A novel technology for harmonizing and analyzing cancer data. Observations from integrating health connect in Newfoundland and Labrador, Canada.

Objective: This article aims to describe the implementation of a new health information technology system called Health Connect that is harmonizing cancer data in the Canadian province of Newfoundland and Labrador; explain high-level technical details of this technology; provide concrete examples of how this technology is helping to improve cancer care in the province, and to discuss its future expansion and implications. Methods: We give a technical description of the Health Connect architecture, how it integrated numerous data sources into a single, scalable health information system for cancer data and highlight its artificial intelligence and analytics capacity. Results: We illustrated two practical achievements of Health Connect. First, an analytical dashboard that was used to pinpoint variations in colon cancer screening uptake in small defined geographic regions of the province; and second, a natural language processing algorithm that provided AI-assisted decision support in interpreting appropriate follow-up action based on assessments of breast mammography reports. Conclusion: Health Connect is a cutting-edge, health systems solution for harmonizing cancer screening data for practical decision-making. The long term goal is to integrate all cancer care data holdings into Health Connect to build a comprehensive health information system for cancer care in the province.

Suicide in services for mental health and substance use: A national hybrid registry surveillance system.

Suicide is strongly associated with mental health and substance use disorders, which makes mental health- and substance misuse services important areas for suicide prevention. The aim of The Norwegian Surveillance System for Suicide in Mental Health and Substance Use Services is to describe all suicide deaths in Norway that occur within one year after contact with mental health and substance misuse services. The study uses a hybrid registry case series design. It consists of a yearly linkage between the Norwegian Cause of Death Registry and the Norwegian Patient Registry, which is linked with a questionnaire. The linkage is conducted by using a cryptographic hash function of the deceased's personal id, thus ensuring that the project can link data across sources without the use of directly identifiable information. This indirect linkage ensures the deceased's confidentiality. Moreover, the The Norwegian Surveillance System for Suicide shows how administratively collected data can be harnessed and used for surveillance. Both use of hybrid registry designs and linkage through cryptographic hash functions might contribute to the development of health informatics as well as quality improvement in health care.

Accelerating artificial intelligence: How federated learning can protect privacy, facilitate collaboration, and improve outcomes.

Cross-institution collaborations are constrained by data-sharing challenges. These challenges hamper innovation, particularly in artificial intelligence, where models require diverse data to ensure strong performance. Federated learning (FL) solves data-sharing challenges. In typical collaborations, data is sent to a central repository where models are trained. With FL, models are sent to participating sites, trained locally, and model weights aggregated to create a master model with improved performance. At the 2021 Radiology Society of North America's (RSNA) conference, a panel was conducted titled "Accelerating AI: How Federated Learning Can Protect Privacy, Facilitate Collaboration and Improve Outcomes." Two groups shared insights: researchers from the EXAM study (EMC CXR AI Model) and members of the National Cancer Institute's Early Detection Research Network's (EDRN) pancreatic cancer working group. EXAM brought together 20 institutions to create a model to predict oxygen requirements of patients seen in the emergency department with COVID-19 symptoms. The EDRN collaboration is focused on improving outcomes for pancreatic cancer patients through earlier detection. This paper describes major insights from the panel, including direct quotes. The panelists described the impetus for FL, the long-term potential vision of FL, challenges faced in FL, and the immediate path forward for FL.

Electronic surveillance of patient safety events using natural language processing.

OBJECTIVE: We describe our approach to surveillance of reportable safety events captured in hospital data including free-text clinical notes. We hypothesize that a) some patient safety events are documented only in the clinical notes and not in any other accessible source; and b) large-scale abstraction of event data from clinical notes is feasible. MATERIALS AND METHODS: We use regular expressions to generate a training data set for a machine learning model and apply this model to the full set of clinical notes and conduct further review to identify safety events of interest. We demonstrate this approach on peripheral intravenous (PIV) infiltrations and extravasations (PIVIEs). RESULTS: During Phase 1, we collected 21,362 clinical notes, of which 2342 were reviewed. We identified 125 PIV events, of which 44 cases (35%) were not captured by other patient safety systems. During Phase 2, we collected 60,735 clinical notes and identified 440 infiltrate events. Our classifier demonstrated accuracy above 90%. CONCLUSION: Our method to identify safety events from the free text of clinical documentation offers a feasible and scalable approach to enhance existing patient safety systems. Expert reviewers, using a machine learning model, can conduct routine surveillance of patient safety events.

Guideline interval: A new time interval in the diagnostic pathway for symptomatic cancer.

BACKGROUND: A standard measure of the cancer diagnostic pathway, diagnostic interval, is the time from "first presentation of cancer" to diagnosis. Cancer presentation may be unclear in patients with multimorbidity or non-specific symptoms, signs or test results ("features"). We propose an alternative, guideline interval, with a more certain start date; namely, when the patient first meets suspected-cancer criteria for investigation or referral. METHODS: This retrospective cohort study used Clinical Practice Research Datalink (CPRD) and English cancer registry data. Participants, aged ≥55 years, had diagnostic codes for oesophagogastric cancers in 1/1/12-31/12/17. Features of oesophagogastric cancer in the year before diagnosis were identified from CPRD codes for dysphagia, haematemesis, upper-abdominal mass or pain, low haemoglobin, reflux, dyspepsia, nausea, vomiting, weight loss or thrombocytosis. Diagnostic interval was the time from first feature to diagnosis; guidance interval, the time from first meeting criteria in NICE suspected-cancer guidance to diagnosis. Multimorbidity burden was quantified using Adjusted Clinical Groups®. Accelerated failure-time models explored associations between multimorbidity burden and length of both diagnostic and guideline interval. RESULTS: There were 3,793 eligible participants (69.0 % male), mean age 74.1 years (SD 10.5). 3,097 (81.7 %) presented with ≥1 feature in the year before diagnosis, and 1,990 (52.5 %) met NICE suspected-cancer criteria. The median for both intervals was 11 days in healthy users, and rose with increasing morbidity burden. At very high multimorbidity burden, diagnostic interval was 5.47 (95%CI 3.25-9.20) times longer and guideline interval was 3.91 (2.63-5.80) times longer than for healthy users. CONCLUSIONS: Guideline interval is proposed as a new measure of the cancer diagnostic pathway. It has a more certain start date than diagnostic interval, and is lengthened less than diagnostic interval in people with a very high multimorbidity burden. Guideline interval has potential for assessing the implementation of suspected-cancer policies.

Predicting nationwide obesity from food sales using machine learning.

The obesity epidemic progresses everywhere across the globe, and implementing frequent nationwide surveys to measure the percentage of obese population is costly. Conversely, country-level food sales information can be accessed inexpensively through different suppliers on a regular basis. This study applies a methodology to predict obesity prevalence at the country-level based on national sales of a small subset of food and beverage categories. Three machine learning algorithms for nonlinear regression were implemented using purchase and obesity prevalence data from 79 countries: support vector machines, random forests and extreme gradient boosting. The proposed method was validated in terms of both the absolute prediction error and the proportion of countries for which the obesity prevalence was predicted satisfactorily. We found that the most-relevant food category to predict obesity is baked goods and flours, followed by cheese and carbonated drinks.

A comparison of logistic regression models with alternative machine learning methods to predict the risk of in-hospital mortality in emergency medical admissions via external validation.

We compare the performance of logistic regression with several alternative machine learning methods to estimate the risk of death for patients following an emergency admission to hospital based on the patients' first blood test results and physiological measurements using an external validation approach. We trained and tested each model using data from one hospital (n = 24,696) and compared the performance of these models in data from another hospital (n = 13,477). We used two performance measures - the calibration slope and area under the receiver operating characteristic curve. The logistic model performed reasonably well - calibration slope: 0.90, area under the receiver operating characteristic curve: 0.847 compared to the other machine learning methods. Given the complexity of choosing tuning parameters of these methods, the performance of logistic regression with transformations for in-hospital mortality prediction was competitive with the best performing alternative machine learning methods with no evidence of overfitting.

Publications search optimization: Comparison of a homegrown-API approach versus manual publication searches at an NCI designated cancer center.

One measure of research productivity within the University of Kansas Cancer Center (KU Cancer Center) is peer-reviewed publications. Considerable effort goes into searching, capturing, reviewing, storing, and reporting cancer-relevant publications. Traditionally, the method of gathering relevant information to the publications is done manually. This manuscript describes the efforts to transition KU Cancer Center's publication gathering process from a heavily manual to a more automated and efficient process. To achieve this transition in the most customized and cost-effective manner, a homegrown, automated system was developed using open source API among other software. When comparing the automated and the manual processes over several years of data, publication search and retrieval time dropped from an average of 59 h to 35 min, which would amount to a cost savings of several thousand dollars per year. The development and adoption of an automated publications search process can offer research centers great potential for less-error prone results with a savings in time and cost.

The qualculative dimension of healthcare data interoperability.

Research on interoperability and information exchange between information technology systems touts the use of secondary data for a variety of purposes, including research, management, quality improvement, and accountability. However, many studies have pointed out that this is difficult to achieve in practice. Hence, this article aims to examine the causes for this by reporting an ethnographic study of the data work performed by medical records coders and birth certificate clerks working in a hospital system to uncover the practices of creating administrative data (e.g. secondary data). The article illustrates that clerks and coders use situated qualitative judgments of the accuracy and authority of different primary medical accounts. Coders and clerks also employ their understandings of the importance of different future uses of data as they make crucial decisions about how much discretion to exercise in producing accurate data and how much effort to put toward clarifying problematic medical data. These findings suggest that information technology systems designed for interoperability and secondary data also need to be designed in ways that support the qualculative practices of data workers in order to succeed, including making future uses of data clear to data workers and finding ways to minimize conflicting data before data workers encounter it.

Veracity in big data: How good is good enough.

Veracity, one of the five V's used to describe big data, has received attention when it comes to using electronic medical record data for research purposes. In this perspective article, we discuss the idea of data veracity and associated concepts as it relates to the use of electronic medical record data and administrative data in research. We discuss the idea that electronic medical record data are "good enough" for clinical practice and, as such, are "good enough" for certain applications. We then propose three primary issues to attend to when establishing data veracity: data provenance, cross validation, and context.

Challenges associated with missing data in electronic health records: A case study of a risk prediction model for diabetes using data from Slovenian primary care.

The increasing availability of data stored in electronic health records brings substantial opportunities for advancing patient care and population health. This is, however, fundamentally dependant on the completeness and quality of data in these electronic health records. We sought to use electronic health record data to populate a risk prediction model for identifying patients with undiagnosed type 2 diabetes mellitus. We, however, found substantial (up to 90%) amounts of missing data in some healthcare centres. Attempts at imputing for these missing data or using reduced dataset by removing incomplete records resulted in a major deterioration in the performance of the prediction model. This case study illustrates the substantial wasted opportunities resulting from incomplete records by simulation of missing and incomplete records in predictive modelling process. Government and professional bodies need to prioritise efforts to address these data shortcomings in order to ensure that electronic health record data are maximally exploited for patient and population benefit.

Detecting hospital-acquired infections: A document classification approach using support vector machines and gradient tree boosting.

Hospital-acquired infections pose a significant risk to patient health, while their surveillance is an additional workload for hospital staff. Our overall aim is to build a surveillance system that reliably detects all patient records that potentially include hospital-acquired infections. This is to reduce the burden of having the hospital staff manually check patient records. This study focuses on the application of text classification using support vector machines and gradient tree boosting to the problem. Support vector machines and gradient tree boosting have never been applied to the problem of detecting hospital-acquired infections in Swedish patient records, and according to our experiments, they lead to encouraging results. The best result is yielded by gradient tree boosting, at 93.7 percent recall, 79.7 percent precision and 85.7 percent F1 score when using stemming. We can show that simple preprocessing techniques and parameter tuning can lead to high recall (which we aim for in screening patient records) with appropriate precision for this task.

Infrastructures for healthcare: From synergy to reverse synergy.

The Danish General Practitioners Database has over more than a decade developed into a large-scale successful information infrastructure supporting medical research in Denmark. Danish general practitioners produce the data, by coding all patient consultations according to a certain set of classifications, on the entire Danish population. However, in the Autumn of 2014, the system was temporarily shut down due to a lawsuit filed by two general practitioners. In this article, we ask why and identify a political struggle concerning authority, control, and autonomy related to a transformation of the fundamental ontology of the information infrastructure. We explore how the transformed ontology created cracks in the inertia of the information infrastructure damaging the long-term sustainability. We propose the concept of reverse synergy as the awareness of negative impacts occurring when uncritically adding new actors or purposes to a system without due consideration to the nature of the infrastructure. We argue that while long-term information infrastructures are dynamic by nature and constantly impacted by actors joining or leaving the project, each activity of adding new actors must take reverse synergy into account, if not to risk breaking down the fragile nature of otherwise successful information infrastructures supporting research on healthcare.

Usability evaluation of the digital anger thermometer app.

The digital anger thermometer is a prototype for a mobile application (app) for use with adults in anger management treatment. The digital anger thermometer incorporates standards of software development in addition to anger management resources from the Substance Abuse and Mental Health Services Administration. The digital anger thermometer underwent a usability study conducted by five expert reviewers. The results indicate that it is easy to learn, efficient, and ergonomically sound. However, it does not offer support features or user-error tolerance. The digital anger thermometer prototype requires additional usability studies and comparative research in order for it to become an actual mental health app.

Governance and oversight of researcher access to electronic health data: the role of the Independent Scientific Advisory Committee for MHRA database research, 2006-2015.

In order to promote understanding of UK governance and assurance relating to electronic health records research, we present and discuss the role of the Independent Scientific Advisory Committee (ISAC) for MHRA database research in evaluating protocols proposing the use of the Clinical Practice Research Datalink. We describe the development of the Committee's activities between 2006 and 2015, alongside growth in data linkage and wider national electronic health records programmes, including the application and assessment processes, and our approach to undertaking this work. Our model can provide independence, challenge and support to data providers such as the Clinical Practice Research Datalink database which has been used for well over 1,000 medical research projects. ISAC's role in scientific oversight ensures feasible and scientifically acceptable plans are in place, while having both lay and professional membership addresses governance issues in order to protect the integrity of the database and ensure that public confidence is maintained.

Is routine hospital episode data sufficient for identifying individuals with chronic kidney disease? A comparison study with laboratory data.

Internationally, investment in the availability of routine health care data for improving health, health surveillance and health care is increasing. We assessed the validity of hospital episode data for identifying individuals with chronic kidney disease compared to biochemistry data in a large population-based cohort, the Grampian Laboratory Outcomes, Morbidity and Mortality Study-II (n = 70,435). Grampian Laboratory Outcomes, Morbidity and Mortality Study-II links hospital episode data to biochemistry data for all adults in a health region with impaired kidney function and random samples of individuals with normal and unmeasured kidney function in 2003. We compared identification of individuals with chronic kidney disease by hospital episode data (based on International Classification of Diseases-10 codes) to the reference standard of biochemistry data (at least two estimated glomerular filtration rates <60 mL/min/1.73 m(2) at least 90 days apart). Hospital episode data, compared to biochemistry data, identified a lower prevalence of chronic kidney disease and had low sensitivity (<10%) but high specificity (>97%). Using routine health care data from multiple sources offers the best opportunity to identify individuals with chronic kidney disease.

A modular informatics platform for effective support of collaborative and multicenter studies in cardiology.

Collaborative and multicenter studies permit a large number of patients to be enrolled within a reasonable time and providing the opportunity to collect different data. Informatics platforms play an important role in management, storage, and exchange of data between the participants involved in the study. In this article, we describe a modular informatics platform designed and developed to support collaborative and multicenter studies in cardiology. In each developed module, data management is implemented following local defined protocols. The modular characteristic of the developed platform allows independent transfer of different kinds of data, such as biological samples, imaging raw data, and patients' digital information. Moreover, it offers safe central storage of the data collected during the study. The developed platform was successfully tested during a European collaborative and multicenter study, focused on evaluating multimodal non-invasive imaging to diagnose and characterize ischemic heart disease.

Improving National Trauma Data Bank® coding data reliability for traumatic injury using a prospective systems approach.

Trauma centers manage an active Trauma Registry from which research, quality improvement, and epidemiologic information are extracted to ensure optimal care of the trauma patient. We evaluated coding procedures using the Relational Trauma Scoring System™ to determine the relative accuracy of the Relational Trauma Scoring System for coding diagnoses in comparison to the standard retrospective chart-based format. Charts from 150 patients admitted to a level I trauma service were abstracted using standard methods. These charts were then randomized and abstracted by trauma nurse clinicians with coding software aide. For charts scored pre-training, percent correct for the trauma nurse clinicians ranged from 52 to 64 percent, while the registrars scored 51 percent correct. After training, percentage correct for the trauma nurse clinicians increased to a range of 80-86 percent. Our research has demonstrated implementable changes that can significantly increase the accuracy of data from trauma centers.

Development of a pseudo/anonymised primary care research database: Proof-of-concept study.

General practice records present a comprehensive source of data that could form a variety of anonymised or pseudonymised research databases to aid identification of potential research participants regardless of location. A proof-of-concept study was undertaken to extract data from general practice systems in 15 practices across the region to form pseudo and anonymised research data sets. Two feasibility studies and a disease surveillance study compared numbers of potential study participants and accuracy of disease prevalence, respectively. There was a marked reduction in screening time and increase in numbers of potential study participants identified with the research repository compared with conventional methods. Accurate disease prevalence was established and enhanced with the addition of selective text mining. This study confirms the potential for development of national anonymised research database from general practice records in addition to improving data collection for local or national audits and epidemiological projects.

Using electronic health record systems to optimize admission decisions: the Creatinine case study.

Many medical organizations have implemented electronic health record (EHR) and health information exchange (HIE) networks to improve medical decision-making. This study evaluated the contribution of EHR and HIE networks to physicians by investigating whether health information technology can lead to more efficient admission decisions by reducing redundant admissions in the stressful environment of emergency. Log-files were retrieved from an integrative and interoperable EHR that serves seven main Israeli hospitals. The analysis was restricted to a group of patients seen in the emergency departments who were administered a Creatinine test. The assessment of the contribution of EHR to admission decisions used various statistical analyses and track log-file analysis. We showed that using the EHR contributes to more efficient admission decisions and reduces the number of avoidable admissions. In particular, there was a reduction in readmissions when patient history was viewed. Using EHR can help respond to the international problem of avoidable hospital readmissions.