Building a better model: abandon kitchen sink regression.

This paper critically examines 'kitchen sink regression', a practice characterised by the manual or automated selection of variables for a multivariable regression model based on p values or model-based information criteria. We highlight the pitfalls of this method, using examples from perinatal/neonatal medicine, and propose more robust alternatives. The concept of directed acyclic graphs (DAGs) is introduced as a tool for describing and analysing causal relationships. We highlight five key issues with 'kitchen sink regression': (1) the disregard for the directionality of variable relationships, (2) the lack of a meaningful causal interpretation of effect estimates from these models, (3) the inflated alpha error rate due to multiple testing, (4) the risk of overfitting and model instability and (5) the disregard for content expertise in model building. We advocate for the use of DAGs to guide variable selection for models that aim to examine associations between a putative risk factor and an outcome and emphasise the need for a more thoughtful and informed use of regression models in medical research.

Health Care Utilization in the First 7 Years in Children with Fetal Growth Abnormalities: A Retrospective Cohort Study.

OBJECTIVE: To investigate the relationship between birth weight for gestational age and health care utilization of term offspring from birth to 7 years. STUDY DESIGN: We used a population-based retrospective cohort study of infants (≥37 weeks' gestational age) born between 2003 and 2007 in the Canadian province of Nova Scotia (n = 42 050). Perinatal records were linked to provincial administrative health data from birth to age 7 years. The primary outcome was health care utilization (physician visits and hospital admissions) and costs. Birth weight was categorized as small for gestational age (SGA, <10th percentile), appropriate for gestational age (AGA), or large for gestational age (LGA, >90th percentile). Regression models adjusted for potential confounders were used to investigate the associations. RESULTS: Children born SGA had a higher number of specialist visits and hospital admissions, a longer length of stay for the birth admission, and, as a result, higher physician and hospital costs amounting to a cost differential of Can $1222 during the first 7 years of life compared with children born AGA. By contrast, health care use and costs did not differ between children born LGA and AGA. CONCLUSION: Former SGA term infants have a moderate increase in health care use and costs in early childhood compared with former AGA infants, and LGA birth at term is not associated with higher health care utilization.

Association between second stage of labour length and risk of obstetrical anal sphincter injury in nulliparous women: a population-based retrospective cohort study.

INTRODUCTION AND HYPOTHESIS: Obstetrical anal sphincter injury (OASIS) is a common consequence of vaginal delivery in nulliparas and carries the risk of short- and long-term morbidity. The objective of this study was to estimate the association between the duration of the second stage of labour and OASIS risk. METHODS: A population-based, retrospective cohort of nulliparas delivering singleton, vertex, non-anomalous fetuses at term in Nova Scotia, Canada, from 2005 to 2019, were identified using the Nova Scotia Atlee Perinatal Database. Poisson regression models were used to estimate risk ratios (RR) with robust 95% confidence intervals (CI) adjusting for confounding variables to investigate the association between the length of the second stage and OASIS in the entire cohort and in operative vaginal deliveries. RESULTS: Of 36,662 participants, 7.6% sustained an OASIS (6.8% third-degree, 0.8% fourth-degree tear). The proportion of participants who sustained an OASIS increased over the study period. For each 30-min increase in the length of second stage, the OASIS risk increased by 11% (RR 1.11, 95% CI 1.10-1.12). When stratified by mode of delivery, second stage length ≥ 90 min was associated with an increased OASIS risk in spontaneous (RR 1.35, 95% CI 1.15-1.58) and vacuum-assisted vaginal deliveries (RR 1.42, 95% CI 1.11-1.81). In forceps-assisted vaginal deliveries, OASIS risk was increased, with shorter and longer durations of the second stage. CONCLUSION: Increasing length of the second stage of labour was associated with increasing risk of OASIS overall, but the association was heterogeneous between modes of delivery. Length of the second stage should be considered in counseling about OASIS risk.

The Association Between Caesarean Section and Inflammatory Bowel Disease in Childhood and Young Adulthood: Findings From 2 Retrospective Cohort Studies.

OBJECTIVE: The aim of the study was to examine the association of Caesarean section (CS) with inflammatory bowel disease (IBD) in Nova Scotian children. METHODS: The study consisted of 2 retrospective cohorts in the Canadian province of Nova Scotia: all births between 1988 and 2014 (n = 262,729) linked with a clinical registry of all children diagnosed with IBD at the IWK Health Centre, Halifax (Clinical Cohort) and all births from 1989 to 1993 (n = 42,999) linked with provincial administrative health data (Administrative Cohort). The primary outcome was a diagnosis of IBD; the outcome in the Administrative Cohort was ascertained using a previously validated algorithm. Information on the exposures and confounding variables was obtained from the Nova Scotia Atlee Perinatal Database. The association between CS and time to diagnosis of IBD was examined using survival analysis. RESULTS: The population incidence of IBD in the Clinical and Administrative Cohort were 13.0 and 20.6, respectively, per 100,000 person-years; 23% and 19% of children were born by CS in the 2 cohorts. There was no association of CS with IBD in the 2 cohorts. CONCLUSIONS: Findings from 2 population-based cohorts in Atlantic Canada did not provide any evidence for an association between CS and IBD in childhood and young adulthood.

Cesarean Delivery and Healthcare Utilization and Costs in the Offspring: A Retrospective Cohort Study.

OBJECTIVE: To examine the association between cesarean delivery and healthcare utilization and costs in offspring from birth until age 7 years. STUDY DESIGN: A retrospective cohort study of singleton term births in the Canadian province of Nova Scotia between 2003 and 2007 followed until age 7 years was conducted using data from the Nova Scotia Atlee Perinatal Database and administrative health data. The main exposure was mode of delivery (cesarean delivery vs vaginal birth); the outcome was healthcare utilization and costs during the first 7 years of life. Associations were modeled using multiple regression adjusting for maternal prepregnancy weight and sociodemographic factors. RESULTS: In total, 32 464 births were included in the analysis. Compared with children born by vaginal birth, children born by cesarean delivery had more physician visits (incidence rate ratio 1.06, 95% CI 1.05-1.08) and longer hospital stays (incidence rate ratio 1.12, 95% CI 1.03-1.21) and were more likely to be high utilizers of physician visits (OR 1.23, 95% CI 1.10-1.37). Physician and hospital costs were $775 higher for children born by cesarean delivery compared with vaginal birth. CONCLUSIONS: Cesarean delivery compared with vaginal birth is associated with small but statistically significant increases in healthcare utilization and costs during the first 7 years of life.

Maternal pre-pregnancy weight status and health care use for mental health conditions in the offspring.

The objective of this study was to examine the relationship between pre-pregnancy maternal weight status and offspring physician visits for mental health conditions in childhood and adolescence. We conducted a population-based retrospective cohort study of singleton infants born between the years of 1989 and 1993 using a linkage of the Nova Scotia Atlee Perinatal Database with administrative health data. Offspring were followed from birth to age 18 years. Maternal weight status was categorized according to WHO body mass index cutoffs. The number of physician visits for any mental health condition, mood, anxiety, and adjustment disorders, conduct disorder, and attention-deficit hyperactivity disorder (ADHD) from age 0-18 years was determined from ICD codes in physician billings and hospital discharge abstract data. Negative binomial regression adjusting for sociodemographics, maternal psychiatric disorders and smoking was used to model the association. In total, 38,211 mother-offspring pairs were included in the cohort. Within the first 18 years of life, offspring of mothers with obesity had significantly more physician visits for any mental health condition [adjusted incidence rate ratio (IRR) 1.26, 95% CI 1.19-1.34], mood, anxiety, and adjustment disorders (IRR 1.16, 95% CI 1.07-1.25), conduct disorder (IRR 1.25, 95% CI 1.08-1.45), and ADHD (IRR 1.45, 95% CI 1.24-1.69) compared to mothers of normal weight. Associations for mood, anxiety, and adjustment disorders and conduct disorder were strongest at 13-18 years. Offspring of mothers with obesity appear to use health care for mental health conditions more frequently than offspring of normal weight mothers.

Correction to: The association of venous thromboembolism with survival in pediatric cancer patients: a population-based cohort study.

The Figure 1 used in the originally published version of this article was incorrect.

The association of venous thromboembolism with survival in pediatric cancer patients: a population-based cohort study.

Venous thromboembolism (VTE) is a well-recognized complication in pediatric oncology patients. Studies in adult oncology patients have suggested a potential negative association between VTE and survival, but this association has not been examined in pediatric patients yet. The aim of this study was to assess the association of VTE with survival in pediatric oncology patients. Data from all pediatric oncology patients treated at the two tertiary care centers in Atlantic Canada were pooled to create a population-based cohort. The association between VTE and survival was analyzed using a Cox proportional hazards model stratified by diagnosis group (leukemia, lymphoma, and other; sarcoma) and adjusted for age at diagnosis and sex. Out of 939 patients included in this study, 73 had a VTE (8%) and 131 (14%) patients died during the study period. Children in the leukemia/lymphoma/other group with a VTE had significantly poorer survival relative to children in the same group who did not have a VTE. Although children with sarcoma and VTE had poorer survival compared to children with sarcoma with no VTE, this association was not statistically significant. In this population-based study, we found a negative association between VTE and survival in pediatric oncology patients. If future studies confirm this association, this finding may have prognostic implications and potentially offer new avenues for the management of pediatric patients with cancer.