Search

Abstract Humans can be exposed to multiple pollutants in the air and surface water. These environments are non-static, trans-boundary and correlated, creating a complex network, and significant challenges for research on environmental hazards, especially in real-world cancer research. This article reports on a large study (377 million people in 30 provinces of China) that evaluated the combined impact of air and surface water pollution on cancer. We formulate a spatial evaluation system and a common grading scale for co-pollution measurement, and validate assumptions that air and surface water environments are spatially connected and that cancers of different types tend to cluster in areas where these environments are poorer. We observe “dose–response” relationships in both the number of affected cancer types and the cancer incidence with an increase in degree of co-pollution. We estimate that 62,847 (7.4%) new cases of cancer registered in China in 2016 were attributable to air and surface water pollution, and the majority (69.7%) of these excess cases occurred in areas with the highest level of co-pollution. The findings clearly show that the environment cannot be considered as a set of separate entities. They also support the development of policies for cooperative environmental governance and disease prevention.

Abstract Background The current guidelines for managing screen‐detected pulmonary nodules offer rule‐based recommendations for immediate diagnostic work‐up or follow‐up at intervals of 3, 6, or 12 months. Customized visit plans are lacking. Purpose To develop individualized screening schedules using reinforcement learning (RL) and evaluate the effectiveness of RL‐based policy models. Methods Using a nested case–control design, we retrospectively identified 308 patients with cancer who had positive screening results in at least two screening rounds in the National Lung Screening Trial. We established a control group that included cancer‐free patients with nodules, matched (1:1) according to the year of cancer diagnosis. By generating 10,164 sequence decision episodes, we trained RL‐based policy models, incorporating nodule diameter alone, combined with nodule appearance (attenuation and margin) and/or patient information (age, sex, smoking status, pack‐years, and family history). We calculated rates of misdiagnosis, missed diagnosis, and delayed diagnosis, and compared the performance of RL‐based policy models with rule‐based follow‐up protocols (National Comprehensive Cancer Network guideline; China Guideline for the Screening and Early Detection of Lung Cancer). Results We identified significant interactions between certain variables (e.g., nodule shape and patient smoking pack‐years, beyond those considered in guideline protocols) and the selection of follow‐up testing intervals, thereby impacting the quality of the decision sequence. In validation, one RL‐based policy model achieved rates of 12.3% for misdiagnosis, 9.7% for missed diagnosis, and 11.7% for delayed diagnosis. Compared with the two rule‐based protocols, the three best‐performing RL‐based policy models consistently demonstrated optimal performance for specific patient subgroups based on disease characteristics (benign or malignant), nodule phenotypes (size, shape, and attenuation), and individual attributes. Conclusions This study highlights the potential of using an RL‐based approach that is both clinically interpretable and performance‐robust to develop personalized lung cancer screening schedules. Our findings present opportunities for enhancing the current cancer screening system.

Abstract Little is known about the association between surface water quality and cancer incidence, especially in China. Drinking water quality has been linked to the incidence of several cancers in individual-level studies. However, few studies have attempted to examine multiple pollutants and multiple cancers at population level. This study used water monitoring and population-level cancer data from across China to examine spatial associations between water pollutants and types of cancer. We found a “dose–response” relationship between the number of pollutants present at high levels and cancer incidence. These results provide evidence of a nationwide spatial association between water quality and cancer in China. The precise relationship varies with cancers and pollutants. However, the overall consistency of the “dose–response” relationship suggests that surface water quality is an important factor in cancer incidence. Our findings highlight new issues such as the changing effects when different pollutants co-exist and an increasing number of new cancer cases partially attributable to poor water quality. Our work also points to some ways to deal with these challenges.

Surface water pollution and digestive cancers are concerning environmental health issues. Multiple pollutants of surface water coexist within unique geospatial heterogeneous contexts of surface water flow, which can lead to complex patterns of impact on cancer incidence. This complexity has posed challenges to comprehensive assessment. In this study, we integrated nationwide surface water data of 21 pollutants collected from 3632 sections of nine large river basins and cancer incidence data for six digestive cancers covering 381.6 million people in China. Local Moran's I method was used to assess the patterns of spatial aggregation of each surface water pollutant separately to observe the local spatial nonstationarity and heterogeneity of the pollutant concentrations in different spatial areas. A geographically weighted random forest (GWRF) model, a spatial machine learning approach, was applied to address the spatial heterogeneity of the pollutants and assess their association with cancer incidence. Goodness-of-fit and overall performance of the GWRF model were assessed by the local statistics in the out-of-bag set, such as the mean local pseudo-R2, increase in mean squared error, and local residuals in each local model. Compared with the aspatial random forest model, the GWRF model had better performance (mean local pseudo-R2 ranged from 0.16 (gallbladder cancer) to 0.60 (pancreatic cancer)). Surface water pollution was associated with digestive cancer incidence mainly in the Songhua and Liaohe, Haihe, and Huaihe River Basins. By focusing on these areas, we identified key pollutants specific to different cancer types. Incidence of oesophageal, stomach, colorectal, gallbladder, and pancreatic cancer was associated with common pollutants such as fluoride and arsenic. Our assessment provides targets for the government and environmental health specialists to take tailored actions to control pollution and effectively prevent cancer incidence.

Abstract Background Numerous observational studies have revealed an increased risk of death and complications with transfusion, but this observation has not been confirmed in randomized controlled trials (RCTs). The “transfusion kills patients” paradox persists in real-world observational studies despite application of analytic methods such as propensity-score matching. We propose a new design to address this long-term existing issue, which if left unresolved, will be deleterious to the healthy generation of evidence that supports optimized transfusion practice. Methods In the new design, we stress three aspects for reconciling observational studies and RCTs on transfusion safety: (1) re-definition of the study population according to a stable hemoglobin range (gray zone of transfusion decision; 7.5–9.5 g/dL in this study); (2) selection of comparison groups according to a trigger value (last hemoglobin measurement before transfusion; nadir during hospital stay for control); (3) dealing with patient heterogeneity according to standardized mean difference (SMD) values. We applied the new design to hospitalized older patients (aged ≥60 years) undergoing general surgery at four academic/teaching hospitals. Four datasets were analyzed: a base population before (Base Match−) and after (Base Match+) propensity-score matching to simulate previous observational studies; a study population before (Study Match−) and after (Study Match+) propensity-score matching to demonstrate effects of our design. Results Of 6141 older patients, 662 (10.78%) were transfused and showed high heterogeneity compared with those not receiving transfusion, particularly regarding preoperative hemoglobin (mean: 11.0 vs. 13.5 g/dL) and intraoperative bleeding (≥500 mL: 37.9% vs. 2.1%). Patient heterogeneity was reduced with the new design; SMD of the two variables was reduced from approximately 100% (Base Match−) to 0% (Study Match+). Transfusion was related to a higher risk of death and complications in Base Match− (odds ratio [OR], 95% confidence interval [CI]: 2.68, 1.86–3.86) and Base Match+ (2.24, 1.43–3.49), but not in Study Match− (0.77, 0.32–1.86) or Study Match+ (0.66, 0.23–1.89). Conclusions We show how choice of study population and analysis could affect real-world study findings. Our results following the new design are in accordance with relevant RCTs, highlighting its value in accelerating the pace of transfusion evidence generation and generalization.

Objective: China is a developing country with urban–rural disparities and accelerating population aging. Therefore, quantifying the effects of population aging on the cancer mortality burden is urgently needed. Methods: Using data from China’s death surveillance datasets (2004–2017), we decomposed and quantified the effects of population aging and factor variations on cancer mortality rates in urban and rural China during 2004–2017 through a decomposition method. R ratios were used to assess the extent of the mortality decreases attributable to factor variations offsetting the increases attributable to population aging for 4 aging-related cancers (lung, colorectal, esophageal, and stomach cancer). Results: Overall, population aging has led to continued increases in cancer mortality rates in China during 2004–2017 (mortality rates attributable to population aging: 8.63/100,000 for urban men, 4.21/100,000 for urban women, 11.95/100,000 for rural men, and 5.66/100,000 for rural women). The 4 cancers displayed 3 patterns. The mortality rates from lung cancer in rural China and from colorectal cancer nationwide increased because of both population aging and factor variations. Population aging was primarily responsible for the growing mortality due to lung cancer in urban areas. However, for esophageal and stomach cancer, the effect of population aging was not dominant, thus resulting in decreases in mortality rates. Conclusions: Health resource allocation should prioritize areas or cancers more adversely affected by population aging. The burden of cancer will continue to increase in the future, because of rapid population aging, but can still be offset or even reversed with enhanced cancer control and prevention.

Abstract Background Radiomics-based image metrics are not used in the clinic despite the rapidly growing literature. We selected eight promising radiomic features and validated their value in decoding lung cancer heterogeneity. Methods CT images of 236 lung cancer patients were obtained from three different institutes, whereupon radiomic features were extracted according to a standardized procedure. The predictive value for patient long-term prognosis and association with routinely used semantic, genetic (e.g., epidermal growth factor receptor (EGFR)), and histopathological cancer profiles were validated. Feature measurement reproducibility was assessed. Results All eight selected features were robust across repeat scans (intraclass coefficient range: 0.81–0.99), and were associated with at least one of the cancer profiles: prognostic, semantic, genetic, and histopathological. For instance, “kurtosis” had a high predictive value of early death (AUC at first year: 0.70–0.75 in two independent cohorts), negative association with histopathological grade (Spearman’s r: − 0.30), and altered expression levels regarding EGFR mutation and semantic characteristics (solid intensity, spiculated shape, juxtapleural location, and pleura tag; all p

Abstract Background The timeliness of diagnostic testing after positive screening remains suboptimal because of limited evidence and methodology, leading to delayed diagnosis of lung cancer and over-examination. We propose a radiomics approach to assist with planning of the diagnostic testing interval in lung cancer screening. Methods From an institute-based lung cancer screening cohort, we retrospectively selected 92 patients with pulmonary nodules with diameters ≥ 3 mm at baseline (61 confirmed as lung cancer by histopathology; 31 confirmed cancer-free). Four groups of region-of-interest-based radiomic features (n = 310) were extracted for quantitative characterization of the nodules, and eight features were proven to be predictive of cancer diagnosis, noise-robust, phenotype-related, and non-redundant. A radiomics biomarker was then built with the random survival forest method. The patients with nodules were divided into low-, middle- and high-risk subgroups by two biomarker cutoffs that optimized time-dependent sensitivity and specificity for decisions about diagnostic workup within 3 months and about repeat screening after 12 months, respectively. A radiomics-based follow-up schedule was then proposed. Its performance was visually assessed with a time-to-diagnosis plot and benchmarked against lung RADS and four other guideline protocols. Results The radiomics biomarker had a high time-dependent area under the curve value (95% CI) for predicting lung cancer diagnosis within 12 months; training: 0.928 (0.844, 0.972), test: 0.888 (0.766, 0.975); the performance was robust in extensive cross-validations. The time-to-diagnosis distributions differed significantly between the three patient subgroups, p

Abstract Background Cancer has become a global health problem, and assessments of cancer mortality are important for effective public health policy‐making and adequate resource allocation. In this study, we aimed to predict the mortality rates and numbers of deaths related to four common cancers (lung, liver, stomach, and esophagus) in China from 2020 to 2030 and to estimate the corresponding cancer burden caused by population aging and tobacco smoking. Methods Cancer mortality data (2004–2017) were extracted from China's death surveillance datasets, and China's population figures (2020–2030) were obtained from the United Nations population projections. Smoking prevalence data were retrieved from a World Health Organization global report, and relative risks of smoking and cancers were derived from large‐scale Asian studies. We predicted the deaths related to the four major cancers and age‐standardized mortality rates using joinpoint regression and linear regression models. The tobacco smoking‐related burden of these four major cancers was estimated using the population attributable fraction. Results Unlike lung cancer mortality which was predicted to continue to increase, the age‐standardized mortality rates for digestive cancers (liver, stomach, and esophageal cancers) are predicted to decline over the next decade. The number of deaths caused by the four major cancers is predicted to increase from 1,490,304 in 2020 to 1,823,960 in 2030. The age‐specific mortality rates of the four major cancers are predicted to increase with age after 40–45 years, peaking in the age groups of 80–84 and ≥85 years. In 2030, the combined number of deaths from the four examined cancers among adults aged ≥65 years is predicted to be 1,167,153, accounting for 64% of all deaths from these cancers. Tobacco smoking is predicted to contribute to nearly 29% of deaths from these cancers, corresponding to 527,577 deaths. Conclusions The overall trend in the combined total mortality from four major cancers is predicted to decline over the next decade; however, the corresponding death toll is expected to surge, in the context of China's population aging and high smoking prevalence. These estimates provide data‐driven evidence for China to implement effective cancer control measures in the future.

Abstract Background Extrahepatic cholangiocarcinoma (ECC) has become one of the most rapidly increasing malignancies in China during recent decades. The relationship between tobacco exposure and ECC epidemics is unclear; this study aimed to explore this relationship. Methods We included 55,806 participants aged 30 years or older from the National Mortality and Smoking Survey of China. Smoking in participants and spouses was defined as 1 cigarette or more per day for up to 1 year. Spouses’ smoking was taken as a measure of exposure to passive smoking. Smoking information in 1980 was ascertained and outcomes were defined as ECC mortality during 1986–1988. Results We found that either passive or active smoking increased the risk of death from ECC by 20% (risk ratio [RR], 1.20; 95% confidence interval [CI], 0.99–1.47), compared with no exposure to any tobacco. This risk was a notable 98% (RR, 1.98; 95% CI, 1.49–2.64) for individuals exposed to passive plus active smoking. These findings were highly consistent among men and women. Pathology-based analyses showed dose-response relationships of ECC with pack-years for all types of smoking exposure (Ps for trend

Abstract Following publication of the results of the National Lung Screening Trial in the United States, a randomized controlled trial in Italy (ITALUNG) and two simulation studies in China reported similar findings in 2017 favoring lung cancer screening with low-dose computed tomography among smokers. With such advances in lung cancer screening, worldwide interest has gradually shifted from evaluating whether refining lung cancer screening protocols is effective in preventing deaths. However, there are several practical problems to be resolved, including the balance of enrollment criteria and cost effectiveness, precise measurements to reduce false positive findings, risk-based optimization of screening frequency, challenges associated with cancer heterogeneity, strategies to combine image screening with novel biomarkers, dynamic monitoring of the natural history of cancer, accurate identification and diagnosis of cases among huge populations, and the impact of tobacco control policy and environment protection. As one in three individuals with lung cancer worldwide resides in China, these questions pose great challenges as well as research opportunities for population screening programs in China.

Abstract Background Mortality outcomes in trials of low-dose computed tomography (CT) screening for lung cancer are inconsistent. This study aimed to evaluate whether CT screening in urban areas of China could reduce lung cancer mortality and to investigate the factors that associate with the screening effect. Methods A decision tree model with three scenarios (low-dose CT screening, chest X-ray screening, and no screening) was developed to compare screening results in a simulated Chinese urban cohort (100,000 smokers aged 45–80 years). Data of participant characteristics were obtained from national registries and epidemiological surveys for estimating lung cancer prevalence. The selection of other tree variables such as sensitivities and specificities of low-dose CT and chest X-ray screening were based on literature research. Differences in lung cancer mortality (primary outcome), false diagnoses, and deaths due to false diagnosis were calculated. Sensitivity analyses were performed to identify the factors that associate with the screening results and to ascertain worst and optimal screening effects considering possible ranges of the variables. Results Among the 100,000 subjects, there were 448, 541, and 591 lung cancer deaths in the low-dose CT, chest X-ray, and no screening scenarios, respectively (17.2% reduction in low-dose CT screening over chest X-ray screening and 24.2% over no screening). The costs of the two screening scenarios were 9387 and 2497 false diagnoses and 7 and 2 deaths due to false diagnosis among the 100,000 persons, respectively. The factors that most influenced death reduction with low-dose CT screening over no screening were lung cancer prevalence in the screened cohort, low-dose CT sensitivity, and proportion of early-stage cancers among low-dose CT detected lung cancers. Considering all possibilities, reduction in deaths (relative numbers) with low-dose CT screening in the worst and optimal cases were 16 (5.4%) and 288 (40.2%) over no screening, respectively. Conclusions In terms of mortality outcomes, our findings favor conducting low-dose CT screening in urban China. However, approaches to reducing false diagnoses and optimizing important screening conditions such as enrollment criteria for screening are highly needed.

ObjectivesTo develop a risk-stratified intervention strategy and evaluate its effect on reducing surgical complications.DesignA multicentre prospective study with preintervention and postintervention stages: period I (January to June 2015) to develop the intervention strategy and period II (January to June 2016) to evaluate its effectiveness.SettingFour academic/teaching hospitals representing major Chinese administrative and economic regions.ParticipantsAll surgical (elective and emergent) inpatients aged ≥14 years with a minimum hospital stay of 24 hours, who underwent a surgical procedure requiring an anesthesiologist.InterventionsTargeted complications were grouped into three categories (common, specific, serious) according to their incidence pattern, severity and preventability. The corresponding expert consensus-generated interventions, which focused on both regulating medical practices and managing inherent patient-related risks, were implemented in a patient-tailored way via an electronic checklist system.Primary and secondary outcomesPrimary outcomes were (1) in-hospital death/confirmed death within 30 days after discharge and (2) complications during hospitalisation. Secondary outcome was length of stay (LOS).ResultsWe included 51 030 patients in this analysis (eligibility rate 87.7%): 23 413 during period I, 27 617 during period II. Patients’ characteristics were comparable during the two periods. After adjustment, the mean number of overall complications per 100 patients decreased from 8.84 to 7.56 (relative change 14.5%; P

OBJECTIVES:To portray the stage characteristics of lung cancers detected in CT screenings, and explore whether there's universal stage superiority over other methods for various pathological types using available data worldwide in a meta-analysis approach. MATERIALS AND METHODS:EMBASE and MEDLINE were searched for studies on lung cancer CT screening in natural populations through July 2015 without language or other filters. Twenty-four studies (8 trials and 16 cohorts) involving 1875 CT-detected lung cancer patients were enrolled and assessed by QUADAS-2. Pathology-confirmed stage information was carefully extracted by two reviewers. Stage I or limited stage proportions were pooled by random effect model with Freeman-Tukey double arcsine transformation. RESULTS:Pooled stage I cancer proportion in CT screenings was 73.2% (95% confidence interval: 68.6%, 77.5%), with a significant rising trend (Ptrend0.05), and 46.5% (P0.05). CONCLUSION:CT screening can detect more early stage non-small cell lung cancers, but not all of them could be beneficial as there are a considerable number of indolent ones such as bronchioloalveolar adenocarcinomas. Still, current evidence is lacking regarding small cell lung cancers.

This study aimed to assess the trend of human fecundity over time in China.This retrospective study was conducted in Tongliao, China. Couples who were married during the time period between January 1, 1981 and December 31, 2003 were considered eligible for this study. A total of 27,413 individuals provided valid information via house-to-house interviews. The 12-month cumulative pregnancy rate (CPR) and annual percentage change were used as the outcome measurements.There was a significant increase in the CPR over the five successive time groups. A break point in 1988 divided the entire study period into two distinct segments: 1981-1988, during which the CPR increased from 72.2% to 84.2%, and 1988-2003, during which the CPR increased from 84.2% to 87.2%.The findings were unlikely to be the result of biases, and could not be explained by increased medical treatment for infertility and changes in the prevalence of sexually transmitted diseases. Dramatic societal and behavioral changes due to the unique family planning policy and economic reform policies in China might have been the plausible reason for the results.

A relationship of blood uric acid (UA) with hypertension and cardiovascular risk is under debate thus salt intake is hypothesized to contribute to such associations.In this cross-sectional study, stratified cluster random sampling elicited a sample of 1805 Kazakhs with 92.4% compliance. Hypertension and moderate-or-high total cardiovascular risk (mTCR) were defined according to guidelines. Sodium intake was assessed by urinary sodium excretion. Prevalence ratios (PRs) were used to express associations of UA with hypertension and mTCR.In the highest tertile of sodium intake in women, the adjusted PRs (95% confidence intervals) of low to high quartiles compared with the lowest quartile of UA, were 1.22(0.78-1.91), 1.18(0.75-1.85), and 1.65(1.09-2.51) for hypertension and 1.19(0.74-1.90), 1.39(0.91-2.11), and 1.65(1.10-2.47) for mTCR (P for trend

We assessed the effect of smoking on death from chronic obstructive pulmonary disease (COPD) in China by employing a large population-based, case-spouse control study design using data from a nationwide survey of mortality. Methods: During 1989-1991, a nationwide retrospective survey of mortality was conducted in China. For approximately 1,000,000 adults dying from all causes during 1986-1988, their surviving spouses or other informants provided detailed information about their own as well as the deceased person's smoking history. For this study, 183,393 individuals who died of COPD at age ≥ 40 years were taken as cases, while 272,984 sex-matched surviving spouses of subjects who died from any cause were taken as controls. Results: COPD death rates for smokers were more than twice as high as those of non-smokers, with a dose-response risk pattern, despite the fact that COPD death rates varied widely by region and age. Tobacco accounted for 41.4% of COPD deaths in men, but only 13.5% of those in women, who had a lower rate of smoking. Conclusion: A case-spouse control study, as an alternative design, is valid and feasible in utilizing information from population-based, retrospective mortality survey data for an analytical epidemiological study of disease etiology.

Purpose: To assess the validation of a novel control selection design by comparing the consistency between the new design and a routine design in a large case-control study that was incorporated into a nationwide mortality survey in China. Methods: A nationwide mortality study was conducted during 1989-1991. Surviving spouses or other relatives of all adults who died during 1986-1988 provided detailed information about their own as well as the deceased person's smoking history. In this study, 130,079 males who died of various smoking-related cancers at age 35 or over were taken as cases, while 103,248 male surviving spouses (same age range with cases) of women who died during the same period and 49,331 males who died from causes other than those related to smoking were used as control group 1 and control group 2, respectively. Consistency in the results when comparing cases with each of the control groups was assessed. Results: Consistency in the results was observed in the analyses using different control groups although cancer deaths varied with region and age. Equivalence could be ascertained using a 15% criterion in most cancer deaths which had high death rates in urban areas, but they were uncertain for most cancers in rural areas irrespective of whether the hypothesis testing showed significant differences or not. Conclusions: Sex-matched living spouse control design as an alternative control selection for a case-control study is valid and feasible, and the basic principles of the equivalence study are also supported by epidemiological survey data.

The purpose of this study was to evaluate the efficacy and safety of three nevirapine-based antiretroviral treatments for adult antiretroviral-naïve Chinese patients with HIV-1 infection.This was a prospective, multicenter study. 198 antiretroviral-naïve HIV-1 positive subjects with CD4 lymphocyte counts between 100/ul and 350/ul and plasma HIV-1 RNA levels more than 500 copies/ml were randomized to start three NVP-based antiretroviral treatments: group A, NVP+AZT+ddI; group B, NVP+3TC+d4T; group C, NVP+AZT+3TC. Viral responses, immunologic responses, adverse events and drug resistance were monitored at baseline and the end of week 4, 12, 24, 36, 52. Viralogical response and immunological response were also compared in different strata of baseline CD4 T lymphocyte counts and plasma HIV-1 RNA concentrations. At baseline, the plasma HIV-1 RNA was 4.44+/-0.68, 4.52+/-0.71 and 4.41+/-0.63 lg copies/ml in group A, B and C respectively (p = 0.628). At the end of the study, the plasma viral load reached 2.54+/-1.11, 1.89+/-0.46 and 1.92+/-0.58 lg copies/ml in group A, B and C respectively (p200/ul (subgroup H). But in subgroup L, viral response rate of three groups has no significant statistic difference. There were no statistically significant differences among three groups in immunological response within any of the CD4 or pVL strata. 3 out of 193 patients with available genotype at baseline showed primary drug resistant. Of 26 patients with virologic failure, 17 patients showed secondary drug resistant, 16 subjects in group A and 1 subject in group B. Logistic regression analysis indicated that presence of hepatotoxicity was associated with HCV-Ab positive (OR = 2.096, 95%CI: 1.106-3.973, P = 0.023) and higher CD4 baseline (CD4 count >250/ul) (OR = 2.096, 95%CI: 1.07-4.107, P = 0.031).Our findings strongly support the use of 3TC+d4T and 3TC+AZT as the nucleoside analogue combination in NVP-based antiretroviral therapy. The regimen of AZT+ddI+NVP produced poor virological response especially in the stratum of CD4 count more than 200/ul. More patients showed secondary drug resistant in this arm too. Patients with HCV-Ab+ and CD4 count >250/ul appear to have significantly high risk of hepatoxicity.ClinicalTrials.gov NCT00618176.

Background: When they encounter various highly related postoperative complications, existing risk evaluation tools that focus on single or any complications are inadequate in clinical practice. This seriously hinders complication management because of the lack of a quantitative basis. An interpretable multilabel model framework that predicts multiple complications simultaneously is urgently needed. Materials and Methods: The authors included 50 325 inpatients from a large multicenter cohort (2014-2017). The authors separated patients from one hospital for external validation and randomly split the remaining patients into training and internal validation sets. A MARKov-EmbeDded (MARKED)multilabelmodel was proposed, and three models were trained for comparison: binary relevance, a fully connected network (FULLNET), and a deep neural network. Performance was mainly evaluated using the area under the receiver operating characteristic curve (AUC). The authors interpreted the model using Shapley Additive Explanations. Complication-specific risk and risk source inference were provided at the individual level. Results: There were 26 292, 6574, and 17 459 inpatients in the training, internal validation, and external validation sets, respectively. For the external validation set, MARKED achieved the highest average AUC (0.818, 95% CI: 0.771-0.864) across eight outcomes [compared with binary relevance, 0.799 (0.748-0.849), FULLNET, 0.806 (0.756-0.856), and deep neural network, 0.815 (0.765-0.866)]. Specifically, the AUCs of MARKED were above 0.9 for cardiac complications [0.927 (0.894-0.960)], neurological complications [0.905 (0.870-0.941)], and mortality [0.902 (0.867-0.937)]. Serum albumin, surgical specialties, emergency case, American Society of Anesthesiologists score, age, and sex were the six most important preoperative variables. The interaction between complications contributed more than the preoperative variables, and formed a hierarchical chain of risk factors, mild complications, and severe complications. Conclusion: The authors demonstrated the advantage of MARKED in terms of performance and interpretability. The authors expect that the identification of high-risk patients and the inference of the risk source for specific complications will be valuable for clinical decision-making. [ABSTRACT FROM AUTHOR]