Pragmatic and Observational Research最新文献

Background: The National Death Index (NDI) is the gold standard for mortality data in the United States (US) but has a time lag and can be operationally intensive. This validation study assesses the accuracy of various mortality data sources with the NDI.

Methods: This validation study is a secondary analysis of an advanced cancer cohort in the US between January 2010 and December 2018, with an established NDI linkage. Mortality data sources, inpatient discharge, disenrollment, death master file (DMF), Center for Medicare and Medicaid Services (CMS), Utilization management data (U.M.), and online obituary data were compared to NDI.

Results: Among 40,692 patients, 25,761 (63.3%) had a death date using NDI; the composite algorithm had a sensitivity of 88.9% (95% CI = 88.5%, 89.3%), specificity was 89.1% (95% CI = 88.6%, 89.6%). At the same time, positive predictive value (PPV) was 93.4% (95% CI = 93.1%, 93.7%), negative predictive value (NPV) was 82.3% (95% CI = 81.7%, 82.9%), and when comparing each individual source, each had a high PPV but limited sensitivity.

Conclusion: The composite algorithm was demonstrated to be a sensitive and precise measure of mortality, while individual database sources were accurate but had limited sensitivity.

In recent years, regulatory authorities have signaled a willingness to consider real-world evidence (RWE) data to support applications for new claims and indications for pharmaceuticals. Historically, RWE studies have been the domain of prescription drugs, driven by the fact that clinical data on patients are routinely captured in medical records, claims databases, registries, etc. However, RWE reports of nonprescription drugs and supplements are relatively sparse due to methodological gaps in this area. The objective of this narrative review is to identify which RWE methodologies have been used to study nonprescription products. A total of 49 articles were included based on literature searches. Label comprehension studies, used to support prescription-to-nonprescription switches, are useful in determining how nonprescription products will be used; however, they provide no actual clinical data. The most common RWE studies of nonprescription products were cross-sectional surveys, which investigated a broad range of indications and were conducted in an array of settings, including online, by phone, point-of-sale (pharmacy), outpatient clinics, and shopping malls. However, while this type of study is effective for identifying use patterns and attitudes in the general population, recall bias limits the ability to collect safety and effectiveness data. Studies of electronic medical records and claims databases are hampered by incomplete or absent capturing of data on nonprescription products. As a result, most RWE studies to date have provided limited useful information. Although case reports and expert opinion should not be discounted, in the absence of other information they provide few actual data. Novel approaches using smartphone apps and artificial intelligence may provide new opportunities to collect RWE for nonprescription products, but these areas of research are in their infancy. Overall, there is a need to develop standards for execution of RWE studies of nonprescription products in terms of endpoints, study design, and study quality.

Purpose: We developed an artificial intelligence (AI) model to detect immunotherapy -related adverse events (irAEs) from clinical narratives of electronic health records (EHRs) at the patient level.

Patients and methods: Training data, used for internal validation of the AI model, comprised 1230 clinical notes from 30 patients at The Ohio State University James Cancer Hospital-20 patients who experienced irAEs and ten who did not. 3256 clinical notes of 50 patients were utilized for external validation of the AI model.

Results: Use of a leave-one-out cross-validation technique for internal validation among those 30 patients yielded accurate identification of 19 of 20 with irAEs (positive patients; 95% sensitivity) and correct dissociation of eight of ten without (negative patients; 80% specificity). External validation on 3256 clinical notes of 50 patients yielded high sensitivity (95%) but moderate specificity (64%). If we improve the model's specificity to 100%, it could eliminate the need to manually review 2500 of those 3256 clinical notes (77%).

Conclusion: Combined use of this AI model with the manual review of clinical notes will improve both sensitivity and specificity in the detection of irAEs, decreasing workload and costs and facilitating the development of improved immunotherapies.

Background: As part of Electrophysiology Predictable and Sustainable Implementation of National Registries (EP PASSION), a multi-stakeholder collaboration between the US Food and Drug Administration (FDA), academic and society partners, and cardiovascular implantable electronic device manufacturers, a 5-year bradycardia lead study transitioned from a traditional post-approval study (PAS) to a real-world data (RWD) approach using a novel method to evaluate chronic cardiac lead complications.

Methods: Lead complications were identified using a combination of diagnosis and procedure codes from 2013 to 2020 fee-for-service Medicare claims data along with BIOTRONIK device registration and Medical Device Reporting data from patients implanted between 2013 and 2015 with a Solia S lead. A proof-of-concept analysis was performed using McNemar's test to compare lead complications reported in the traditional PAS with lead complications identified in the RWD. Kaplan-Meier survival and incidence rates were evaluated to determine real-world long-term safety.

Results: The proof-of-concept analysis of 896 patients found in both traditional PAS and RWD sources demonstrated a 99.7% proportion of overall agreement in identifying lead complications (p = 0.0833). Following this validation, 1841 study leads from 1015 Medicare patients were analyzed. A total of 33 lead complications (attributable or possibly attributable to the study lead) were identified for a rate of 0.005 complications per lead-year. The complication-free rate at 5-years post-implant was 97.2% (95% CI: 96.07%, 98.06%).

Conclusion: These results led to the first FDA approval for transition of a cardiac lead PAS to long-term safety reporting using RWD, paving the way for future real-world cardiac lead and device surveillance studies.

Clinicaltrialsgov identifier: NCT01791127.

Background: Preserved ratio impaired spirometry (PRISm) represents a population with spirometry results that do not meet standardized COPD obstruction criteria, yet present with high respiratory symptom burden and might benefit from respiratory management and treatment. We aimed to determine prevalence of PRISm in US primary care patients diagnosed with COPD, describe their demographic, clinical, and CT scan characteristics.

Methods: An observational registry study utilizing the US APEX COPD registry, composed of patients diagnosed with COPD aged 35+ years. Demographic and clinical data were collected from EHRs and complemented by questionnaires. Multivariable logistic regression was performed to assess whether PRISm predicts lung function decline.

Results: Prevalence of PRISm within a primary care population clinically diagnosed with COPD was 23.6% (678/2866, 95% CI 22.0-25.1). Those with PRISm were more likely female (55.9% vs 46.9%), younger (66.3±11.1 vs 69.2±10.3 years), with a greater mean BMI (33.5±9.2 vs 27.8±7.2 kg/m²), more often African American or Hispanic (37.2% vs 26.3%), and with fewer current smokers (33.1% vs 36.8%) when compared to those meeting COPD spirometry criteria (all p<0.05). Compared to COPD GOLD 0 patients, individuals with PRISm had greater BMI (33.5±9.2 vs 30.6±7.8), and were more likely current smokers (33.1% vs 23.4%), both p<0.05. Patients with PRISm had similar respiratory symptoms (chronic bronchitis, CAT, and mMRC) to overall COPD patients, but more frequently than GOLD 0 COPD patients (p<0.01). Emphysema was more commonly reported in CT scans from patients with PRISm 70.3% (260/369, 95% CI 65.8-75.3) than those with GOLD 0 COPD 64.1% (218/340, 95% CI 58.8-69.2) (p<0.05). PRISm status was not predictive of lung function decline.

Interpretation: One in four primary care patients with clinically diagnosed COPD in a large US registry fulfil the spirometric definition of PRISm rather than COPD, but suffers from emphysema in CT and significant respiratory symptoms.

Purpose: Patients with asthma in Singapore often have complex patient journeys, with diagnosis and management across various primary and speciality care settings. Real-world population health data is needed to identify care gaps and inform policies.

Patients and methods: This retrospective, longitudinal cohort study assessed real-world data from adults (aged ≥18 years) with asthma in the SingHealth Chronic Obstructive Pulmonary Disease and Asthma Data Mart, an integrated database of electronic medical records of patients who attended primary and/or speciality care clinics in the SingHealth Regional Health System 01/01/2015-12/31/2020. Patients were indexed by first asthma diagnosis and categorized into cohorts of index year. Patient characteristics, asthma management and outcomes were described during baseline (1-year pre-index) and follow-up periods (1-year post-index).

Results: Overall, 21,215 patients were included across 4 cohorts: 2016, N=12,947; 2017, N=3419; 2018, N=2816; 2019, N=2033. Most common baseline asthma medication changed from inhaled corticosteroids (ICS) alone in the 2016 cohort (32.8% [n=4252]) to ICS/long-acting β₂-agonist in the 2019 cohort (33.3% [n=677]). Asthma symptom control (mean [SD] Asthma Control Test scores) improved from 2016 to 2019 during baseline (18.38 [4.93] vs 19.87 [4.56]; p<0.001) and follow-up (18.34 [4.23] vs 21.07 [3.51]; p<0.001). Mean (standard deviation [SD]) number of exacerbations per patient during follow-up decreased from 2016 to 2019 (1.91 [3.11] vs 0.89 [2.07]; p<0.001). Mean (SD) number of emergency department visits per patient during follow-up decreased from 0.21 (0.75) in 2016 to 0.17-0.18 (0.60-0.65; p<0.001) between 2017 and 2019.

Conclusion: Health status at first asthma diagnosis improved for each succeeding cohort from 2016 to 2019, along with improvements in patient management and outcomes. This reflects greater awareness of the condition and improved use of medication and referrals in recent years, suggesting policy changes and their implementation, including promotion of disease awareness and adoption of guideline recommendations, may improve asthma outcomes in Singapore.

Background: Although the estimated prevalence of non-alcoholic steatohepatitis (NASH) in Italy is 4-6%, little is known about patient characteristics and care pathways.

Aim: To describe patient characteristics and management approaches for patients with NASH or suspected NASH in Italy.

Methods: Data were drawn from the Adelphi Real World NASH Disease Specific Programme™, a cross-sectional survey of endocrinologists and gastroenterologists in Italy from January to March 2018. Physicians completed questionnaires for their next five consecutively consulting patients with NASH or suspected NASH. Analyses were descriptive.

Results: Seventy-six physicians provided data on 380 patients. The mean age was 58.5 ± 11.1 years and the mean body mass index was 31.8 ± 5.5 kg/m². A total of 231 patients (61%) had no/non-advanced fibrosis as evaluated by liver biopsy or non-invasive tests. Common diagnostic assessments were cholesterol, hemoglobin A1c, absence of viral hepatitis, and alcohol assessment. At diagnosis, 87% (n=322/372) and 45% (n=169/372) of patients received an ultrasound and liver biopsy, respectively. Overall, 88% of patients were referred from primary to secondary care. Obesity (81%) and type 2 diabetes (62%) were the most commonly recorded comorbidities, with 70% of patients having ≥3 comorbidities. Vitamin E (13%) and GLP-1 receptor agonists (13%) were the most prescribed guideline-recommended treatments for all patients.

Conclusion: Patients with NASH in Italy had high levels of obesity and comorbidities, while diagnosis and treatment frequently were not according to guidelines. Our data show an unmet need for more targeted diagnosis and treatment in Italian patients with NASH, in order to optimize outcomes.

Purpose: Observational research provides valuable insights into treatments used in patient populations in real-world settings. However, confounding is likely to occur if there are differences in patient characteristics associated with both the exposure and outcome between the groups being evaluated. One approach to reduce confounding and facilitate unbiased comparisons is inverse probability of treatment weighting (IPTW) using propensity scores. Machine learning (ML) and entropy balancing can potentially be used in generating propensity scores for IPTW, but there is limited literature on this application. We aimed to assess the feasibility of applying these methods for reducing confounding in observational studies. These methods were assessed in a study comparing cardiovascular outcomes in adults with type 2 diabetes and established atherosclerotic cardiovascular disease taking once-weekly glucagon-like peptide-1 receptor agonists or dipeptidyl peptidase-4 inhibitors.

Methods: We applied advanced methods to generate the propensity scores compared to the original logistic regression method in terms of covariate balance. After calculating weights, a weighted Cox proportional hazards model was used to calculate the sample average treatment effect. Support Vector Classification, Support Vector Regression, XGBoost, and LightGBM were the ML models used. Entropy balancing was also performed on features identified in the original cardiovascular outcomes study.

Results: Accuracy (range: 0.71 to 0.73), area under the curve (0.77 to 0.79), precision (0.53 to 0.60), recall (0.66 to 0.68), and F1 score (0.60 to 0.64) were similar between all of the advanced propensity score methods and traditional logistic regression. Among ML models, only XGBoost achieved balance in all measured baseline characteristics between the two treatment groups, closely approximating the performance of the original logistic regression. Entropy balancing weights provided the best performance among all models in balancing baseline characteristics, achieving near perfect balancing.

Conclusion: Among the advanced methods examined, entropy balancing weights performed the best for optimizing balancing and can produce similar results compared to traditional logistic regression.