Electronic Health Record Patient Research Articles

Key interoperability Factors for patient portals and Electronic health Records: A scoping review

AimTo identify the key requirements and challenges to interoperability between patient portals and electronic health records (EHRs). IntroductionPatient portals provide patients with access to their health information directly from EHRs within hospitals, primary care centres and general practices (GPs). Patient portals offer many benefits to patients including improved communication with healthcare providers and care coordination. However, many challenges exist with the integration and automatic and secure sharing of information between EHRs and patient portals. It is critical that countries learn from international experiences to successfully develop interoperable national patient portals. MethodsA scoping review methodology was undertaken. A search strategy using index terms and keywords was applied across four key databases, an additional grey literature search was also run. The identified studies were screened by two reviewers to determine eligibility against defined inclusion criteria. Data were abstracted from the eligible studies and reviewed to identify the key requirements and challenges to interoperability of patient portals with EHRs. ResultsAfter screening 3,462 studies, 34 were included across 11 countries. Of the 29 unique patient portals studied, few offered patients access to their entire healthcare record across multiple sites and a number of different functionalities were available. Key interoperability requirements and challenges identified were: Data Sharing Incentives & Supports; Heterogenous Organisations & Information Systems; Data Storage & Management; Available Information & Functionalities; Data Formats & Standards; Identification of Individuals; User Access, Control & Consent; and Security & Privacy. ConclusionSeamless exchange of health information across patient portals and EHRs required organisational and individual factors, as well as technical considerations. Interorganisational collaboration and engagement of key stakeholders to determine standards and guidelines for consent and sharing of information, as well as technical standards and security measures were recommended.

Beyond Hospital-Level Aggregated Data: A Methodology to Adapt Clinical Data From the Electronic Health Record for Nursing Unit-Level Research.

Studies of nurse staffing frequently use data aggregated at the hospital level that do not provide the appropriate context to inform unit-level decisions, such as nurse staffing. Describe a method to link patient data collected during the provision of routine care and recorded in the electronic health record (EHR) to the nursing units where care occurred in a national dataset. We identified all Veterans Health Administration acute care hospitalizations in the calendar year 2019 nationwide. We linked patient-level EHR and bar code medication administration data to nursing units using a crosswalk. We divided hospitalizations into segments based on the patient's time-stamped location (ward stays). We calculated the number of ward stays and medication administrations linked to a nursing unit and the unit-level and facility-level mean patient risk scores. We extracted data on 1117 nursing units, 3782 EHR patient locations associated with 1,137,391 ward stays, and 67,772 bar code medication administration locations associated with 147,686,996 medication administrations across 125 Veterans Health Administration facilities. We linked 89.46% of ward stays and 93.10% of medication administrations to a nursing unit. The average (standard deviation) unit-level patient severity across all facilities is 4.71 (1.52), versus 4.53 (0.88) at the facility level. Identification of units is indispensable for using EHR data to understand unit-level phenomena in nursing research and can provide the context-specific information needed by managers making frontline decisions about staffing.

Quantifying and Reducing Errors in Vascular Imaging Examination Orders Through a Multistage Quality Improvement Intervention

PurposeThe aims of this study were to quantify order error rates for vascular imaging examinations and to assess the effects of a multistage quality improvement intervention on those rates. MethodsIn this prospective, institutional review board–exempt project at a large academic quaternary care hospital, the authors aimed to quantify and reduce the order error rate by 50%. The authors analyzed 844 orders for all vascular imaging examinations placed before the intervention (July 19 to August 1, 2021, and September 13 to September 26, 2021), after an intervention in the cardiac surgery department consisting of a new customized order option in the electronic health record for routine preoperative patients (postintervention 1, February 28 to March 27, 2022); and after an educational and feedback campaign (postintervention 2, May 23 to June 5, 2022). Incorrect orders were identified by a radiology trainee during protocoling if the reasons for ordered examination and imaging examination were discordant and subsequently confirmed with the ordering provider. The primary outcome, order error rate, was compared across the project periods using the χ2 test and by ordering department using the χ2 and Fisher exact tests. ResultsThe preintervention order error rate of 16% (50 of 306) decreased by 83% to 3% (10 of 353) at postintervention 1 (P < .001) and was durable at 3% (6 of 185) by project end. Chest CT with or without contrast constituted the majority of incorrect orders (44%, 22 of 50); “Pre-Op” was the most common examination reason (32% [16 of 50]). Cardiac surgery orderers were responsible for the most incorrect orders (32% [16 of 50]). All four most common ordering departments, including cardiac surgery, reduced their order error rates after the intervention (P < .001). ConclusionsIncorrect orders for imaging examinations can be reduced through targeted quality improvement interventions combining tailored electronic health record order options with education and feedback on practice habits.

Real world challenges in maintaining data integrity in electronic health records in a cancer program

Electronic Health Record (EHR) systems increase clerical workload, promote copy-paste and error propagation. Documentation error rate in cancer diagnosis and treatment was examined in 776 patient records. Fifteen percent of the charts contained an error. Modern EHR systems, patient portals and engagement tools may facilitate the maintenance of accurate information.

An Access Control Scheme With Privacy-Preserving Authentication and Flexible Revocation for Smart Healthcare.

IoT and 5G-enabled smart healthcare allows medical practitioners to diagnose patients from any location via electronic health records (EHRs) by wireless body area network (WBAN) devices. Privacy, including the medical practitioner's identity and the patient's EHR, can easily be leaked from hospitals or cloud servers, and secret keys used to access EHRs must be revoked after diagnosis. In response to the challenges associated with user authentication and secret key revocation, this paper proposes an access control scheme with privacy-preserving authentication and flexible revocation for smart healthcare using attribute-based encryption (ABE), named PAFR-ABE, which provides access control to prevent malicious users from decrypting EHRs. Meanwhile, PAFR-ABE ensures privacy-preserving authentication for users during secret key generation, which safeguards users' identities and prevents unauthorized requests for secret keys. In addition, PAFR-ABE achieves flexible revocation and recovery of secret keys, which eliminates the need to update secret keys for unrevoked users. Security analysis indicates that PAFR-ABE meets the security requirements of an access control scheme for smart healthcare, especially in terms of forward security and backward security. Performance analysis shows that PAFR-ABE is efficient in the key generation and revocation algorithms compared with typical access control schemes.

Risk of systemic lupus erythematosus flare after COVID-19 hospitalization: A matched cohort study.

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.

Towards Reducing Diagnostic Errors with Interpretable Risk Prediction.

Many diagnostic errors occur because clinicians cannot easily access relevant information in patient Electronic Health Records (EHRs). In this work we propose a method to use LLMs to identify pieces of evidence in patient EHR data that indicate increased or decreased risk of specific diagnoses; our ultimate aim is to increase access to evidence and reduce diagnostic errors. In particular, we propose a Neural Additive Model to make predictions backed by evidence with individualized risk estimates at time-points where clinicians are still uncertain, aiming to specifically mitigate delays in diagnosis and errors stemming from an incomplete differential. To train such a model, it is necessary to infer temporally fine-grained retrospective labels of eventual "true" diagnoses. We do so with LLMs, to ensure that the input text is from before a confident diagnosis can be made. We use an LLM to retrieve an initial pool of evidence, but then refine this set of evidence according to correlations learned by the model. We conduct an in-depth evaluation of the usefulness of our approach by simulating how it might be used by a clinician to decide between a pre-defined list of differential diagnoses.

Merkle-Damgård hash functions and blockchains : Securing electronic health records

The integration of Blockchain (BC) technology with the Merkle-Damgård Structure (M-DS) presents a practical model for securing Electronic Health Records (EHR). This research study examines how these cryptographic mechanisms can be functional to address confidentiality, integrity, and availability concerns when working with specific clinical data. Secure EHR is developed based on the M-DS, which is notable for being capable of generating Collision-Resistant Hash Functions (CRHF). The summary of BC-based technology in the EHR environment enhances access control, allows it to be fixed, and simplifies decentralized access. A decentralized BC system stores the values that have been hashed, resulting in patient EHR using the M-DC, as per the proposed method. As an outcome of smart contracts’ ability to manage access, only persons who have been authorized to access can analyze and alter clinical information. The strategy of the model’s cryptographic mechanism highlights openness and accountability by securing data confidentiality by encryption and by regularly generating a verification record of every transaction. This work discusses an exhaustive method to secure EHR in a clinical system that develops more computer-aided and data-driven over performing a complete assessment of the mathematical models that are enabling MDC and BC Technology within the overall system of EHR security.

A COVID-19 Prediction Using a Blockchain Approach Integrated With UV-NET for Urban Management

Growing cities contributed to the increase in the COVID-19 virus. Consequently, many researchers are exploring the dynamics of the pandemic and to analyse the impacts of pandemic on such cities. The main aim of this research is to understand impacts of the pandemic in Vellore city, Tamil Nadu, India by building a viable solution which can be beneficial for the medical fraternity. A blockchain based approach integrating Artificial Intelligence (AI) is proposed for secure access and storage of Electronic Health Records (EHR) of COVID-19 patients in Vellore city. The blockchain follows the principle of absolute privacy and anonymity of medical records. The decentralized architecture is built to secure from different attacks as the hash of the records are stored in the blockchain. The proposed approach consists of a U-net and V-net model, one for segmenting lungs from the x-rays and second one for segmenting COVID-19 infection patches. The UV-net model is a Convolutional Neural Network (CNN) for fast and precise image segmentation. Experimental analysis is provided on nearly 33,920 chest X-ray images and text records gathered from a hospital in the Vellore city. The proposed model resulted in a precision, recall and F-score of 0.91, 0.87 and 0.89 respectively. The predicted results are manually analysed by the doctor in their login to finally cross verify and conclude the results which are stored in blockchain. This will aid doctors to centrally diagnose the patients and assist in proper treatment for faster recovery.

A machine learning model to predict therapeutic inertia in type 2 diabetes using electronic health record data.

To estimate the therapeutic inertia prevalence for patients with type 2 diabetes, develop and validate a machine learning model predicting therapeutic inertia, and determine the added predictive value of area-level social determinants of health (SDOH). This prognostic study with a retrospective cohort design used OneFlorida data (linked electronic health records (EHRs) from 1240 practices/clinics in Florida). The study cohort included adults (aged ≥ 18) with type 2 diabetes, HbA1C ≥ 7% (53mmol/mol), ≥one ambulatory visit, and ≥one antihyperglycemic medication prescribed (excluded patients prescribed insulin before HbA1C). The outcome was therapeutic inertia, defined as absence of treatment intensification within six months after HbA1C ≥ 7% (53mmol/mol). The predictors were patient, provider, and healthcare system factors. Machine learning methods included gradient boosting machines (GBM), random forests (RF), elastic net (EN), and least absolute shrinkage and selection operator (LASSO). The DeLong test compared the discriminative ability (represented by C-statistics) between models. The cohort included 31,087 patients with type 2 diabetes (mean age = 58.89 (SD = 13.27)years, 50.50% male, 58.89% White). The therapeutic inertia prevalence was 39.80% among the 68,445 records. GBM outperformed (C-statistic from testing sample = 0.84, 95% CI = 0.83-0.84) RF (C-statistic = 0.80, 95% CI = 0.79-0.80), EN (C-statistic = 0.80, 95% CI = 0.80-0.81), and LASSO (C-statistic = 0.80, 95% CI = 0.80-0.81), p < 0.05. Area-level SDOH significantly increased the discriminative ability versus models without SDOH (C-statistic for GBM = 0.84, 95% CI = 0.84-0.85 vs. 0.84, 95% CI = 0.83-0.84), p < 0.05. Using EHRs of patients with type 2 diabetes from a large state, machine learning predicted therapeutic inertia (prevalence = 40%). The model's ability to predict patients at high risk of therapeutic inertia is clinically applicable to diabetes care.

ACDNet: Attention-guided Collaborative Decision Network for effective medication recommendation

Medication recommendation using Electronic Health Records (EHR) is challenging due to complex medical data. Current approaches extract longitudinal information from patient EHR to personalize recommendations. However, existing models often lack sufficient patient representation and overlook the importance of considering the similarity between a patient’s medication records and specific medicines. Therefore, an Attention-guided Collaborative Decision Network (ACDNet) for medication recommendation is proposed in this paper. Specifically, ACDNet utilizes attention mechanism and Transformer to effectively capture patient health conditions and medication records by modeling their historical visits at both global and local levels. ACDNet also employs a collaborative decision framework, utilizing the similarity between medication records and medicine representation to facilitate the recommendation process. The experimental results on two extensive medical datasets, MIMIC-III and MIMIC-IV, clearly demonstrate that ACDNet outperforms state-of-the-art models in terms of Jaccard, PR-AUC, and F1 score, reaffirming its superiority. Moreover, the ablation experiments provide solid evidence of the effectiveness of each module in ACDNet, validating their contribution to the overall performance. Furthermore, a detailed case study reinforces the effectiveness of ACDNet in medication recommendation based on EHR data, showcasing its practical value in real-world healthcare scenarios.

Incomplete and possibly selective recording of signs, symptoms and measurements in free text fields of primary care electronic health records of adults with lower respiratory tract infections

ObjectiveTo assess the completeness of recording of relevant signs, symptoms, and measurements in Dutch free text fields of primary care electronic health records (EHR) of adults with lower respiratory tract infections (LRTI). Study design and settingRetrospective cohort study embedded in a prediction modelling project using routine health care data of the Julius General Practitioners’ Network of adult patients with LRTIs. Free text fields of 1,000 primary care consultations of LRTI episodes between 2016-2019 were manually annotated to retrieve data on the recording of sixteen relevant signs, symptoms, and measurements. ResultsFor 12/16 (75%) of the relevant signs, symptoms, and measurements, more than 50% of the values was not recorded. The patterns of recorded values indicated selective recording of positive or abnormal values. Recording rates varied across consultation type (physical consultation vs. home visit), diagnosis (acute bronchitis vs. pneumonia), antibiotic prescription issued (yes vs. no), and between practices. ConclusionIn EHR of primary care LRTI patients, recording of signs, symptoms, and measurements in free text fields is incomplete and possibly selective. When using free text data in EHR-based research, careful consideration of its recording patterns and appropriate missing data handling techniques is therefore required.

The development of an explainable machine learning system for predicting brain atrophies from electronic health records

AbstractBackgroundThe number of elderly living with dementia is projected to dramatically increase, especially in low‐income and middle‐income countries (LMICs). While obtaining brain scans to detect atrophies and significant radiology findings related to dementia might be costly, the use of electronic health record (EHR) for elderly patients could facilitate identifying patients at a high risk based on routinely collected data. This works, evaluates the use of an interpretable machine learning‐based system that predicts brain atrophies and small vessel disease based on EHR data.MethodWe develop a machine learning predictive system to predict atrophies or small vessel disease from retrospective anonymized EHR data recorded for patients above the age of 50 years from a multispecialty outpatient medical center from Ramallah, Palestine. The patient labelling involved using natural language processing to identify cases in unstructured radiology reports for structural brain Magnetic Resonance Imaging (MRI). For input features, we extracted and preprocessed 30 features for the most commonly collected clinical features such as demographics and laboratory results. The dataset was split into training and test sets, where data preprocessing and augmentation, and hyperparameter tuning were applied before training multiple models including Xgboost, Support Vector Machines, Random Forest, and Logistic Regression. To better understand the relative impact each feature had on the model’s prediction, shapely feature importance analysis was conducted for the top performing model.ResultThe top preforming model was Random Forest, with a performance of Area Under the Precision‐Recall Curve (AUPRC) of 0.989, and 0.842 for Area Under Receiving Operating Curve (AUROC). The feature importance analysis revealed that age, sex, Vitamin D, Vitamin B12 and red blood cell distribution width were the features with highest impact on the model’s predictionConclusionPredicting atrophies from EHRs for elderly visiting outpatient clinics could assist in case‐finding for dementia patients and ultimately help identify patients that would benefit from an MRI. In future works, we aim to further investigate building low‐cost diagnostic tools based on a smaller feature set, which could be valuable for facilities from LMICs.

Social Determinants of Health in Pediatric Ophthalmology Patients: Availability of Data in the Electronic Health Record and Association With Clinic Attendance.

To 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. This 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. The 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. At 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. This study highlights a need for improved documentation of social determinants of health variables in electronic health records.

Benefit of Inpatient Pain Plan Utilization on Hospital Length of Stay in Adolescents and Young Adults with Sickle Cell Disease

Introduction Vaso-occlusive crisis (VOC) is the most common manifestation of sickle cell disease (SCD) and the most common cause for emergency department (ED) visits and hospital admissions. VOC accounts for approximately 230,000 annual admissions in children with SCD. Patients with SCD face inconsistent analgesic management, leading to high inpatient healthcare utilization and financial burden for healthcare institutions. Furthermore, patients describe their pain management as the area of their care in most need of improvement. A prospective longitudinal cohort study showed that implementation of individualized pain plans in adult patients with SCD significantly decreased the duration of VOC hospital stays by 1.2 days. A similar study in children showed no change in duration of hospital stay, however the study was limited by a small sample size. We hypothesized parallel results to the prior adult study in adolescent and young adult (AYA) SCD patients. Methods At UCSF Benioff Children's Hospital Oakland, we implemented inpatient pain plans for SCD patients with frequent VOC admissions (i.e., &amp;gt;1 VOC admission in the prior year or complicated VOC admission history). Pain plans are individualized and embedded into the patient's electronic health record. These plans include detailed daily analgesic recommendations with the general recommendation for discharge on day 5 of the pain plan. After IRB approval, we performed a retrospective chart review of SCD patients with individualized inpatient pain plans initiated between November 2017 and April 2022. We recorded baseline demographic information including age at pain plan initiation, SCD genotype and sex. We assessed length of inpatient hospitalization for VOC admissions in the year prior and the year after pain plan initiation for each individual patient. We also noted number of admissions secondary to VOC during this two year time period. Results Twenty-six patients with SCD and individualized pain plans were identified (Table 1). Average length of hospital stay for VOC was significantly decreased in the year after inpatient pain plan initiation (Table 2). There was a trend towards increased admissions due to VOC in the year after inpatient pain plan initiation (Table 2). Conclusions Similar to the prior adult study in SCD patients, we found significantly reduced inpatient length of hospitalization after initiation of an inpatient pain plan for AYA patients with SCD and VOC. Although there was a trend towards increased hospitalizations in the year after inpatient pain plan initiation, this was likely secondary to the average patient age with an expected increased in VOC year over year. The National Heart, Lung, and Blood Institute (NHLBI) SCD guidelines recommend an individualized prescribing and monitoring protocol be written by a SCD provider and used in patients withVOC to promote effective, timely, and safe analgesic management. We further demonstrate that individualized pain plans for AYA patients with SCD can decrease length of hospital stay due to VOC, likely secondary to more effective pain management recommendations that have been prior agreed upon by the patient/family and SCD provider. We currently are calculating morphine equivalents and hypothesize total opioid utilization will be concomitantly decreased due to this significantly decreased length of hospitalization.

Impact of an Individualized Pain Plan (IPP) for Adults with Sickle Cell Disease Experiencing a Vaso-Occlusive Episode (VOE) on Patient Satisfaction and ED Provider Confidence in Treating Pain

Introduction The leading cause of Emergency Department (ED) visits among individuals with sickle cell disease (SCD) is acute pain vaso-occlusive episodes (VOE's). NHLBI guidelines recommend treatment with individualized pain plans (IPP's). To facilitate the uptake of IPP's, the National Heart Lung and Blood Institute funded eight centers across the US to participate in the Sickle Cell Disease Implementation Consortium (SCDIC). We report findings from one of the consortium's implementation projects: An Individualized Pain Plan with Patient and Provider Access for Emergency Department Care of Sickle Cell Disease (ALIGN). In ALIGN, the SCD specialist developed individualized pain plans (IPPs) for ED providers via the electronic health record (EHR) and patients via their patient portal in a phone app. We assessed changes in patient satisfaction with ED care and ED provider comfort in managing VOE after a recent visit between baseline and after IPP implementation for ED providers and patients. We hypothesized an improvement of 0.5 standard deviations in patient satisfaction from baseline to after the implementation of IPPs. We also assessed patient and provider willingness to use IPPs in the EHR and patient portal. Methods Patients and providers were enrolled from the eight SCDIC sites; six sites had IPPs already available in the EHR for ED providers, and none had IPPs available in the patient portal. Patients were eligible if they were 18-45 years old, had SCD, had access to a cell phone, had an ED VOE visit 90 days before enrollment, and had an ED VOE visit to the study site's ED within one year. Provider inclusion criteria were ED physicians, nurse practitioners, and physician's assistants with access to the EHR. The SCD specialist wrote the IPP after patient enrollment. Patients completed surveys at baseline about an ED visit within the last 90 days and again after enrollment, within 96 hours after an ED visit for VOE. The surveys collected data on patient-perceived quality of care and satisfaction with the ED. Providers completed surveys at baseline and after managing an enrolled patient in the ED. The surveys collected data on provider comfort in managing VOE. Both patients and providers answered questions about their likelihood of using IPPs. Descriptive statistics were utilized to examine sociodemographic characteristics and summarize patient and provider satisfaction. Results One hundred fifty-nine patients completed the baseline and follow-up survey. Almost all were African American (95.4%), female (64.7%), and had Hb SS/Sβ 0 genotype (71.9%). Majority (57.6%) reported they were quite a bit or very much satisfied with the care they received at baseline. There was no difference in satisfaction at baseline vs post IPP implementation (P=0.3). About 77.2% of patients responded either very much or quite a bit when asked how much the ED doctors and nurses believed they had very bad SCD pain after implementing IPPs compared with 65.4% at baseline (p=0.02). Most (64.1%) patients were more or extremely likely to show the doctor their pain plan when they returned to the ED. One hundred thirty-five providers completed the baseline and at least one follow-up survey. There was no difference is provider comfortability in their ability to manage the acute VOE post IPP implementation compared to baseline (p=0.6). More than half (62.2%) could see the pain plan in the EHR and 52.38% found the IPPs extremely easy to use. Almost all (97.6%) ED providers ordered the opioids as suggested in the IPPs. Almost all (94.7%) reported they intended to use the IPPs most or all the time. Conclusion In this implementation science protocol, all eight sites successfully implemented IPPs in the EHR for ED providers and patients in their patient portal. Both patients and ED providers enthusiastically endorsed the use of IPP's, which can help guide management of VOE in any ED, regardless of the center.

917. Development and Validation of an Automated Natural Language Processing Method for Identifying Postherpetic Neuralgia from Electronic Health Records

Abstract Background Postherpetic neuralgia (PHN) is a debilitating complication of herpes zoster (HZ). PHN diagnosis codes tend to be inaccurate or incomplete. The criterion standard for identifying PHN in electronic health records (EHR) is manual chart review, which can be costly and time-consuming. We developed and validated a method to automatically identify PHN from free-text EHR data using natural language processing (NLP). Methods This study utilized EHR data of patients aged ≥50 years who had an incident HZ diagnosis and associated antiviral prescription between 2018-2022 at Kaiser Permanente Southern California. Among patients with ≥1 encounter during the 90-180 days after the incident HZ diagnosis, PHN was defined as pain/discomfort consistent with the HZ episode between 90-180 days after the initial HZ diagnosis; the symptoms were at the location of the initial HZ rash and were not due to other obvious causes. Trained research associates and an infectious disease physician manually reviewed and identified PHN cases from EHR. From these reviewed HZ cases, we randomly selected 500 and 800 cases for NLP development and validation, respectively. We developed and validated an NLP algorithm to identify PHN based on our case definition. Using chart-reviewed results as the criterion standard, the accuracy of NLP-based results was compared with that of diagnosis code-based results (ICD-10 codes B02.22, B02.23, B02.29 from all clinical encounters during the 90-180-day period). Results Compared to the criterion standard, the NLP algorithm achieved 100% sensitivity (code-based 64.8%), 99.9% specificity (code-based 97.0%), 98.6% positive predictive value (code-based 67.7%), and 100% negative predictive value (code-based 96.6%) in identifying PHN cases. In the validation data, the prevalence of PHN was 8.9% based on manual chart review. Although the prevalence of PHN identified by the two methods was similar (NLP-based 9.0%, code-based 8.5%), the code-based method misclassified more than one-third of chart-confirmed PHN cases. Accuracy measurements of NLP and diagnosis codes in identifying PHN, as compared with chart-confirmed validation data. Conclusion We developed and validated an automated method to identify PHN cases using the clinical text from EHR of HZ patients with high accuracy. This method can be a valuable tool to facilitate population-based studies of PHN. Disclosures Bradley Ackerson, MD, Dynavax: Grant/Research Support|Genentech: Grant/Research Support|GlaxoSmithKline: Grant/Research Support|Moderna: Grant/Research Support|Pfizer: Grant/Research Support Leticia I. Vega Daily, MSW, GlaxoSmithKline: Grant/Research Support Jeannie Song, MPH, GlaxoSmithKline: Grant/Research Support|Moderna: Grant/Research Support|Pfizer: Grant/Research Support Lina S. Sy, MPH, Dynavax: Grant/Research Support|GlaxoSmithKline: Grant/Research Support|Moderna: Grant/Research Support Lei Qian, PhD, Dynavax: Grant/Research Support|GlaxoSmithKline: Grant/Research Support|Moderna: Grant/Research Support Yi Luo, PhD, GlaxoSmithKline: Grant/Research Support|Moderna: Grant/Research Support|Pfizer: Grant/Research Support Ana Florea, PhD MPH, Gilead: Grant/Research Support|GlaxoSmithKline: Grant/Research Support|Moderna: Grant/Research Support|Pfizer: Grant/Research Support Jennifer H. Ku, PhD MPH, GlaxoSmithKline: Grant/Research Support|Moderna: Grant/Research Support Hung Fu Tseng, PhD MPH, GSK: Grant/Research Support|Moderna: Grant/Research Support

Assessing, updating and utilising primary care smoking records for lung cancer screening

BackgroundLung cancer screening with low-dose computed tomography for high-risk populations is being implemented in the UK. However, inclusive identification and invitation of the high-risk population is a major challenge for equitable lung screening implementation. Primary care electronic health records (EHRs) can be used to identify lung screening-eligible individuals based on age and smoking history, but the quality of EHR smoking data is limited. This study piloted a novel strategy for ascertaining smoking status in primary care and tested EHR search combinations to identify those potentially eligible for lung cancer screening.MethodsSeven primary care General Practices in South Wales, UK were included. Practice-level data on missing tobacco codes in EHRs were obtained. To update patient EHRs with no tobacco code, we developed and tested an algorithm that sent a text message request to patients via their GP practice to update their smoking status. The patient’s response automatically updated their EHR with the relevant tobacco code. Four search strategies using different combinations of tobacco codes for the age range 55-74+ 364 were tested to estimate the likely impact on the potential lung screening-eligible population in Wales. Search strategies included: BROAD (wide range of ever smoking codes); VOLUME (wide range of ever-smoking codes excluding “trivial” former smoking); FOCUSED (cigarette-related tobacco codes only), and RECENT (current smoking within the last 20 years).ResultsTobacco codes were not recorded for 3.3% of patients (n = 724/21,956). Of those with no tobacco code and a validated mobile telephone number (n = 333), 55% (n = 183) responded via text message with their smoking status. Of the 183 patients who responded, 43.2% (n = 79) had a history of smoking and were potentially eligible for lung cancer screening. Applying the BROAD search strategy was projected to result in an additional 148,522 patients eligible to receive an invitation for lung cancer screening when compared to the RECENT strategy.ConclusionAn automated text message system could be used to improve the completeness of primary care EHR smoking data in preparation for rolling out a national lung cancer screening programme. Varying the search strategy for tobacco codes may have profound implications for the size of the population eligible for lung-screening invitation.

Abstract 15291: Early Outcomes With Left Cardiac Sympathetic Denervation for Medically Refractory Idiopathic Ventricular Fibrillation

Introduction: Treatment options for patients with recurrent ventricular arrhythmias (VAs) refractory to pharmacotherapy and/or ablation are minimal. Although left cardiac sympathetic denervation (LCSD) is well established in long QT syndrome (LQTS) and catecholaminergic polymorphic ventricular tachycardia (CPVT), LCSD’s role in idiopathic ventricular fibrillation (IVF) is less clear. Here, we report our single-center experience regarding the early outcomes of LCSD in patients with IVF. Hypothesis: To determine the efficacy of LCSD on reducing VAs in IVF patients refractory to conventional therapy. Methods: A retrospective analysis of the electronic health record of patients with IVF between January 2000 to April 2023 was performed. Cardiac event rates [defined as arrhythmic syncope/seizure, appropriate implantable cardioverter defibrillator (ICD) therapy, sudden cardiac arrest (SCA), or sudden cardiac death (SCD)] before and after LCSD for each patient were calculated and a Wilcoxon signed rank test was performed. Results: In total, 94 patients (41% female) were classified as IVF. Overall, 15/94 patients (16%) had LCSD due to refractory VAs (60% male; mean age at LCSD 36 ± 20 years) despite beta-blocker treatment (100%) or combined drug therapy (67%). In total, 14 patients (15%) had an electrophysiology study performed before LCSD of which 10 (67%) had VAs ablation. All patients had a structurally normal heart and genetic testing performed in 14/15 patients (93%) confirmed absence of genetic cardiomyopathies and channelopathies. SCA was the sentinel event in 10 of 15 patients (67%); 14 patients (93%) had an ICD in place. The average follow-up time after LCSD was 3 ± 4 years. The VA event rate before LCSD was 4 ± 3 events per year and was reduced to 1 ± 3 per year (p&lt;0.04) after LCSD. Patients had a median of 3.2 less events per year after LCSD (p=0.02) with 7 patients (47%) having no post-LCSD recurrences thus far while 3 (20%) patients were deemed LCSD non-responders with no reduction of event rate. Conclusions: LCSD represents a substrate-independent, anti-fibrillatory treatment option for patients with life-threatening, genetic arrhythmias. While not curative, the early outcomes of LCSD for patients with medically refractory IVF are promising.

Abstract 14276: Using Artificial Intelligence and Natural Language Processing on Electronic Health Records of Heart Failure Patients to Examine the Impact of Estimated Glomerular Filtration Rate Trend on Mortality

Introduction and Hypothesis: Heart failure (HF) often leads to an imbalance between heart and kidney function, with a poor prognosis in HF patients linked to dynamic changes in renal function as well as chronic kidney disease (CKD). This study aimed to determine the incidence of reduced renal function upon HF diagnosis and to explore the potential for dynamic changes in kidney function to predict short- and long-term outcomes in a real-world cohort of HF patients. Methods: Clinical data of patients with a diagnosis of HF (index event) were gathered retrospectively using natural language processing. Patients were classified as having CKD based on their eGFR (&lt;60 mL/min/1.73m2) and a history of CKD and/or dialysis. eGFR trajectories were analyzed using linear mixed effects models. Cox proportional hazard models were used to analyze the relationship between baseline variables and mortality. Joint modeling was employed to assess the relationship between the eGFR trend and mortality. Results: There were 1986 patients in the study population, with a mean (SD) age of 74.8 ± 11.7 years and 58% of males. At the index event, 58% (n = 1156) had CKD, and 17% (n = 339) had HF with preserved ejection fraction (HFpEF). Median follow-up was 3.16 years. After 2 years, 10% (n = 198) of the patients had passed away. The proportion of CKD at index was considerably lower in the 2-year survivors group (55% vs 76%, p &lt; 0.001) in this group. The calculated mean annual eGFR reduction was 4.2 mL/min/1.73m 2 . Age, CVA, diastolic blood pressure, hemoglobin, HDL, potassium, and NT-proBNP levels in blood serum were found to be the primary predictors of mortality in multivariate Cox PH model. Joint modeling revealed that a current eGFR value that was 10 mL/min/1.73m 2 lower between patients corresponded with a higher mortality hazard of 1.22 ± 0.05 (p &lt; 0.001). In addition, a drop in eGFR of 10 mL/min/1.73m 2 over the previous year showed an the elevated mortality hazard of 1.97 ± 0.16 (p &lt; 0.001). Conclusions: We show that the prevalence of CKD is significant in a real-world population of HF patients, and CKD is a standalone predictor of mortality. Current eGFR value and the eGFR slope from the previous year have a potential to be used for individual mortality hazard assessment in the clinical follow-up of HF patients.