Dataset Information

Assessing risk prediction models using individual participant data from multiple studies.

ABSTRACT: Individual participant time-to-event data from multiple prospective epidemiologic studies enable detailed investigation into the predictive ability of risk models. Here we address the challenges in appropriately combining such information across studies. Methods are exemplified by analyses of log C-reactive protein and conventional risk factors for coronary heart disease in the Emerging Risk Factors Collaboration, a collation of individual data from multiple prospective studies with an average follow-up duration of 9.8 years (dates varied). We derive risk prediction models using Cox proportional hazards regression analysis stratified by study and obtain estimates of risk discrimination, Harrell's concordance index, and Royston's discrimination measure within each study; we then combine the estimates across studies using a weighted meta-analysis. Various weighting approaches are compared and lead us to recommend using the number of events in each study. We also discuss the calculation of measures of reclassification for multiple studies. We further show that comparison of differences in predictive ability across subgroups should be based only on within-study information and that combining measures of risk discrimination from case-control studies and prospective studies is problematic. The concordance index and discrimination measure gave qualitatively similar results throughout. While the concordance index was very heterogeneous between studies, principally because of differing age ranges, the increments in the concordance index from adding log C-reactive protein to conventional risk factors were more homogeneous.

SUBMITTER: Pennells L

PROVIDER: S-EPMC3927974 | biostudies-literature | 2014 Mar

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Assessing risk prediction models using individual participant data from multiple studies.

Pennells Lisa L Kaptoge Stephen S White Ian R IR Thompson Simon G SG Wood Angela M AM

American journal of epidemiology 20131222 5

Individual participant time-to-event data from multiple prospective epidemiologic studies enable detailed investigation into the predictive ability of risk models. Here we address the challenges in appropriately combining such information across studies. Methods are exemplified by analyses of log C-reactive protein and conventional risk factors for coronary heart disease in the Emerging Risk Factors Collaboration, a collation of individual data from multiple prospective studies with an average f ...[more]

PMID: 24366051

Similar Datasets

Project description:BackgroundDiarrhoeal disease is a leading cause of childhood illness and death globally, and Shigella is a major aetiological contributor for which a vaccine might soon be available. The primary objective of this study was to model the spatiotemporal variation in paediatric Shigella infection and map its predicted prevalence across low-income and middle-income countries (LMICs).MethodsIndividual participant data for Shigella positivity in stool samples were sourced from multiple LMIC-based studies of children aged 59 months or younger. Covariates included household-level and participant-level factors ascertained by study investigators and environmental and hydrometeorological variables extracted from various data products at georeferenced child locations. Multivariate models were fitted and prevalence predictions obtained by syndrome and age stratum.Findings20 studies from 23 countries (including locations in Central America and South America, sub-Saharan Africa, and south and southeast Asia) contributed 66 563 sample results. Age, symptom status, and study design contributed most to model performance followed by temperature, wind speed, relative humidity, and soil moisture. Probability of Shigella infection exceeded 20% when both precipitation and soil moisture were above average and had a 43% peak in uncomplicated diarrhoea cases at 33°C temperatures, above which it decreased. Compared with unimproved sanitation, improved sanitation decreased the odds of Shigella infection by 19% (odds ratio [OR]=0·81 [95% CI 0·76-0·86]) and open defecation decreased them by 18% (OR=0·82 [0·76-0·88]).InterpretationThe distribution of Shigella is more sensitive to climatological factors, such as temperature, than previously recognised. Conditions in much of sub-Saharan Africa are particularly propitious for Shigella transmission, although hotspots also occur in South America and Central America, the Ganges-Brahmaputra Delta, and the island of New Guinea. These findings can inform prioritisation of populations for future vaccine trials and campaigns.FundingNASA, National Institutes of Health-The National Institute of Allergy and Infectious Diseases, and Bill & Melinda Gates Foundation.

Project description:Birth size, perhaps a proxy for prenatal environment, might be a correlate of subsequent breast cancer risk, but findings from epidemiological studies have been inconsistent. We re-analysed individual participant data from published and unpublished studies to obtain more precise estimates of the magnitude and shape of the birth size-breast cancer association.Studies were identified through computer-assisted and manual searches, and personal communication with investigators. Individual participant data from 32 studies, comprising 22,058 breast cancer cases, were obtained. Random effect models were used, if appropriate, to combine study-specific estimates of effect. Birth weight was positively associated with breast cancer risk in studies based on birth records (pooled relative risk [RR] per one standard deviation [SD] [= 0.5 kg] increment in birth weight: 1.06; 95% confidence interval [CI] 1.02-1.09) and parental recall when the participants were children (1.02; 95% CI 0.99-1.05), but not in those based on adult self-reports, or maternal recall during the woman's adulthood (0.98; 95% CI 0.95-1.01) (p for heterogeneity between data sources = 0.003). Relative to women who weighed 3.000-3.499 kg, the risk was 0.96 (CI 0.80-1.16) in those who weighed < 2.500 kg, and 1.12 (95% CI 1.00-1.25) in those who weighed > or = 4.000 kg (p for linear trend = 0.001) in birth record data. Birth length and head circumference from birth records were also positively associated with breast cancer risk (pooled RR per one SD increment: 1.06 [95% CI 1.03-1.10] and 1.09 [95% CI 1.03-1.15], respectively). Simultaneous adjustment for these three birth size variables showed that length was the strongest independent predictor of risk. The birth size effects did not appear to be confounded or mediated by established breast cancer risk factors and were not modified by age or menopausal status. The cumulative incidence of breast cancer per 100 women by age 80 y in the study populations was estimated to be 10.0, 10.0, 10.4, and 11.5 in those who were, respectively, in the bottom, second, third, and top fourths of the birth length distribution.This pooled analysis of individual participant data is consistent with birth size, and in particular birth length, being an independent correlate of breast cancer risk in adulthood.

Project description:ObjectiveTo externally validate various prognostic models and scoring rules for predicting short term mortality in patients admitted to hospital for covid-19.DesignTwo stage individual participant data meta-analysis.SettingSecondary and tertiary care.Participants46 914 patients across 18 countries, admitted to a hospital with polymerase chain reaction confirmed covid-19 from November 2019 to April 2021.Data sourcesMultiple (clustered) cohorts in Brazil, Belgium, China, Czech Republic, Egypt, France, Iran, Israel, Italy, Mexico, Netherlands, Portugal, Russia, Saudi Arabia, Spain, Sweden, United Kingdom, and United States previously identified by a living systematic review of covid-19 prediction models published in The BMJ, and through PROSPERO, reference checking, and expert knowledge.Model selection and eligibility criteriaPrognostic models identified by the living systematic review and through contacting experts. A priori models were excluded that had a high risk of bias in the participant domain of PROBAST (prediction model study risk of bias assessment tool) or for which the applicability was deemed poor.MethodsEight prognostic models with diverse predictors were identified and validated. A two stage individual participant data meta-analysis was performed of the estimated model concordance (C) statistic, calibration slope, calibration-in-the-large, and observed to expected ratio (O:E) across the included clusters.Main outcome measures30 day mortality or in-hospital mortality.ResultsDatasets included 27 clusters from 18 different countries and contained data on 46 914patients. The pooled estimates ranged from 0.67 to 0.80 (C statistic), 0.22 to 1.22 (calibration slope), and 0.18 to 2.59 (O:E ratio) and were prone to substantial between study heterogeneity. The 4C Mortality Score by Knight et al (pooled C statistic 0.80, 95% confidence interval 0.75 to 0.84, 95% prediction interval 0.72 to 0.86) and clinical model by Wang et al (0.77, 0.73 to 0.80, 0.63 to 0.87) had the highest discriminative ability. On average, 29% fewer deaths were observed than predicted by the 4C Mortality Score (pooled O:E 0.71, 95% confidence interval 0.45 to 1.11, 95% prediction interval 0.21 to 2.39), 35% fewer than predicted by the Wang clinical model (0.65, 0.52 to 0.82, 0.23 to 1.89), and 4% fewer than predicted by Xie et al's model (0.96, 0.59 to 1.55, 0.21 to 4.28).ConclusionThe prognostic value of the included models varied greatly between the data sources. Although the Knight 4C Mortality Score and Wang clinical model appeared most promising, recalibration (intercept and slope updates) is needed before implementation in routine care.

Dataset Information

Assessing risk prediction models using individual participant data from multiple studies.

Publications

Assessing risk prediction models using individual participant data from multiple studies.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets