Interpretable machine learning in predicting drug-induced liver injury among tuberculosis patients: model development and validation study.

BACKGROUND: The objective of this research was to create and validate an interpretable prediction model for drug-induced liver injury (DILI) during tuberculosis (TB) treatment. METHODS: A dataset of TB patients from Ningbo City was used to develop models employing the eXtreme Gradient Boosting (XGBoost), random forest (RF), and the least absolute shrinkage and selection operator (LASSO) logistic algorithms. The model's performance was evaluated through various metrics, including the area under the receiver operating characteristic curve (AUROC) and the area under the precision recall curve (AUPR) alongside the decision curve. The Shapley Additive exPlanations (SHAP) method was used to interpret the variable contributions of the superior model. RESULTS: A total of 7,071 TB patients were identified from the regional healthcare dataset. The study cohort consisted of individuals with a median age of 47 years, 68.0% of whom were male, and 16.3% developed DILI. We utilized part of the high dimensional propensity score (HDPS) method to identify relevant variables and obtained a total of 424 variables. From these, 37 variables were selected for inclusion in a logistic model using LASSO. The dataset was then split into training and validation sets according to a 7:3 ratio. In the validation dataset, the XGBoost model displayed improved overall performance, with an AUROC of 0.89, an AUPR of 0.75, an F1 score of 0.57, and a Brier score of 0.07. Both SHAP analysis and XGBoost model highlighted the contribution of baseline liver-related ailments such as DILI, drug-induced hepatitis (DIH), and fatty liver disease (FLD). Age, alanine transaminase (ALT), and total bilirubin (Tbil) were also linked to DILI status. CONCLUSION: XGBoost demonstrates improved predictive performance compared to RF and LASSO logistic in this study. Moreover, the introduction of the SHAP method enhances the clinical understanding and potential application of the model. For further research, external validation and more detailed feature integration are necessary.

Incidence, risk factors, and outcomes of chylothorax after cardiac procedure in the United States.

Background: To examine the epidemiology and risk factors of chylothorax after cardiac procedure in the United States using a contemporary nationally representative database. Methods: We identified postoperative chylothorax events through National Inpatient Sample database (2016-2019) and compared baseline demographics, comorbidities, and in-hospital outcomes between hospitalizations with and without postoperative chylothorax. The Cochrane-Armitage test was used to analyze trends in incidence rates. Multivariable Poisson regression models were used to identify potential risk factors for postoperative chylothorax after cardiac procedure. Results: A total of 819 (0.24%) admissions were associated with postoperative chylothorax. The crude and standardized incidence rates of chylothorax were 23.7 (95%CI, 22.1-25.4) and 61.5 per 10,000 cardiac procedure-related admissions, respectively, with no significant temporal change in incidence rate over the study period (Ptrend = 0.5249). Infants [adjusted rate ratio (aRR), 117.3, 95% confidence interval (CI), 94.5-145.5] and children (aRR, 60.2, 95%CI, 48.0-75.5) were more likely to develop chylothorax compared to adults. Heart and great vessel procedures (aRR, 4.36, 95%CI, 3.61-5.26), septal repair (aRR, 1.91, 95%CI, 1.58-2.29), heart transplant (aRR, 5.68, 95%CI, 4.55-7.10) and pericardial procedures (aRR, 4.04, 95%CI, 3.32-4.91) were associated with elevated risk for chylothorax. Admissions with chylothorax were associated with higher inpatient mortality (4.9% vs. 3.0%, p<0.0001), longer inpatient stay, higher costs and greater perioperative complication burden. Conclusions: Following cardiac procedures, chylothorax is an uncommon but serious complication that affects the prognosis. The analysis reveals varying incidence rates across age groups and specific surgical procedures, with infants at elevated risk.

Association between hospital racial composition and aortic valve replacement outcomes: A national inpatients sample database analysis.

BACKGROUND: Racial and ethnic disparities exist in the outcomes following surgical aortic valve replacement (SAVR) and transcatheter aortic valve implantation (TAVI). However, it is unclear whether hospital racial composition contributes to these racial disparities. METHODS: We analyzed the National Inpatient Sample (NIS) database from 2015 to 2019 to identify patients with aortic stenosis (AS) who received SAVR and TAVI. The Racial/Ethnic Diversity Index (RDI) was used to assess hospital racial composition as the proportion of nonwhite patients to total hospital admissions. Hospitals were categorized into RDI quintiles. Textbook outcome (TO) was defined as no in-hospital mortality, no postoperative complications and no prolonged length of stay (LOS). Multivariable mixed generalized linear models were conducted to assess the association between RDI and post-SAVR and post-TAVI outcomes. Moreover, quantile regression was used to assess the additional cost and length of stay associated with the RDI quintile. RESULTS: The study included 82,502 SAVR or TAVI performed across 3285 hospitals, with 47.4% isolated SAVR and 52.5% isolated TAVI. After adjustment, quintiles 4 and 5 demonstrated significantly lower odds of TO than the lowest RDI quintile in both the SAVR cohort (quintile 4, 0.79 [95% CI, 0.73-0.85]; quintile 5, 0.79 [95% CI, 0.73-0.86]) and TAVI cohort (quintile 4, 0.88 [95% CI, 0.82-0.95]; quintile 5, 0.80 [95% CI, 0.74-0.86]). Despite non-observable differences in in-hospital mortality across all RDI quintiles, the rate of AKI and blood transfusion increased with increasing RDI for both cohorts. Further, Higher RDI quintiles were associated with increased costs and longer LOS. From 2015 to 2019, post-TAVI outcomes improved across all RDI quintiles. CONCLUSIONS: Hospitals with a higher RDI experienced lower TO achievements, increased AKI, and blood transfusion, along with extended LOS and higher costs. Importantly, post-TAVI outcomes improved from 2015 to 2019 across all RDI groups.

Impacts of national volume-based drug procurement policy on the utilization and costs of antihypertensive drugs in a Chinese medicine hospital: an interrupted time series analysis of 5138 patients.

Objectives: The study aimed to estimate the effects of National Volume-based Drug Procurement (NVBP) policy on drug utilization and medical expenditures of hypertension patients in public medical institutions in mainland China. Methods: This study used patient-level data based on electronic health records retrieved from the hospital information system of Nanjing Hospital of Chinese Medicine. Data on patients with hypertension who received care at this institution between 2016 and 2021 was used for analysis. Segmented linear regression models incorporating Interrupted Time Series (ITS) analysis were adopted to examine the effects of NVBP policy on drug utilization and health expenditures of eligible patients. Drug utilization volume and health expenditures were the primary outcomes used to assess the policy effects, and were measured using the prescription proportion of each drug class and the overall per-encounter treatment costs. Results: After the implementation of NVBP policy, the volume of non-winning drugs decreased from 54.42% to 36.25% for outpatient care and from 35.62% to 15.65% for inpatient care. The ITS analysis showed that the volume of bid-winning drugs in outpatient and inpatient settings increased by 9.55% (p < 0.001) and 6.31% (p < 0.001), respectively. The volume changes in non-volume based purchased (non-VBP) drugs differed between outpatients and inpatients. The proportion of non-VBP drugs immediately increased by 5.34% (p = 0.002) overall, and showed an upward trend in the outpatient setting specially (p < 0.001) during the post-intervention period. However, no significant differences were observed in the proportion of non-VBP drugs in inpatient setting (p > 0.05) in term of level change (p > 0.05) or trend change (p > 0.05). The average per-visit expenditures of outpatients across all drug groups exhibited an upward trend (p < 0.05) post policy intervention. In addition, a similar increase in the overall costs for chemical drugs were observed in inpatient settings (coefficient = 2,599.54, p = 0.036), with no statistically significant differences in the regression slope and level (p = 0.814). Conclusion: The usage proportion of bid-winning drugs increased significantly post policy intervention, indicating greater use of bid-winning drugs and the corresponding substitution of non-winning hypertensive drugs. Drug expenditures for outpatients and health expenditures per visit for inpatients also exhibited an upward trend, suggesting the importance of enhanced drug use management in Traditional Chinese Medicine hospital settings.

Elevated respiratory complications following robotic-assisted lobectomy: A national cohort analysis of 26,140 cases.

BACKGROUND: Robotic-assisted lobectomy (RL) is increasingly used nationally, but little comparative data exist on its safety compared with open lobectomy (OL) or video-assisted lobectomy (VL). This study aimed to estimate the risk of perioperative complications for RL, VL, or OL. METHODS: Admissions were identified from the hospital administrative data collected between 2015 and 2019. Propensity score matching and inverse probabilistic weighting were used to account for selection bias. Logistic and quantile regression models were applied to determine perioperative outcome differences. RESULTS: We identified 26,140 cases of which 5337 (20.4%), 12,680 (48.5%), and 8123 (31.1%) underwent RL, VL, and OL. RL and VL were associated with lower complication rates, shorter lengths of stay, and fewer mortality risks. RL was associated with significantly elevated risks for perioperative respiratory complications (adjusted odds ratio 1.10, p = 0.010). CONCLUSION: Relatively low rates of perioperative complications for VL and RL, and higher respiratory complication rates in RL are concerning.

Development and validation of a patient satisfaction survey for pharmaceutical service at primary care settings.

Patient satisfaction is a key quality indicator of pharmacy service. However, there are few studies that develop and validate patient satisfaction surveys applied to pharmaceutical services in primary care settings. It is imperative to establish a well-validated multidimensional instrument for evaluating the viability and sustainability of pharmacy service across geographically diverse regions in low- and middle-income countries. To develop and validate a patient satisfaction instrument for community pharmaceutical services, we carried out a cross-sectional survey in seven provinces across China. The study was conducted in four phases: (i) literature review-based item generation, (ii) expert panel-endorsed questionnaire refinement, (iii) pilot questionnaire development, and (iv) psychometric validation. Survey respondents were standard patients recruited locally and trained to conduct unannounced visits to preselected primary care centers. Between December 2020 and November 2021, the pilot survey comprised a total of 166 unannounced standard patient visits from 125 health-care facilities. The final 24-item Likert-type instrument encompassed five domains: relationship, medication counseling, empathy, accessibility, and overall satisfaction. The satisfactory survey revealed excellent internal consistency. Factor analyses resulted in a 4-factor solution that accounted for 70.7% variance. The results suggest that the questionnaire is a valid and reliable instrument, which has been taken an important step to evaluate patient satisfaction with pharmaceutical services in Chinese primary care settings. Further research on its cross-culture adaptation and applicability in urban retail pharmacy settings is warranted.

Effectiveness of Prophylactic Use of Hepatoprotectants for Tuberculosis Drug-Induced Liver Injury: A Population-Based Cohort Analysis Involving 6,743 Chinese Patients.

Background: Tuberculosis drug-induced liver injury (TB-DILI) is a common and potentially severe adverse drug reaction leading to treatment interruption and treatment failure. The real-world preventive effectiveness of hepatoprotective agents for DILI is not well described. The aim of the study was to evaluate the patterns of prophylactic therapies in real-world settings and risks of DILI among adult TB patients without known risk factors for DILI. Methods: This is a population-based retrospective cohort study of patients receiving first-line anti-tuberculosis drugs in the Chinese Center for Disease Control and Prevention (CDC) TB registry linked to the Ningbo Regional Health Care Database (NRHCD) between 2015 and 2020. The primary exposure was any use of chemopreventive agents including silymarin and/or glycyrrhetinic acid during the 30-day period prior to TB diagnosis (index date). The main outcome measure was the occurrence of newly onset DILI following TB treatment. Eligible patients were followed until the earliest of any DILI, treatment discontinuation, death, or end of the study period (30 June 2020). Marginal structural competing risk models and Cox models via inverse probability treatment weights using high-dimensional propensity scores were used to estimate subdistribution hazard risks (SHR) and 95% confidence intervals (CIs) for DILI risks, with adjustment for age, sex, TB-related characteristics, and comorbidities. Results: We identified a cohort of 6,743 adult patients with TB (mean age of 47.1 [SD 18.7] years; 65.80% male), of whom 2,886 (42.8%) patients received hepatoprotective agents. A total of 895 DILI events and 111 all-cause death events without DILI were observed over a median follow-up of 367 days post-TB diagnosis. The incidence rates of composite outcomes combining DILI and all-cause mortality were 248.9 and 222.3 per 1,000 person-years in the hepatoprotective agent exposed and unexposed groups (relative hazard ratio 1.35, 95% CI 1.11-1.64), respectively. The incidence rates of DILI were 223.7 and 196.1 per 1,000 person-years in the hepatoprotective agent exposed and unexposed groups (relative hazard ratio 1.38, 95% CI 1.12-1.71), respectively. Patients with any chemopreventive agent use had comparable liver function changes as evidenced by laboratory tests. Conclusion: A non-trivial number of adult patients received chemopreventive agents for TB-DILI. However, prophylactic utilization of hepatoprotective agents was not associated with a reduction in TB-DILI risks.

Intraoperative venous congestion rather than hypotension is associated with acute adverse kidney events after cardiac surgery: a retrospective cohort study.

BACKGROUND: The pathophysiological mechanisms by which venous congestion and hypotension lead to acute adverse kidney events after cardiac surgery with cardiopulmonary bypass have not been elucidated. We tested the hypothesis that intraoperative hypotension and venous congestion are associated with acute kidney injury and acute kidney disease. METHODS: Primary exposures were venous congestion and intraoperative hypotension defined by central venous pressure ≥12, 16, or 20 mm Hg or mean arterial pressure ≤55, 65, or 75 mm Hg. The primary outcomes were acute kidney injury and acute kidney disease. Multivariable logistic regression and Cox proportional hazard models were used, adjusted for relevant confounding factors and multiple comparisons. RESULTS: Of 5127 eligible subjects, 1070 (20.9%) and 327 (7.2%) developed acute kidney injury and acute kidney disease, respectively. The occurrence of acute kidney injury was statistically associated with both venous congestion and intraoperative hypotension. The cumulative incidence rate for new onset acute kidney disease was 1.34 (95% confidence interval [CI], 1.21-1.60) per 100 person-days. Acute kidney disease was significantly associated with each 10 min epoch of central venous pressure ≥12 mm Hg (hazard ratio [HR]=1.03; 99% CI, 1.01-1.06; P<0.001), ≥16 mm Hg (HR=1.04; 99% CI, 1.01-1.07; P<0.001), and ≥20 mm Hg (HR=1.07; 99% CI, 1.02-1.13; P<0.001). Venous congestion was associated with an 8-17% increased risk for de novo renal replacement therapy. In contrast, intraoperative hypotension was not associated with development of acute kidney disease. CONCLUSION: Although both venous congestion and intraoperative hypotension are associated with acute kidney injury, only venous congestion correlates with acute kidney disease among patients undergoing cardiac surgery requiring cardiopulmonary bypass. The reported associations are suggestive of a pathophysiological role of venous congestion in acute kidney disease.

[Effect of tetramethylpyrazine on endothelin, von Willebrand factor and thromboxane A2 during cardiopulmonary bypass in patients of congenital heart disease with pulmonary hypertension].

OBJECTIVE: To study the effect of tetramethylpyrazine (TMP) on the vascular endothelial cell (VEC) related humoral factors, including endothelin (ET), factor VIII related antigen (i.e. von Willebrand factor, vWF) and thromboxane A2(TXA2) in patients of congenital heart disease with pulmonary hypertension (CHD-PH) during cardiopulmonary bypass (CPB), and explore the clinical physiopathologic significance of them. METHODS: Thirty non-cyanotic patients of CHD-PH were randomly divided into the control group and the treated group. TMP was given to the treated group by intravenous dripping 3 mg/kg after anesthesia induction and adding 1 mg/kg in oxygenator during CPB. Blood samples were collected from radial artery at the time points of after anesthesia induction, 15 min after beginning CPB, 5 min after opening aorta, 20 min, 6 hrs and 24 hrs after stopping CPB, to determine the plasma contents of ET and vWF, as well as TXB2, the stable metabolite of TXA2. The pulmonary vascular reactivity 6 hrs (6h-PVR) after CPB and the mechanical ventilatory support time (VST) after operation were calculated. RESULTS: Levels of ET, vWF and TXB2 increased obviously during CPB, but the degree of increasing in the treated group was lower than that in the control group (P < 0.05), and the 6h-PVR and VST in the former were also lower than those in the latter respectively. CONCLUSION: TMP could obviously reduce the production of ET, vWF and TXB2 during CPB and relieve the pulmonary vascular reactivity after operation, indicating that TMP could reduce the injury of CPB on VEC, and is benefit to enhance the efficacy of treatment.