Application of clinical text data for phenome-wide association studies (PheWASs)

Hebbring, Scott J.; Rastegar-Mojarad, Majid; Ye, Zhan; Mayer, John; Jacobson, Crystal; Lin, Simon

doi:10.1093/bioinformatics/btv076

Cited by 34 publications

(31 citation statements)

References 37 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In the first demonstration of the value of EHR text data, Hebbring et al (27) identified 23,384 one- to four-word phrases occurring in clinical narratives that matched to medical concepts in the UMLS. In this approach, which they called a TextWAS, they replicated known associations for five SNPs with similar performance to using ICD9 codes for these diseases.…”

Section: Future Directionsmentioning

confidence: 99%

Phenome-Wide Association Studies as a Tool to Advance Precision Medicine

Denny

Bastarache

Roden

2016

Annu. Rev. Genom. Hum. Genet.

207

205

View full text Add to dashboard Cite

Beginning in the early 2000s, the accumulation of biospecimens linked to electronic health records (EHRs) made possible genome-phenome studies (i.e., comparative analyses of genetic variants and phenotypes) using only data collected as a by-product of typical health care. In addition to disease and trait genetics, EHRs proved a valuable resource for analyzing pharmacogenetic traits and developing reverse genetics approaches such as phenome-wide association studies (PheWASs). PheWASs are designed to survey which of many phenotypes may be associated with a given genetic variant. PheWAS methods have been validated through replication of hundreds of known genotype-phenotype associations, and their use has differentiated between true pleiotropy and clinical comorbidity, added context to genetic discoveries, and helped define disease subtypes, and may also help repurpose medications. PheWAS methods have also proven to be useful with research-collected data. Future efforts that integrate broad, robust collection of phenotype data (e.g., EHR data) with purpose-collected research data in combination with a greater understanding of EHR data will create a rich resource for increasingly more efficient and detailed genome-phenome analysis to usher in new discoveries in precision medicine.

show abstract

Section: Future Directionsmentioning

confidence: 99%

Phenome-Wide Association Studies as a Tool to Advance Precision Medicine

Denny

Bastarache

Roden

2016

Annu. Rev. Genom. Hum. Genet.

207

205

View full text Add to dashboard Cite

show abstract

“…Using automated phenotyping algorithms, investigators have identified cases and controls for diseases of interest to replicate known phenotype-genotype associations and make novel discoveries, [12][13][14][15][16][17] potentially with decreased cost 18 and faster execution than traditional trials.…”

Section: Background and Significancementioning

confidence: 99%

Evaluating electronic health record data sources and algorithmic approaches to identify hypertensive individuals

Teixeira

Wei

Cronin

et al. 2016

Journal of the American Medical Informatics Association

View full text Add to dashboard Cite

Objective: Phenotyping algorithms applied to electronic health record (EHR) data enable investigators to identify large cohorts for clinical and genomic research. Algorithm development is often iterative, depends on fallible investigator intuition, and is time-and labor-intensive. We developed and evaluated 4 types of phenotyping algorithms and categories of EHR information to identify hypertensive individuals and controls and provide a portable module for implementation at other sites. Materials and Methods: We reviewed the EHRs of 631 individuals followed at Vanderbilt for hypertension status. We developed features and phenotyping algorithms of increasing complexity. Input categories included International Classification of Diseases, Ninth Revision (ICD9) codes, medications, vital signs, narrative-text search results, and Unified Medical Language System (UMLS) concepts extracted using natural language processing (NLP). We developed a module and tested portability by replicating 10 of the best-performing algorithms at the Marshfield Clinic. Results: Random forests using billing codes, medications, vitals, and concepts had the best performance with a median area under the receiver operator characteristic curve (AUC) of 0.976. Normalized sums of all 4 categories also performed well (0.959 AUC). The best non-NLP algorithm combined normalized ICD9 codes, medications, and blood pressure readings with a median AUC of 0.948. Blood pressure cutoffs or ICD9 code counts alone had AUCs of 0.854 and 0.908, respectively. Marshfield Clinic results were similar. Conclusion: This work shows that billing codes or blood pressure readings alone yield good hypertension classification performance. However, even simple combinations of input categories improve performance. The most complex algorithms classified hypertension with excellent recall and precision.

show abstract

“…Based on ICD9 codes to define cases and controls [29–31], no phenotype passed a conservative Bonferroni threshold (p<7.2E-6, assuming α < 0.05 and 6910 tests/phenotypes). The top associations included ICD9 616.3 defining abscess of Bartholin’s gland (p= 0.00020, OR=2.0[1.4–2.9]) followed by ICD9 379.92 defining swelling or mass of eye (p= 0.00021, OR=1.7[1.3–2.3]) (Figure 2, Supplementary Table 2); the relevance of these association is uncertain.…”

Section: Resultsmentioning

confidence: 99%

“…The phenome was defined by ICD9 coding extracted from patient EHR data using standard methods as described previously [29–31]. Individuals whose medical records contained ICD9 codes inclusive of three levels of resolution defined by ICD9 code suffix (for example, ICD9 720, 720.8, 720.89) were designated as a case for a particular condition, whereas individuals with no record of the broadest code (e.g., 720) were classified as controls.…”

Section: Methodsmentioning

confidence: 99%

“…Special attention was given toward two codes defining breast cancer (ICD9 174, malignant neoplasm of breast; and ICD9 233.0, carcinoma in situ of breast) in females only and 115 codes that define adverse drug events (E930–E949, E850–E858, except E850.1, E854.1) [35]. As an alternative to PheWAS, a TextWAS was conducted with methodologies reported previously [29]. Briefly, all clinical notes were broken down into four possible combinations of word strings which included unigrams (one word), bigrams (two adjacent words), trigrams (three adjacent words), and quad-grams (four adjacent words).…”

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

Relationship of SULT1A1 copy number variation with estrogen metabolism and human health

Liu

Zhao

et al. 2017

The Journal of Steroid Biochemistry and Molecular Biology

Self Cite

View full text Add to dashboard Cite

Human cytosolic sulfotransferase 1A1 (SULT1A1) is considered to be one of the most important SULT isoforms for metabolism, detoxification, and carcinogenesis. This theory is driven by observations that SULT1A1 is widely expressed in multiple tissues and acts on a wide range of phenolic substrates. SULT1A1 is subject to functional common copy number variation (CNV) including deletions or duplications. However, it is less clear how SULT1A1 CNV impacts health and disease. To better understand the biological role of SULT1A1 in human health, we genotyped CNV in 14,275 Marshfield Clinic patients linked to an extensive electronic health record. Since SULT1A1 is linked to steroid metabolism, select serum steroid hormones were measured in 100 individuals with a wide spectrum of SULT1A1 CNV genotypes. Furthermore, comprehensive phenome-wide association studies (PheWAS) were conducted using diagnostic codes and clinical text data. For the first time, individuals homozygous null for SULT1A1 were identified in a human population. Thirty-six percent of the population carried >2 copies of SULT1A1 whereas 4% had ≤ 1 copy. Results indicate SULT1A1 CNV was negatively correlated with estrone-sulfate to estrone ratio predominantly in males (E1S/E1; p=0.03, r=−0.21) and may be associated with increased risk for common allergies. The effect of SULT1A1 CNV on circulating estrogen metabolites was opposite to the predicted CNV-metabolite trend based on enzymatic function. This finding, and the potential association with common allergies reported herein, warrants future studies.

show abstract

Application of clinical text data for phenome-wide association studies (PheWASs)

Cited by 34 publications

References 37 publications

Phenome-Wide Association Studies as a Tool to Advance Precision Medicine

Phenome-Wide Association Studies as a Tool to Advance Precision Medicine

Evaluating electronic health record data sources and algorithmic approaches to identify hypertensive individuals

Relationship of SULT1A1 copy number variation with estrogen metabolism and human health

Contact Info

Product

Resources

About