Comparing LASSO and random forest models for predicting neurological dysfunction among fluoroquinolone users.

BACKGROUND: Fluoroquinolones are associated with central (CNS) and peripheral (PNS) nervous system symptoms, and predicting the risk of these outcomes may have important clinical implications. Both LASSO and random forest are appealing modeling methods, yet it is not clear which method performs better for clinical risk prediction. PURPOSE: To compare models developed using LASSO versus random forest for predicting neurological dysfunction among fluoroquinolone users. METHODS: We developed and validated risk prediction models using claims data from a commercially insured population. The study cohort included adults dispensed an oral fluoroquinolone, and outcomes were CNS and PNS dysfunction. Model predictors included demographic variables, comorbidities and medications known to be associated with neurological symptoms, and several healthcare utilization predictors. We assessed the accuracy and calibration of these models using measures including AUC, calibration curves, and Brier scores. RESULTS: The underlying cohort contained 16 533 (1.18%) individuals with CNS dysfunction and 46 995 (3.34%) individuals with PNS dysfunction during 120 days of follow-up. For CNS dysfunction, LASSO had an AUC of 0.81 (95% CI: 0.80, 0.82), while random forest had an AUC of 0.80 (95% CI: 0.80, 0.81). For PNS dysfunction, LASSO had an AUC of 0.75 (95% CI: 0.74, 0.76) versus an AUC of 0.73 (95% CI: 0.73, 0.74) for random forest. Both LASSO models had better calibration, with Brier scores 0.17 (LASSO) versus 0.20 (random forest) for CNS dysfunction and 0.20 (LASSO) versus 0.25 (random forest) for PNS dysfunction. CONCLUSIONS: LASSO outperformed random forest in predicting CNS and PNS dysfunction among fluoroquinolone users, and should be considered for modeling when the cohort is modest in size, when the number of model predictors is modest, and when predictors are primarily binary.

Comparative neurological safety of fluoroquinolones versus therapeutic alternatives.

BACKGROUND: Fluoroquinolones, one of the most commonly prescribed antibiotic classes, have been implicated in cases of central nervous system (CNS) and peripheral nervous system (PNS) adverse events, which highlights the need for epidemiologic studies of the neurological safety of fluoroquinolones. PURPOSE: To evaluate the safety of fluoroquinolones with regard to risk of diagnosed neurological dysfunction. METHODS: We conducted a propensity score-matched inception cohort study using claims data from a commercially insured population. Our study included adults prescribed an oral fluoroquinolone or comparator antibiotic between January 2000 and September 2015 for acute bacterial sinusitis, acute bacterial exacerbation of chronic bronchitis, uncomplicated urinary tract infection, or acute bronchitis. Our outcomes were CNS dysfunction, and four separate but complementary PNS dysfunction outcomes. Cox proportional hazards models were estimated after matching on propensity scores fitted using the variables age, sex, epilepsy, hereditary peripheral neuropathy, renal dysfunction, diabetes, gabapentinoid use, statin use, isoniazid use, and chemotherapy use. RESULTS: Our cohort contained 976 568 individuals exposed to a fluoroquinolone antibiotic matched 1:1 with a comparator. Matching produced balance (standardized mean difference <0.1) on all variables included in the propensity score. The hazard ratio associated with fluoroquinolone exposure was 1.08 (95% confidence interval 1.05-1.11) for CNS dysfunction, and 1.09 (95% CI 1.07-1.11) for the most commonly occurring PNS dysfunction outcome. CONCLUSIONS: Fluoroquinolone antibiotic use was associated with the development of neurological dysfunction versus comparator antibiotic use in the adult population.

Qualitative study of system-level factors related to genomic implementation.

PURPOSE: Research on genomic medicine integration has focused on applications at the individual level, with less attention paid to implementation within clinical settings. Therefore, we conducted a qualitative study using the Consolidated Framework for Implementation Research (CFIR) to identify system-level factors that played a role in implementation of genomic medicine within Implementing GeNomics In PracTicE (IGNITE) Network projects. METHODS: Up to four study personnel, including principal investigators and study coordinators from each of six IGNITE projects, were interviewed using a semistructured interview guide that asked interviewees to describe study site(s), progress at each site, and factors facilitating or impeding project implementation. Interviews were coded following CFIR inner-setting constructs. RESULTS: Key barriers included (1) limitations in integrating genomic data and clinical decision support tools into electronic health records, (2) physician reluctance toward genomic research participation and clinical implementation due to a limited evidence base, (3) inadequate reimbursement for genomic medicine, (4) communication among and between investigators and clinicians, and (5) lack of clinical and leadership engagement. CONCLUSION: Implementation of genomic medicine is hindered by several system-level barriers to both research and practice. Addressing these barriers may serve as important facilitators for studying and implementing genomics in practice.

Comparative risk of serious hypoglycemia among persons dispensed a fluoroquinolone versus a non-fluoroquinolone antibiotic.

AIM: Fluoroquinolone antibiotics have been implicated in cases of metabolic adverse events. This study investigated the causal association between fluoroquinolones and serious hypoglycemia in those with and without diabetes. METHODS: We conducted a propensity score-matched cohort study using Optum claims data. We included adults dispensed an oral fluoroquinolone or comparator antibiotic between January 2000 and September 2015 for specific infections of interest. The outcome was serious hypoglycemia, defined using a validated algorithm. Conditional logistic regression was used to estimate odds ratios (ORs) in diabetes and non-diabetes cohorts after matching on propensity scores fitted using confounding variables of interest. RESULTS: Our cohort contained 119,112 individuals with diabetes and 917,867 individuals without diabetes exposed to a fluoroquinolone, matched 1:1 with a comparator. Matching produced balance (standardized mean difference < 0.1) on all variables included in the propensity score. The OR for the association between fluoroquinolones and serious hypoglycemia was 1.28 (95% confidence interval [CI]: 1.04-1.57) in the entire cohort, 1.30 (95% CI: 1.05-1.62) in individuals with diabetes, and 1.06 (95% CI: 0.53-2.13) in individuals without diabetes. CONCLUSION: Fluoroquinolone users are at an increased risk of serious hypoglycemia relative to comparator antibiotic users. This association was evident only among persons diagnosed with diabetes.

Readmissions after hospital care for meningitis in the United States.

BACKGROUND: Our objectives were to (1) characterize patient and clinical characteristics of adults hospitalized with meningitis; (2) describe meningitis hospitalization outcomes, including 30- and 90-day readmissions; and (3) determine whether clinical, patient, or index hospitalization characteristics are associated with readmission and readmission outcomes. METHODS: This retrospective study of the 2014 National Readmissions Database extracted data on hospitalized adults with a principal diagnosis of meningitis and examined hospitalization outcomes using descriptive statistics. Logistic regression models were built to determine whether characteristics were associated with 30- or 90-day readmissions. RESULTS: For the 30-day readmission analyses, 18,883 adults qualified. Meningitis hospitalizations commonly involved adults 25 to 54 years of age who were insured by private carriers. The readmission rates were 7.0% at 30 days and 11.4% at 90 days. Readmission was associated with greater comorbidity burden (2 conditions: adjusted odds ratio [AOR] = 1.60, range 1.24-2.08; 3 conditions: AOR = 1.92, range 1.43-2.58; 4+ conditions: AOR = 2.68, range 2.04-3.51 vs 0 or 1 condition), public insurance (Medicare: AOR = 1.85, range 1.30-2.62; Medicaid: AOR = 1.48, range 1.16-1.90 vs private insurance), and medical error (AOR = 1.43, range 1.07-1.91). Readmissions were most often for meningitis, septicemia, or medical complications. CONCLUSIONS: Readmission after hospitalization for meningitis is associated with both fixed and modifiable factors. More research is needed to determine which post-meningitis readmissions are preventable.