Weighing data and models
New partnerships over data will drive transparency and enable individuals to unlock value and insights for better medical and financial health. But a figure is only as good as the source that made it, and a model is only as good as the world it assumes — here a model means a simulation of reality, not any statistical analysis.
Just in
Weighing data and models posts, most recent first.
-
Item new preprint
England's ECHILD linkage under-covers deprived and minoritised ethnic children
ECHILD, which links hospital, school and social care records for 21 million children in England, found lower coverage in London, deprived areas and minoritised ethnic groups, with 90% of birth records linked to school records and 88% of pupils linked to a patient record.
-
Item new preprint
Gaussian process models fill gaps in Canada's patchy measles vaccine coverage data
A Canadian preprint uses Gaussian process models to impute missing measles vaccine coverage by province, territory and age, creating a complete modelled coverage dataset for outbreak response work and testing the approach against England's more complete records.
-
Item confirms peer-reviewed
A random forest beat logistic regression at predicting death in US diabetic adults
In 2428 US adults with both type 2 diabetes and hypertension from NHANES 1999-2018, a random forest model predicted all-cause mortality with an AUC of 0.873, against 0.783 for logistic regression, with no external validation reported.
-
Item new peer-reviewed
Later Chinese cohorts spend more years happy, with gaps widening by education
A cohort-based multistate life table analysis of the Chinese Longitudinal Healthy Longevity Survey for 2002 to 2018 found later-born cohorts of older adults spent a higher proportion of life happy at every age range, with larger gains for women, urban residents and the formally educated.
-
Item refines preprint
AI on breast biopsy slides flagged intermediate-risk patients who gained from chemotherapy
Applied without retraining to 6,735 TAILORx trial participants, an AI model reading H&E slides alongside clinical variables predicted disease-free interval (C-index 0.736) and, within the intermediate-risk group, separated patients who gained from chemotherapy from those who did not.
-
Item new peer-reviewed
Care-experienced Scottish children had roughly triple the mental health admission risk
Linking Scottish social care and hospital records for 663,601 children born between 1990 and 2004 and followed to 2016, care-experienced girls had a hazard ratio of 3.22 and boys 2.64 for a first mental health hospitalisation, with higher ratios still for repeat admissions.
-
Item new peer-reviewed
Higher particulate exposure before surgery tracked worse gastric cancer survival in China
In 4,441 patients who had radical surgery for stage I-III gastric cancer in Zhejiang, China, each standard-deviation rise in pre-surgery PM2.5 exposure was associated with 15% higher all-cause mortality and 31% higher gastric cancer-specific mortality.
-
Item refines peer-reviewed
Austrian all-cause mortality in 2024 sat at or below 2019 at every age over 60
An Austrian analysis of 2024 death certificates found all-cause mortality equal to or below 2019 in every group aged 60 and over despite 1,212 reported COVID-19 deaths, while weighting contributing causes put the COVID death rate at 0.51 to 0.58 of the unweighted figure.
-
Item new peer-reviewed
Aumolertinib with radiotherapy gave median PFS of 34 vs 7.4 months in stage III NSCLC
In a Chinese phase III trial stopped early after 43 of a planned 98 patients were randomised, chemotherapy-free aumolertinib with definitive radiotherapy gave a median progression-free survival of 34.0 months against 7.4 months for chemoradiotherapy in EGFR-mutated unresectable stage III NSCLC.
-
Item refines peer-reviewed
Variability in blood pressure tracks aortic aneurysm risk in UK Biobank
In 189,770 UK Biobank participants with linked primary care records, visit-to-visit systolic blood pressure variability was associated with 33% to 54% higher risk of incident abdominal aortic aneurysm over a median 10.7 years, independent of mean systolic pressure.
-
Item confirms peer-reviewed
US life expectancy ranked 38th in 2023 as the gap with peer countries widened
An analysis of US and comparator mortality data finds US life expectancy ranked 38th globally in 2023, with the gap against 16 high-income countries widening and the rising death rates concentrated at younger and working ages.
-
Item confirms peer-reviewed
AI models predict death after hip fracture but are seldom externally validated
A systematic review of 40 studies published between 2010 and 2026 found that machine learning models for outcomes after hip fracture, most often death, achieved moderate to strong discrimination but were limited by inconsistent external validation and sparse calibration.
