Random forests for survival data: which methods work best and under what conditions?

Few systematic comparisons of methods for constructing survival trees and forests exist in the literature. Importantly, when the goal is to predict a survival time or estimate a survival function, the optimal choice of method is unclear. We use an extensive simulation study to systematically investigate various factors that influence survival forest performance - forest construction method, censoring, sample size, distribution of the response, structure of the linear predictor, and presence of correlated or noisy covariates. In particular, we study 11 methods that have recently been proposed in the literature and identify 6 top performers. We find that all the factors that we investigate have significant impact on the methods' relative accuracy of point predictions of survival times and survival function estimates. We use our results to make recommendations for which methods to use in a given context and offer explanations for the observed differences in relative performance.

Uveal melanoma incidence and survival analysis in Portugal between 2013 and 2022.

PURPOSE: This study aims to provide a comprehensive overview of the clinical characteristics and epidemiology of uveal melanoma (UM) in the Portuguese population, evaluated at the National Reference Centre (NRC). METHODS: A prospective observational study was conducted, involving patients consecutively diagnosed with UM at the Portuguese NRC between July 2013 and December 2022. The study collected data on demographic and tumour characteristics, clinical staging according to the American Joint Committee on Cancer (AJCC), treatment approaches, local disease control, patient survival, and the occurrence of distant metastases. RESULTS: The study included a total of 316 patients, 53.8% female. The mean age at diagnosis was 61.8±14.2 years, and 75.0% of patients presented with symptoms. The mean annual age-adjusted incidence of uveal melanoma in Portugal between 2014 and 2022 was 2.4 cases per million (95% confidence interval [CI]: 2.1-2.8). For choroidal/ciliary body tumours, the overall cumulative survival and distant metastases-free survival (DMFS) rates at 5 years were 84.9% (95% CI: 78.7-91.1) and 79.4% (95%CI: 72.8-86.0), respectively. Notably, higher AJCC stages at presentation, the need for enucleation, and increased tumour thickness were associated with lower DSS and DMFS rates. CONCLUSION: This study represents the most extensive analysis of UM epidemiology within the Portuguese population. The findings underscore the importance of early diagnosis and treatment in UM, as lower AJCC stages and smaller tumour thickness at diagnosis correlate with improved DSS and DMFS.

Risk Factors of Revision Operation and Early Revision for Adjacent Segment Degeneration After Lumbar Fusion Surgery: A Case-Control Study.

BACKGROUND CONTEXT: Adjacent segment degeneration (ASD) following lumbar fusion operation is common and can occur at varying timepoints after index surgery. An early revision operation for ASD, however, signifies a short symptom-free period and might increase the risk of successive surgeries. PURPOSE: We aimed to elucidate the overall risk factors associated with revision surgeries for ASD with distinct attention to early revisions. STUDY DESIGN/SETTING: Retrospective, case-control study. PATIENT SAMPLE: The study included 86 patients who underwent revision operations for ASD after lumbar fusion in the revision group and 166 patients who did not for at least 5 years after index surgery. OUTCOME MEASURES: Sagittal parameters, Pfirrmann grading, facet degeneration grading, and disc space height (DSH) of adjacent segments were assessed. METHODS: Revision operations within 5 years post-surgery were defined as early revision. We compared the revision and no-revision groups as well as the early- and late-revision groups. RESULTS: The revision group demonstrated a significantly greater preoperative C7-S1 sagittal vertical axis (SVA) (p=0.001), postoperative C7-S1 SVA (p<0.001), and postoperative pelvic incidence (PI)-lumbar lordosis (LL) (p<0.001) than those in the no-revision group. Preoperative DSH of the proximal adjunct segment (p=0.001), postoperative PI-LL (p=0.014), and postoperative C7-S1 SVA (p=0.037) exhibited significant association with ASD in logistic regression analysis. The early-revision group had a significantly higher patient age (p=0.001) and a greater number of levels fused (p=0.030) than those in the late-revision group. Multivariate Cox regression analysis demonstrated that old age (p=0.045), a significant number of levels fused (p=0.047), and a narrow preoperative DSH of the proximal adjacent level (p=0.011) were risk factors for early revision. CONCLUSIONS: Postoperative sagittal imbalance, including significant PI-LL and C7-S1 SVA were risk factors for revision operation for ASD but not for early revision. These factors are likely to affect the long-term risk of revision operation due to ASD and thus are not considered risk factors for early revision. Narrow DSH of the proximal adjacent level increased the risks of both revision and early revision surgeries. Moreover, old age and a significant number of levels fused further increased the risk for early revision for ASD.

Predictors of chronic kidney disease survival in type 2 diabetes: a 12-year retrospective cohort study utilizing estimated glomerular filtration rate.

Predicting the course of kidney disease in individuals with both type 1 and type 2 diabetes mellitus (DM) is a significant clinical and policy challenge. In several regions, DM is now the leading cause of end-stage renal disease. The aim of this study to identify both modifiable and non-modifiable risk factors, along with clinical markers and coexisting conditions, that increase the likelihood of stage 3-5 chronic kidney disease (CKD) development in individuals with type 2 DM in the United Arab Emirates (UAE). This was a single-center retrospective cohort study based on data derived from electronic medical records of UAE patients with DM who were registered at outpatient clinics at Tawam Hospital in Al Ain, UAE, between January 2011 and December 2021. Type 2 DM patients aged ≥ 18 years who had serum HbA1c levels ≥ 6.5% were included in the study. Patients with type 1 DM, who had undergone permanent renal replacement therapy, who had under 1 year of follow-up, or who had missing or incomplete data were excluded from the study. Factors associated with diabetic patients developing stage 3-5 CKD were identified through Cox regression analysis and a fine and gray competing risk model to account for competing events that could potentially hinder the development of CKD. A total of 1003 patients were recruited for the study. The mean age of the study cohort at baseline was 70.6 ± 28.2 years. Several factors were found to increase the risk of developing stage 3-5 CKD: advancing age (HR 1.005, 95% CI 1.002-1.009, p = 0.026), a history of hypertension (HR 1.69, 95% CI 1.032-2.8, p = 0.037), a history of heart disease (HR 1.49, 95% CI 1.16-1.92, p = 0.002), elevated levels of serum creatinine (HR 1.006, 95% CI 1.002-1.010, p = 0.003), decreased levels of estimated glomerular filtration rate (eGFR) (HR 0.943, 95% CI, 0.938-0.947; p < 0.001), and the use of beta-blockers (HR 139, 95% CI 112-173, p = 0.003). Implementing preventative measures, initiating early interventions, and developing personalized care plans tailored to address specific risk factors are imperative for reducing the impact of CKD. Additionally, the unforeseen findings related to eGFR highlight the ongoing need for research to deepen our understanding of the complexities of kidney disease.

Bioinformatics-driven identification of prognostic biomarkers in kidney renal clear cell carcinoma.

Renal cell carcinoma (RCC), particularly the clear cell subtype (ccRCC), poses a significant global health concern due to its increasing prevalence and resistance to conventional therapies. Early detection of ccRCC remains challenging, resulting in poor patient survival rates. In this study, we employed a bioinformatic approach to identify potential prognostic biomarkers for kidney renal clear cell carcinoma (KIRC). By analyzing RNA sequencing data from the TCGA-KIRC project, differentially expressed genes (DEGs) associated with ccRCC were identified. Pathway analysis utilizing the Qiagen Ingenuity Pathway Analysis (IPA) tool elucidated key pathways and genes involved in ccRCC dysregulation. Prognostic value assessment was conducted through survival analysis, including Cox univariate proportional hazards (PH) modeling and Kaplan-Meier plotting. This analysis unveiled several promising biomarkers, such as MMP9, PIK3R6, IFNG, and PGF, exhibiting significant associations with overall survival and relapse-free survival in ccRCC patients. Cox multivariate PH analysis, considering gene expression and age at diagnosis, further confirmed the prognostic potential of MMP9, IFNG, and PGF genes. These findings enhance our understanding of ccRCC and provide valuable insights into potential prognostic biomarkers that can aid healthcare professionals in risk stratification and treatment decision-making. The study also establishes a foundation for future research, validation, and clinical translation of the identified prognostic biomarkers, paving the way for personalized approaches in the management of KIRC.

Derivative survival analyses: Analysis methods to derive survival outcomes for the remainder patient cohort without individual patient data.

It is not uncommon for industry-sponsored randomized controlled trials to publish survival curves/data for the overall patient cohort("A+B") and for a favorable subgroup ("A") pre-specified or post hoc, but not the survival curves/data for the remainder cohort("B"). Consequently, following regulatory approval of the intervention treatment for the overall patient population if the primary endpoint is met, it is common for cancer patients representing the remainder cohort (B) to be treated as per the results of the overall cohort (A+B). To overcome this important issue in clinical decision-making, this study aimed to identify methods to accurately derive the survival curves and/or hazard ratio (95% confidence interval) for the remainder cohort (B), utilizing published curves and hazard ratios (95% confidence intervals) of the overall (A+B) and favorable subgroup (A) cohorts. The analysis methods (method I and method II) presented here, termed "derivative survival analyses," enable accurate assessment of survival outcomes in the remainder cohort without individual patient data.

The quasi-xgamma frailty model with survival analysis under heterogeneity problem, validation testing, and risk analysis for emergency care data.

Frailty models are important for survival data because they allow for the possibility of unobserved heterogeneity problem. The problem of heterogeneity can be existed due to a variety of factors, such as genetic predisposition, environmental factors, or lifestyle choices. Frailty models can help to identify these factors and to better understand their impact on survival. In this study, we suggest a novel quasi xgamma frailty (QXg-F) model for the survival analysis. In this work, the test of Rao-Robson and Nikulin is employed to test the validity and suitability of the probabilistic model, we examine the distribution's properties and evaluate its performance in comparison with many relevant cox-frailty models. To show how well the QXg-F model captures heterogeneity and enhances model fit, we use simulation studies and real data applications, including a fresh dataset gathered from an emergency hospital in Algeria. According to our research, the QXg-F model is a viable replacement for the current frailty modeling distributions and has the potential to improve the precision of survival analyses in a number of different sectors, including emergency care. Moreover, testing the ability and the importance of the new QXg-F model in insurance is investigated using simulations via different methods and application to insurance data.

Time to initiate complementary feeding and its predictors among primipara mothers who have infants aged 6-12 months in the Awi zone, northwest Ethiopia.

Introduction: The timely initiation of complementary feeding is essential to reduce infant mortality. In Ethiopia, 37.5 % of mothers did not initiate complementary feeding to their infants in time. However, previous studies could not identify the time to initiate complementary feeding among primipara mothers. Therefore, this study aims to identify the time to initiate complementary feeding and its predictors among primipara mothers with infants aged 6-12 months in the Awi zone, northwest Ethiopia. Methods: A community-based retrospective follow-up study was conducted among 732 primipara mothers who had infants aged 6-12 months from January 1, 2022, to December 30, 2022. A multistage sampling technique was used to select study participants through questionnaires administered by interviewers. Data were entered into EPI-data 3.1 and exported to STATA 17 for further analysis. The Kaplan-Meier survival curve together with the log-rank test was used to assess the survival experience of the infant at specific times and to compare the survival of the infant in it between different categorical independent variables. Bivariable and multivariable Cox proportional hazard regression models were used to identify significant predictors. Model fitness was also assessed using the Schoenfield residual and the Cox-Snell global residual test. Statistical significance was declared at the p-value <0.05. Result: The median time to initiate complementary feeding among primipara mother-infant pairs was 6 ± 2 months. The overall incidence rate of complementary feeding initiation before, at and after 6 months of age of the infant was 7.28 (95%CI: 6.44, 8.25), 41.41 (95%CI: 36.05, 47.56), and 42.97 (95%CI: 36.89, 50.05) per 100 person-month observations, respectively. Among those mothers who initiated complementary feeding for their infants, 249, 200, and 165 initiated complementary feeding before, at and after 6 months of age of the infants. Furthermore, the age (15-24 years) of mothers (AHR: 1.63, 95%CI: 1.16, 2.29), rich wealth (AHR: 1.35, 95%CI: 1.05, 1.75), and richest wealth (AHR: 1.43, 95%CI: 1.10, 1.84) were identified as statistically significant predictors of the time to initiate complementary feeding. Conclusions: The median time to initiate complementary feeding among primipara mother-infant pairs was 6 months. The age of the mother and the wealth status of the household were found to be statistically independent predictors of the time to initiate complementary feeding. Therefore, community health professionals are better able to work on when to initiate complementary feeding to infants of rich and richest young primipara mothers.

[Prognosis Analysis of Additional Surgical Treatment for High-Risk T1 Colorectal Cancer Patients After Endoscopic Resection]. / è¿½åŠ å¤–ç§‘æ‰‹æœ¯æ²»ç–—å¯¹é«˜é£Žé™©T1æœŸç»“ç›´è‚ ç™Œå† é•œåˆ‡é™¤æ‚£è€ çš„é¢„åŽå½±å“.

Objective: To analyze the effect of additional surgery on the survival and prognosis of high-risk T1 colorectal cancer patients who have undergone endoscopic resection. Methods: The clinical data of patients with high-risk T1 colorectal cancer were retrospectively collected. The patients were divided into the endoscopic resection (ER) plus additional surgical resection (SR) group, or the ER+SR group, and the ER group according to whether additional SR were performed after ER. Baseline data of the patients and information on the location, size, and postoperative pathology of the lesions were collected. Patient survival-related information was obtained through the medical record system and patient follow-up. The primary outcome indicators were the overall survival and the colorectal cancer-specific survival. Univariate Cox regression analysis was used to screen survival-related risk factors and hazard ratio (HR) was calculated. Multivariate Cox regression analysis was used to analyze the independent influencing factors. Results: The data of 109 patients with T1 high-risk colorectal cancer were collected, with 52 patients in the ER group and 57 patients in the ER+SR group. The mean age of patients in the ER group was higher than that in the ER+SR group (65.21 years old vs. 60.54 years old, P=0.035), and the median endoscopic measurement of the size of lesions in the ER group was slightly lower than that in the ER+SR group (2.00 cm vs. 2.50 cm, P=0.026). The median follow-up time was 30.00 months, with the maximum follow-up time being 119 months, in the ER+SR group and there were 4 patients deaths, including one colorectal cancer-related death. Whereas the median follow-up time in the ER group was 28.50 months, with the maximum follow-up time being 78.00 months, and there were 4 patient deaths, including one caused by colorectal cancer. The overall 5-year cumulative survival rates in the ER+SR group and the ER group were 94.44% and 81.65%, respectively, and the cancer-specific 5-year cumulative survival rates in the ER+SR group and the ER group were 97.18% and 98.06%, respectively. The Kaplan-Meier analysis showed no significant difference in the overall cumulative survival or cancer-specific cumulative survival between the ER+SR and the ER groups. Univariate Cox regression analysis showed that age and the number of reviews were the risk factors of overall survival (HR=1.16 and HR=0.27, respectively), with age identified as an independent risk factor of overall survival in the multivariate Cox regression analysis (HR=1.10, P=0.045). Conclusion: For T1 colorectal cancer patients with high risk factors after ER, factors such as patient age and their personal treatment decisions should not be overlooked. In clinical practice, additional caution should be exercised in decision-making concerning additional surgery.

Investigating Causal Genetic Effects on Overall Survival of Glioblastoma Patients using Normalizing Flow and Structural Causal Model.

Glioblastoma (GBM) is the most common and aggressive brain tumor with short overall survival (OS) of about 15 months. Understanding the causal factors affecting the patient survival is crucial for disease prognosis and treatment planning. Although previous efforts on survival prediction using multi-omics data has yielded useful predictive models, the causation of the correlated genetic risk factors has not been addressed. Recent advances in causal deep learning models enable the study of causality from complex dataset. In this paper, we leverage the recently proposed structural causal model (SCM) with normalizing flows parameterized by deep networks to perform the counterfactual query to investigate the causal relationship between gene mutation and OS with the presence of other confounders including sex, age and radiomics features. The query amounts to the question that what the survival days will be if the gene mutation status has been changed, i.e., from mutant to non-mutant and vice versa. The trained causal model will infer the counterfactual outcome given the intervention on specific gene mutation. We apply multivariate Cox-PH model to find the genes associated with survival, and investigate the causal genetic effect by comparing the original and counterfactual survival days in a bi-directional fashion. Particularly, the following two scenarios are considered: (1) intervention on a specific gene with non-mutant status to generate the counterfactual survival days as if the gene is mutant, with which the original survival days of the subjects with that mutant gene will be compared; (2) intervention on the gene with mutant status and perform the comparison with survival days of subjects with that non-mutant gene. Our experimental results show that no causation of two correlated genes (NF1, RB1) was revealed in the cohort (n=181), while their genetic effects on OS in terms of prolonging or shortening are generally in accordance with clinical findings.

Survival and safety analysis of COVID-19 vaccine in Chinese patients with non-small cell lung cancer.

BACKGROUND: Severe acute respiratory syndrome coronavirus 2 disease (COVID-19) has caused a worldwide challenging and threatening pandemic. We aimed to assess the safety and efficacy of the COVID-19 vaccines in Non-Small Cell Lung Cancer (NSCLC) patients. METHODS: Patient self-reported adverse events related to vaccines were recorded by follow-up through a uniform questionnaire. Survival analysis was performed by Kaplan-Meier method. A multivariate analysis was performed by the Cox proportional hazard regression model to determine the effect of each variable on the survival of lung cancer patients. RESULTS: A total of 860 patients with NSCLC on treatment were enrolled. Mean age was 57 years in patients with early stage group and 62 years in advanced stage group. The vaccination rate was 71.11% for early-stage patients and 19.48% for advanced-stage patients; most of them (86.5%) received the COVID-19 inactivated virus (Vero cell) vaccine (Coronavac; Sinovac). The most common systemic adverse reaction was weakness. The main reason for vaccine refusal in those unvaccinated patients was concern about the safety of vaccination in the presence of a tumor and undergoing treatment (56.9% and 53.4%). The 1-year disease-free survival (DFS) rate was 100% for vaccinated and 97.4% for unvaccinated early-stage patients. Then we compared the progression-free survival (PFS) of vaccinated (median PFS 9.0 months) and unvaccinated (median PFS 7.0 months) advanced stage patients (p = 0.815). Advanced NSCLC patients continued to be divided into groups receiving radio-chemotherapy, immunotherapy, and targeted therapy, with no statistical difference in PFS between the groups (p > 0.05). The median overall survival (OS) of vaccinated patients was 20.5 months, and that of unvaccinated patients was 19.0 months (p = 0.478) in advanced NSCLC patients. CONCLUSIONS: COVID-19 vaccination is safe for Chinese NSCLC patients actively receiving different antitumor treatments without increasing the incidence of adverse reactions, and vaccination does not affect cancer patient survival.

Phase 1/2 multicenter trial of acalabrutinib in Chinese patients with relapsed/refractory mantle cell lymphoma.

Acalabrutinib studies have limited Asian participation. This phase 1/2 study (NCT03932331) assessed acalabrutinib in Chinese patients with relapsed/refractory (R/R) mantle cell lymphoma (MCL). Primary endpoint was blinded independent central review (BICR)-assessed overall response rate (ORR). Overall, 34 patients were enrolled. Most patients were men (88%); median age was 63 years and 59% had ≥3 prior treatments. Median treatment duration was 14 months (range, 1-24). Any-grade adverse events (AEs) and grade ≥3 AEs occurred in 85.3% and 44.1% of patients, respectively. AEs causing treatment discontinuation were aplastic anemia, thrombocytopenia, and gastrointestinal infection (n = 1 each). Fatal AEs occurred in 2 patients (aplastic anemia and multiple organ dysfunction syndrome [n = 1 each]). BICR-assessed ORR was 82.4% (95% confidence interval [CI]: 65.5, 93.2); 12 (35.3%) patients achieved complete response. Estimated 12-month OS was 84.5% (95% CI: 66.6, 93.3). Acalabrutinib yielded tolerable safety and high response rates in Chinese patients with R/R MCL.

Personalized prognosis stratification of newly diagnosed glioblastoma applying a statistical decision tree model.

PURPOSE: Glioblastoma (GBM) is the most frequent glioma in adults with a high treatment resistance resulting into limited survival. The individual prognosis varies depending on individual prognostic factors, that must be considered while counseling patients with newly diagnosed GBM. The aim of this study was to elaborate a risk stratification algorithm based on reliable prognostic factors to facilitate a personalized prognosis estimation early on after diagnosis. METHODS: A consecutive patient cohort with confirmed GBM treated between 2010 and 2021 was retrospectively analyzed. Clinical, radiological, and molecular parameters were assessed and included in the analysis. Overall survival (OS) was the primary outcome parameter. After identifying the strongest prognostic factors, a risk stratification algorithm was elaborated with estimated odds of survival. RESULTS: A total of 462 GBM patients were analyzed. The strongest prognostic factors were Charlson Comorbidity Index (CCI), extent of tumor resection, and adjuvant treatment. Patients with CCI ≤ 1 receiving tumor resection had the highest survival odds (88% for 10 months). On the contrary, patients with CCI > 3 receiving no adjuvant treatment had the lowest survival odds (0% for 10 months). The 10-months survival rate in patients with CCI > 3 receiving adjuvant treatment was 56% for patients younger than 70 years and 22% for patients older than 70 years. CONCLUSION: A risk stratification algorithm based on significant prognostic factors allowed a personalized early prognosis estimation at the time of GBM diagnosis, that can contribute to a more personalized patient counseling.

Maximizing matching, equity and survival in kidney transplantation using molecular HLA immunogenicity quantitation.

HLA matching improves long-term outcomes of kidney transplantation, yet implementation challenges persist, particularly within the African American (Black) patient demographic due to donor scarcity. Consequently, kidney survival rates among Black patients significantly lag behind those of other racial groups. A refined matching scheme holds promise for improving kidney survival, with prioritized matching for Black patients potentially bolstering rates of HLA-matched transplants. To facilitate quantity, quality and equity in kidney transplants, we propose two matching algorithms based on quantification of HLA immunogenicity using the hydrophobic mismatch score (HMS) for prospective transplants. We mined the national transplant patient database (SRTR) for a diverse group of donors and recipients with known racial backgrounds. Additionally, we use novel methods to infer survival assessment in the simulated transplants generated by our matching algorithms, in the absence of actual target outcomes, utilizing modified unsupervised clustering techniques. Our allocation algorithms demonstrated the ability to match 87.7% of Black and 86.1% of White recipients under the HLA immunogenicity threshold of 10. Notably, at the lowest HMS threshold of 0, 4.4% of Black and 12.1% of White recipients were matched, a marked increase from the 1.8% and 6.6% matched under the prevailing allocation scheme. Furthermore, our allocation algorithms yielded similar or improved survival rates, as illustrated by Kaplan-Meier (KM) curves, and enhanced survival prediction accuracy, evidenced by C-indices and Integrated Brier Scores.

Kaplan-Meier Survival Analysis: Practical Insights for Clinicians.

This article aims to provide a guide that will help healthcare professionals and clinical researchers from all fields that deal with Kaplan-Meier curves. Survival analysis methods are among the most frequently used in the medical sciences and in clinical research. Overall survival, progression free survival, time to recurrence, or any other clinically relevant parameter represented by a Kaplan-Meier curve will be discussed. We will present a practical and straightforward interpretation of these curves, setting aside intricate mathematical considerations. Our focus will be on essential concepts that interface with biological sciences and medicine in order to guarantee proficiency in one of the most popular yet frequently misunderstood methods in clinical research. Being familiar with these concepts is not only essential for designing new clinical studies but also for critically assessing and interpreting published data.

Adjustment for duration of employment in occupational epidemiology.

PURPOSE: In occupational epidemiology, the healthy worker survivor effect can manifest as a time-dependent confounder because healthier workers can accrue greater amounts of exposure over longer periods of employment. For example, in occupational studies of radiation exposure that focus on cumulative annualized radiation dose, workers can accrue greater amounts of cumulative radiation exposure over longer periods of employment, while workers with longer periods of employment can transition into jobs with a reduced potential for annualized radiation exposure. The extent to which confounding arising from the healthy worker survivor effect impacts radiation risk estimates is unknown. METHODS: We assessed the impact of the healthy worker survivor effect on estimates of radiation risk among nuclear workers in a Million Person Study cohort. In simulation studies, we contrasted the ability of marginal structural Cox models with inverse probability weighting and Cox proportional hazards models to account for time-dependent confounding arising from the healthy worker survivor effect. RESULTS: Marginal structural Cox models and Cox proportional hazards models with flexible functional forms for duration of employment provided reliable results. CONCLUSIONS: It is crucial to flexibly adjust for duration of employment to account for confounding arising from the healthy worker survivor effect in occupational epidemiology.

MixEHR-SurG: A joint proportional hazard and guided topic model for inferring mortality-associated topics from electronic health records.

Survival models can help medical practitioners to evaluate the prognostic importance of clinical variables to patient outcomes such as mortality or hospital readmission and subsequently design personalized treatment regimes. Electronic Health Records (EHRs) hold the promise for large-scale survival analysis based on systematically recorded clinical features for each patient. However, existing survival models either do not scale to high dimensional and multi-modal EHR data or are difficult to interpret. In this study, we present a supervised topic model called MixEHR-SurG to simultaneously integrate heterogeneous EHR data and model survival hazard. Our contributions are three-folds: (1) integrating EHR topic inference with Cox proportional hazards likelihood; (2) integrating patient-specific topic hyperparameters using the PheCode concepts such that each topic can be identified with exactly one PheCode-associated phenotype; (3) multi-modal survival topic inference. This leads to a highly interpretable survival topic model that can infer PheCode-specific phenotype topics associated with patient mortality. We evaluated MixEHR-SurG using a simulated dataset and two real-world EHR datasets: the Quebec Congenital Heart Disease (CHD) data consisting of 8211 subjects with 75,187 outpatient claim records of 1767 unique ICD codes; the MIMIC-III consisting of 1458 subjects with multi-modal EHR records. Compared to the baselines, MixEHR-SurG achieved a superior dynamic AUROC for mortality prediction, with a mean AUROC score of 0.89 in the simulation dataset and a mean AUROC of 0.645 on the CHD dataset. Qualitatively, MixEHR-SurG associates severe cardiac conditions with high mortality risk among the CHD patients after the first heart failure hospitalization and critical brain injuries with increased mortality among the MIMIC-III patients after their ICU discharge. Together, the integration of the Cox proportional hazards model and EHR topic inference in MixEHR-SurG not only leads to competitive mortality prediction but also meaningful phenotype topics for in-depth survival analysis. The software is available at GitHub: https://github.com/li-lab-mcgill/MixEHR-SurG.

Shape restricted additive hazards models: Monotone, unimodal, and U-shape hazard functions.

We consider estimation of the semiparametric additive hazards model with an unspecified baseline hazard function where the effect of a continuous covariate has a specific shape but otherwise unspecified. Such estimation is particularly useful for a unimodal hazard function, where the hazard is monotone increasing and monotone decreasing with an unknown mode. A popular approach of the proportional hazards model is limited in such setting due to the complicated structure of the partial likelihood. Our model defines a quadratic loss function, and its simple structure allows a global Hessian matrix that does not involve parameters. Thus, once the global Hessian matrix is computed, a standard quadratic programming method can be applicable by profiling all possible locations of the mode. However, the quadratic programming method may be inefficient to handle a large global Hessian matrix in the profiling algorithm due to a large dimensionality, where the dimension of the global Hessian matrix and number of hypothetical modes are the same order as the sample size. We propose the quadratic pool adjacent violators algorithm to reduce computational costs. The proposed algorithm is extended to the model with a time-dependent covariate with monotone or U-shape hazard function. In simulation studies, our proposed method improves computational speed compared to the quadratic programming method, with bias and mean square error reductions. We analyze data from a recent cardiovascular study.

The effect of first-lactation calving season, milk production, and morphology on the survival of Simmental cows.

Longevity in dairy and dual-purpose cattle is a complex trait which depends on many individual and managerial factors. The purpose of the present study was to investigate the survival (SURV) rate of Italian Simmental dual-purpose cows across different parities. Data of this study referred to 2 173 primiparous cows under official milk recording that calved between 2002 and 2020. Only cows linearly classified for type traits, including muscularity (MU) and body condition score (BCS) were kept. Survival analysis was carried out, through the Cox regression model, for different pairwise combinations of classes of milk productivity MU, BCS, and calving season. Herd-year of first calving was also considered in the model. SURV (0 = culled; 1 = survived) at each lactation up to the 6th were the dependent variables, so that, for example, SURV2 equal to 1 was attributed to cows that entered the 2nd lactation. Survival rates were 98, 71, 63, 56, and 53% for 2nd, 3rd, 4th, 5th, and 6th lactation, respectively. Results revealed that SURV2 was not dependent on milk yield, while in subsequent parities, low-producing cows were characterized by higher SURV compared to high-producing ones. Additionally, cows starting the lactation in autumn survived less (47.38%) than those starting in spring (53.49%), suggesting that facing the late gestation phase in summer could increase the culling risk. The present study indicates that SURV in Italian Simmental cows is influenced by various factors in addition to milk productivity. However, it is important to consider that in this study all first-calving cows culled before the linear evaluation - carried out between mid- and late lactation in this breed - were not accounted for. Finding can be transferred to other dual-purpose breeds, where the cows' body conformation and muscle development - i.e. meat-related features - are often considered as important as milk performance by farmers undertaking culling decisions.

Statistical primer: individual patient data meta-analysis and meta-analytic approaches in case of non-proportional hazards.

Individual patient data (IPD) meta-analyses build upon traditional (aggregate data) meta-analyses by collecting IPD from the individual studies rather than using aggregated summary data. Although both traditional and IPD meta-analyses produce a summary effect estimate, IPD meta-analyses allow for the analysis of data to be performed as a single dataset. This allows for standardization of exposure, outcomes, and analytic methods across individual studies. IPD meta-analyses also allow the utilization of statistical methods typically used in cohort studies, such as multivariable regression, survival analysis, propensity score matching, uniform subgroup and sensitivity analyses, better management of missing data, and incorporation of unpublished data. However, they are more time-intensive, costly, and subject to participation bias. A separate issue relates to the meta-analytic challenges when the proportional hazards assumption is violated. In these instances, alternative methods of reporting time-to-event estimates, such as restricted mean survival time should be used. This statistical primer summarizes key concepts in both scenarios and provides pertinent examples.