Trends in Neonatal Intensive Care Unit Utilization in a Large Integrated Health Care System.

Importance: There are few population-based studies addressing trends in neonatal intensive care unit (NICU) admission and NICU patient-days, especially in the subpopulation that, by gestational age (GA) and birth weight (BW), might otherwise be able to stay in the room with their mothers. Objective: To describe population-based trends in NICU admissions, NICU patient-days, readmissions, and mortality in the birth population of a large integrated health care system. Design, Setting, and Participants: This cohort study was conducted using data extracted from electronic medical records at Kaiser Permanente Southern California (KPSC) health care system. Participants included all women who gave birth at KPSC hospitals and their newborns from January 1, 2010, through December 31, 2018. Data extraction was limited to data entry fields whose contents were either numbers or fixed categorical choices. Rates of NICU admission, NICU patient-days, readmission rates, and mortality rates were measured in the total population, in newborns with GA 35 weeks or greater and BW 2000 g or more (high GA and BW group), and in the remaining newborns (low GA and BW group). Admissions to the NICU and NICU patient-days were risk adjusted with a machine learning model based on demographic and clinical characteristics before NICU admission. Changes in the trends were assessed with 2-sided correlated seasonal Mann-Kendall test. Data analysis was performed in August 2019. Exposures: Admission to the NICU and NICU patient-days among the birth cohort. Main Outcomes and Measures: The primary outcomes were NICU admission and NICU patient-days in the total neonatal population and GA and BW subgroups. The secondary outcomes were readmission and mortality rates. Results: Over the study period there were 320 340 births (mean [SD] age of mothers, 30.1 [5.7] years; mean [SD] gestational age, 38.6 [1.97] weeks; mean [SD] birth weight, 3302 [573] g). The risk-adjusted NICU admission rate decreased from a mean of 14.5% (95% CI, 14.2%-14.7%) to 10.9% (95% CI, 10.7%-11.7%) (P for trend = .002); 92% of the change was associated with changes in the care of newborns in the high GA and BW group. The number of risk-adjusted NICU patient-days per birth decreased from a mean of 1.50 patient-days (95% CI, 1.43-1.54 patient-days) to 1.40 patient-days (95% CI, 1.36-1.48 patient-days) (P for trend = .03); 70% of the change was associated with newborns in the high GA and BW group. The unadjusted 30-day readmission rates and mortality rates did not change. Conclusions and Relevance: Admission rates to the NICU and numbers of NICU patient-days decreased over the study period without an increase in readmissions or mortality. The observed decrease was associated with the high GA and BW newborn population. How much of this decrease is attributable to intercurrent health care systemwide quality improvement initiatives would require further investigation. The remaining unexplained variation suggests that further changes are also possible.

French Health Technology Assessment of Antineoplastic Drugs Indicated in the Treatment of Solid Tumours: Perspective for Future Trends.

BACKGROUND: France is one of the European countries that spend the most on oncology drugs. To keep pharmaceutical expenditure under control, Health Authorities highly scrutinize market access of costly medicines. OBJECTIVE: To assess current and future trends in French health technology assessment (HTA) of antineoplastic drugs indicated in the treatment of solid tumours. METHODS: A review of the SMR and ASMR drivers of the Transparency Committee (CT) opinions issued for antineoplastic drugs indicated in the treatment of solid tumours and approved between 2009 and 2014 was performed to assess current trends in French health technology assessment (HTA), complemented by an expert board consultation to capture the critical issues on the future of antineoplastic drugs HTA. RESULTS: Thirty-one drugs indicated for the treatment of solid tumours were identified (77 % targeted therapies). Initial CT assessments were available for 26 drugs. Four key items in the CT assessment were identified: 1) Clinical trial methodology; 2) Acceptance of progression-free survival (PFS) as a valuable endpoint; 3) Transferability of clinical trials in clinical practice; 4) Unpredictability of CT decisions. Experts raised the important development of personalised medicines in oncology and key challenges for oncology products to generate information expected from HTA perspective. CONCLUSION: The French system remains committed to its values and philosophy (access of all innovations for everybody) which are threatened by the increasing launch of innovative therapies and budget constraint. Both HTA decision framework evolution and revision of the current pricing process should be considered in France to cope with these new challenges.

Medico-economic evaluation of healthcare products. Methodology for defining a significant impact on French health insurance costs and selection of benchmarks for interpreting results.

Decree No. 2012-1116 of 2 October 2012 on medico-economic assignments of the French National Authority for Health (Haute autorité de santé, HAS) significantly alters the conditions for accessing the health products market in France. This paper presents a theoretical framework for interpreting the results of the economic evaluation of health technologies and summarises the facts available in France for developing benchmarks that will be used to interpret incremental cost-effectiveness ratios. This literature review shows that it is difficult to determine a threshold value but it is also difficult to interpret then incremental cost effectiveness ratio (ICER) results without a threshold value. In this context, round table participants favour a pragmatic approach based on "benchmarks" as opposed to a threshold value, based on an interpretative and normative perspective, i.e. benchmarks that can change over time based on feedback.

Automated diagnosis of epilepsy using EEG power spectrum.

Interictal electroencephalography (EEG) has clinically meaningful limitations in its sensitivity and specificity in the diagnosis of epilepsy because of its dependence on the occurrence of epileptiform discharges. We have developed a computer-aided diagnostic (CAD) tool that operates on the absolute spectral energy of the routine EEG and has both substantially higher sensitivity and negative predictive value than the identification of interictal epileptiform discharges. Our approach used a multilayer perceptron to classify 156 patients admitted for video-EEG monitoring. The patient population was diagnostically diverse; 87 were diagnosed with either generalized or focal seizures. The remainder of the patients were diagnosed with nonepileptic seizures. The sensitivity was 92% (95% confidence interval [CI] 85-97%) and the negative predictive value was 82% (95% CI 67-92%). We discuss how these findings suggest that this CAD can be used to supplement event-based analysis by trained epileptologists.

Parameter Selection in Mutual Information-Based Feature Selection in Automated Diagnosis of Multiple Epilepsies Using Scalp EEG.

Developing EEG-based computer aided diagnostic (CAD) tools would allow identification of epilepsy in individuals who have experienced possible seizures, yet such an algorithm requires efficient identification of meaningful features out of potentially more than 35,000 features of EEG activity. Mutual information can be used to identify a subset of minimally-redundant and maximally relevant (mRMR) features but requires a priori selection of two parameters: the number of features of interest and the number of quantization levels into which the continuous features are binned. Here we characterize the variance of cross-validation accuracy with respect to changes in these parameters for four classes of machine learning (ML) algorithms. This assesses the efficiency of combining mRMR with each of these algorithms by assessing when the variance of cross-validation accuracy is minimized and demonstrates how naive parameter selection may artificially depress accuracy. Our results can be used to improve the understanding of how feature selection interacts with four classes of ML algorithms and provide guidance for better a priori parameter selection in situations where an overwhelming number of redundant, noisy features are available for classification.