Statistical Learning in Palaeolimnology

Machine Learning Algorithms Predict Successful Weaning From Mechanical Ventilation Before Intubation: Retrospective Analysis From the Medical Information Mart for Intensive Care IV Database

BACKGROUND

The prediction of successful weaning from mechanical ventilation (MV) in advance of intubation can facilitate discussions regarding end-of-life care before unnecessary intubation.

OBJECTIVE

We aimed to develop a machine learning-based model that predicts successful weaning from ventilator support based on routine clinical and laboratory data taken before or immediately after intubation.

METHODS

We used the Medical Information Mart for Intensive Care IV database, which is an open-access database covering 524,740 admissions of 382,278 patients in Beth Israel Deaconess Medical Center, United States, from 2008 to 2019. We selected adult patients who underwent MV in the intensive care unit (ICU). Clinical and laboratory variables that are considered relevant to the prognosis of the patient in the ICU were selected. Data collected before or within 24 hours of intubation were used to develop machine learning models that predict the probability of successful weaning within 14 days of ventilator support. Developed models were integrated into an ensemble model. Performance metrics were calculated by 5-fold cross-validation for each model, and a permutation feature importance and Shapley additive explanations analysis was conducted to better understand the impacts of individual variables on outcome prediction.

RESULTS

Of the 23,242 patients, 19,025 (81.9%) patients were successfully weaned from MV within 14 days. Using the preselected 46 clinical and laboratory variables, the area under the receiver operating characteristic curve of CatBoost classifier, random forest classifier, and regularized logistic regression classifier models were 0.860 (95% CI 0.852-0.868), 0.855 (95% CI 0.848-0.863), and 0.823 (95% CI 0.813-0.832), respectively. Using the ensemble voting classifier using the 3 models above, the final model revealed the area under the receiver operating characteristic curve of 0.861 (95% CI 0.853-0.869), which was significantly better than that of Simplified Acute Physiology Score II (0.749, 95% CI 0.742-0.756) and Sequential Organ Failure Assessment (0.588, 95% CI 0.566-0.609). The top features included lactate and anion gap. The model's performance achieved a plateau with approximately the top 21 variables.

CONCLUSIONS

We developed machine learning algorithms that can predict successful weaning from MV in advance to intubation in the ICU. Our models can aid the appropriate management for patients who hesitate to decide on ventilator support or meaningless end-of-life care.

Linking 19th century European settlement to the disruption of a seabird's natural population dynamics

Recent estimates indicate that ∼70% of the world's seabird populations have declined since the 1950s due to human activities. However, for almost all bird populations, there is insufficient long-term monitoring to understand baseline (i.e., preindustrial) conditions, which are required to distinguish natural versus anthropogenically driven changes. Here, we address this lack of long-term monitoring data with multiproxy paleolimnological approaches to examine the long-term population dynamics of a major colony of Leach's Storm-petrel (Hydrobates leucorhous) on Grand Colombier Island in the St. Pierre and Miquelon archipelago-an overseas French territory in the northwest Atlantic Ocean. By reconstructing the last ∼5,800 y of storm-petrel dynamics, we demonstrate that this colony underwent substantial natural fluctuations until the start of the 19th century, when population cycles were disrupted, coinciding with the establishment and expansion of a European settlement. Our paleoenvironmental data, coupled with on-the-ground population surveys, indicate that the current colony is only ∼16% of the potential carrying capacity, reinforcing concerning trends of globally declining seabird populations. As seabirds are sentinel species of marine ecosystem health, such declines provide a call to action for global conservation. In response, we emphasize the need for enlarged protected areas and the rehabilitation of disturbed islands to protect ecologically critical seabird populations. Furthermore, long-term data, such as those provided by paleoecological approaches, are required to better understand shifting baselines in conservation to truly recognize current rates of ecological loss.

Bottom-Up Forces Drive Increases in the Abundance of Large Daphnids in Four Small Lakes Stocked with Rainbow Trout (Oncorhynchus mykiss), Interior British Columbia, Canada

The introduction of salmonids into lakes of western North America for sport fishing is a widespread phenomenon. While numerous investigations have documented cascading trophic interactions upon the introduction of fish into naturally fishless systems, little research has been done to investigate the importance of natural fish status (fishless vs. fish bearing) in modulating historical food web response to dual forcing by bottom-up (resource regulation from nutrients) and top-down (planktivory from stocked fish) processes. We used the paleolimnological record to reconstruct food web changes in four lakes in interior British Columbia that have been stocked with rainbow trout since the early to mid-1900s. Analysis of pigments, diatoms, and Cladocera was undertaken in cores from all lakes. We predicted that if fish were important in structuring cladoceran abundance and composition, we would document a decline in the abundance of large daphnids post-stocking in our two naturally fishless lakes, and little change in the two fish-bearing lakes. Instead, we documented increased abundance of large daphnids after stocking in all lakes in the early to mid-1900s, a finding inconsistent with size-selective predation from planktivorous fish. Further, our data suggest that deep, low-oxygen refugia may be important in sustaining populations of large Daphnia, a process which was enhanced by increased nutrients and lake production according to sub-fossil diatom and pigment analyses. This study shows that fish stocking does not invariably result in a decrease in large-bodied Cladocera and that nutrients and lake type can modulate the response of invertebrate planktivores.