Developing an expert panel process to refine health outcome definitions in observational data

Genotypes and phenotypes of motor neuron disease: an update of the genetic landscape in Scotland

BACKGROUND

Using the Clinical Audit Research and Evaluation of Motor Neuron Disease (CARE-MND) database and the Scottish Regenerative Neurology Tissue Bank, we aimed to outline the genetic epidemiology and phenotypes of an incident cohort of people with MND (pwMND) to gain a realistic impression of the genetic landscape and genotype-phenotype associations.

METHODS

Phenotypic markers were identified from the CARE-MND platform. Sequence analysis of 48 genes was undertaken. Variants were classified using a structured evidence-based approach. Samples were also tested for C9orf72 hexanucleotide expansions using repeat-prime PCR methodology.

RESULTS

339 pwMND donated a DNA sample: 44 (13.0%) fulfilled criteria for having a pathogenic variant/repeat expansion, 53.5% of those with a family history of MND and 9.3% of those without. The majority (30 (8.8%)) had a pathogenic C9orf72 repeat expansion, including two with intermediate expansions. Having a C9orf72 expansion was associated with a significantly lower Edinburgh Cognitive and Behavioural ALS Screen ALS-Specific score (p = 0.0005). The known pathogenic SOD1 variant p.(Ile114Thr), frequently observed in the Scottish population, was detected in 9 (2.7%) of total cases but in 17.9% of familial cases. Rare variants were detected in FUS and NEK1. One individual carried both a C9orf72 expansion and SOD1 variant.

CONCLUSIONS

Our results provide an accurate summary of MND demographics and genetic epidemiology. We recommend early genetic testing of people with cognitive impairment to ensure that C9orf72 carriers are given the best opportunity for informed treatment planning. Scotland is enriched for the SOD1 p.(Ile114Thr) variant and this has significant implications with regards to future genetically-targeted treatments.

MetaLAB-HOI: Template standardization of health outcomes enable massive and accurate detection of adverse drug reactions from electronic health records

PURPOSE

This study aimed to advance the MetaLAB algorithm and verify its performance with multicenter data to effectively detect major adverse drug reactions (ADRs), including drug-induced liver injury.

METHODS

Based on MetaLAB, we created an optimal scenario for detecting ADRs by considering demographic and clinical records. MetaLAB-HOI was developed to identify ADR signals using common model-based multicenter electronic health record (EHR) data from the clinical health outcomes of interest (HOI) template and design for drug-exposed and nonexposed groups. In this study, we calculated the odds ratio of 101 drugs for HOI in Konyang University Hospital, Seoul National University Hospital, Chungbuk National University Hospital, and Seoul National University Bundang Hospital.

RESULTS

The overlapping drugs in four medical centers are amlodipine, aspirin, bisoprolol, carvedilol, clopidogrel, clozapine, digoxin, diltiazem, methotrexate, and rosuvastatin. We developed MetaLAB-HOI, an algorithm that can detect ADRs more efficiently using EHR. We compared the detection results of four medical centers, with drug-induced liver injuries as representative ADRs.

CONCLUSIONS

MetaLAB-HOI's strength lies in fully utilizing the patient's clinical information, such as prescription, procedure, and laboratory results, to detect ADR signals. Considering changes in the patient's condition over time, we created an algorithm based on a scenario that accounted for each drug exposure and onset period supervised by specialists for HOI. We determined that when a template capable of detecting ADR based on clinical evidence is developed and manualized, it can be applied in medical centers for new drugs with insufficient data.

Translational Biomedical Informatics and Pharmacometrics Approaches in the Drug Interactions Research

Drug interaction is a leading cause of adverse drug events and a major obstacle for current clinical practice. Pharmacovigilance data mining, pharmacokinetic modeling, and text mining are computation and informatic tools on integrating drug interaction knowledge and generating drug interaction hypothesis. We provide a comprehensive overview of these translational biomedical informatics methodologies with related databases. We hope this review illustrates the complementary nature of these informatic approaches and facilitates the translational drug interaction research.

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.

How well do various health outcome definitions identify appropriate cases in observational studies?

BACKGROUND

Observational data can be useful for drug safety research, but accurate measurement of adverse health outcomes is paramount. Best practices for identifying important health outcomes of interest (HOI) are needed.

OBJECTIVES

To evaluate the extent to which health outcome definitions commonly used in observational database studies identify cases that are consistent with expert panel assessment of the underlying data.

METHODS

Competing HOI definitions were used to identify potential cases of acute liver injury (ALI; n = 208), acute kidney injury (AKI; n = 200), and myocardial infarction (MI; n = 204) in the Truven MarketScan Lab Database (MSLR). Panelists reviewed patient-level data and answered questions about whether they believed the case actually reflected the HOI and their certainty of case classification on a 10-point scale (1 = unlikely to 10 = likely). Each patient was reviewed independently by two panelists. Case disagreements were resolved through consensus meetings. Positive predictive value (PPV) was calculated as the number of cases deemed to be true over the total number of sampled cases. Kappa statistics assessed inter-rater agreement.

RESULTS

PPV ranged from 0 to 52 % across ALI definitions, 12 to 82 % across AKI definitions, and 1 to 56 % across MI definitions. Certainty scores on the 10-point scale paralleled the PPV, with a range of mean values from 1.7 to 4.8 across ALI definitions, 3.1 to 6.0 across AKI definitions, and 2.8 to 5.7 across MI definitions. Inter-rater agreement was low to moderate (Kappa range 0.0-0.6).

IMPLICATIONS/CONCLUSIONS

Existing HOI definitions had relatively low PPV based on expert panel review. Experts commonly disagreed on case classification. Additional work is needed to refine HOI case definitions if observational data are to be reliably used for health outcome assessment.

Clinical research informatics and electronic health record data

OBJECTIVES

The goal of this survey is to discuss the impact of the growing availability of electronic health record (EHR) data on the evolving field of Clinical Research Informatics (CRI), which is the union of biomedical research and informatics.

RESULTS

Major challenges for the use of EHR-derived data for research include the lack of standard methods for ensuring that data quality, completeness, and provenance are sufficient to assess the appropriateness of its use for research. Areas that need continued emphasis include methods for integrating data from heterogeneous sources, guidelines (including explicit phenotype definitions) for using these data in both pragmatic clinical trials and observational investigations, strong data governance to better understand and control quality of enterprise data, and promotion of national standards for representing and using clinical data.

CONCLUSIONS

The use of EHR data has become a priority in CRI. Awareness of underlying clinical data collection processes will be essential in order to leverage these data for clinical research and patient care, and will require multi-disciplinary teams representing clinical research, informatics, and healthcare operations. Considerations for the use of EHR data provide a starting point for practical applications and a CRI research agenda, which will be facilitated by CRI's key role in the infrastructure of a learning healthcare system.

Application of a 5-tiered scheme for standardized classification of 2,360 unique mismatch repair gene variants in the InSiGHT locus-specific database

The clinical classification of hereditary sequence variants identified in disease-related genes directly affects clinical management of patients and their relatives. The International Society for Gastrointestinal Hereditary Tumours (InSiGHT) undertook a collaborative effort to develop, test and apply a standardized classification scheme to constitutional variants in the Lynch syndrome-associated genes MLH1, MSH2, MSH6 and PMS2. Unpublished data submission was encouraged to assist in variant classification and was recognized through microattribution. The scheme was refined by multidisciplinary expert committee review of the clinical and functional data available for variants, applied to 2,360 sequence alterations, and disseminated online. Assessment using validated criteria altered classifications for 66% of 12,006 database entries. Clinical recommendations based on transparent evaluation are now possible for 1,370 variants that were not obviously protein truncating from nomenclature. This large-scale endeavor will facilitate the consistent management of families suspected to have Lynch syndrome and demonstrates the value of multidisciplinary collaboration in the curation and classification of variants in public locus-specific databases.