Development and internal validation of clinical prediction models for outcomes of complicated intra-abdominal infection.

BACKGROUND: Complicated intra-abdominal infections (cIAIs) are associated with significant morbidity and mortality. The aim of this study was to describe the clinical characteristics of patients with cIAI in a multicentre study and to develop clinical prediction models (CPMs) to help identify patients at risk of mortality or relapse. METHODS: A multicentre observational study was conducted from August 2016 to February 2017 in the UK. Adult patients diagnosed with cIAI were included. Multivariable logistic regression was performed to develop CPMs for mortality and cIAI relapse. The c-statistic was used to test model discrimination. Model calibration was tested using calibration slopes and calibration in the large (CITL). The CPMs were then presented as point scoring systems and validated further. RESULTS: Overall, 417 patients from 31 surgical centres were included in the analysis. At 90 days after diagnosis, 17.3 per cent had a cIAI relapse and the mortality rate was 11.3 per cent. Predictors in the mortality model were age, cIAI aetiology, presence of a perforated viscus and source control procedure. Predictors of cIAI relapse included the presence of collections, outcome of initial management, and duration of antibiotic treatment. The c-statistic adjusted for model optimism was 0.79 (95 per cent c.i. 0.75 to 0.87) and 0.74 (0.73 to 0.85) for mortality and cIAI relapse CPMs. Adjusted calibration slopes were 0.88 (95 per cent c.i. 0.76 to 0.90) for the mortality model and 0.91 (0.88 to 0.94) for the relapse model; CITL was -0.19 (95 per cent c.i. -0.39 to -0.12) and - 0.01 (- 0.17 to -0.03) respectively. CONCLUSION: Relapse of infection and death after complicated intra-abdominal infections are common. Clinical prediction models were developed to identify patients at increased risk of relapse or death after treatment, these now require external validation.

Climate response to orbital forcing across the Oligocene-Miocene boundary.

Spectral analyses of an uninterrupted 5.5-million-year (My)-long chronology of late Oligocene-early Miocene climate and ocean carbon chemistry from two deep-sea cores recovered in the western equatorial Atlantic reveal variance concentrated at all Milankovitch frequencies. Exceptional spectral power in climate is recorded at the 406-thousand-year (ky) period eccentricity band over a 3.4-million-year period [20 to 23.4 My ago (Ma)] as well as in the 125- and 95-ky bands over a 1.3-million-year period (21.7 to 23.0 Ma) of suspected low greenhouse gas levels. Moreover, a major transient glaciation at the epoch boundary ( approximately 23 Ma), Mi-1, corresponds with a rare orbital congruence involving obliquity and eccentricity. The anomaly, which consists of low-amplitude variance in obliquity (a node) and a minimum in eccentricity, results in an extended period ( approximately 200 ky) of low seasonality orbits favorable to ice-sheet expansion on Antarctica.

Multiresolution forecasting for futures trading using wavelet decompositions.

We investigate the effectiveness of a financial time-series forecasting strategy which exploits the multiresolution property of the wavelet transform. A financial series is decomposed into an over complete, shift invariant scale-related representation. In transform space, each individual wavelet series is modeled by a separate multilayer perceptron (MLP). We apply the Bayesian method of automatic relevance determination to choose short past windows (short-term history) for the inputs to the MLPs at lower scales and long past windows (long-term history) at higher scales. To form the overall forecast, the individual forecasts are then recombined by the linear reconstruction property of the inverse transform with the chosen autocorrelation shell representation, or by another perceptron which learns the weight of each scale in the prediction of the original time series. The forecast results are then passed to a money management system to generate trades.

Weight perturbation: an optimal architecture and learning technique for analog VLSI feedforward and recurrent multilayer networks.

Previous work on analog VLSI implementation of multilayer perceptrons with on-chip learning has mainly targeted the implementation of algorithms such as back-propagation. Although back-propagation is efficient, its implementation in analog VLSI requires excessive computational hardware. It is shown that using gradient descent with direct approximation of the gradient instead of back-propagation is more economical for parallel analog implementations. It is shown that this technique (which is called ;weight perturbation') is suitable for multilayer recurrent networks as well. A discrete level analog implementation showing the training of an XOR network as an example is presented.