Contrastive Transfer Learning for Prediction of Adverse Events in Hospitalized Patients.

OBJECTIVE: Deterioration index (DI) is a computer-generated score at a specific frequency that represents the overall condition of hospitalized patients using a variety of clinical, laboratory and physiologic data. In this paper, a contrastive transfer learning method is proposed and validated for early prediction of adverse events in hospitalized patients using DI scores. METHODS AND PROCEDURES: An unsupervised contrastive learning (CL) model with a classifier is proposed to predict adverse outcome using a single temporal variable (DI scores). The model is pretrained on an unsupervised fashion with large-scale time series data and fine-tuned with retrospective DI score data. RESULTS: The performance of this model is compared with supervised deep learning models for time series classification. Results show that unsupervised contrastive transfer learning with a classifier outperforms supervised deep learning solutions. Pretraining of the proposed CL model with large-scale time series data and fine-tuning that with DI scores can enhance prediction accuracy. CONCLUSION: A relationship exists between longitudinal DI scores of a patient and the corresponding outcome. DI scores and contrastive transfer learning can be used to predict and prevent adverse outcomes in hospitalized patients. CLINICAL IMPACT: This paper successfully developed an unsupervised contrastive transfer learning algorithm for prediction of adverse events in hospitalized patients. The proposed model can be deployed in hospitals as an early warning system for preemptive intervention in hospitalized patients, which can mitigate the likelihood of adverse outcomes.

Clinical Implementation of an AI Early Warning System Algorithm: Lessons Learned.

The Deterioration Index (DI) is an automatic early warning system that utilizes a machine learning algorithm integrated into the electronic health record and was implemented to improve risk stratification of inpatients. Our pilot implementation showed superior diagnostic accuracy than standard care. A score >60 had a specificity of 88.5% and a sensitivity of 59.8% (PPV 0.1758, NPP 0.9817). However, acceptance in the clinical workflow was divided; nurses preferred standard care, while providers found it helpful.

Gram negative pyomyositis: Two case reports and a review of the literature.

Pyomyositis due to Gram negative bacteria is rare. Here we describe two cases in immunocompromised hosts. Both were bacteremic with a Gram-negative bacterium and had impaired immunity related to prolonged and ongoing chemotherapy for hematologic malignancies. Both eventually cleared the infection with a combination of local drainage and systemic antibiotics. This unusual diagnosis should be considered in an immunocompromised patient with muscle pain and fever.

Novel machine learning model to improve performance of an early warning system in hospitalized patients: a retrospective multisite cross-validation study.

Background: Threshold-based early warning systems (EWS) are used to predict adverse events (Aes). Machine learning (ML) algorithms that incorporate all EWS scores prior to an event may perform better in hospitalized patients. Methods: The deterioration index (DI) is a proprietary EWS. A threshold of DI >60 is used to predict a composite AE: all-cause mortality, cardiac arrest, transfer to intensive care, and evaluation by the rapid response team in practice. The DI scores were collected for adult patients (≥18 y-o) hospitalized on medical or surgical services during 8-23-2021 to 3-31-2022 from four different Mayo Clinic sites in the United States. A novel ML model was developed and trained on a retrospective cohort of hospital encounters. DI scores were represented in a high-dimensional space using random convolution kernels to facilitate training of a classifier and the area under the receiver operator characteristics curve (AUC) was calculated. Multiple time intervals prior to an AE were analyzed. A leave-one-out cross-validation protocol was used to evaluate performance across separate clinic sites. Findings: Three different classifiers were trained on 59,617 encounter-derived DI scores in high-dimensional feature space and the AUCs were compared to two threshold models. All three tested classifiers improved the AUC over the threshold approaches from 0.56 and 0.57 to 0.76, 0.85 and 0.94. Time interval analysis of the top performing classifier showed best accuracy in the hour before an event occurred (AUC 0.91), but prediction held up even in the 12 h before an AE (AUC 0.80 at minus 12 h, 0.81 at minus 9 h, 0.85 at minus 6 h, and 0.88 at minus 3 h before an AE). Multisite cross-validation using leave-one-out approach on data from four different clinical sites showed broad generalization performance of the top performing ML model with AUC of 0.91, 0.91, 0.95, and 0.91. Interpretation: A novel ML model that incorporates all the longitudinal DI scores prior to an AE in a hospitalized patient performs better at outcome prediction than the currently used threshold model. The use of clinical data, a generalized ML technique, and successful multisite cross-validation demonstrate the feasibility of our model in clinical implementation. Funding: No funding to report.

Incidence and Outcomes of Life-Threatening Events During Hospitalization: A Retrospective Study of Patients Treated with Naloxone.

BACKGROUND: We describe the clinical course of medical and surgical patients who received naloxone on general hospital wards for suspected opioid-induced respiratory depression (OIRD). METHODS: From May 2018 through October 2020, patients who received naloxone on hospital wards were identified and their records reviewed for incidence and clinical course. RESULTS: There were 86,030 medical and 106,807 surgical admissions. Naloxone was administered to 99 (incidence 11.5 [95% confidence interval 9.4-14.0] per 10,000 admissions) medical and 63 (5.9 [95% confidence interval 4.5-7.5]) surgical patients (P < 0.001). Median oral morphine equivalents administered within 24 hours before naloxone were 32 [15, 64] and 60 [32, 88] mg for medical and surgical patients, respectively (P = 0.002). The rapid response team was activated in 69 (69.7%) vs 42 (66.7%) and critical care transfers in 51 (51.5%) vs 30 (47.6%) medical and surgical patients respectively. The number of in-hospital deaths was 21 (21.2%) vs two (3.2%) and the number of discharges to hospice 12 (12.1%) vs one (1.6%) for medical and surgical patients, respectively (P = 0.001). Naloxone did not reverse OIRD in 38 (23%) patients, and these patients had more transfers to the intensive care unit and a higher 30-day mortality rate. CONCLUSION: Medical inpatients are more likely to suffer OIRD than are surgical inpatients despite lower opioid doses. Definitive OIRD was confirmed in 77% of patients because of immediate naloxone response, whereas 23% of patients did not respond, and this subset was more likely to need a higher level of care and had a higher 30-day mortality rate. Careful monitoring of mental and respiratory variables is necessary when opiates are used in hospital.

EVALI: A Mimicker of COVID-19.

E-cigarette or vaping product use-associated lung injury (EVALI) is a respiratory illness that has significant overlap with the symptoms of coronavirus disease 2019 (COVID-19). In the current pandemic, diagnosis of EVALI may be delayed because of anchoring bias when patients present with symptoms consistent with COVID-19. We present 3 cases of patients who were hospitalized with a presumed diagnosis of COVID-19 but were later diagnosed with EVALI.

Outcomes of COVID-19 With the Mayo Clinic Model of Care and Research.

OBJECTIVE: To report the Mayo Clinic experience with coronavirus disease 2019 (COVID-19) related to patient outcomes. METHODS: We conducted a retrospective chart review of patients with COVID-19 diagnosed between March 1, 2020, and July 31, 2020, at any of the Mayo Clinic sites. We abstracted pertinent comorbid conditions such as age, sex, body mass index, Charlson Comorbidity Index variables, and treatments received. Factors associated with hospitalization and mortality were assessed in univariate and multivariate models. RESULTS: A total of 7891 patients with confirmed COVID-19 infection with research authorization on file received care across the Mayo Clinic sites during the study period. Of these, 7217 patients were adults 18 years or older who were analyzed further. A total of 897 (11.4%) patients required hospitalization, and 354 (4.9%) received care in the intensive care unit (ICU). All hospitalized patients were reviewed by a COVID-19 Treatment Review Panel, and 77.5% (695 of 897) of inpatients received a COVID-19-directed therapy. Overall mortality was 1.2% (94 of 7891), with 7.1% (64 of 897) mortality in hospitalized patients and 11.3% (40 of 354) in patients requiring ICU care. CONCLUSION: Mayo Clinic outcomes of patients with COVID-19 infection in the ICU, hospital, and community compare favorably with those reported nationally. This likely reflects the impact of interprofessional multidisciplinary team evaluation, effective leveraging of clinical trials and available treatments, deployment of remote monitoring tools, and maintenance of adequate operating capacity to not require surge adjustments. These best practices can help guide other health care systems with the continuing response to the COVID-19 pandemic.

A Collaborative Multidisciplinary Approach to the Management of Coronavirus Disease 2019 in the Hospital Setting.

The novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) causes coronavirus disease 2019 (COVID-19), which presents an unprecedented challenge to medical providers worldwide. Although most SARS-CoV-2-infected individuals manifest with a self-limited mild disease that resolves with supportive care in the outpatient setting, patients with moderate to severe COVID-19 will require a multidisciplinary collaborative management approach for optimal care in the hospital setting. Laboratory and radiologic studies provide critical information on disease severity, management options, and overall prognosis. Medical management is mostly supportive with antipyretics, hydration, oxygen supplementation, and other measures as dictated by clinical need. Among its medical complications is a characteristic proinflammatory cytokine storm often associated with end-organ dysfunction, including respiratory failure, liver and renal insufficiency, cardiac injury, and coagulopathy. Specific recommendations for the management of these medical complications are discussed. Despite the issuance of emergency use authorization for remdesivir, there are still no proven effective antiviral and immunomodulatory therapies, and their use in COVID-19 management should be guided by clinical trial protocols or treatment registries. The medical care of patients with COVID-19 extends beyond their hospitalization. Postdischarge follow-up and monitoring should be performed, preferably using telemedicine, until the patients have fully recovered from their illness and are released from home quarantine protocols.

Cancer and older adult patient care.

Cancer occurs most frequently in patients aged 65 and older. With the increasing age of the world's population, there will be a significant increase in cancer diagnoses in older adults. Aging imposes a wide variety of physiological responses, comorbidities, and ailments, but older patients are less represented in clinical studies. Specific needs of older patients with cancer often go under-recognized and consequently unmet. In this review, common diagnoses that can affect the outcomes of this population, including frailty, malnutrition, and delirium, are discussed. Areas that need further research to improve the care of geriatric cancer patients, particularly in the hospital settings, are also identified.

Stercoral Ulcer-Associated Perforation and Chemotherapy.

Stercoral ulcer perforation is a life-threatening surgical condition which is thought to result from necrosis of the bowel wall due to an ischemic pressure by stool. This condition usually afflicts patients with chronic constipation. CT scan can identify most of the cases and emergent surgery is usually indicated.

A cyclophilin homology domain-independent role for Nup358 in HIV-1 infection.

The large nucleoporin Nup358/RanBP2 forms eight filaments that project from the nuclear pore into the cytoplasm where they function as docking platforms for nucleocytoplasmic transport receptors. RNAi screens have implicated Nup358 in the HIV-1 life cycle. The 164 C-terminal amino acids of this 3,224 amino acid protein are a cyclophilin homology domain (Nup358Cyp), which has potential to bind the HIV-1 capsid and regulate viral progress to integration. Here we examined the virological role of Nup358 in conditional knockout mouse cells and in RNAi-depleted human CD4⁺ T cells. Cre-mediated gene knockout was toxic and diminished HIV-1 infectivity. However, cellular health and HIV-1 susceptibility were coordinately preserved if, prior to gene inactivation, a transposon was used to express all of Nup358 or only the N-terminal 1340 amino acids that contain three FG repeats and a Ran-binding domain. HIV-1, but not N74D capsid-mutant HIV-1, was markedly sensitive to TNPO3 depletion, but they infected 1-1340 segment-complemented Nup358 knockout cells equivalently. Human and mouse CypA both rescued HIV-1 in CypA gene⁻/⁻ Jurkat cells and TRIM-Nup358Cyp fusions derived from each species were equally antiviral; each also inhibited both WT and N74D virus. In the human CD4⁺T cell line SupT1, abrupt Nup358 depletion reduced viral replication but stable Nup358-depleted cells replicated HIV-1 normally. Thus, human CD4⁺ T cells can accommodate to loss of Nup358 and preserve HIV-1 susceptibility. Experiments with cylosporine, viruses with capsids that do not bind cyclophilins, and growth arrest did not uncover viral dependency on the C-terminal domains of Nup358. Our data reinforce the virological importance of TNPO3 and show that Nup358 supports nuclear transport functions important for cellular homeostasis and for HIV-1 nuclear import. However, the results do not suggest direct roles for the Nup358 cyclophilin or SUMO E3 ligase domains in engaging the HIV-1 capsid prior to nuclear translocation.

LEDGF dominant interference proteins demonstrate prenuclear exposure of HIV-1 integrase and synergize with LEDGF depletion to destroy viral infectivity.

Target cell overexpression of the integrase binding domain (IBD) of LEDGF/p75 (LEDGF) inhibits HIV-1 replication. The mechanism and protein structure requirements for this dominant interference are unclear. More generally, how and when HIV-1 uncoating occurs postentry is poorly defined, and it is unknown whether integrase within the evolving viral core becomes accessible to cellular proteins prior to nuclear entry. We used LEDGF dominant interference to address the latter question while characterizing determinants of IBD antiviral activity. Fusions of green fluorescent protein (GFP) with multiple C-terminal segments of LEDGF inhibited HIV-1 replication substantially, but minimal chimeras of either polarity (GFP-IBD or IBD-GFP) were most effective. Combining GFP-IBD expression with LEDGF depletion was profoundly antiviral. CD4(+) T cell lines were rendered virtually uninfectable, with single-cycle HIV-1 infectivity reduced 4 logs and high-input (multiplicity of infection = 5.0) replication completely blocked. We restricted GFP-IBD to specific intracellular locations and found that antiviral activity was preserved when the protein was confined to the cytoplasm or directed to the nuclear envelope. The life cycle block triggered by the cytoplasm-restricted protein manifested after nuclear entry, at the level of integration. We conclude that integrase within the viral core becomes accessible to host cell protein interaction in the cytoplasm. LEDGF dominant interference and depletion impair HIV-1 integration at distinct postentry stages. GFP-IBD may trigger premature or improper integrase oligomerization.

Chromatin tethering and retroviral integration: recent discoveries and parallels with DNA viruses.

Permanent integration of the viral genome into a host chromosome is an essential step in the life cycles of lentiviruses and other retroviruses. By archiving the viral genetic information in the genome of the host target cell and its progeny, integrated proviruses prevent curative therapy of HIV-1 and make the development of antiretroviral drug resistance irreversible. Although the integration reaction is known to be catalyzed by the viral integrase (IN), the manner in which retroviruses engage and attach to the chromatin target is only now becoming clear. Lens epithelium-derived growth factor (LEDGF/p75) is a ubiquitously expressed nuclear protein that binds to lentiviral IN protein dimers at its carboxyl terminus and to host chromatin at its amino terminus. LEDGF/p75 thus tethers ectopically expressed IN to chromatin. It also protects IN from proteosomal degradation and can stimulate IN catalysis in vitro. HIV-1 infection is inhibited at the integration step in LEDGF/p75-deficient cells, and the characteristic lentiviral preference for integration into active genes is also reduced. A model in which LEDGF/p75 acts to tether the viral preintegration complex to chromatin has emerged. Intriguingly, similar chromatin tethering mechanisms have been described for other retroelements and for large DNA viruses. Here we review the evidence supporting the LEDGF/p75 tethering model and consider parallels with these other viruses.

LEDGF/p75 proteins with alternative chromatin tethers are functional HIV-1 cofactors.

LEDGF/p75 can tether over-expressed lentiviral integrase proteins to chromatin but how this underlies its integration cofactor role for these retroviruses is unclear. While a single integrase binding domain (IBD) binds integrase, a complex N-terminal domain ensemble (NDE) interacts with unknown chromatin ligands. Whether integration requires chromatin tethering per se, specific NDE-chromatin ligand interactions or other emergent properties of LEDGF/p75 has been elusive. Here we replaced the NDE with strongly divergent chromatin-binding modules. The chimeras rescued integrase tethering and HIV-1 integration in LEDGF/p75-deficient cells. Furthermore, chromatin ligands could reside inside or outside the nucleosome core, and could be protein or DNA. Remarkably, a short Kaposi's sarcoma virus peptide that binds the histone 2A/B dimer converted GFP-IBD from an integration blocker to an integration cofactor that rescues over two logs of infectivity. NDE mutants were corroborative. Chromatin tethering per se is a basic HIV-1 requirement and this rather than engagement of particular chromatin ligands is important for the LEDGF/p75 cofactor mechanism.

Severe pleuropulmonary paragonimiasis 8 years after emigration from a region of endemicity.

A Laotian man who had resided only in the north-central United States for 8 years sought care for an acute, progressive syndrome of severe dyspnea, chest pain, bilateral pneumothoraces, lung and liver nodules, and marked peripheral blood eosinophilia. He habitually ate raw crabmeat imported pickled or frozen from Southeast Asia; he denied eating local crustaceans. Ova consistent with the lung fluke Paragonimus westermani were identified in a bronchoalveolar lavage specimen, and the eosinophilia and pulmonary symptoms resolved with praziquantel therapy.