Dataset Information

A data driven learning approach for the assessment of data quality

ABSTRACT:

Background

Data quality assessment is important but complex and task dependent. Identifying suitable measurement methods and reference ranges for assessing their results is challenging. Manually inspecting the measurement results and current data driven approaches for learning which results indicate data quality issues have considerable limitations, e.g. to identify task dependent thresholds for measurement results that indicate data quality issues.

Objectives

To explore the applicability and potential benefits of a data driven approach to learn task dependent knowledge about suitable measurement methods and assessment of their results. Such knowledge could be useful for others to determine whether a local data stock is suitable for a given task.

Methods

We started by creating artificial data with previously defined data quality issues and applied a set of generic measurement methods on this data (e.g. a method to count the number of values in a certain variable or the mean value of the values). We trained decision trees on exported measurement methods’ results and corresponding outcome data (data that indicated the data’s suitability for a use case). For evaluation, we derived rules for potential measurement methods and reference values from the decision trees and compared these regarding their coverage of the true data quality issues artificially created in the dataset. Three researchers independently derived these rules. One with knowledge about present data quality issues and two without.

Results

Our self-trained decision trees were able to indicate rules for 12 of 19 previously defined data quality issues. Learned knowledge about measurement methods and their assessment was complementary to manual interpretation of measurement methods’ results.

Conclusions

Our data driven approach derives sensible knowledge for task dependent data quality assessment and complements other current approaches. Based on labeled measurement methods’ results as training data, our approach successfully suggested applicable rules for checking data quality characteristics that determine whether a dataset is suitable for a given task.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12911-021-01656-x.

SUBMITTER: Tute E

PROVIDER: S-EPMC8561935 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:BackgroundMental health disorders affect multiple aspects of patients' lives, including mood, cognition, and behavior. eHealth and mobile health (mHealth) technologies enable rich sets of information to be collected noninvasively, representing a promising opportunity to construct behavioral markers of mental health. Combining such data with self-reported information about psychological symptoms may provide a more comprehensive and contextualized view of a patient's mental state than questionnaire data alone. However, mobile sensed data are usually noisy and incomplete, with significant amounts of missing observations. Therefore, recognizing the clinical potential of mHealth tools depends critically on developing methods to cope with such data issues.ObjectiveThis study aims to present a machine learning-based approach for emotional state prediction that uses passively collected data from mobile phones and wearable devices and self-reported emotions. The proposed methods must cope with high-dimensional and heterogeneous time-series data with a large percentage of missing observations.MethodsPassively sensed behavior and self-reported emotional state data from a cohort of 943 individuals (outpatients recruited from community clinics) were available for analysis. All patients had at least 30 days' worth of naturally occurring behavior observations, including information about physical activity, geolocation, sleep, and smartphone app use. These regularly sampled but frequently missing and heterogeneous time series were analyzed with the following probabilistic latent variable models for data averaging and feature extraction: mixture model (MM) and hidden Markov model (HMM). The extracted features were then combined with a classifier to predict emotional state. A variety of classical machine learning methods and recurrent neural networks were compared. Finally, a personalized Bayesian model was proposed to improve performance by considering the individual differences in the data and applying a different classifier bias term for each patient.ResultsProbabilistic generative models proved to be good preprocessing and feature extractor tools for data with large percentages of missing observations. Models that took into account the posterior probabilities of the MM and HMM latent states outperformed those that did not by more than 20%, suggesting that the underlying behavioral patterns identified were meaningful for individuals' overall emotional state. The best performing generalized models achieved a 0.81 area under the curve of the receiver operating characteristic and 0.71 area under the precision-recall curve when predicting self-reported emotional valence from behavior in held-out test data. Moreover, the proposed personalized models demonstrated that accounting for individual differences through a simple hierarchical model can substantially improve emotional state prediction performance without relying on previous days' data.ConclusionsThese findings demonstrate the feasibility of designing machine learning models for predicting emotional states from mobile sensing data capable of dealing with heterogeneous data with large numbers of missing observations. Such models may represent valuable tools for clinicians to monitor patients' mood states.

Project description:Background Preterm birth (PTB) is a growing health issue worldwide, currently considered the leading cause of newborn deaths. To address this challenge, the present work aims to develop an algorithm capable of accurately predicting the week of delivery supporting the identification of a PTB in Brazil. Methods This a population-based study analyzing data from 3,876,666 mothers with live births distributed across the 3,929 Brazilian municipalities. Using indicators comprising delivery characteristics, primary care work processes, and physical infrastructure, and sociodemographic data we applied a machine learning-based approach to estimate the week of delivery at the point of care level. We tested six algorithms: eXtreme Gradient Boosting, Elastic Net, Quantile Ordinal Regression - LASSO, Linear Regression, Ridge Regression and Decision Tree. We used the root-mean-square error (RMSE) as a precision. Findings All models obtained RMSE indexes close to each other. The lower levels of RMSE were obtained using the eXtreme Gradient Boosting approach which was able to estimate the week of delivery within a 2.09 window 95%IC (2.090–2.097). The five most important variables to predict the week of delivery were: number of previous deliveries through Cesarean-Section, number of prenatal consultations, age of the mother, existence of ultrasound exam available in the care network, and proportion of primary care teams in the municipality registering the oral care consultation. Interpretation Using simple data describing the prenatal care offered, as well as minimal characteristics of the pregnant, our approach was capable of achieving a relevant predictive performance regarding the week of delivery. Funding Bill and Melinda Gates Foundation, and National Council for Scientific and Technological Development – Brazil, (Conselho Nacional de Desenvolvimento Científico e Tecnológico - CNPQ acronym in portuguese) Support of the research project named: Data-Driven Risk Stratification for Preterm Birth in Brazil: Development of a Machine Learning-Based Innovation for Health Care- Grant: OPP1202186

Project description:INTRODUCTION:Adverse event (AE) under-reporting has been a recurrent issue raised during health authorities Good Clinical Practices (GCP) inspections and audits. Moreover, safety under-reporting poses a risk to patient safety and data integrity. The current clinical Quality Assurance (QA) practices used to detect AE under-reporting rely heavily on investigator site and study audits. Yet several sponsors and institutions have had repeated findings related to safety reporting, and this has led to delays in regulatory submissions. Recent developments in data management and IT systems allow data scientists to apply techniques such as machine learning to detect AE under-reporting in an automated fashion. OBJECTIVE:In this project, we developed a predictive model that enables Roche/Genentech Quality Program Leads oversight of AE reporting at the program, study, site, and patient level. This project was part of a broader effort at Roche/Genentech Product Development Quality to apply advanced analytics to augment and complement traditional clinical QA approaches. METHOD:We used a curated data set from 104 completed Roche/Genentech sponsored clinical studies to train a machine learning model to predict the expected number of AEs. Our final model used 54 features built on patient (e.g., demographics, vitals) and study attributes (e.g., molecule class, disease area). RESULTS:In order to evaluate model performance, we tested how well it would detect simulated test cases based on data not used for model training. For relevant simulation scenarios of 25%, 50%, and 75% under-reporting on the site level, our model scored an area under the curve (AUC) of the receiver operating characteristic (ROC) curve of 0.62, 0.79, and 0.92, respectively. CONCLUSION:The model has been deployed to evaluate safety reporting performance in a set of ongoing studies in the form of a QA/dashboard cockpit available to Roche Quality Program Leads. Applicability and production performance will be assessed over the next 12-24 months in which we will develop a validation strategy to fully integrate our model into Roche QA processes.

Project description:Advances in the field of targeted proteomics and mass spectrometry have significantly improved assay sensitivity and multiplexing capacity. The high-throughput nature of targeted proteomics experiments has increased the rate of data production, which requires development of novel analytical tools to keep up with data processing demand. Currently, development and validation of targeted mass spectrometry assays require manual inspection of chromatographic peaks from large datasets to ensure quality, a process that is time consuming, prone to inter- and intra-operator variability and limits the efficiency of data interpretation from targeted proteomics analyses. To address this challenge, we have developed TargetedMSQC, an R package that facilitates quality control and verification of chromatographic peaks from targeted proteomics datasets. This tool calculates metrics to quantify several quality aspects of a chromatographic peak, e.g. symmetry, jaggedness and modality, co-elution and shape similarity of monitored transitions in a peak group, as well as the consistency of transitions' ratios between endogenous analytes and isotopically labeled internal standards and consistency of retention time across multiple runs. The algorithm takes advantage of supervised machine learning to identify peaks with interference or poor chromatography based on a set of peaks that have been annotated by an expert analyst. Using TargetedMSQC to analyze targeted proteomics data reduces the time spent on manual inspection of peaks and improves both speed and accuracy of interference detection. Additionally, by allowing the analysts to customize the tool for application on different datasets, TargetedMSQC gives the users the flexibility to define the acceptable quality for specific datasets. Furthermore, automated and quantitative assessment of peak quality offers a more objective and systematic framework for high throughput analysis of targeted mass spectrometry assay datasets and is a step towards more robust and faster assay implementation.

Project description:BACKGROUND:Using big data and the theory of cumulative deficits to develop the multimorbidity frailty index (mFI) has become a widely accepted approach in public health and health care services. However, constructing the mFI using the most critical determinants and stratifying different risk groups with dose-response relationships remain major challenges in clinical practice. OBJECTIVE:This study aimed to develop the mFI by using machine learning methods that select variables based on the optimal fitness of the model. In addition, we aimed to further establish 4 entities of risk using a machine learning approach that would achieve the best distinction between groups and demonstrate the dose-response relationship. METHODS:In this study, we used Taiwan's National Health Insurance Research Database to develop a machine learning multimorbidity frailty index (ML-mFI) using the theory of cumulative diseases/deficits of an individual older person. Compared to the conventional mFI, in which the selection of diseases/deficits is based on expert opinion, we adopted the random forest method to select the most influential diseases/deficits that predict adverse outcomes for older people. To ensure that the survival curves showed a dose-response relationship with overlap during the follow-up, we developed the distance index and coverage index, which can be used at any time point to classify the ML-mFI of all subjects into the categories of fit, mild frailty, moderate frailty, and severe frailty. Survival analysis was conducted to evaluate the ability of the ML-mFI to predict adverse outcomes, such as unplanned hospitalizations, intensive care unit (ICU) admissions, and mortality. RESULTS:The final ML-mFI model contained 38 diseases/deficits. Compared with conventional mFI, both indices had similar distribution patterns by age and sex; however, among people aged 65 to 69 years, the mean mFI and ML-mFI were 0.037 (SD 0.048) and 0.0070 (SD 0.0254), respectively. The difference may result from discrepancies in the diseases/deficits selected in the mFI and the ML-mFI. A total of 86,133 subjects aged 65 to 100 years were included in this study and were categorized into 4 groups according to the ML-mFI. Both the Kaplan-Meier survival curves and Cox models showed that the ML-mFI significantly predicted all outcomes of interest, including all-cause mortality, unplanned hospitalizations, and all-cause ICU admissions at 1, 5, and 8 years of follow-up (P<.01). In particular, a dose-response relationship was revealed between the 4 ML-mFI groups and adverse outcomes. CONCLUSIONS:The ML-mFI consists of 38 diseases/deficits that can successfully stratify risk groups associated with all-cause mortality, unplanned hospitalizations, and all-cause ICU admissions in older people, which indicates that precise, patient-centered medical care can be a reality in an aging society.

Project description:INTRODUCTION:Previous reports in European populations demonstrated the existence of five data-driven adult-onset diabetes subgroups. Here, we use self-normalizing neural networks (SNNN) to improve reproducibility of these data-driven diabetes subgroups in Mexican cohorts to extend its application to more diverse settings. RESEARCH DESIGN AND METHODS:We trained SNNN and compared it with k-means clustering to classify diabetes subgroups in a multiethnic and representative population-based National Health and Nutrition Examination Survey (NHANES) datasets with all available measures (training sample: NHANES-III, n=1132; validation sample: NHANES 1999-2006, n=626). SNNN models were then applied to four Mexican cohorts (SIGMA-UIEM, n=1521; Metabolic Syndrome cohort, n=6144; ENSANUT 2016, n=614?and CAIPaDi, n=1608) to characterize diabetes subgroups in Mexicans according to treatment response, risk for chronic complications and risk factors for the incidence of each subgroup. RESULTS:SNNN yielded four reproducible clinical profiles (obesity related, insulin deficient, insulin resistant, age related) in NHANES and Mexican cohorts even without C-peptide measurements. We observed in a population-based survey a high prevalence of the insulin-deficient form (41.25%, 95%?CI 41.02% to 41.48%), followed by obesity-related (33.60%, 95%?CI 33.40% to 33.79%), age-related (14.72%, 95%?CI 14.63% to 14.82%) and severe insulin-resistant groups. A significant association was found between the SLC16A11 diabetes risk variant and the obesity-related subgroup (OR 1.42, 95%?CI 1.10 to 1.83, p=0.008). Among incident cases, we observed a greater incidence of mild obesity-related diabetes (n=149, 45.0%). In a diabetes outpatient clinic cohort, we observed increased 1-year risk (HR 1.59, 95%?CI 1.01 to 2.51) and 2-year risk (HR 1.94, 95%?CI 1.13 to 3.31) for incident retinopathy in the insulin-deficient group and decreased 2-year diabetic retinopathy risk for the obesity-related subgroup (HR 0.49, 95%?CI 0.27 to 0.89). CONCLUSIONS:Diabetes subgroup phenotypes are reproducible using SNNN; our algorithm is available as web-based tool. Application of these models allowed for better characterization of diabetes subgroups and risk factors in Mexicans that could have clinical applications.

Project description:IntroductionAlzheimer's disease (AD) is highly heterogeneous, with substantial individual variabilities in clinical progression and neurobiology. Amyloid deposition has been thought to drive cognitive decline and thus a major contributor to the variations in cognitive deterioration in AD. However, the clinical heterogeneity of patients with early symptomatic AD (mild cognitive impairment or mild dementia due to AD) already with evidence of amyloid abnormality in the brain is still unknown.MethodsParticipants with a baseline diagnosis of mild cognitive impairment or mild dementia, a positive amyloid-PET scan, and more than one follow-up Alzheimer's Disease Assessment Scale-Cognitive Subscale-13 (ADAS-Cog-13) administration within a period of 5-year follow-up were selected from the Alzheimer's Disease Neuroimaging Initiative database (n = 421; age = 73±7; years of education = 16 ± 3; percentage of female gender = 43%; distribution of APOE4 carriers = 68%). A non-parametric k-means longitudinal clustering analysis in the context of the ADAS-Cog-13 data was performed to identify cognitive subtypes.ResultsWe found a highly variable profile of cognitive decline among patients with early AD and identified 4 clusters characterized by distinct rates of cognitive progression. Among the groups there were significant differences in the magnitude of rates of changes in other cognitive and functional outcomes, clinical progression from mild cognitive impairment to dementia, and changes in markers presumed to reflect neurodegeneration and neuronal injury. A nomogram based on a simplified logistic regression model predicted steep cognitive trajectory with an AUC of 0.912 (95% CI: 0.88 - 0.94). Simulation of clinical trials suggested that the incorporation of the nomogram into enrichment strategies would reduce the required sample sizes from 926.8 (95% CI: 822.6 - 1057.5) to 400.9 (95% CI: 306.9 - 516.8).DiscussionOur findings show usefulness in the stratification of patients in early AD and may thus increase the chances of finding a treatment for future AD clinical trials.