Large Data Repositories Research Articles

Algorithmic Identification of Multiple Sclerosis in the SAIL databank

BackgroundMultiple Sclerosis (MS) is a challenging disease to identify within large repositories of healthcare data. Diagnosis can be protracted requiring many modalities that may not form part of the repository data collection. Accurate case finding would have multiple applications for studying epidemiology. AimDevelop a case finding algorithm within the Secure Anonymised Information Linkage (SAIL) Databank that can reliably elicit Multiple Sclerosis. MethodUtilising a cross sectional cohort study we used multiple datasets within SAIL; General Practice, Inpatient, Outpatient and Office of National Statistics to develop our case finding algorithm based on the coding nomenclatures and timing within these datasets. The results of this algorithm were then tested against two patient data sets: The UK MS Register (n=836), consisting of mostly self confirmed disease and a clinical cohort from South West Wales (n=713). ResultsFrom 4,757,428 records, the algorithm identified 6,194 cases of MS within Wales on 31st December 2020 (prevalence 221.65 [95%CI 216.17–227.24] per 100,000). Case finding sensitivity and specificity was 96.8% and 99.9% for the clinically validated population-based cohort and sensitivity was 96.7% from the self-declared registry population. ConclusionWe successfully identified MS cases within Wales and verified this within two independent data sets.

The risk of second malignancies following prostate cancer radiotherapy in the era of conformal radiotherapy: astatement of the Prostate Cancer Working Group of the German Society of Radiation Oncology (DEGRO).

Asignificant number of prostate cancer patients are long-term survivors after primary definitive therapy, and the occurrence of late side effects, such as second primary cancers, has gained interest. The aim of this editorial is to discuss the most current evidence on second primary cancers based on six retrospective studies published in 2021-2024 using large data repositories not accounting for all possible confounding factors, such as smoking or pre-existing comorbidities. Overall, prostate cancer patients treated with curative radiotherapy have an increased risk (0.7-1%) of the development of second primary cancers compared to patients treated with surgery up to 25years after treatment. However, current evidence suggests that the implementation of intensity modulated radiation therapy is not increasing the risk of second primary cancers compared to conformal 3D-planned radiotherapy. Furthermore, increasing evidence indicates that highly conformal radiotherapy techniques may not increase the probability of second primary cancers compared to radical prostatectomy. Consequently, future studies should consider the radiotherapy technique and other confounding factors to provide amore accurate estimation of the occurrence of second primary cancers.

Mapping the structure-function relationship along macroscale gradients in the human brain

Functional coactivation between human brain regions is partly explained by white matter connections; however, how the structure-function relationship varies by function remains unclear. Here, we reference large data repositories to compute maps of structure-function correspondence across hundreds of specific functions and brain regions. We use natural language processing to accurately predict structure-function correspondence for specific functions and to identify macroscale gradients across the brain that correlate with structure-function correspondence as well as cortical thickness. Our findings suggest structure-function correspondence unfolds along a sensory-fugal organizational axis, with higher correspondence in primary sensory and motor cortex for perceptual and motor functions, and lower correspondence in association cortex for cognitive functions. Our study bridges neuroscience and natural language to describe how structure-function coupling varies by region and function in the brain, offering insight into the diversity and evolution of neural network properties.

Using Data Anonymization in big data analytics security and privacy

Big Data and Analytics mean an enormous and complex collection of very diverse information, which is processed with various technologies and methods to produce and deliver useful and valuable insights. Analytics is the science of using data, or information to extract useful and actionable insights, facts and knowledge from a collection of data it could be stated that Big Data Analytics is the best thing since every commercial data system ever built, although everybody with a more optimistic vision of technology would like to take note that there is a fine line where Everything Data crosses the boundary to something else, especially with regard to privacy and security of the world as we know it. Privacy and security are two distinct but closely related phenomena. Whereas privacy refers to the control over access to the individual, security refers to the stability or strength of controls designed to protect the individual’s privacy. There are many obvious considerations and obstacles when attempting to securely share data. During big data analytics, many invasive techniques such as data fusion, cross-correlation, and algorithm training are often conducted over shared data, which can lead to severe privacy leaks. This means that every enterprise, organization, and individual maintaining large data repositories are in danger of being breached. Our study teaches us that security, privacy, and ethical concerns in big data analytics do not exist in parallel to the business cycle, but must be wisely and ethically managed in coherence throughout all emerging processes of the big data and information systems.

Building Research Infrastructure to Develop Greater Learning Efficiencies (BRIDGE).

In this manuscript, we outline our developed version of a Learning Health System (LHS) in oncology implemented at the Department of Veterans Affairs (VA). Transferring healthcare into an LHS framework has been one of the spearpoints of VA's Central Office and given the general lack of evidence generated through randomized control clinical trials to guide medical decisions in oncology, this domain is one of the most suitable for this change. We describe our technical solution, which includes a large real-world data repository, a data science and algorithm development framework, and the mechanism by which results are brought back to the clinic and to the patient. Additionally, we propose the need for a bridging framework that requires collaboration between informatics specialists and medical professionals to integrate knowledge generation into the clinical workflow at the point of care.

From Scalp to Ear-EEG: A Generalizable Transfer Learning Model for Automatic Sleep Scoring in Older People.

Sleep monitoring has extensively utilized electroencephalogram (EEG) data collected from the scalp, yielding very large data repositories and well-trained analysis models. Yet, this wealth of data is lacking for emerging, less intrusive modalities, such as ear-EEG. The current study seeks to harness the abundance of open-source scalp EEG datasets by applying models pre-trained on data, either directly or with minimal fine-tuning; this is achieved in the context of effective sleep analysis from ear-EEG data that was recorded using a single in-ear electrode, referenced to the ipsilateral mastoid, and developed in-house as described in our previous work. Unlike previous studies, our research uniquely focuses on an older cohort (17 subjects aged 65-83, mean age 71.8 years, some with health conditions), and employs LightGBM for transfer learning, diverging from previous deep learning approaches. Results show that the initial accuracy of the pre-trained model on ear-EEG was 70.1%, but fine-tuning the model with ear-EEG data improved its classification accuracy to 73.7%. The fine-tuned model exhibited a statistically significant improvement (p < 0.05, dependent t-test) for 10 out of the 13 participants, as reflected by an enhanced average Cohen's kappa score (a statistical measure of inter-rater agreement for categorical items) of 0.639, indicating a stronger agreement between automated and expert classifications of sleep stages. Comparative SHAP value analysis revealed a shift in feature importance for the N3 sleep stage, underscoring the effectiveness of the fine-tuning process. Our findings underscore the potential of fine-tuning pre-trained scalp EEG models on ear-EEG data to enhance classification accuracy, particularly within an older population and using feature-based methods for transfer learning. This approach presents a promising avenue for ear-EEG analysis in sleep studies, offering new insights into the applicability of transfer learning across different populations and computational techniques. An enhanced ear-EEG method could be pivotal in remote monitoring settings, allowing for continuous, non-invasive sleep quality assessment in elderly patients with conditions like dementia or sleep apnea.

Evaluating statin-induced protection against liver cancer in patients with chronic liver disease: insights from a large regional health data repository

Evaluating statin-induced protection against liver cancer in patients with chronic liver disease: insights from a large regional health data repository

Secure Authentication and Data Transmission for Patients Healthcare Data in Internet of Medical Things

Currently, data transmission is an expanding area in healthcare, enabling health practitioners to examine, assess, and manage patients using mobile communication technologies. To identify and analyze a patient, healthcare providers need to access the physician's Electronic Medical Record (EMR), which may contain extensive audiovisual big data such as MRIs, CT scans, PET scans, X-rays, and more. To ensure accessibility and scalability for healthcare workers and consumers, the EMR needs to be stored in large data repositories on cloud servers. However, due to the sensitive nature of medical information stored in the cloud, the healthcare profession faces numerous security challenges, with data theft attacks being one of the most critical vulnerabilities. This research focuses on protecting medically sensitive data in the cloud by leveraging cloud computing facilities. The upgraded AES approach ensures that confidential data is securely accessible and stored. In addition, improved Elliptic Curve Cryptography (ECC) is utilized for key generation and validation. A hybrid optimization approach, combining robust optimization and genetic algorithms, is employed to select unique and distinct keys. Decryption is performed using deep neural networks, and Convolutional Neural Networks (CNN) enable batch encryption of multiple documents. The comparison between old methods and the proposed approach is based on encryption time, decryption time, and security strength.

Individual Patient Data Analysis of 17 Randomized Trials vs. Real-World Data for Men with Localized Prostate Cancer Receiving Radiotherapy.

Prior work has demonstrated poor correlation between the results of randomized controlled trials (RCTs) and real-world evidence (RWD). However, patients enrolled in RCTs are often considered to poorly represent the real-world population. Herein, we utilize multiple large data repositories to determine differences in baseline characteristics and long-term outcomes between patients enrolled in RCTs and RWD that received radiotherapy for localized prostate cancer. Meta-Analysis of Randomized trials in Cancer of the Prostate (MARCAP) Consortium was leveraged, and 17 phase III randomized trials were included. RWD were accessed through the Staging Collaboration for Cancer of the Prostate (STAR-CAP) cohort, a cohort that is comprised of >60 centers across the United States and Europe. Additionally, RWD was assessed via the Surveillance, Epidemiology, and End Results (SEER) database. MARCAP and STAR-CAP both contain outcomes for distant metastasis (DM), metastasis-free survival (MFS), prostate cancer-specific mortality (PCSM), and overall survival (OS). SEER only contains PCSM and OS. Wilcoxon signed-rank test and chi-square test were used to compare continuous and categorical variables, respectively. Inverse probability of treatment weighting (IPTW) analysis was conducted, balancing for age, PSA, Gleason score, T stage, and treatment year in the three cohorts. Cox and Fine-Gray regression models were used to compare disease outcomes between RCTs vs. RWD. Data from 10,666 patients from RCTs, 6,530 patients in STAR-CAP, and 117,586 patients in SEER were included. SEER patients were slightly younger (p<0.001, median age 68 (IQR 62-73) than those in RCTs (70, IQR 65-74) and in STAR-CAP (70, IQR 64-74). 10-year OS in RCTs was 65.4%, STAR-CAP 70.2%, SEER 64.1%. OS was superior in STAR-CAP (RCTs as reference; HR 0.91, 95% CI 0.85-0.96, p<0.0001), but there was no significant difference between SEER and RCTs (HR 0.96, 95% CI 0.91-1.02, p = 0.22). 10-year PCSM cumulative incidence was 7.4% in RCTs, 8.1% in STAR-CAP, and 11.0% in SEER. There was no significant difference in PCSM between STAR-CAP RWD and RCTs (HR 0.88, 95% CI 0.78-1.01, p = 0.08), whereas PCSM was worse in SEER than RCTs (HR 1.37, 95% CI 1.21-1.55, p<0.0001). There was no significant difference in DM between STAR-CAP RWD and RCTs (HR 0.93, 95% CI 0.83-1.04, p = 0.2). While baseline differences exist in patients enrolled on localized prostate cancer RCTs and real-world datasets, there were small if any significant relative differences in oncologic outcomes. This provides reassurance that RCT results are generally applicable to patients in routine practice.

Tackling the lack of diversity in cancer research.

Despite the clear benefit of studying biological samples from diverse genetic backgrounds and geographical locations, our current knowledge of disease is mostly derived from the study of European-descent individuals. In the cancer field, this is reflected in the poor representation of African and Amerindian/Latino samples in most large public data repositories. This lack of diversity is due to several reasons, but here we focus on (1) the lack of support for studies on non-European populations that are performed in low- and middle-income countries (LMICs), and (2) unequal partnerships between scientists in LMICs and those in high-income countries. We argue that expanding access to research funding, increasing the participation of underrepresented scientists in editorial boards and international conferences, facilitating the publication of studies conducted in these countries, and properly acknowledging LMIC researchers' contributions in publications and grant applications will promote equity for scientists working in LMICs. We envisage that this will translate to more impactful research in these countries, which will include more samples from diverse populations. For the cancer field, this will broaden our understanding of pathomechanisms and may help to improve the treatment of patients from all backgrounds.

Methods Used in the Development of Common Data Models for Health Data: Scoping Review.

Common data models (CDMs) are essential tools for data harmonization, which can lead to significant improvements in the health domain. CDMs unite data from disparate sources and ease collaborations across institutions, resulting in the generation of large standardized data repositories across different entities. An overview of existing CDMs and methods used to develop these data sets may assist in the development process of future models for the health domain, such as for decision support systems. This scoping review investigates methods used in the development of CDMs for health data. We aim to provide a broad overview of approaches and guidelines that are used in the development of CDMs (ie, common data elements or common data sets) for different health domains on an international level. This scoping review followed the PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) checklist. We conducted the literature search in prominent databases, namely, PubMed, Web of Science, Science Direct, and Scopus, starting from January 2000 until March 2022. We identified and screened 1309 articles. The included articles were evaluated based on the type of adopted method, which was used in the conception, users' needs collection, implementation, and evaluation phases of CDMs, and whether stakeholders (such as medical experts, patients' representatives, and IT staff) were involved during the process. Moreover, the models were grouped into iterative or linear types based on the imperativeness of the stages during development. We finally identified 59 articles that fit our eligibility criteria. Of these articles, 45 specifically focused on common medical conditions, 10 focused on rare medical conditions, and the remaining 4 focused on both conditions. The development process usually involved stakeholders but in different ways (eg, working group meetings, Delphi approaches, interviews, and questionnaires). Twenty-two models followed an iterative process. The included articles showed the diversity of methods used to develop a CDM in different domains of health. We highlight the need for more specialized CDM development methods in the health domain and propose a suggestive development process that might ease the development of CDMs in the health domain in the future.

Annotated Coadds: Concise Metrics for Characterizing Survey Cadence and for Discovering Variable and Transient Sources

In order to study transient phenomena in the Universe, existing and forthcoming imaging surveys are covering wide areas of sky repeatedly over time, with a range of cadences, point spread functions, and depths. We describe here a framework that allows an efficient search for different types of time-varying astrophysical phenomena in current and future, large data repositories. We first present a methodology to generate and store key survey parameters that enable researchers to determine if a survey, or a combination of surveys, allows specific time-variable astrophysical phenomena to be discovered. To facilitate further exploration of sources in regions of interest, we then generate a few sample metrics that capture the essential brightness characteristics of a sky pixel at a specific wavelength. Together, we refer to these as “annotated coadds.” The techniques presented here for WISE/NEOWISE-R data are sensitive to 10% brightness variations at around 12th Vega magnitude at 4.5 μm wavelength. Application of the technique to Zwicky Transient Factory data also enabled the detection of 0.5 mag variability at 20 AB mag in the r-band. We demonstrate the capabilities of these metrics for different classes of sources: high proper-motion stars, periodic variable stars, and supernovae, and find that each metric has its advantages depending on the nature of variability. We also present a data structure which will ease the search for temporally varying phenomena in future surveys.

Total Hip Arthroplasty in Patients Aged 40 to 60 Years Old: A Population-Based Study

BackgroundContemporary total hip arthroplasty (THA) has resolved many implant longevity concerns in younger patients. Patients in their fourth and fifth decades of life are projected to be the fastest-growing demographic of THA patients. We aimed to assess this demographic to: 1) evaluate the rate of THA over time; 2) evaluate the cumulative incidence of revision; and 3) identify risk factors for revision. MethodsA retrospective population-based study of patients between 40 and 60 years old undergoing primary THA was conducted using administrative data from a large clinical data repository. A total of 28,414 patients were included for analysis with a mean age of 53 years (range, 40-60 years) and median follow-up of 9 years (range, 0-17 years). Linear regressions were used to assess annual rates of THA in this cohort over time. Kaplan-Meier analysis was used to determine cumulative incidence of revision. Multivariate Cox proportional hazards models were used to determine association of variables with revision risk. ResultsThe annual rate of THA in our population increased by 60.7% over the study period (P < .0001). Cumulative incidence of revision was 2.9% at 5 years and 4.8% at 10 years. Younger age, women, non-osteoarthritis diagnosis, medical complications, and annual surgeon volume ≤ 60 THA were associated with increased revision risk. ConclusionDemand for THA continues to dramatically increase in this cohort. Risk of revision was low but multiple risk factors were identified. Future studies will help delineate the effect of these variables on revision risk and assess implant survivorship beyond 10 years.

Label-Free Quantification Mass Spectrometry Identifies Protein Markers of Chemotherapy Response in High-Grade Serous Ovarian Cancer

Eighty percent of ovarian cancer patients initially respond to chemotherapy, but the majority eventually experience a relapse and die from the disease with acquired chemoresistance. In addition, 20% of patients do not respond to treatment at all, as their disease is intrinsically chemotherapy resistant. Data-independent acquisition nano-flow liquid chromatography-mass spectrometry (DIA LC-MS) identified the three protein markers: gelsolin (GSN), calmodulin (CALM1), and thioredoxin (TXN), to be elevated in high-grade serous ovarian cancer (HGSOC) tissues from patients that responded to chemotherapy compared to those who did not; the differential expression of the three protein markers was confirmed by immunohistochemistry. Analysis of the online GENT2 database showed that mRNA levels of GSN, CALM1, and TXN were decreased in HGSOC compared to fallopian tube epithelium. Elevated levels of GSN and TXN mRNA expression correlated with increased overall and progression-free survival, respectively, in a Kaplan-Meier analysis of a large online repository of HGSOC patient data. Importantly, differential expression of the three protein markers was further confirmed when comparing parental OVCAR-5 cells to carboplatin-resistant OVCAR-5 cells using DIA LC-MS analysis. Our findings suggest that GSN, CALM1, and TXN may be useful biomarkers for predicting chemotherapy response and understanding the mechanisms of chemotherapy resistance. Proteomic data are available via ProteomeXchange with identifier PXD033785.

Minimizing Data Waste: Conservation in the Big Data Era

Minimizing Data Waste: Conservation in the Big Data Era

Citizen science aids the quantification of the distribution and prediction of present and future temporal variation in habitat suitability at species’ range edges

Citizen science programs are effective methods to collect large volumes of data to assist researchers in monitoring ecological environments. As species shift their distributions globally due to climate change, the use of citizen science data to detect these shifts is increasing. Using targeted citizen science programs to collect data on these species could provide information on range edges to inform species distribution modelling. Currently, species distribution models (SDMs) often rely on large data repositories that may lack observations, and hence ability, to detect changes at the range edge. Here, we developed a SDM to compare traditional data repository observations with targeted citizen science data at the southern distribution limit of two recreationally important marine fish in Tasmania, Australia to investigate the potential change in spatial predictions at their range edge. The SDM using the targeted citizen science data in addition to traditional observation data improved the representation of species by 2.3 and 52.7% and increased the southern distribution of the species by 277 and 438 km, for snapper and King George whiting, respectively. Future (centred around 2050 under IPCC RCP 8.5) habitat suitability was predicted to increase more over the winter season, with implications for species overwintering and persistence of populations. The use of citizen science data allowed for the modelling of historical and future change for two range-extending species, an outcome possible due to the collaboration of two citizen science programs that collected observational data on the target species. Species range shifts will require ongoing monitoring and we have demonstrated that complimentary citizen science initiatives are effective in capturing occurrences of species at their range edge. Increasing collaboration between programs may further increase data collection efforts and provide the knowledge to create a hub for these data to be used more efficiently in the future.

Latent classes of neurodevelopmental profiles and needs in children and adolescents with prenatal alcohol exposure.

Fetal alcohol spectrum disorder (FASD) resulting from prenatal alcohol exposure (PAE) is a common neurodevelopmental disorder, but substantial interindividual heterogeneity complicates timely and accurate assessment, diagnosis, and intervention. The current study aimed to identify classes of children and adolescents with PAE assessed for FASD according to their pattern of significant neurodevelopmental functioning across 10 domains using latent class analysis (LCA), and to characterize these subgroups across clinical features. Data from the Canadian National FASD Database, a large ongoing repository of anonymized clinical data received from diagnostic clinics across Canada, was analyzed using a retrospective cross-sectional cohort design. The sample included 1440 children and adolescents ages 6 to 17 years (M=11.0, SD=3.5, 41.7% female) with confirmed PAE assessed for FASD between 2016 and 2020. Results revealed an optimal four-class solution. The Global needs group was characterized by high overall neurodevelopmental impairment considered severe in nature. The Regulation and Cognitive needs groups presented with moderate but substantively distinguishable patterns of significant neurodevelopmental impairment. The Attention needs group was characterized by relatively low probabilities of significant neurodevelopmental impairment. Both the Global and Regulation needs groups also presented with the highest probabilities of clinical needs, further signifying potential substantive differences in assessment and intervention needs across classes. Four relatively distinct subgroups were present in a large heterogeneous sample of children and adolescents with PAE assessed for FASD in Canada. These findings may inform clinical services by guiding clinicians to identify distinct service pathways for these subgroups, potentially increasing access to a more personalized treatment approach and improving outcomes.

Natural Language Processing Basics.

Natural Language Processing Basics.

Automatic Indexing of Digital Objects Through Learning from User Data

Digital data objects increasingly take the form of a non-textual nature, and the effective retrieval of these objects using their intrinsic contents largely depends on the underlying indexing mechanism. Since current multimedia objects are created with ever-increasing speed and ease, they often form the bulk of the data contents in large data repositories. In this study, we provide an effective automatic indexing mechanism based on learning reinforcement by systematically exploiting the big data obtained from different user interactions. Such human interaction with the search system is able to encode the human intelligence in assessing the relevance of a data object against user retrieval intentions and expectations. By methodically exploiting the big data and learning from such interactions, we establish an automatic indexing mechanism that allows multimedia data objects to be gradually indexed in the normal course of their usage. The proposed method is especially efficient for the search of multimedia data objects such as music, photographs and movies, where the use of straightforward string-matching algorithms are not applicable. The method also permits the index to respond to change in relation to user feedback, which at the same time avoids the system landing in a local optimum. Through the use of the proposed method, the accuracy of searching and retrieval of multimedia objects and documents may be significantly enhanced.

Health Care Disparities in Diabetes and Diabetic Retinopathy

ABSTRACT Purpose To investigate prevalence of diabetes (DM), diabetic retinopathy (DR), and areas with highest rates of undetected DR. To quantify and map locations of disparities as they relate to poverty and minority populations. Methods Retrospective cohort study from large regional health data repository (HealthLNK). Geographic Information System (GIS) analysis mapped rates of DM and DR in Chicago area ZIP Codes. Results Of 1,086,921 adults who met the inclusion criteria, 143,790 with DM were identified. ZIP Codes with higher poverty rates were correlated with higher prevalence of DM and DR (Pearson’s correlation coefficient 0.614, p < .05, 0.333, p < .05). Poverty was negatively correlated with likelihood of DR diagnosis (−0.638, p < .05). Relative risks of DM and DR were calculated in each ZIP Code and compared to actual rates. 36 high-risk ZIP Codes had both high-risk of DM and low DR detection. In high-risk ZIP Codes 85.4% of households self-identified as ethnic minority and 33.0% were below the Federal Poverty Level (FPL). Both percentages were significantly higher than the Chicago average of 50.5% minority and 19.9% below FPL (p < .05). 67 ideal ZIP Codes had both low risk of DM and high DR detection. In ideal ZIP Codes 32.6% of households self-identified as minority, and 10.2% were below the FPL (p < .05). Conclusions A health care disparity exists with regards to DM and DR. High-risk ZIP Codes are associated with higher poverty and higher minority population, and they are highly concentrated in just 17% of the ZIP codes in the Chicago area.