Background Prognostic modelling using standard methods is well-established, particularly for predicting risk of single diseases. Machine-learning may offer potential to explore outcomes of even greater complexity, such as premature death. This study aimed to develop novel prediction algorithms using machine-learning, in addition to standard survival modelling, to predict premature all-cause mortality. Methods A prospective population cohort of 502,628 participants aged 40–69 years were recruited to the UK Biobank from 2006–2010 and followed-up until 2016. Participants were assessed on a range of demographic, biometric, clinical and lifestyle factors. Mortality data by ICD-10 were obtained from linkage to Office of National Statistics. Models were developed using deep learning, random forest and Cox regression. Calibration was assessed by comparing observed to predicted risks; and discrimination by area under the ‘receiver operating curve’ (AUC). Findings 14,418 deaths (2.9%) occurred over a total follow-up time of 3,508,454 person-years. A simple age and gender Cox model was the least predictive (AUC 0.689, 95% CI 0.681–0.699). A multivariate Cox regression model significantly improved discrimination by 6.2% (AUC 0.751, 95% CI 0.748–0.767). The application of machine-learning algorithms further improved discrimination by 3.2% using random forest (AUC 0.783, 95% CI 0.776–0.791) and 3.9% using deep learning (AUC 0.790, 95% CI 0.783–0.797). These ML algorithms improved discrimination by 9.4% and 10.1% respectively from a simple age and gender Cox regression model. Random forest and deep learning achieved similar levels of discrimination with no significant difference. Machine-learning algorithms were well-calibrated, while Cox regression models consistently over-predicted risk. Conclusions Machine-learning significantly improved accuracy of prediction of premature all-cause mortality in this middle-aged population, compared to standard methods. This study illustrates the value of machine-learning for risk prediction within a traditional epidemiological study design, and how this approach might be reported to assist scientific verification.

Introduction:Previous studies have found partners’ smoking status, multiparity, and nicotine dependence to be associated with smoking cessation in pregnancy. However, no studies have investigated influences on cessation among women using nicotine replacement therapy (NRT). We analyzed data from a trial of NRT in pregnancy to determine factors associated with shorter- and longer-term cessation.Methods:Data were collected at baseline, 1 month, and delivery from 1,050 pregnant women. Two multivariable logistic models for validated cessation at 1 month and delivery were created with a systematic strategy for selection of included factors.Results:All findings are from multivariable analyses. At 1 month, odds of cessation were greater among those who completed full time education at >16 years of age (odds ratio [OR] = 1.82, 95% confidence interval CI = 1.24–2.67, p = .002) but they were lower in women with higher baseline cotinine levels (OR = 0.93, 95% CI = 0.90–0.95, p < .001). At delivery, the odds of cessation were greater among those who completed full time education at >16 years of age (OR = 1.89, 95% CI = 1.16–3.07, p = 0.010) but were inversely associated with higher baseline cotinine levels (OR = 0.96, 95% CI = 0.92–0.99, p = .010).Conclusions:Women who are better educated and have lower pretreatment cotinine concentrations had higher odds of stopping smoking and factors associated with shorter and longer term cessation were similar.

Background Limited research exists on interest in and use of smoking cessation support in pregnancy and postpartum. Methods A longitudinal cohort of pregnant smokers and recent ex-smokers were recruited in Nottinghamshire, United Kingdom (N = 850). Data were collected at 8–26 weeks gestation, 34–36 weeks gestation, and 3 months postpartum and used as three cross-sectional surveys. Interest and use of cessation support and belief and behavior measures were collected at all waves. Key data were adjusted for nonresponse and analyzed descriptively, and multiple regression was used to identify associations. Results In early and late pregnancy, 44% (95% CI 40% to 48%) and 43% (95% CI 37% to 49%) of smokers, respectively, were interested in cessation support with 33% (95% CI 27% to 39%) interested postpartum. In early pregnancy, 43% of smokers reported discussing cessation with a midwife and, in late pregnancy, 27% did so. Over one-third (38%) did not report discussing quitting with a health professional during pregnancy. Twenty-seven percent of smokers reported using any National Health Service (NHS) cessation support and 12% accessed NHS Stop Smoking Services during pregnancy. Lower quitting confidence (self-efficacy), higher confidence in stopping with support, higher quitting motivation, and higher age were associated with higher interest in support (ps ≤ .001). A recent quit attempt and greater interest in support was associated with speaking to a health professional about quitting and use of NHS cessation support (ps ≤ .001). Conclusions When asked in early or late pregnancy, about half of pregnant smokers were interested in cessation support, though most did not engage. Cessation support should be offered throughout pregnancy and after delivery. Implications There is relatively high interest in cessation support in early and late pregnancy and postpartum among smokers; however, a much smaller proportion of pregnant or postpartum women access any cessation support, highlighting a gap between interest and engagement. Reflecting women’s interest, offers of cessation support should be provided throughout pregnancy and after delivery. Increasing motivation to quit and confidence in quitting with assistance may enhance interest in support, and promoting the discussion of stopping smoking between women and health practitioners may contribute to higher support engagement rates.