Interpretable Machine Learning–Based Prediction of Intraoperative Cerebrospinal Fluid Leakage in Endoscopic Transsphenoidal Pituitary Surgery: A Pilot Study

Purpose Transsphenoidal surgery (TSS) for pituitary adenomas can be complicated by the occurrence of intraoperative cerebrospinal fluid (CSF) leakage (IOL). IOL significantly affects the course of surgery predisposing to the development of postoperative CSF leakage, a major source of morbidity and mortality in the postoperative period. The authors trained and internally validated the Random Forest (RF) prediction model to preoperatively identify patients at high risk for IOL. A locally interpretable model-agnostic explanations (LIME) algorithm is employed to elucidate the main drivers behind each machine learning (ML) model prediction. Methods The data of 210 patients who underwent TSS were collected; first, risk factors for IOL were identified via conventional statistical methods (multivariable logistic regression). Then, the authors trained, optimized, and audited a RF prediction model. Results IOL reported in 45 patients (21.5%). The recursive feature selection algorithm identified the following variables as the most significant determinants of IOL: Knosp's grade, sellar Hardy's grade, suprasellar Hardy's grade, tumor diameter (on X, Y, and Z axes), intercarotid distance, and secreting status (nonfunctioning and growth hormone [GH] secreting). Leveraging the predictive values of these variables, the RF prediction model achieved an area under the curve (AUC) of 0.83 (95% confidence interval [CI]: 0.78; 0.86), significantly outperforming the multivariable logistic regression model (AUC = 0.63). Conclusion A RF model that reliably identifies patients at risk for IOL was successfully trained and internally validated. ML-based prediction models can predict events that were previously judged nearly unpredictable; their deployment in clinical practice may result in improved patient care and reduced postoperative morbidity and healthcare costs.

Relationship between Eating Alone and Poor Appetite Using the Simplified Nutritional Appetite Questionnaire

One prominent factor associated with malnutrition is poor appetite. In Japan, the number of older adults living alone has increased annually. Those living alone tended to eat alone, which may lead to poor appetite. This study aimed to investigate the association between eating alone and poor appetite using an index called the Simplified Nutritional Appetite Questionnaire (SNAQ). We surveyed 818 people aged 70 and over in Takashimadaira, Itabashi-ku, Tokyo, Japan, in 2016. Comparisons were made between two groups, a poor appetite group (n = 295) and a good appetite group (n = 523), and results indicate that the poor appetite group had a higher rate of eating alone than the good appetite group (38.0% vs. 20. 1%: p < 0.001). Multivariable logistic regression (OR; 95%CI) was performed and poor appetite was significantly associated with the Geriatric Depression Scale (GDS) score (1.707; 1.200–2.427), the number of medications (1.061; 1.007–1.118), JST score (0.894; 0.841–0.950), the indication of “very healthy” on a self-rated health scale (0.343; 0.152–0.774), and reports of eating alone (1.751; 1.130–2.712). Our results suggest that eating alone is associated with a poor appetite.

Fluid Overload Phenotypes in Critical Illness—A Machine Learning Approach

Background: The detrimental impact of fluid overload (FO) on intensive care unit (ICU) morbidity and mortality is well known. However, research to identify subgroups of patients particularly prone to fluid overload is scarce. The aim of this cohort study was to derive “FO phenotypes” in the critically ill by using machine learning techniques. Methods: Retrospective single center study including adult intensive care patients with a length of stay of ≥3 days and sufficient data to compute FO. Data was analyzed by multivariable logistic regression, fast and frugal trees (FFT), classification decision trees (DT), and a random forest (RF) model. Results: Out of 1772 included patients, 387 (21.8%) met the FO definition. The random forest model had the highest area under the curve (AUC) (0.84, 95% CI 0.79–0.86), followed by multivariable logistic regression (0.81, 95% CI 0.77–0.86), FFT (0.75, 95% CI 0.69–0.79) and DT (0.73, 95% CI 0.68–0.78) to predict FO. The most important predictors identified in all models were lactate and bicarbonate at admission and postsurgical ICU admission. Sepsis/septic shock was identified as a risk factor in the MV and RF analysis. Conclusion: The FO phenotypes consist of patients admitted after surgery or with sepsis/septic shock with high lactate and low bicarbonate.

Acceptance of a Third Dose of COVID-19 Vaccine and Associated Factors in China Based on Health Belief Model: A National Cross-Sectional Study

COVID-19 infections are returning to many countries because of the emergence of variants or declining antibody levels provided by vaccines. An additional dose of vaccination is recommended to be a considerable supplementary intervention. We aim to explore public acceptance of the third dose of the COVID-19 vaccine and related influencing factors in China. This nationwide cross-sectional study was conducted in the general population among 31 provinces in November, 2021. We collected information on basic characteristics, vaccination knowledge and attitudes, and vaccine-related health beliefs of the participants. Univariable and multivariable logistic regression models were used to assess factors associated with the acceptance of a third COVID-19 vaccine. A total of 93.7% (95% CI: 92.9–94.6%) of 3119 Chinese residents were willing to receive a third dose of the COVID-19 vaccine. Individuals with low level of perceived susceptibility, perceived benefit, cues to action cues, and high level of perceived barriers, old age, low educational level, low monthly household income, and low knowledge score on COVID-19 were less likely to have the acceptance of a third dose of COVID-19 (all p < 0.05). In the multivariable logistic regression model, acceptance of the third dose of COVID-19 vaccine was mainly related to previous vaccination history [Sinopharm BBIP (aOR = 6.55, 95% CI 3.30–12.98), Sinovac (aOR = 5.22, 95% CI:2.72–10.02), Convidecia (aOR = 5.80, 95% CI: 2.04–16.48)], high level of perceived susceptibility (aOR = 2.48, 95% CI: 1.48–4.31) and high level of action cues (aOR = 23.66, 95% CI: 9.97–56.23). Overall, residents in China showed a high willingness to accept the third dose of COVID-19 vaccines, which can help vaccine manufacturers in China to manage the vaccine production and distribution for the huge domestic and international vaccine demand. Relevant institutions could increase people’s willingness to booster shots by increasing initial COVID-19 vaccination rates, public’s perception of COVID-19 susceptibility and cues to action through various strategies and channels. Meanwhile, it also has certain reference significance for other countries to formulate vaccine promotion strategies.

Radiomics nomogram for the preoperative prediction of lymph node metastasis in pancreatic ductal adenocarcinoma

Purpose To develop and validate a radiomics nomogram for the preoperative prediction of lymph node (LN) metastasis in pancreatic ductal adenocarcinoma (PDAC). Materials and methods In this retrospective study, 225 patients with surgically resected, pathologically confirmed PDAC underwent multislice computed tomography (MSCT) between January 2014 and January 2017. Radiomics features were extracted from arterial CT scans. The least absolute shrinkage and selection operator method was used to select the features. Multivariable logistic regression analysis was used to develop the predictive model, and a radiomics nomogram was built and internally validated in 45 consecutive patients with PDAC between February 2017 and December 2017. The performance of the nomogram was assessed in the training and validation cohort. Finally, the clinical usefulness of the nomogram was estimated using decision curve analysis (DCA). Results The radiomics signature, which consisted of 13 selected features of the arterial phase, was significantly associated with LN status (p < 0.05) in both the training and validation cohorts. The multivariable logistic regression model included the radiomics signature and CT-reported LN status. The individualized prediction nomogram showed good discrimination in the training cohort [area under the curve (AUC), 0.75; 95% confidence interval (CI), 0.68–0.82] and in the validation cohort (AUC, 0.81; 95% CI, 0.69–0.94) and good calibration. DCA demonstrated that the radiomics nomogram was clinically useful. Conclusions The presented radiomics nomogram that incorporates the radiomics signature and CT-reported LN status is a noninvasive, preoperative prediction tool with favorable predictive accuracy for LN metastasis in patients with PDAC.

Predictors of Mortality in Hemodialyzed Patients after SARS-CoV-2 Infection

Introduction: The determinants of COVID-19 mortality are well-characterized in the general population. Less numerous and inconsistent data are among the maintenance hemodialysis (HD) patients, who are the population most at risk of an unfavorable prognosis. Methods: In this retrospective cohort study we included all adult HD patients from the Pomeranian Voivodeship, Poland, with laboratory-confirmed SARS-CoV-2 infection hospitalized between 6 October 2020 and 28 February 2021, both those who survived, and also those who died. Demographic, clinical, treatment, and laboratory data on admission, were extracted from the electronic medical records of the dedicated hospital and patients’ dialysis unit, and compared between survivors and non-survivors. We used univariable and multivariable logistic regression methods to explore the risk factors associated with 3-month all-cause mortality. Results: The 133 patients (53.38% males) aged 73.0 (67–79) years, with a median duration of hemodialysis of 42.0 (17–86) months, were included in this study. At diagnosis, the majority were considered to have a mild course (34 of 133 patients were asymptomatic, another 63 subjects presented mild symptoms), while 36 (27.07%) patients had low blood oxygen saturation and required oxygen supplementation. Three-month mortality was 39.08% including an in-hospital case fatality rate of 33.08%. Multivariable logistic regression showed that the frailty clinical index of 4 or greater (OR 8.36, 95%CI 1.81–38.6; p < 0.01), D-Dimer of 1500 ng/mL or greater (6.00, 1.94–18.53; p < 0.01), and CRP of >118 mg/L at admission (3.77 1.09–13.01; p = 0.04) were found to be predictive of mortality. Conclusion: Very high 3-month all-cause mortality in hospitalized HD patients was determined mainly by frailty. High CRP and D-dimer levels upon admission further confer mortality risk.

The Construction and Development of a Clinical Prediction Model to Assess Lymph Node Metastases in Osteosarcoma

Background: This study aimed to construct a clinical prediction model for osteosarcoma patients to evaluate the influence factors for the occurrence of lymph node metastasis (LNM).Methods: In our retrospective study, a total of 1,256 patients diagnosed with chondrosarcoma were enrolled from the SEER (Surveillance, Epidemiology, and End Results) database (training cohort, n = 1,144) and multicenter dataset (validation cohort, n = 112). Both the univariate and multivariable logistic regression analysis were performed to identify the potential risk factors of LNM in osteosarcoma patients. According to the results of multivariable logistic regression analysis, A nomogram were established and the predictive ability was assessed by calibration plots, receiver operating characteristics (ROCs) curve, and decision curve analysis (DCA). Moreover, Kaplan-Meier plot of overall survival (OS) was plot and a web calculator visualized the nomogram.Results: Five independent risk factors [chemotherapy, surgery, lung metastases, lymphatic metastases (M-stage) and tumor size (T-stage)] were identified by multivariable logistic regression analysis. What's more, calibration plots displayed great power both in training and validation group. DCA presented great clinical utility. ROCs curve provided the predictive ability in the training cohort (AUC = 0.805) and the validation cohort (AUC = 0.808). Moreover, patients in LNN group had significantly better survival than that in LNP group both in training and validation group.Conclusion: In this study, we constructed and developed a nomogram with risk factors, which performed well in predicting risk factors of LNM in osteosarcoma patients. It may give a guide for surgeons and oncologists to optimize individual treatment and make a better clinical decision.

The discrepancy between radiographically-assessed and self-recognized hallux valgus in a large population-based cohort

Background There has been a paucity of literature revealing the discrepancy between self-recognition about hallux valgus (HV) and radiographically-evaluated foot configuration. Knowing this discrepancy will help to make a comparative review of the findings of previous literatures about epidemiological studies about the prevalence of HV. Questions/purposes (1) Is there a discrepancy between radiographically-assessed and self-recognized HV in the general population? (2) What factors affect the self-recognition of HV in the general population? Methods The fifth survey of the Research on Osteoarthritis/Osteoporosis against Disability study involved 1996 participants who had undergone anterior-posterior radiography of bilateral feet and answered a simple dichotomous questionnaire on self-recognition of HV. Measurements of the HV angle (HVA), interphalangeal angle of the hallux (IPA), and intermetatarsal angle between 1st and 2nd metatarsals (IMA) were performed using radiographs. Radiographic diagnosis of HV was done using the definition of hallux valgus angle of 20° or more. After univariate comparison of the participant backgrounds and radiographic measurements between participants with or without self-recognition of HV, multivariable logistic regression analysis was conducted in order to reveal independent factors affecting self-recognition. Results Significant difference was found between the prevalence of radiographically-assessed and self-recognized HV (29.8% vs. 16.5%, p < 0.0001). The prevalence of self-recognized HV increased with the progression of HV severity from a single-digit percentage (normal grade, HVA < 20°) up to 100% (severe grade, HVA ≥ 40°). A multivariable logistic regression analysis demonstrated that HVA, IMA, and female sex were independent positive factors for self-recognition of HV (HVA [per 1° increase]: OR, 1.18; 95% CI, 1.15–1.20; p < 0.0001; IMA [per 1° increase]: OR, 1.15; 95% CI, 1.09–1.20; p < 0.0001; and female sex [vs. male sex]: OR, 3.47; 95% CI, 2.35–5.18; p < 0.0001). Conclusions There was a significant discrepancy between radiographically-assessed and self-recognized HV which narrowed with the progressing severity of HV. HVA, IMA, and female sex were independent positive factors for self-recognition of HV. Attention needs to be paid to potentially lowered prevalence of HV in epidemiological studies using self-reporting based on self-recognition.

Effectiveness of COVID-19 vaccines against Omicron or Delta infection

Background The incidence of SARS-CoV-2 infection, including among those who have received 2 doses of COVID-19 vaccines, has increased substantially since Omicron was first identified in the province of Ontario, Canada. Methods Applying the test-negative design to linked provincial data, we estimated vaccine effectiveness against infection (irrespective of symptoms or severity) caused by Omicron or Delta between November 22 and December 19, 2021. We included individuals who had received at least 2 COVID-19 vaccine doses (with at least 1 mRNA vaccine dose for the primary series) and used multivariable logistic regression to estimate the effectiveness of two or three doses by time since the latest dose. Results We included 3,442 Omicron-positive cases, 9,201 Delta-positive cases, and 471,545 test-negative controls. After 2 doses of COVID-19 vaccine, vaccine effectiveness against Delta infection declined steadily over time but recovered to 93% (95%CI, 92-94%) ≥7 days after receiving an mRNA vaccine for the third dose. In contrast, receipt of 2 doses of COVID-19 vaccines was not protective against Omicron. Vaccine effectiveness against Omicron was 37% (95%CI, 19-50%) ≥7 days after receiving an mRNA vaccine for the third dose. Conclusions Two doses of COVID-19 vaccines are unlikely to protect against infection by Omicron. A third dose provides some protection in the immediate term, but substantially less than against Delta. Our results may be confounded by behaviours that we were unable to account for in our analyses. Further research is needed to examine protection against severe outcomes.