Evaluating eligibility criteria of oncology trials using real-world data and AI.

There is a growing focus on making clinical trials more inclusive but the design of trial eligibility criteria remains challenging1-3. Here we systematically evaluate the effect of different eligibility criteria on cancer trial populations and outcomes with real-world data using the computational framework of Trial Pathfinder. We apply Trial Pathfinder to emulate completed trials of advanced non-small-cell lung cancer using data from a nationwide database of electronic health records comprising 61,094 patients with advanced non-small-cell lung cancer. Our analyses reveal that many common criteria, including exclusions based on several laboratory values, had a minimal effect on the trial hazard ratios. When we used a data-driven approach to broaden restrictive criteria, the pool of eligible patients more than doubled on average and the hazard ratio of the overall survival decreased by an average of 0.05. This suggests that many patients who were not eligible under the original trial criteria could potentially benefit from the treatments. We further support our findings through analyses of other types of cancer and patient-safety data from diverse clinical trials. Our data-driven methodology for evaluating eligibility criteria can facilitate the design of more-inclusive trials while maintaining safeguards for patient safety.

Novel Health Information Technology to Aid Provider Recognition and Treatment of Major Depressive Disorder and Posttraumatic Stress Disorder in Primary Care.

BACKGROUND: Millions of traumatized refugees worldwide have resettled in the United States. For one of the largest, the Cambodian community, having their mental health needs met has been a continuing challenge. A multicomponent health information technology screening tool was designed to aid provider recognition and treatment of major depressive disorder and posttraumatic stress disorder (PTSD) in the primary care setting. METHODS: In a clustered randomized controlled trial, 18 primary care providers were randomized to receive access to a multicomponent health information technology mental health screening intervention, or to a minimal intervention control group; 390 Cambodian American patients empaneled to participating providers were assigned to the providers' randomized group. RESULTS: Electronic screening revealed that 65% of patients screened positive for depression and 34% screened positive for PTSD. Multilevel mixed effects logistic models, accounting for clustering structure, indicated that providers in the intervention were more likely to diagnose depression [odds ratio (OR), 6.5; 95% confidence interval (CI), 1.48-28.79; P=0.013] and PTSD (OR, 23.3; 95% CI, 2.99-151.62; P=0.002) among those diagnosed during screening, relative to the control group. Providers in the intervention were more likely to provide evidence-based guideline (OR, 4.02; 95% CI, 1.01-16.06; P=0.049) and trauma-informed (OR, 15.8; 95% CI, 3.47-71.6; P<0.001) care in unadjusted models, relative to the control group. Guideline care, but not trauma-informed care, was associated with decreased depression at 12 weeks in both study groups (P=0.003), and neither was associated with PTSD outcomes at 12 weeks. CONCLUSIONS: This innovative approach offers the potential for training primary care providers to diagnose and treat traumatized patients, the majority of whom seek mental health care in primary care (ClinicalTrials.gov number, NCT03191929).

Systematic analysis of off-label and off-guideline cancer therapy usage in a real-world cohort of 165,912 US patients.

Patients with cancer may be given treatments that are not officially approved (off-label) or recommended by guidelines (off-guideline). Here we present a data science framework to systematically characterize off-label and off-guideline usages using real-world data from de-identified electronic health records (EHR). We analyze treatment patterns in 165,912 US patients with 14 common cancer types. We find that 18.6% and 4.4% of patients have received at least one line of off-label and off-guideline cancer drugs, respectively. Patients with worse performance status, in later lines, or treated at academic hospitals are significantly more likely to receive off-label and off-guideline drugs. To quantify how predictable off-guideline usage is, we developed machine learning models to predict which drug a patient is likely to receive based on their clinical characteristics and previous treatments. Finally, we demonstrate that our systematic analyses generate hypotheses about patients' response to treatments.

Quality of Life and Healthcare Resource Use in a Real-world Patient Population with Idiopathic Pulmonary Fibrosis: The PROOF Registry.

INTRODUCTION: The PROOF registry is a prospective, observational study that aimed to monitor disease progression in a real-world cohort of patients with idiopathic pulmonary fibrosis (IPF). Here, longitudinal quality-of-life (QoL) outcomes, healthcare resource use (HCRU), and the association between QoL and mortality in patients enrolled in the PROOF registry are presented. METHODS: QoL outcomes (St. George's Respiratory Questionnaire [SGRQ], EuroQoL-5 dimensions-5 levels Health Questionnaire [EQ-5D-5L], EuroQoL-5 dimensions Health Questionnaire [EQ-5D] visual analogue scale [VAS] and cough VAS) and HCRU were collected for all patients. Associations between baseline QoL and mortality were assessed using univariate and multivariate analyses. During multivariate analyses, individual QoL measures were adjusted for the following covariates: age, sex, percent predicted forced vital capacity, percent predicted diffusing capacity of the lungs for carbon monoxide, smoking status, and supplementary oxygen use at registry inclusion. RESULTS: In total, 277 patients were enrolled in the PROOF registry. During the follow-up period, worsening in cough VAS score, SGRQ symptom score, and SGRQ activity score was observed, while EQ-5D VAS, SGRQ total score, and SGRQ impact score remained stable. During univariate analyses, EQ-5D VAS and all SGRQ sub-scores and total score at baseline were associated with mortality; however, during multivariate analyses, only the SGRQ total score, SGRQ impact score, and SGRQ symptom score at baseline were associated with mortality. During the follow-up period, 261 (94.2%) patients required an outpatient consultation (IPF- or non-IPF-related) and there were 182 hospitalizations in total, most of which were respiratory related (66.5%). CONCLUSIONS: The PROOF registry provided valuable, real-world data on the association between baseline QoL and mortality, and longitudinal HCRU and QoL outcomes in patients with IPF over 24 months and identified that SGRQ may be an independent prognostic factor in IPF.

Systematic pan-cancer analysis of mutation-treatment interactions using large real-world clinicogenomics data.

Quantifying the effectiveness of different cancer therapies in patients with specific tumor mutations is critical for improving patient outcomes and advancing precision medicine. Here we perform a large-scale computational analysis of 40,903 US patients with cancer who have detailed mutation profiles, treatment sequences and outcomes derived from electronic health records. We systematically identify 458 mutations that predict the survival of patients on specific immunotherapies, chemotherapy agents or targeted therapies across eight common cancer types. We further characterize mutation-mutation interactions that impact the outcomes of targeted therapies. This work demonstrates how computational analysis of large real-world data generates insights, hypotheses and resources to enable precision oncology.

Machine Learning Prediction of Clinical Trial Operational Efficiency.

Clinical trials are the gatekeepers and bottlenecks of progress in medicine. In recent years, they have become increasingly complex and expensive, driven by a growing number of stakeholders requiring more endpoints, more diverse patient populations, and a stringent regulatory environment. Trial designers have historically relied on investigator expertise and legacy norms established within sponsor companies to improve operational efficiency while achieving study goals. As such, data-driven forecasts of operational metrics can be a useful resource for trial design and planning. We develop a machine learning model to predict clinical trial operational efficiency using a novel dataset from Roche containing over 2,000 clinical trials across 20 years and multiple disease areas. The data includes important operational metrics related to patient recruitment and trial duration, as well as a variety of trial features such as the number of procedures, eligibility criteria, and endpoints. Our results demonstrate that operational efficiency can be predicted robustly using trial features, which can provide useful insights to trial designers on the potential impact of their decisions on patient recruitment success and trial duration.

Reduction in fluoroquinolone use following introduction of ertapenem into a hospital formulary is associated with improvement in susceptibility of Pseudomonas aeruginosa to group 2 carbapenems: a 10-year study.

We examined the effect of the addition of ertapenem to our hospital formulary on the resistance of nosocomial Pseudomonas aeruginosa to group 2 carbapenems (imipenem, meropenem, and doripenem). This was a retrospective, observational study conducted between 1 January 2000 and 31 January 2009 at a large, tertiary-care hospital. Autoregressive integrated moving average (ARIMA) regression models were used to evaluate the effect of ertapenem use on the susceptibility of Pseudomonas aeruginosa to group 2 carbapenems as well as on the use of the group 2 carbapenems, ciprofloxacin, and other antipseudomonal drugs (i.e., tobramycin, cefepime, and piperacillin-tazobactam). Resistance was expressed as a percentage of total isolates as well as the number of carbapenem-resistant bacterial isolates per 10,000 patient days. Pearson correlation was used to assess the relationship between antibiotic use and carbapenem resistance. Following the addition of ertapenem to the formulary, there was a statistically significant decrease in the percentage of Pseudomonas aeruginosa isolates resistant to the group 2 carbapenems (P = 0.003). Group 2 carbapenem use and the number of carbapenem-resistant Pseudomonas aeruginosa isolates per 10,000 patient days did not change significantly over the time period. There was a large decrease in the use of ciprofloxacin (P = 0.0033), and there was a correlation of ciprofloxacin use with the percentage of isolates resistant to the group 2 carbapenems (ρ = 0.47, P = 0.002). We suspect that the improvement in susceptibility of Pseudomonas aeruginosa to group 2 carbapenems was related to a decrease in ciprofloxacin use.

Sustained reduction in antimicrobial use and decrease in methicillin-resistant Staphylococcus aureus and Clostridium difficile infections following implementation of an electronic medical record at a tertiary-care teaching hospital.

OBJECTIVES: we evaluated the effect of implementation of an electronic medical record (EMR) on the use of antimicrobial agents and on the rates of infections with Clostridium difficile and methicillin-resistant Staphylococcus aureus (MRSA). METHODS: this was a retrospective, observational study conducted between 1 January 2005 and 31 December 2009. Antimicrobial drug use, rates of nosocomial C. difficile infection (CDI) and MRSA infection, the number of medical charts reviewed and number of antimicrobial recommendations made and accepted were compared before and after implementing the EMR utilizing interrupted time-series analysis. RESULTS: compared with the 10 quarters prior to implementing the EMR, there was a 36.6% increase in the number of charts reviewed (P < 0.0001), a 98.1% increase in the number of antimicrobial recommendations made (P < 0.0001) and a 124% increase in the number of recommendations accepted (P < 0.0001). There was a 28.8% decrease in the use of 41 commonly used antibacterial agents (P < 0.0001). Nosocomial CDI decreased by 18.7% (P = 0.07) and nosocomial MRSA infections decreased by 45.2% (P < 0.0001) following implementation of the EMR. CONCLUSIONS: adoption of an EMR facilitated a significant increase in chart reviews and antimicrobial recommendations, which resulted in a sustained decrease in antimicrobial use. There were decreased nosocomial infections with MRSA and a trend towards decreasing CDIs following implementation of the EMR.

Factors associated with COVID-19 viral and antibody test positivity and assessment of test concordance: a retrospective cohort study using electronic health records from the USA.

OBJECTIVES: To identify factors associated with COVID-19 test positivity and assess viral and antibody test concordance. DESIGN: Observational retrospective cohort study. SETTING: Optum de-identified electronic health records including over 700 hospitals and 7000 clinics in the USA. PARTICIPANTS: There were 891 754 patients who had a COVID-19 test identified in their electronic health record between 20 February 2020 and 10 July 2020. PRIMARY AND SECONDARY OUTCOME MEASURES: Per cent of viral and antibody tests positive for COVID-19 ('positivity rate'); adjusted ORs for factors associated with COVID-19 viral and antibody test positivity; and per cent concordance between positive viral and subsequent antibody test results. RESULTS: Overall positivity rate was 9% (70 472 of 771 278) and 12% (11 094 of 91 741) for viral and antibody tests, respectively. Positivity rate was inversely associated with the number of individuals tested and decreased over time across regions and race/ethnicities. Antibody test concordance among patients with an initial positive viral test was 91% (71%-95% depending on time between tests). Among tests separated by at least 2 weeks, discordant results occurred in 7% of patients and 9% of immunocompromised patients. Factors associated with increased odds of viral and antibody positivity in multivariable models included: male sex, Hispanic or non-Hispanic black or Asian race/ethnicity, uninsured or Medicaid insurance and Northeast residence. We identified a negative dose effect between the number of comorbidities and viral and antibody test positivity. Paediatric patients had reduced odds (OR=0.60, 95% CI 0.57 to 0.64) of a positive viral test but increased odds (OR=1.90, 95% CI 1.62 to 2.23) of a positive antibody test compared with those aged 18-34 years old. CONCLUSIONS: This study identified sociodemographic and clinical factors associated with COVID-19 test positivity and provided real-world evidence demonstrating high antibody test concordance among viral-positive patients.

Prognostic model to identify and quantify risk factors for mortality among hospitalised patients with COVID-19 in the USA.

OBJECTIVES: To develop a prognostic model to identify and quantify risk factors for mortality among patients admitted to the hospital with COVID-19. DESIGN: Retrospective cohort study. Patients were randomly assigned to either training (80%) or test (20%) sets. The training set was used to fit a multivariable logistic regression. Predictors were ranked using variable importance metrics. Models were assessed by C-indices, Brier scores and calibration plots in the test set. SETTING: Optum de-identified COVID-19 Electronic Health Record dataset including over 700 hospitals and 7000 clinics in the USA. PARTICIPANTS: 17 086 patients hospitalised with COVID-19 between 20 February 2020 and 5 June 2020. MAIN OUTCOME MEASURE: All-cause mortality while hospitalised. RESULTS: The full model that included information on demographics, comorbidities, laboratory results, and vital signs had good discrimination (C-index=0.87) and was well calibrated, with some overpredictions for the most at-risk patients. Results were similar on the training and test sets, suggesting that there was little overfitting. Age was the most important risk factor. The performance of models that included all demographics and comorbidities (C-index=0.79) was only slightly better than a model that only included age (C-index=0.76). Across the study period, predicted mortality was 1.3% for patients aged 18 years old, 8.9% for 55 years old and 28.7% for 85 years old. Predicted mortality across all ages declined over the study period from 22.4% by March to 14.0% by May. CONCLUSION: Age was the most important predictor of all-cause mortality, although vital signs and laboratory results added considerable prognostic information, with oxygen saturation, temperature, respiratory rate, lactate dehydrogenase and white cell count being among the most important predictors. Demographic and comorbidity factors did not improve model performance appreciably. The full model had good discrimination and was reasonably well calibrated, suggesting that it may be useful for assessment of prognosis.