Predictive Modeling with Temporal Graphical Representation on Electronic Health Records.

DyHDGE: Dynamic heterogeneous transaction graph embedding for safety-centric fraud detection in financial scenarios

Clinicians usually intend to get overall health status as well as specific health care information from patient's electronic health record (EHR). However, the amount of EHR data can be incredibly large, which makes it a very tedious and inefficient work for clinicians to extract valuable information. This paper first investigates the clinical relevance of EHR data content. Then it presents a Level of Detail (LOD) information navigation model, which classified patient's whole EHR into different detail levels according to their clinical relevance. A novel navigation and visualization method is brought forward based on this model. Finally, an application is designed and implemented based on this method. It can guide clinicians from summarized overview to detailed contents of the EHR smoothly. Both the evolution of specific health problems and other related EHR information can be visualized and located quickly.

With the rapid development of information technologies, which facilitates the perfection of healthcare systems, a variety of clinical data is becoming available. The patient Electronic Health Records (EHR) is one of important sources in healthcare data on which conducts personalized medicine. However, it is challenging if the raw EHRs are directly used to conduct related medical prediction due to its heterogeneity, sparsity and the existence of noise. To address this issue, this paper proposes an integrative data driven medical prediction approach called Medical Temporal Profile Graph (MTPGraph). The approach consists of two parts, first of which is a unified representation for each patient raw EHRs, namely patient temporal profile graph. Secondly, based on this representation, an algorithm TRApriori to obtain temporal feature graphs is further developed which is used to reconstruct each patient temporal profiling. The generated coefficient can be efficiently used for executing medical risk prediction. Finally, we validate MTPGraph through two real world clinical scenarios. The experimental results show that the predicted performance of the approach can be improved significantly in both tasks.

PurposeTo characterize the availability of social determinants of health data in the electronic health record of pediatric ophthalmology patients and to examine the association of social determinants of health with attendance at scheduled operating room and clinic visits.MethodsThis was a retrospective cohort study of pediatric ophthalmology patients seen at The Hospital for Sick Children between June 1, 2018, and May 23, 2022. Data were collected on demographics, diagnosis, and management-plan. The χ2 tests and multivariable regression were used to examine associations between social determinants of health and attendance at scheduled operating room and clinic visits.ResultsThe cohort consisted of 26,102 study subjects with 31,288 unique eye-related diagnoses representing 57 unique ICD-10 codes. Availability of data in the electronic health record ranged from 100% for sex, age and postal code to 0.1% for ethnic group. Female sex (P = 0.004) and urbanicity (P = 0.05) were associated with higher operating room visit cancellations. Female sex (P = 0.002), age group 0–13 (P ≤ 0.001), low-medium neighborhood income quintile (P ≤ 0.001), residence of Northern Ontario (P ≤ 0.001), and urbanicity (P ≤ 0.001) were associated with higher clinic visit cancellations and no-shows.ConclusionsAt a major tertiary-care hospital in Canada, key social determinant data such as ethnicity are not consistently available in the electronic health record of pediatric ophthalmology patients. Female sex, younger age, and living in a rural area or neighborhood with low-medium income quintile may be predictors of missed visits and require further study.Translational RelevanceThis study highlights a need for improved documentation of social determinants of health variables in electronic health records.

The primary care setting is the ideal location for identifying the condition of frailty in older adults. The aim of this pragmatic study was twofold: (1) to identify data items to extract the data required for an electronic Frailty Index (eFI) from electronic health records (EHRs); and (2) test the ability of an eFI to accurately and feasibly identify frailty in older adults. In a rural South Australian primary care clinic, we derived an eFI from routinely collected EHRs using methodology described by Clegg et al. We assessed feasibility and accuracy of the eFI, including complexities in data extraction. The reference standard for comparison was Fried's frailty phenotype. The mean (SD) age of participants was 80.2 (4.8) years, with 36 (60.0%) female (n = 60). Frailty prevalence was 21.7% by Fried's frailty phenotype, and 35.0% by eFI (scores > 0.21). When deriving the eFI, 85% of EHRs were perceived as easy or neutral difficulty to extract the required data from. Complexities in data extraction were present in EHRs of patients with multiple health problems and/or where the majority of data items were located other than on the patient's summary problem list. This study demonstrated that it is entirely feasible to extract an eFI from routinely collected Australian primary care data. We have outlined a process for extracting an eFI from EHRs without needing to modify existing infrastructure. Results from this study can inform the development of automated eFIs, including which data items to best access data from.

Few smokers receive evidence-based tobacco treatment during healthcare visits. Electronic health records (EHRs) present an opportunity to efficiently identify and refer smokers to state tobacco quitlines. The purpose of this case study is to develop and evaluate a secure, closed-loop EHR referral system linking patients visiting healthcare clinics with a state tobacco quitline. A regional health system, EHR vendor, tobacco cessation telephone quitline vendor, and university research center collaborated to modify a health system's EHR to create an eReferral system. Modifications included the following: clinic workflow adjustments, EHR prompts, and return of treatment delivery information from the quitline to the patient's EHR. A markedly higher percentage of adult tobacco users were referred to the quitline using eReferral than using the previous paper fax referral (14 vs. 0.3%). The eReferral system increased the referral of tobacco users to quitline treatment. This case study suggests the feasibility and effectiveness of a secure, closed-loop EHR-based eReferral system.

Estimating population-level obesity rates is important for informing policy and targeting treatment. The current gold standard for obesity measurement in the United States-the National Health and Nutrition Examination Survey (NHANES)-samples <0.1% of the population and does not target state-level or health system-level measurement. To assess the feasibility of using body mass index (BMI) data from the electronic health record (EHR) to assess rates of overweight and obesity and compare these rates to national NHANES estimates. Using outpatient data from 42 clinics, we studied 388,762 patients in a large health system with at least 1 primary care visit in 2011-2012. We compared crude and adjusted overweight and obesity rates by age category and ethnicity (white, black, Hispanic, Other) between EHR and NHANES participants. Adjusted overweight (BMI≥25) and obesity rates were calculated by a 2-step process. Step 1 accounted for missing BMI data using inverse probability weighting, whereas step 2 included a poststratification correction to adjust the EHR population to a nationally representative sample. Adjusted rates of obesity (BMI≥30) for EHR patients were 37.3% [95% confidence interval (95% CI), 37.1-37.5] compared with 35.1% (95% CI, 32.3-38.1) for NHANES patients. Among the 16 different obesity class, ethnicity, and sex strata that were compared between EHR and NHANES patients, 14 (87.5%) contained similar obesity estimates (ie, overlapping 95% CIs). EHRs may be an ideal tool for identifying and targeting patients with obesity for implementation of public health and/or individual level interventions.

Healthcare providers need to communicate and exchange the clinical and administrative data from Electronic Health Records of patients. The Health Insurance Portability and Accountability Act (HIPAA) of 1996 requires the safeguarding and protection of Electronic Health Records when in transit and in storage. This paper presents a solution to protect Electronic Health Records in both states, in transit and at rest. Our secure data container guarantees data confidentiality and integrity and supports different data formats, including text and images. It can work in environments with limited Internet connectivity, which is important at times of disease pandemics or natural disasters. The potential benefit of this solution for healthcare providers would allow emailing data to patients in the form of our data container without using a HIPAA-compliant email server. This technology supports the ability to send and receive medical record documents, for a single encounter or encounters related to a condition, (1) allowing patient to download and keep their personal health record; (2) sharing continuity of care information with a provider referral to ensure medication reconciliation, for instance, with providers who are outside the information exchange network; and (3) delivering medical records securely to an insurance company who requires notes for a submitted claim. To address the problem of clinical and administrative data exchange between different healthcare providers across multiple networks, states and countries, our methodology allows patients to download and store their Electronic Health Records locally.

Constant interruptions and continual data flow result in information overload for clinicians and become barriers to identification and extraction of relevant patient data and its correct interpretation. The aim of the study was to describe the types, frequencies, and impact of intensive care unit (ICU) interruptions on patient safety event occurrences and electronic health records (EHR) use. We conducted a live observational study for 6 weeks, observing critical care physicians' and other providers' communication while recording interruptions, patient safety events, and EHR use. Across 55 hours, the researchers observed 7515 ICU tasks, 15.7% of which were interrupted. We found that technological interruptions directly influences the occurrence of patient safety events: an increase in technological interruptions directly contributes to patient safety event occurrence (P = 0.004). Technological interruptions had a direct effect on human interruptions, as the frequency of technological interruptions increase, human interruptions also increase (P = 0.02). A prospective, observational study was conducted to understand the relationship between interruptions and patient safety events and EHR use, in a time-sensitive, activity-based study in a large academic medical center with a certified EHR system. We found that technological interruptions were statistically correlated to the occurrence of patient safety events, and human interruptions significantly affected the level of EHR use. This study recommends that ICUs adopt a safety culture that promotes minimizing unnecessary interruptions, such as side conversations during rounds, for improved quality of care.

The recent proliferation of cloud computing has leveraged its use in different areas. One of the striking areas is healthcare in which the patient's electronic health care data is maintained by the cloud providers. However, the main concern when using the cloud in sharing the health-care information is the security and privacy of the patients' data. An Electronic Health Record (EHR) system must ensure the security and privacy in both storing and distributing the EHR data to/from the cloud. In this paper, we propose an EHR system, which preserves the privacy of the patient personal information as well as provides the confidentiality to the sensitive medical information using cryptographic organization based access control. Also, the security elements of the proposed system are analyzed to increase the patient's trust on this system.

Transthyretin amyloid cardiomyopathy (ATTR-CM), a progressive and fatal cardiomyopathy, is frequently misdiagnosed or entails diagnostic delays, hindering patients from timely treatment. This study aimed to generate a systematic framework based on data from electronic health records (EHRs) to assess patients with ATTR-CM in a real-world population of heart failure (HF) patients. Predictive factors or combinations of predictive factors related to ATTR-CM in a European population were also assessed. Retrospective unstructured and semi-structured data from EHRs of patients from OLV Hospital Aalst, Belgium (2012-20), were processed using natural language processing (NLP) to generate an Observational Medical Outcomes Partnership Common Data Model database. NLP model performance was assessed on a random subset of EHRs by comparing algorithm outputs to a physician-generated standard (using precision, recall, and their harmonic mean, or F1-score). Of the 3127 HF patients, 103 potentially had ATTR-CM (age 78±9years; male 55%; ejection fraction of 48%±16). The mean diagnostic delay between HF and ATTR-CM diagnosis was 1.8years. Besides HF and cardiomyopathy-related phenotypes, the strongest cardiac predictor was atrial fibrillation (AF; 72% in ATTR-CM vs. 60% in non-ATTR-CM, P=0.02), whereas the strongest non-cardiac predictor was carpal tunnel syndrome (21% in ATTR-CM vs. 3% in non-ATTR-CM, P<0.001). The strongest combination predictor was AF, joint disorders, and HF with preserved ejection fraction (29% in ATTR-CM vs. 18% in non-ATTR-CM: odds ratio=2.03, 95% confidence interval=1.28-3.22). Not only well-known variables associated with ATTR-CM but also unique combinations of cardiac and non-cardiac phenotypes are able to predict ATTR-CM in a real-world HF population, aiding in early identification of ATTR-CM patients.

147 Background: Distress has a significant impact on the quality of life for patients with cancer, however, implementation of such screening is variable. There is limited data in this area for patients who have limited English proficiency (LEP). Given data suggesting worsening disparities in medical care with telemedicine use, we evaluated the completion of distress screening and interpreter use for LEP patients with telemedicine. Methods: We analyzed assigned (complete and incomplete) questionnaires from 10/2019-3/2021 at Stanford Cancer Center (n = 181,105). We defined LEP as a patient electronic health record (EHR) demographic listing either a non-English preferred language or a request for an interpreter. We defined telemedicine as any video or phone visit. Data was analyzed with Fisher's test for differences. Given limitations in EHR data, we conducted a validation subset chart review for LEP patients in Thoracic and Gastrointestinal (GI) oncology which have the two highest % of LEP patients (n = 177 patients). The subset list was obtained for the above timeframe from the interpreter services group. We analyzed EHR notes from visit day for written mention of professional vs family interpreter use. Results: Overall, 14% of our cohort had LEP, highest prevalence in the Thoracic (21%) and GI groups (16%), with a total of 48 languages represented (Spanish, Mandarin &amp; Vietnamese as the 3 most common). There was a significant difference in the English and Non-English groups in overall completion rates of the screening questionnaire (62% in English population vs 49% for LEP, p &lt; 0.001). Completion rates for telemedicine vs. in person visits were overall higher for both English (78.9% vs. 55%, p &lt; 0.001) and LEP (65.1% vs. 43.7%, p &lt; 0.001) groups respectively. The overall screening completion rate for all visits was 57%, with a 62% completion rate with telemedicine and 51% completion with in-person visits (p = 0.2). In the LEP validation subset, there were 18 languages represented with the similar distribution as the larger cohort. Of all clinical notes reviewed, 48.8% included written mention of professional interpreter use, while 25.5% noted interpretation by a family member. There was no difference in professional interpreter use between visit types (̃50%), but with higher rate of family interpretation with telemedicine (35%) vs with in person visits (21%) (p = 0.04). Conclusions: Patients who have some level of limited English proficiency complete distress screening questionnaires less frequently compared to English speaking patients, which highlights the need to improve access to distress screening and supportive care. However, we found similar completion rates with in person vs telemedicine visits, with a signal for increased use of family members as interpreter with telemedicine. This warrants further analysis of the family role and patient understanding during these virtual visits.

Overly restrictive eligibility criteria for clinical trials may limit the generalizability of the trial results to their target real-world patient populations. We developed a novel machine learning approach using large collections of real-world data (RWD) to better inform clinical trial eligibility criteria design. We extracted patients’ clinical events from electronic health records (EHRs), which include demographics, diagnoses, and drugs, and assumed certain compositions of these clinical events within an individual’s EHRs can determine the subphenotypes—homogeneous clusters of patients, where patients within each subgroup share similar clinical characteristics. We introduced an outcome-guided probabilistic model to identify those subphenotypes, such that the patients within the same subgroup not only share similar clinical characteristics but also at similar risk levels of encountering severe adverse events (SAEs). We evaluated our algorithm on two previously conducted clinical trials with EHRs from the OneFlorida+ Clinical Research Consortium. Our model can clearly identify the patient subgroups who are more likely to suffer or not suffer from SAEs as subphenotypes in a transparent and interpretable way. Our approach identified a set of clinical topics and derived novel patient representations based on them. Each clinical topic represents a certain clinical event composition pattern learned from the patient EHRs. Tested on both trials, patient subgroup (#SAE=0) and patient subgroup (#SAE>0) can be well-separated by k-means clustering using the inferred topics. The inferred topics characterized as likely to align with the patient subgroup (#SAE>0) revealed meaningful combinations of clinical features and can provide data-driven recommendations for refining the exclusion criteria of clinical trials. The proposed supervised topic modeling approach can infer the clinical topics from the subphenotypes with or without SAEs. The potential rules for describing the patient subgroups with SAEs can be further derived to inform the design of clinical trial eligibility criteria.

ChatGPT: the future of discharge summaries?

To analyze the risk of systemic lupus erythematosus (SLE) flare after admission for COVID-19. We performed a matched cohort study using the Assistance Publique-Hôpitaux de Paris Clinical Data Warehouse which collects structured medical, biological and administrative information from 11 million patients in Paris area, France. Each SLE patient hospitalized with a COVID-19 diagnosis code between March 2020 and December 2021 was matched to one SLE control patient with an exact matching procedure using age ±3 years, gender, chronic kidney disease, end-stage renal disease, and serological activity. The main outcome was a lupus flare during the 6 months follow-up. A flare was considered if a) documented by the treating physician in the patient's EHR and b) justifying a change in SLE treatment. The electronic health records (EHRs) were individually checked for data accuracy. Among 4,533 SLE patients retrieved from the database, 81 (2.8%) have been admitted for COVID-19 between March 2020 and December 31, 2021, and 79 (n = 79/81,97.5%) were matched to a unique unexposed SLE. During follow-up, a flare occurred in 14 (17.7%) patients from the COVID-19 group as compared to 5 (6.3%) in the unexposed control group, including 4 lupus nephritis in the exposed group and 1 in the control group. After adjusting for HCQ use at index date and history of lupus nephritis, the risk of flare was higher in exposed SLE patients (hazard ratio [95% confidence interval] of 3.79 [1.49-9.65]). COVID-19 hospitalization is associated with an increased risk of flare in SLE.