Asthma innovations from the second International Collaborative Asthma Network (ICAN) forum.

Asthma continues to cause morbidity and mortality despite advances in treatment that include biologics targeting Type 2 inflammation. The International Collaborative Asthma Network (ICAN) forum was developed with the primary goal of promoting innovative, collaborative research that focuses on mechanisms and treatment for asthma that does not respond or that responds poorly to currently available treatments. The mission of ICAN is innovation, collaboration, translation, and increasing high quality research. At the second ICAN meeting, presenters covered a broad scope and depth of asthma-related topics in the categories of complex data, novel therapeutics and diagnostics, breath analysis and microbiome, disease mechanisms, systemic effects, and circadian rhythm. Key actionable needs and research topics were identified during the group discussions. The presentations and discussions that occurred at the second ICAN had an immediate impact on asthma research in the form of new collaborations and implementation of new research ideas and techniques. The forum also served to connect early-stage investigators with investigators who are well established, thereby fostering innovation, translation, and collaboration well into the future. A third ICAN meeting is planned for 2025 to further the innovations and collaborations that will translate into novel therapies and diagnostics to improve the lives of patients with asthma.

The Pediatric Data Science and Analytics Subgroup of the Pediatric Acute Lung Injury and Sepsis Investigators Network: Use of Supervised Machine Learning Applications in Pediatric Critical Care Medicine Research.

OBJECTIVE: Perform a scoping review of supervised machine learning in pediatric critical care to identify published applications, methodologies, and implementation frequency to inform best practices for the development, validation, and reporting of predictive models in pediatric critical care. DESIGN: Scoping review and expert opinion. SETTING: We queried CINAHL Plus with Full Text (EBSCO), Cochrane Library (Wiley), Embase (Elsevier), Ovid Medline, and PubMed for articles published between 2000 and 2022 related to machine learning concepts and pediatric critical illness. Articles were excluded if the majority of patients were adults or neonates, if unsupervised machine learning was the primary methodology, or if information related to the development, validation, and/or implementation of the model was not reported. Article selection and data extraction were performed using dual review in the Covidence tool, with discrepancies resolved by consensus. SUBJECTS: Articles reporting on the development, validation, or implementation of supervised machine learning models in the field of pediatric critical care medicine. INTERVENTIONS: None. MEASUREMENTS AND MAIN RESULTS: Of 5075 identified studies, 141 articles were included. Studies were primarily (57%) performed at a single site. The majority took place in the United States (70%). Most were retrospective observational cohort studies. More than three-quarters of the articles were published between 2018 and 2022. The most common algorithms included logistic regression and random forest. Predicted events were most commonly death, transfer to ICU, and sepsis. Only 14% of articles reported external validation, and only a single model was implemented at publication. Reporting of validation methods, performance assessments, and implementation varied widely. Follow-up with authors suggests that implementation remains uncommon after model publication. CONCLUSIONS: Publication of supervised machine learning models to address clinical challenges in pediatric critical care medicine has increased dramatically in the last 5 years. While these approaches have the potential to benefit children with critical illness, the literature demonstrates incomplete reporting, absence of external validation, and infrequent clinical implementation.

Incidence rates of childhood asthma with recurrent exacerbations in the US Environmental influences on Child Health Outcomes (ECHO) program.

BACKGROUND: Descriptive epidemiological data on incidence rates (IRs) of asthma with recurrent exacerbations (ARE) are sparse. OBJECTIVES: This study hypothesized that IRs for ARE would vary by time, geography, age, and race and ethnicity, irrespective of parental asthma history. METHODS: The investigators leveraged data from 17,246 children born after 1990 enrolled in 59 US with 1 Puerto Rican cohort in the Environmental Influences on Child Health Outcomes (ECHO) consortium to estimate IRs for ARE. RESULTS: The overall crude IR for ARE was 6.07 per 1000 person-years (95% CI: 5.63-6.51) and was highest for children aged 2-4 years, for Hispanic Black and non-Hispanic Black children, and for those with a parental history of asthma. ARE IRs were higher for 2- to 4-year-olds in each race and ethnicity category and for both sexes. Multivariable analysis confirmed higher adjusted ARE IRs (aIRRs) for children born 2000-2009 compared with those born 1990-1999 and 2010-2017, 2-4 versus 10-19 years old (aIRR = 15.36; 95% CI: 12.09-19.52), and for males versus females (aIRR = 1.34; 95% CI 1.16-1.55). Black children (non-Hispanic and Hispanic) had higher rates than non-Hispanic White children (aIRR = 2.51; 95% CI 2.10-2.99; and aIRR = 2.04; 95% CI: 1.22-3.39, respectively). Children born in the Midwest, Northeast and South had higher rates than those born in the West (P < .01 for each comparison). Children with a parental history of asthma had rates nearly 3 times higher than those without such history (aIRR = 2.90; 95% CI: 2.43-3.46). CONCLUSIONS: Factors associated with time, geography, age, race and ethnicity, sex, and parental history appear to influence the inception of ARE among children and adolescents.

Breast Cancer Screening, Diagnosis, and Surgery during the Pre- and Peri-pandemic: Experience of Patients in a Statewide Health Information Exchange.

BACKGROUND: Measures taken to address the COVID-19 pandemic interrupted routine diagnosis and care for breast cancer. The aim of this study was to characterize the effects of the pandemic on breast cancer care in a statewide cohort. PATIENTS AND METHODS: Using data from a large health information exchange, we retrospectively analyzed the timing of breast cancer screening, and identified a cohort of newly diagnosed patients with any stage of breast cancer to further access the information available about their surgical treatments. We compared data for four subgroups: pre-lockdown (preLD) 25 March to 16 June 2019; lockdown (LD) 23 March to 3 May 2020; reopening (RO) 4 May to 14 June 2020; and post-lockdown (postLD) 22 March to 13 June 2021. RESULTS: During LD and RO, screening mammograms in the cohort decreased by 96.3% and 36.2%, respectively. The overall breast cancer diagnosis and surgery volumes decreased up to 38.7%, and the median time to surgery was prolonged from 1.5 months to 2.4 for LD and 1.8 months for RO. Interestingly, higher mean DCIS diagnosis (5.0 per week vs. 3.1 per week, p < 0.05) and surgery volume (14.8 vs. 10.5, p < 0.05) were found for postLD compared with preLD, while median time to surgery was shorter (1.2 months vs. 1.5 months, p < 0.0001). However, the postLD average weekly screening and diagnostic mammogram did not fully recover to preLD levels (2055.3 vs. 2326.2, p < 0.05; 574.2 vs. 624.1, p < 0.05). CONCLUSIONS: Breast cancer diagnosis and treatment patterns were interrupted during the lockdown and still altered 1 year after. Screening in primary care should be expanded to mitigate possible longer-term effects of these interruptions.

Social determinants of health and pediatric cancer survival: A systematic review.

Despite treatment advancements and improved survival, approximately 1800 children in the United States will die of cancer annually. Survival may depend on nonclinical factors, such as economic stability, neighborhood and built environment, health and health care, social and community context, and education, otherwise known as social determinants of health (SDoH). Extant literature reviews have linked socioeconomic status (SES) and race to disparate outcomes; however, these are not inclusive of all SDoH. Thus, we conducted a systematic review on associations between SDoH and survival in pediatric cancer patients. Of the 854 identified studies, 25 were included in this review. In addition to SES, poverty and insurance coverage were associated with survival. More studies that include other SDoH, such as social and community factors, utilize prospective designs, and conduct analyses with more precise SDoH measures are needed.

Chromosome 17q12-21 Variants Are Associated with Multiple Wheezing Phenotypes in Childhood.

Rationale: Birth cohort studies have identified several temporal patterns of wheezing, only some of which are associated with asthma. Whether 17q12-21 genetic variants, which are closely associated with asthma, are also associated with childhood wheezing phenotypes remains poorly explored.Objectives: To determine whether wheezing phenotypes, defined by latent class analysis (LCA), are associated with nine 17q12-21 SNPs and if so, whether these relationships differ by race/ancestry.Methods: Data from seven U.S. birth cohorts (n = 3,786) from the CREW (Children's Respiratory Research and Environment Workgroup) were harmonized to represent whether subjects wheezed in each year of life from birth until age 11 years. LCA was then performed to identify wheeze phenotypes. Genetic associations between SNPs and wheeze phenotypes were assessed separately in European American (EA) (n = 1,308) and, for the first time, in African American (AA) (n = 620) children.Measurements and Main Results: The LCA best supported four latent classes of wheeze: infrequent, transient, late-onset, and persistent. Odds of belonging to any of the three wheezing classes (vs. infrequent) increased with the risk alleles for multiple SNPs in EA children. Only one SNP, rs2305480, showed increased odds of belonging to any wheezing class in both AA and EA children.Conclusions: These results indicate that 17q12-21 is a "wheezing locus," and this association may reflect an early life susceptibility to respiratory viruses common to all wheezing children. Which children will have their symptoms remit or reoccur during childhood may be independent of the influence of rs2305480.

An evaluation of an Extension for Community Healthcare Outcomes (ECHO) intervention in cancer prevention and survivorship care.

To improve cancer care in Indiana, a telementoring program using the Extension for Community Healthcare Outcomes (ECHO) model was introduced in September 2019 to promote best-practice cancer prevention, screening, and survivorship care by primary care providers (PCPs). The aim of this study was to evaluate the program's educational outcomes in its pilot year, using Moore's Evaluation Framework for Continuing Medical Education and focusing on the program's impact on participants' knowledge, confidence, and professional practice. We collected data in 22 semi-structured interviews (13 PCPs and 9 non-PCPs) and 30 anonymous one-time surveys (14 PCPs and 16 non-PCPs) from the program participants (hub and spoke site members), as well as from members of the target audience who did not participate. In the first year, average attendance at each session was 2.5 PCPs and 12 non-PCP professionals. In spite of a relatively low PCP participation, the program received very positive satisfaction scores, and participants reported improvements in knowledge, confidence, and practice. Both program participants and target audience respondents particularly valued three features of the program: its conversational format, the real-life experiences gained, and the support received from a professional interdisciplinary community. PCPs reported preferring case discussions over didactics. Our results suggest that the Cancer ECHO program has benefits over other PCP-targetted cancer control interventions and could be an effective educational means of improving cancer control capacity among PCPs and others. Further study is warranted to explain the discrepancies among study participants' perceptions of the program's strengths and the relatively low PCP participation before undertaking a full-scale effectiveness study.

The Impact of Diagnostic Decisions on Patient Experience in the Pediatric Emergency Department.

OBJECTIVE: Patient experience serves as both a subjective measure of value-based health care delivery and a metric to inform operational decision making. The objective of this study was to determine if specific diagnostic and therapeutic interventions affect patient experience scores for children seen in the emergency department. METHODS: We performed a retrospective observational study in the emergency department of a large quaternary care children's hospital on patients who were discharged to home and later completed a National Research Corporation Health patient experience survey. We matched the survey results to electronic health record (EHR) data and were able to extract demographics, operational metrics, and order information for each patient. We performed multiple logistic regression analyses to determine the association of image acquisition, laboratory test ordering, medication administration, and discharge prescribing with likelihood to recommend the facility as our measure of patient experience. RESULTS: Of the 4103 patients who met inclusion criteria for the study, 75% strongly recommended the facility. Longer wait times were associated with lower patient experience scores [odds ratio (OR) per waiting room hour increase, 0.72; 95% confidence interval (CI), 0.65-0.81]. Significant diagnostic factors associated with higher patient experience included magnetic resonance imaging ordering (OR, 2.38; 95% CI, 1.00-5.67), x-ray ordering (OR, 1.19; 95% CI, 1.00-1.42), and electrocardiogram ordering (OR, 1.62; 95% CI, 1.07-2.44). Of the treatment factors studied, only antibiotic prescribing at discharge was found to have a significant positive association with patient experience (OR, 1.32; 95% CI, 1.08-1.63). CONCLUSION: The positive association between more intensive diagnostic workups and patient experience could have implications on the utility of patient experience scores to evaluate pediatric care teams. Consideration should be taken to interpret patient experience scores in the context of compliance with approaches in evidence-based medicine.

Sensitivity and Specificity of Real-World Social Factor Screening Approaches.

Health care organizations are increasingly documenting patients for social risk factors in structured data. Two main approaches to documentation, ICD-10 Z codes and screening questions, face limited adoption and conceptual challenges. This study compared estimates of social risk factors obtained via screening questions and ICD-10 Z diagnoses coding, as used in clinical practice, to estiamtes from validated survey instruments in a sample of adult primary care and emergency department patients at an urban safety-net health system. Financial strain, transportation barriers, food insecurity, and housing instability were independently assessed using instruments with published reliability and validity. These four social factors were also being collected by the health system in screening questions or could be mapped to ICD-10 Z code diagnosis code concepts. Neither the screening questions nor ICD-10 Z codes performed particularly well in terms of accuracy. For the screening questions, the Area Under the Curve (AUC) scores were 0.609 for financial strain, 0.703 for transportation, 0.698 for food insecurity, and 0.714 for housing instability. For the ICD-10 Z codes, AUC scores tended to be lower in the range of 0.523 to 0.535. For both screening questions and ICD-10 Z codes, the measures were much more specific than sensitive. Under real world conditions, ICD-10 Z codes and screening questions are at the minimal, or below, threshold for being diagnostically useful approaches to identifying patients' social risk factors. Data collection support through information technology or novel approaches combining data sources may be necessary to improve the usefulness of these data.

Identifying patterns and predictors of lifestyle modification in electronic health record documentation using statistical and machine learning methods.

Just under half of the 85.7 million US adults with hypertension have uncontrolled blood pressure using a hypertension threshold of systolic pressure ≥ 140 or diastolic pressure ≥ 90. Uncontrolled hypertension increases risks of death, stroke, heart failure, and myocardial infarction. Guidelines on hypertension management include lifestyle modification such as diet and exercise. In order to improve hypertension control, it is important to identify predictors of lifestyle modification assessment or advice to tailor future interventions using these effective, low-risk interventions. Electronic health record data from 14,360 adult hypertension patients at an academic medical center were analyzed using statistical and machine learning methods to identify predictors and timing of lifestyle modification. Multiple variables were statistically significant in analysis of lifestyle modification documentation at multiple time points. Random Forest was the best machine learning method to classify lifestyle modification documentation at any time with Area Under the Receiver Operator Curve (AUROC) 0.831. Logistic regression was the best machine learning method for classifying lifestyle modification documentation at ≤3 months with an AUROC of 0.685. Analyzing narrative and coded data from electronic health records can improve understanding of timing of lifestyle modification and patient, clinic and provider characteristics that are correlated with or predictive of documentation of lifestyle modification for hypertension. This information can inform improvement efforts in hypertension care processes, treatment implementation, and ultimately hypertension control.

Survivor and Clinician Assessment of Survivorship Care Plans for Hematopoietic Stem Cell Transplantation Patients: An Engineering, Primary Care, and Oncology Collaborative for Survivorship Health.

The long-term care of hematopoietic stem cell transplantation (HSCT) survivors poses special challenges owing to a myriad of possible chronic and/or late complications. Survivorship care plans (SCPs) have been proposed as tools to communicate information on the late effects of treatment and recommended follow-up care to clinicians and survivors. The primary aims of this study were to determine SCP content and format, as well as to assess the preferred timing of SCP provision following HSCT. HSCT survivors and nontransplantation clinicians (oncologists and primary care physicians) were invited to participate in a survey evaluating the usefulness and utility of a sample HSCT-specific SCP with a treatment summary generated by autopopulation from an electronic health record (EHR). All participating HSCT survivors (n = 29) and clinicians (n = 18) indicated a desire to receive an SCP. More than 85% of the participants perceived information about treatments received, recommended follow-up and health maintenance including vaccinations, survivor and clinician resources, and graft-versus-host disease and other late/chronic side effects to be useful. The majority of survivors also believed that care team contact information was useful. In addition, >85% of survivors and clinicians agreed that the SCP increased their understanding of treatments and chronic/late side effects, improved health care provided, and were satisfied with the SCP and found it understandable and easy to use. The majority of survivors indicated that additional information should be added to the SCP, whereas some clinicians deemed the SCP too long. Survivors preferred to receive the SCP as a paper document at the end of a regular follow-up visit and review it with a cancer clinician, whereas clinicians preferred to receive the SCP through the EHR. These findings will help improve the design of future SCPs for use by HSCT survivors and clinicians. Future work will include leveraging the EHR to ease the burden of creating user-centered documents.

Training and Interpreting Machine Learning Algorithms to Evaluate Fall Risk After Emergency Department Visits.

BACKGROUND: Machine learning is increasingly used for risk stratification in health care. Achieving accurate predictive models do not improve outcomes if they cannot be translated into efficacious intervention. Here we examine the potential utility of automated risk stratification and referral intervention to screen older adults for fall risk after emergency department (ED) visits. OBJECTIVE: This study evaluated several machine learning methodologies for the creation of a risk stratification algorithm using electronic health record data and estimated the effects of a resultant intervention based on algorithm performance in test data. METHODS: Data available at the time of ED discharge were retrospectively collected and separated into training and test datasets. Algorithms were developed to predict the outcome of a return visit for fall within 6 months of an ED index visit. Models included random forests, AdaBoost, and regression-based methods. We evaluated models both by the area under the receiver operating characteristic (ROC) curve, also referred to as area under the curve (AUC), and by projected clinical impact, estimating number needed to treat (NNT) and referrals per week for a fall risk intervention. RESULTS: The random forest model achieved an AUC of 0.78, with slightly lower performance in regression-based models. Algorithms with similar performance, when evaluated by AUC, differed when placed into a clinical context with the defined task of estimated NNT in a real-world scenario. CONCLUSION: The ability to translate the results of our analysis to the potential tradeoff between referral numbers and NNT offers decisionmakers the ability to envision the effects of a proposed intervention before implementation.

Development and validation of a pragmatic natural language processing approach to identifying falls in older adults in the emergency department.

BACKGROUND: Falls among older adults are both a common reason for presentation to the emergency department, and a major source of morbidity and mortality. It is critical to identify fall patients quickly and reliably during, and immediately after, emergency department encounters in order to deliver appropriate care and referrals. Unfortunately, falls are difficult to identify without manual chart review, a time intensive process infeasible for many applications including surveillance and quality reporting. Here we describe a pragmatic NLP approach to automating fall identification. METHODS: In this single center retrospective review, 500 emergency department provider notes from older adult patients (age 65 and older) were randomly selected for analysis. A simple, rules-based NLP algorithm for fall identification was developed and evaluated on a development set of 1084 notes, then compared with identification by consensus of trained abstractors blinded to NLP results. RESULTS: The NLP pipeline demonstrated a recall (sensitivity) of 95.8%, specificity of 97.4%, precision of 92.0%, and F1 score of 0.939 for identifying fall events within emergency physician visit notes, as compared to gold standard manual abstraction by human coders. CONCLUSIONS: Our pragmatic NLP algorithm was able to identify falls in ED notes with excellent precision and recall, comparable to that of more labor-intensive manual abstraction. This finding offers promise not just for improving research methods, but as a potential for identifying patients for targeted interventions, quality measure development and epidemiologic surveillance.

Potential yield of imminent death kidney donation.

About 99 000 people are waiting for a kidney in the United States, and many will die waiting. The concept of "imminent death" donation, a type of living donation, has been gaining attention among physicians, patients, and ethicists. We estimated the number of potential imminent death kidney donors at the University of Wisconsin Hospital and Clinics by assessing the number of annual deaths in individuals with normal kidney function. Based on a previous survey suggesting that one-third of patients might be willing to donate at imminent death, we estimate that between 76 and 396 people in the state of Wisconsin would be medically eligible and willing to donate each year at the time of imminent death. We extrapolated these numbers to all transplant centers in the United States, estimating that between 5925 and 31 097 people might be eligible and willing to donate each year. Our results suggest that allowing donation at imminent death and including discussions about organ donation in end-of-life planning could substantially reduce the nation's kidney waiting list while providing many more donors the opportunity to give this gift.

Killer immunoglobulin-like receptor (KIR) and KIR-ligand genotype do not correlate with clinical outcome of renal cell carcinoma patients receiving high-dose IL2.

NK cells play a role in many cancer immunotherapies. NK cell activity is tightly regulated by killer immunoglobulin-like receptor (KIR) and KIR-ligand interactions. Inhibitory KIR-ligands have been identified as HLA molecules, while activating KIR-ligands are largely unknown. Individuals that have not inherited the corresponding KIR-ligand for at least one inhibitory KIR gene are termed the "KIR-ligand missing" genotype, and they are thought to have a subset of NK cells that express inhibitory KIRs for which the corresponding KIR-ligand is missing on autologous tissue, and thus will not be inhibited through KIR-ligand recognition. In some settings where an anticancer immunotherapeutic effect is likely mediated by NK cells, individuals with a KIR-ligand missing genotype have shown improved clinical outcome compared to individuals with an "all KIR-ligands present" genotype. In addition, patients receiving hematopoietic stem cell transplants for leukemia may do better if their donor has more activating KIR genes (i.e., KIR haplotype-B). In a recent multi-institution clinical trial of patients with metastatic renal cell carcinoma receiving high-dose IL2 (HD-IL2), 25 % of patients showed a complete or partial tumor response to this therapy. We genotyped KIR and KIR-ligand genes for these patients (n = 107) and tested whether KIR/KIR-ligand genotypes correlated with patient clinical outcomes. In these analyses, we did not find any significant association of KIR/KIR-ligand genotype (either KIR-ligand missing or the presence of KIR haplotype-B) with patient outcome in response to the HD-IL2 therapy.

The Association Between Use of Brain CT for Atraumatic Headache and 30-Day Emergency Department Revisitation.

OBJECTIVE: The purpose of this article is to describe the association between initial CT for atraumatic headache and repeat emergency department (ED) visitation within 30 days of ED discharge. MATERIALS AND METHODS: A retrospective observational study was performed at an academic urban ED with more than 85,000 annual visits. All adult patients with a chief complaint of headache from January through December 2010 who were discharged after ED evaluation were included in the analysis. Patients were excluded if they were transferred, died in the ED, or had a diagnosis indicating a traumatic mechanism. A propensity score-matched logistic regression model was used to determine whether the use of brain CT was associated with the primary outcome of ED revisitation within 30 days, controlling for potential confounding variables. RESULTS: Of 80,619 total patient visits to the ED during the study period, 922 ED discharges with a chief complaint of headache were included. A total of 139 (15.1%) patients revisited within 30 days. The return rate was 11.2% among patients who underwent CT at their initial visit and 21.1% among those who did not. In the adjusted analysis, controlling for age, race, sex, insurance status, triage vital signs, laboratory values, and triage pain level, the odds ratio for revisitation given CT performance was 0.49 (95% CI, 0.27-0.86). CONCLUSION: After adjustment for clinical factors, we found that patients who underwent a brain CT examination for atraumatic headache at an initial ED visit were less likely to return to the ED within 30 days. Future appropriate use quality metrics regarding ED imaging use may need to incorporate downstream health care use.

Documented lifestyle education among young adults with incident hypertension.

BACKGROUND: Only 38% of young adults with hypertension have controlled blood pressure. Lifestyle education is a critical initial step for hypertension control. Previous studies have not assessed the type and frequency of lifestyle education in young adults with incident hypertension. OBJECTIVE: The purpose of this study was to determine patient, provider, and visit predictors of documented lifestyle education among young adults with incident hypertension. DESIGN: We conducted a retrospective analysis of manually abstracted electronic health record data. PARTICIPANTS: A random selection of adults 18-39 years old (n = 500), managed by a large academic practice from 2008 to 2011 and who met JNC 7 clinical criteria for incident hypertension, participated in the study. MAIN MEASURES: The primary outcome was the presence of any documented lifestyle education during one year after meeting criteria for incident hypertension. Abstracted topics included documented patient education for exercise, tobacco cessation, alcohol use, stress management/stress reduction, Dietary Approaches to Stop Hypertension (DASH) diet, and weight loss. Clinic visits were categorized based upon a modified established taxonomy to characterize patients' patterns of outpatient service. We excluded patients with previous hypertension diagnoses, previous antihypertensive medications, or pregnancy. Logistic regression was used to identify predictors of documented education. KEY RESULTS: Overall, 55% (n = 275) of patients had documented lifestyle education within one year of incident hypertension. Exercise was the most frequent topic (64%). Young adult males had significantly decreased odds of receiving documented education. Patients with a previous diagnosis of hyperlipidemia or a family history of hypertension or coronary artery disease had increased odds of documented education. Among visit types, chronic disease visits predicted documented lifestyle education, but not acute or other/preventive visits. CONCLUSIONS: Among young adults with incident hypertension, only 55% had documented lifestyle education within one year. Knowledge of patient, provider, and visit predictors of education can help better target the development of interventions to improve young adult health education and hypertension control.

Varied manifestations of persistent hyperplastic primary vitreous with graded somatic mosaic deletion of a single gene.

PURPOSE: Persistent hyperplastic primary vitreous (PHPV) represents a developmental eye disease known to have diverse manifestations ranging from a trivial remnant of hyaloid vessels to a dense fibrovascular mass causing lens opacity and retinal detachment. PHPV can be modeled in mice lacking individual genes, but certain features of such models differ from the clinical realm. For example, mice lacking the Arf gene have uniformly severe disease with consistent autosomal recessive disease penetrance. We tested whether the graded somatic loss of Arf in a subset of cells in chimeric mice mimics the range of disease in a non-heritable manner. METHODS: Wild type ↔ Arf(-/-) mouse chimeras were generated by morulae fusion, and when the mice were 10 weeks old, fundoscopic, slit-lamp, and histological evaluations were performed. The relative fraction of cells of the Arf(-/-) lineage was assessed with visual, molecular genetic, and histological analysis. Objective quantification of various aspects of the phenotype was correlated with the genotype. RESULTS: Sixteen chimeras were generated and shown to have low, medium, and high contributions of Arf(-/-) cells to tail DNA, the cornea, and the retinal pigment epithelium (RPE), with excellent correlation between chimerism in the tail DNA and the RPE. Phenotypic differences (coat color and severity of eye disease) were evident, objectively quantified, and found to correlate with the contribution of Arf(-/-) cells to the RPE and tail-derived DNA, but not the cornea. CONCLUSIONS: Generating animals composed of different numbers of Arf(-/-) cells mimicked the range of disease severity observed in patients with PHPV. This establishes the potential for full manifestations of PHPV to be caused by somatic mutations of a single gene during development.

Relational machine learning for electronic health record-driven phenotyping.

OBJECTIVE: Electronic health records (EHR) offer medical and pharmacogenomics research unprecedented opportunities to identify and classify patients at risk. EHRs are collections of highly inter-dependent records that include biological, anatomical, physiological, and behavioral observations. They comprise a patient's clinical phenome, where each patient has thousands of date-stamped records distributed across many relational tables. Development of EHR computer-based phenotyping algorithms require time and medical insight from clinical experts, who most often can only review a small patient subset representative of the total EHR records, to identify phenotype features. In this research we evaluate whether relational machine learning (ML) using inductive logic programming (ILP) can contribute to addressing these issues as a viable approach for EHR-based phenotyping. METHODS: Two relational learning ILP approaches and three well-known WEKA (Waikato Environment for Knowledge Analysis) implementations of non-relational approaches (PART, J48, and JRIP) were used to develop models for nine phenotypes. International Classification of Diseases, Ninth Revision (ICD-9) coded EHR data were used to select training cohorts for the development of each phenotypic model. Accuracy, precision, recall, F-Measure, and Area Under the Receiver Operating Characteristic (AUROC) curve statistics were measured for each phenotypic model based on independent manually verified test cohorts. A two-sided binomial distribution test (sign test) compared the five ML approaches across phenotypes for statistical significance. RESULTS: We developed an approach to automatically label training examples using ICD-9 diagnosis codes for the ML approaches being evaluated. Nine phenotypic models for each ML approach were evaluated, resulting in better overall model performance in AUROC using ILP when compared to PART (p=0.039), J48 (p=0.003) and JRIP (p=0.003). DISCUSSION: ILP has the potential to improve phenotyping by independently delivering clinically expert interpretable rules for phenotype definitions, or intuitive phenotypes to assist experts. CONCLUSION: Relational learning using ILP offers a viable approach to EHR-driven phenotyping.