach (PheRe) used to ascertain social determinants of health and suicidality across entire healthcare records. To demonstrate generalizability, we validated this approach on two separate phenotypes that share common challenges with respect to accurate ascertainment: 1) suicide attempt; 2) sleep-related behaviors. With samples of 89,428 records and 35,863 records for suicide attempt and sleep-related behaviors, respectively, we conducted silver standard (diagnostic coding) and gold standard (manual chart review) validation. We showed Area Under the Precision-Recall Curve of ∼ 0.77 (95% CI 0.75-0.78) for suicide attempt and AUPR ∼ 0.31 (95% CI 0.28-0.34) for sleep-related behaviors. We also evaluated performance by coded race and demonstrated differences in performance by race were dissimilar across phenotypes and require algorithmovigilance and debiasing prior to implementation.

Switching patterns of immediate-release forms of generic mixed amphetamine salts products among privately and publicly insured individuals aged 15-64 years in the United States, 2013-2019.

PURPOSE: Immediate-release forms of generic mixed amphetamine salts (MAS) have been the subject of passive surveillance reports signaling lack of effectiveness. We examined switching patterns that might suggest whether long-term users of specific MAS are more likely to switch away or switch back after use of the MAS of interest in the FDA's Sentinel Distributed Database. METHODS: We required at least 60-day continuous supply of selected MAS grouped by Abbreviated New Drug Application (ANDA) to describe patterns of switching away from and to generics approved under the ANDAs of interest among individuals ages 15-64 years with attention deficit hyperactivity disorder or narcolepsy during 2013-2019. RESULTS: We observed the greatest number of treatment episodes for ANDA 040422 (n = 525 771), followed by ANDA 202424 (n = 181 693), ANDA 040439 (n = 62 363), ANDA 040440 (n = 21 143), and ANDA 040480 (n = 8792). Of those with switches away from their original ANDA, episodes initiated on generic products under ANDA 040422 (48.6%) and ANDA 202424 (43.0%) were most likely to switch back, while those initiated on generic product under ANDA 040480 were least likely (24.1%). Of those episodes with switches to a generic under an ANDA of interest, about one-third (range 27.1% to 37.0%) switched back to the same product. These switches back had a median time to switch of about 30 days. CONCLUSIONS: These descriptive analyses, although subject to limitations, did not suggest increased switching away or switching back after use of the generics of interest. Continued post-marketing surveillance is warranted.

A COVID-19-ready public health surveillance system: The Food and Drug Administration's Sentinel System.

The US Food and Drug Administration's Sentinel System was established in 2009 to use routinely collected electronic health data for improving the national capability to assess post-market medical product safety. Over more than a decade, Sentinel has become an integral part of FDA's surveillance capabilities and has been used to conduct analyses that have contributed to regulatory decisions. FDA's role in the COVID-19 pandemic response has necessitated an expansion and enhancement of Sentinel. Here we describe how the Sentinel System has supported FDA's response to the COVID-19 pandemic. We highlight new capabilities developed, key data generated to date, and lessons learned, particularly with respect to working with inpatient electronic health record data. Early in the pandemic, Sentinel developed a multi-pronged approach to support FDA's anticipated data and analytic needs. It incorporated new data sources, created a rapidly refreshed database, developed protocols to assess the natural history of COVID-19, validated a diagnosis-code based algorithm for identifying patients with COVID-19 in administrative claims data, and coordinated with other national and international initiatives. Sentinel is poised to answer important questions about the natural history of COVID-19 and is positioned to use this information to study the use, safety, and potentially the effectiveness of medical products used for COVID-19 prevention and treatment.

Mediation analyses of socioeconomic factors determining racial differences in the treatment of diffuse large B-cell lymphoma in a cohort of older adults.

Despite near universal health coverage under Medicare, racial disparities persist in the treatment of diffuse large B-cell lymphoma (DLBCL) among older patients in the United States. Studies evaluating DLBCL outcomes often treat socioeconomic status (SES) measures as confounders, potentially introducing biases when SES factors are mediators of disparities in cancer treatment.To examine differences in DLBCL treatment, we performed causal mediation analyses of SES measures, including: metropolitan statistical area (MSA) of residence; census-tract poverty level; and private Medicare supplementation using the Surveillance, Epidemiology and End Results-Medicare linked database between 2001 and 2011. In this retrospective cohort study of DLBCL patients ages 66+ years, we conducted a series of multivariable logistic regression analyses estimating odds ratios (OR) and 95% confidence intervals (CI) relating chemo- and/or immuno-therapy treatment and each SES measure, comparing non-Hispanic (NH)-black, Hispanic/Latino, and Asian/Pacific Islander (API) to NH-white patients.Compared to NH-white patients, racial/ethnic minority patients had lower odds of receiving chemo- and/or immuno-therapy treatment (NH-black: OR 0.84, 95% CI 0.65, 1.08; API: OR 0.80, 95% CI 0.64, 1.01; Hispanic/Latino: OR 0.78, 95% CI 0.64, 0.96) and higher odds of lacking private Medicare supplementation and residence within an urban MSA and poor census tracts. Adjustment for SES measures as confounders nullified observed racial differences. In causal mediation analyses, between 31% and 38% of race/ethnicity differences were mediated by having private Medicare supplementation.Providing equitable access to Medicare supplementation may reduce disparities in receipt of chemo- and/or immuno-therapy treatment in older DLBCL patients.

Bevacizumab Use and the Risk of Arterial and Venous Thromboembolism in Patients with High-Grade Gliomas: A Nested Case-Control Study.

STUDY OBJECTIVE: Bevacizumab is used in the treatment of recurrent glioblastoma, but evidence is limited on the incidence of thromboembolic complications regarding the use of this drug in real-world settings. We evaluated the risk of arterial thromboembolism (ATE) and venous thromboembolism (VTE) associated with the use of bevacizumab among adults diagnosed with high-grade gliomas in a commercially insured U.S. DESIGN: Nested case-control study. DATA SOURCE: Truven Health MarketScan Commercial and Medicare Supplemental health claims databases (2009-2015). PATIENTS: A total of 2157 patients with high-grade gliomas who underwent incident (first-time) craniotomy, radiation, and concurrent temozolomide treatment between 2009 and 2015 were identified. Overall, 25 cases of ATE and 99 cases of VTE were each identified in this cohort, and each case was matched to up to 10 controls (170 for ATE and 819 for VTE) based on sex, age, quarter year of index time, and follow-up duration by using incidence density sampling without replacement from the overall cohort. Controls were at risk for the outcome of interest (ATE or VTE) at the time of case occurrence and survived at least as long as their referent case. MEASUREMENTS AND MAIN RESULTS: Exposure to bevacizumab was determined during inpatient or outpatient encounters between the index date (date of the incident craniotomy) and the ATE or VTE event or corresponding matched control date. Multivariable conditional logistic regression models were used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the risk of ATE and VTE separately. A higher proportion of patients with ATE received bevacizumab compared with controls (28% vs 17%; adjusted OR 1.51, 95% CI 0.54-4.24), but this excess in odds was not statistically significant. Similarly, bevacizumab was not significantly associated with VTE (13% vs 9%; adjusted OR 1.40, 95% CI 0.71-2.75). CONCLUSION: We found no significant association between the use of bevacizumab and the occurrence of thromboembolic events in patients with high-grade gliomas, although our study was limited by the small number of ATE events. Because the potential for complications from arterial thrombosis cannot be completely ruled out, further research is needed to confirm the thromboembolic safety of bevacizumab in a larger sample of patients with high-grade gliomas.

Myelodysplastic syndrome and acute myeloid leukemia after receipt of granulocyte colony-stimulating factors in older patients with non-Hodgkin lymphoma.

BACKGROUND: Granulocyte colony-stimulating factors (G-CSFs), which are used for the prevention of complications from chemotherapy-related neutropenia, are linked to the risk of developing second primary myelodysplastic syndrome (MDS) and acute myeloid leukemia (AML). The objective of this study was to examine the correlation between using a specific G-CSF agent and the risk of MDS/AML among older patients with non-Hodgkin lymphoma (NHL). METHODS: This was a retrospective cohort study of adults aged >65 years who were diagnosed with first primary NHL between 2001 and 2011. With data from the Surveillance, Epidemiology, and End Results-Medicare-linked database, adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated for the risk of MDS/AML associated with the receipt of G-CSF(filgrastim and pegfilgrastim) in Cox proportional-hazards models, which were stratified according to treatment accounting for confounding by indication. RESULTS: Among 18,245 patients with NHL patients who had a median follow-up of 3.5 years, 56% received chemotherapy and/or immunotherapy, and G-CSF was most commonly used in those who received rituximab plus multiple chemotherapy regimens (77%). Subsequent MDS/AML diagnoses were identified in 666 patients (3.7%). A modest increased risk of MDS/AML was observed with the receipt of G-CSF (HR, 1.28; 95% CI, 1.01-1.62) and a trend was observed with increasing doses (Ptrend < .01). When specific agents were analyzed, an increased risk of MDS/AML was consistently observed with filgrastim (≥10 doses: HR, 1.67; 95% CI, 1.25-2.23), but not with pegfilgrastim (≥10 + doses: HR, 1.11; 95% CI, 0.84-1.45). CONCLUSIONS: A higher of MDS/AML was observed in patients with NHL risk among those who received G-CSF that was specific to the use of filgrastim (≥10 doses), but not pegfilgrastim. Neutropenia prophylaxis is an essential component of highly effective NHL treatment regimens. The differential risk related to the types of G-CSF agents used warrants further study given their increasing use and newly available, US Food and Drug Administration-approved, biosimilar products.