Predicting Extubation Readiness in Preterm Infants Utilizing Machine Learning: A Diagnostic Utility Study

OBJECTIVE

The objective of this study was to predict extubation readiness in preterm infants using machine learning analysis of bedside pulse oximeter and ventilator data.

STUDY DESIGN

This is an observational study with prospective recordings of oxygen saturation (SpO2) and ventilator data from infants <30 weeks of gestation age. Research pulse oximeters collected SpO2 (1 Hz sampling rate) to quantify intermittent hypoxemia (IH). Continuous ventilator metrics were collected (4-5-minute sampling) from bedside ventilators. Data modeling was completed using unbiased machine learning algorithms. Three model sets were created using the following data source combinations: (1) IH and ventilator (IH + SIMV), (2) IH, and (3) ventilator (SIMV). Infants were also analyzed separated by postnatal age (infants <2 or ≥2 weeks of age). Models were compared by area under the receiver operating characteristic curve (AUC).

RESULTS

A total of 110 extubation events from 110 preterm infants were analyzed. Infants had a median gestation age and birth weight of 26 weeks and 825 g, respectively. Of the 3 models presented, the IH + SIMV model achieved the highest AUC of 0.77 for all infants. Separating infants by postnatal age increased accuracy further achieving AUC of 0.94 for <2 weeks of age group and AUC of 0.83 for ≥2 weeks group.

CONCLUSIONS

Machine learning analysis has the potential to enhance prediction accuracy of extubation readiness in preterm infants while utilizing readily available data streams from bedside pulse oximeters and ventilators.

Aerosolized Surfactant for Preterm Infants with Respiratory Distress Syndrome

Currently, the administration of surfactant to preterm infants with respiratory distress syndrome (RDS) mainly relies on intratracheal instillation; however, there is increasing evidence of aerosolized surfactant being an effective non-invasive strategy. We present a historical narrative spanning sixty years of development of aerosolization systems. We also offer an overview of the pertinent mechanisms needed to create and manage the ideal aerosolization system, with a focus on delivery, distribution, deposition, and dispersion in the context of the human lung. More studies are needed to optimize treatment with aerosolized surfactants, including determination of ideal dosages, nebulizer types, non-invasive interfaces, and breath synchronization. However, the field is rapidly evolving, and widespread clinical use may be achieved in the near future.

Response to first dose of inhaled albuterol in mechanically ventilated preterm infants

BACKGROUND

Bronchodilator responses among preterm infants are heterogeneous. Bedside measurements may identify responders.

STUDY DESIGN

Respiratory measurements (Resistance, Compliance, FiO₂) and pulse oximetry (SpO₂) patterns were downloaded from infants <30 weeks gestational age during the first 2 months of life. Mechanically ventilated infants who received albuterol were included (n = 33). Measurements were compared before and after first albuterol. Secondary analyses assessed subsequent doses.

RESULTS

Median gestation and birthweight were 25 3/7 weeks and 730 g, respectively. Mean Resistance decreased post-albuterol (p = 0.007). Sixty-eight percent of infants were responders based on decreased Resistance. Compliance and FiO₂ did not significantly differ. Percent time in hypoxemia (SpO₂< 85%) decreased post albuterol (p < 0.02). In responders, Resistance changes diminished with subsequent administration (all p = 0.01).

CONCLUSIONS

Ventilator resistance decreased in two-thirds of preterm infants, consistent with studies that utilized formal pulmonary function testing. Albuterol had a variable effect on delivered FiO₂; however, hypoxemia may be useful in evaluating albuterol response.

HTLV-1 infection in tropical spastic paraparesis: lymphocyte culture and serologic response

All 17 patients with tropical spastic paraparesis (TSP) in a series seen in the United Kingdom have antibodies to the human T cell leukemia virus type 1 (HTLV-1). Cultured peripheral blood lymphocytes from these patients formed multinucleated giant cells and reacted with sera and monoclonal antibodies to HTLV-1 in a manner identical to adult T cell leukemia-lymphoma (ATLL) patient lymphocytes. Western blot analysis failed to reveal any marked difference in the antigens recognized by sera from TSP and ATLL patients. The sera from TSP patients, their asymptomatic relatives and ATLL patients were titrated using the following assays: enzyme-linked immunosorbent assays (ELISA), particle agglutination, antibody-dependent cell-mediated cytotoxicity, and pseudotype neutralization. There were significantly stronger serologic responses in the TSP patients than in their relatives or ATLL patients. High antibody titers in the presence of replicating virus often reflect the antigen load; however, these data are also consistent with the suggestion that neurologic damage in TSP may be immunologically mediated.

Antibodies to poliovirus detected by immunoradiometric assay with a monoclonal antibody

An immunoradiometric assay (IRMA) for the assay of antibodies to poliovirus antigens is described. Dilutions of the test sera or whole (finger prick) blood samples were incubated with the Poliovirus antigen bound to a solid phase and the specific antibody was detected by the addition of a mouse anti-human IgG monoclonal antibody (McAb), which was itself revealed by iodinated sheep IgG anti-mouse F(ab). We have shown that this technique is suitable for the estimation of IgG anti-poliovirus antibodies induced in children following polio vaccine. The present study shows that SPRIA provides a simple and inexpensive method for serological studies with poliovirus particularly for use in large-scale surveys.

Activation of plasminogen as a feature in its assay

Seven laboratories collaborating in a study of two intermediate purity plasminogen preparations (64/23, 63/6) observed that the amount of activator (urokinase or streptokinase) and the time of activation of plasminogen influenced the amount of plasmin generated. Using casein and a synthetic polypeptide (S-2251) as substrates, the authors subsequently showed that complete activation of plasminogen was difficult to achieve without acitivity losses due to plasmin autodigestion. Comparison of the polypeptide subunits (on SDS electrophoresis) of the various plasminogen activation mixtures with their plasmin activity allowed the conclusion that at maximum generation of plasmin from plasminogen, some plasminogen remains in the form of an inactive plasminogen intermediate (PLG-i).

Fibrin subunits in venous and arterial thromboembolism

The subunit fibrin composition of thrombi of both venous and arterial origin was examined by sodium dodecyl sulphate gel electrophoresis. The thrombi were recovered by surgical intervention and all had the same fibrin subunit composition. The alpha chains were cross-linked as alpha-chain polymers alpha (p), the gamma chains as gamma-chain dimers (gamma-gamma) while the beta chains were not crosslinked; a further subunit of molecular weight 33 000 was shown to be present in all the fibrins examined and was a degradation fragment of the beta or gamma chains. This data suggests that the crosslinked alpha chains are rate limiting to the lysis of thrombi in vivo. The digestion of pulmonary emboli by plasmin yielded soluble degradation products which were identified as D dimer and E, the latter fragments being the major products obtained by the lysis of in-vitro made plasma clots. The similarity of the composition and lysis of thrombus fibrin to that formed in vitro augurs well for the justification of in-vitro research on mechanisms in thrombolysis.

Soluble high-molecular-weight E fragments in the plasmin-induced degradation products of cross-linked human fibrin

1. The factor XIII-mediated cross-linked alpha chains in fibrin have no effect on the nature of the fragments released during the solubilization of fibrin by plasmin. 2. Besides the known D dimer and E fragments solubilized during the lysis of cross-linked fibrin, other fragments have been observed on sodium dodecyl sulphate-polyacrylamide gel electrophoresis which have a molecular weight of about 135 000. After prolonged plasmin digestion, these fragments (U fragments) were no longer evident on the gels and the high-molecular-weight E antigen was absent. It is assumed that the E antigen was associated with the U fragments. These fragments also cross-reacted with an anti-D serum. 3. The U fragments have been tentatively presumed to be a factor XIII-mediated cross-linked D-E complex since they degrade only after prolonged degradation with plasmin. Whereas it is known that the fibrin D dimer fragment contains the cross-linked gamma chain residues of the originating fibrin, the presumed covalent cross-linking of the D-E fragments has not been proved. 4. The presence of these high-molecular-weight fragments, containing the E antigen, in cross-linked human fibrin digests should be taken into account in the development of D dimer assays to monitor fibrin lysis in vivo.