Validating a proposed technical desiderata– call for particpation

As part of my PhD work, I’m conducting an IRB-approved anonymous survey entitled “Validating a proposed technical desiderata for the integration of genomic information with clinical decision support among domain experts”. We developed a set of requirements that must be met for a clinical decision support (CDS) architecture to support whole genome sequence information at the point of care. Our goal with this survey is to validate these proposed requirements among CDS and genome experts. If you are a domain expert in either of these fields, I invite you to take this survey located at:

https://redcap01.brisc.utah.edu/ccts/redcap/surveys/?s=7Rp3As 

The text below is the background text included in the survey.

 

BACKGROUND
The application of whole genome sequencing (WGS) in routine clinical care is quickly approaching. Having a patient’s genomic information could allow clinician’s to identify health risks, improve diagnosis, and tailor treatment, all under the paradigm of personalized medicine. However, were the WGS to be widely available in the clinic today, many of these capabilities could remain unachieved for a number of reasons, including the complexity of genetic analysis, practicing clinicians’ limited proficiency in genetics, and the insufficient number of genetics professionals in the workforce (1).

Nevertheless, computerized clinical decision support (CDS) within the electronic health record (EHR) offers a reasonable solution to overcome these barriers and help clinicians leverage a patient’s WGS information at the point of care (2). CDS entails providing clinicians, patients, and other healthcare stakeholders with pertinent knowledge and person-specific information, intelligently filtered or presented at appropriate times, to enhance health and healthcare (3). Examples of point-of-care CDS include medication dosing support, order facilitators, alerts and reminders, relevant information display, expert systems, and workflow support (4).

Masys et al. developed a set of requirements to incorporate genomic variation into the EHR for the provision of healthcare services (5). The requirements include the following:

1. Maintain separation of primary molecular observations from the clinical interpretations of those data
2. Support lossless data compression from primary molecular observations to clinically manageable subsets
3. Maintain linkage of molecular observations to the laboratory methods used to generate them
4. Support compact representation of clinically actionable subsets for optimal performance
5. Simultaneously support human-viewable formats and machine-readable formats in order to facilitate implementation of decision support rules
6. Anticipate fundamental changes in the understanding of human molecular variation
7. Support both individual clinical care and discovery science

This set of requirements provides a strong set of guiding principles to integrate genomic information into the EHR. However, in our effort in developing CDS capabilities for WGS data, we have identified ADDITIONAL REQUIREMENTS specific to CDS to augment the original Masys desiderata. These additional requirements are as follows (further described in the survey):

8. Keep CDS knowledge separate from variant classification
9. Support a large number of gene variants while keeping CDS logic as simple as possible
10. CDS knowledge must be able to incorporate multiple genes and clinical information
11. Access and transmit only the genomic information necessary for CDS
12. Leverage current and developing CDS and genomics infrastructure and standards
13. Support a CDS knowledge base deployed at and developed by multiple independent organizations
14. Have the capacity to support multiple EHR platforms with various data representations with minimal modification

OBJECTIVE
The intent of this survey is to VALIDATE THIS ADDITIONAL SET OF REQUIREMENTS AMONG DOMAIN EXPERTS in CDS and/or genomics. On subsequent pages of this survey, each of these requirements are described in further detail with a Likert scale question to assess the importance of that particular requirement. If desired, there is an optional comment field after each requirement to provide suggestions for improvement and/or reasoning for your response on the importance of the requirement. Finally, general comments can be provided regarding any overarching recommendations or ideas for additional requirements. Survey responses will be reviewed and incorporated into the requirements by the following panel of experts:

Brandon M Welch, MS
Karen Eilbeck, MS, PhD
Laurence J. Meyer, MD, PhD
Guilherme Del Fiol, MD, PhD
Kensaku Kawamoto, MD, PhD

We request that you keep the content confidential until survey results have been published. Thank you for your participation.

DEFINITIONS OF SELECT TERMS USED IN THE SURVEY
Molecular observations: The patient-specific information obtained from sequencing technologies which include the identification and location of variants in DNA, RNA, and protein sequence. Example representation in Human Genome Variation Society (HGVS) format: MSH2 c.1452_1455delAATG

Variant interpretations: Disease causing genes can have one of many known variants present. However, not all variants are pathogenic. Some variants are benign, and others have unknown pathogenicity (also known as variants of unknown significance [VUS]). The American College of Medical Genetics and other groups have put forth various recommendations for variant classification, including pathogenic, likely pathogenic, likely benign, benign, or variant of unknown significance (18). An example variant classification for the MSH2 variant in the above example is Pathogenic.

CDS knowledge: CDS (described above) provides guidance to clinician on how to care for a patient when a particular variant is present or absent. Continuing the previous example of a patient with a pathogenic mutation in the MSH2 gene, a CDS alert for this person could say the following: “This patient is at increased risk for colorectal cancer (Lynch syndrome I). The patient should be placed on an NSAID and receive a colonoscopy annually. This patient should be referred for genetic counseling and to an oncology specialist.”

REFERENCES
1- http://www.ncbi.nlm.nih.gov/pubmed/22918138
2- http://www.ncbi.nlm.nih.gov/pubmed/22922173
3- http://www.ncbi.nlm.nih.gov/pubmed/17213487
4- http://www.ncbi.nlm.nih.gov/pubmed/21415065
5- http://www.ncbi.nlm.nih.gov/pubmed/22223081

3073 disease-causing genes in ClinVar

ClinVar is a new public service provided by the NCBI, to catalogue all genetic mutations and associated phenotypes. I’m looking into using ClinVar to provide such information in support of the CDS architecture for genome information that I’m working on. They make this information freely available to the public through their online interface and have started providing this information through their FTP site. One such document, gene_condition_source_id I found to be really interesting and informative. The tab deliminated file contains HGVS gene symbols and associated disease names. As some genes can cause several diseases and diseases can be caused by several genes, this file captures all these associations.

I was curious about a few things. For example, within ClinVar how many genes are known to cause disease? Likewise, how many diseases are caused by genetic mutations? So I uploaded the file into a MySQL database I created and did some queries to find out. The original file contains 8719 rows of information, however most of the entries are duplicates because the information comes from different sources (other fields in the file).

To find out the number of distinct gene symbols in the ClinVar knowledge base, I performed the following SQL query:.

USE gene_cond_dictionary;
SELECT DISTINCT GeneSymbol FROM gene_cond LIMIT 0,8721;

The result returned a result of 3,077 of which I removed 4 rows because of row headers and space leaving a total of 3,073 genes known to cause disease as provided by ClinVar. Very interesting stuff. I presume this number will increase as more participants contribute to ClnVar and research uncovers more gene associations are made. Still, >3,000 known genes is a substantial number of disease causing genes. Here is an excel document of genes known to cause diseases from ClinVar.

Now my next question, how many diseases are associated with gene mutations? On the same database I ran the following SQL query:

USE gene_cond_dictionary;
SELECT DISTINCT DiseaseName FROM gene_cond LIMIT 0,8721;

As before I received 3,886 results which turns out to be 3,882 diseases known to be associated with genes as provided by ClinVar. I was surprised that there are so many diseases caused by genes but there are more diseases than genes (attributed to the fact that a single gene can produce multiple phenotypes). It is known that multiple genes can cause a single phenotype and presumed that there would be more genes than phenotypes. Here is an excel document of diseases known to be caused by genes.

 

Clinical decision support trends in personalized medicine

As previously mentioned, I participated in a systematic review on CDS for personalized medicine. I’d like to highlight some of the trends from that paper.

Figure 1- Increase in CDS for genetically-guided personalized medicine manuscripts. Distribution of CDS using genotype and/or family history over the years.

In our systematic review we found relatively few (38) manuscripts describing the development and/or evaluation of CDS for genetically-guided personalized medicine. (The term ‘genetically-guided’ was requested by one of the reviewers because we were only looking at CDS that uses family history and or genetics.) The number of manuscripts on CDS for personalized medicine remained fairly constant but started to increase in the mid-2000s and are currently on an upward trend (Figure 1). This upward trend coincides with the use of genetic information to guide CDS. This makes since as family health history was traditionally used and the primary source of data traditionally, however as genetic tests have become more widely available, the need to also support clinicians in utilizing genetic information is imperative. Notice that family health history is still a prominent attribute used for CDS. I don’t expect this to change as family health history is still a good, and among the best, predictors of disease risk.

CDS for GPM integrated with primary health information system or as a stand-alone application.

Also of interest is the recent trend to integrate CDS for these use cases within the primary
information systems. (Figure 2) This strongly correlates with findings by Kawamoto et al regarding CDS effectiveness by directly tied to the provision of care during clinical workflow at the point and time of clinical decision making. This is important as the success of CDS for personalized medicine will be dependent upon its integration within primary health information systems.

Another important finding is that there were no manuscripts describing CDS for whole genome sequencing (WGS). This is important as WGS is beginning to make its way into health care, the use of CDS to support WGS at the point of care will become a necessity. It’s also not surprising given that, until recently, WGS has not been used very extensively in clinical care. Nevertheless, as this changes, CDS will be vital to the success of whole genome sequencing. This is an area of particular interest for me.

Reference:
Welch, BM, Kawamoto, K. Clinical decision support for genetically guided personalized medicine: a systematic review. J Am Med Inform Assoc 2013;20:388–400

 

Introduction to Biomedical Informatics for Personalized Health Care

 

Last night, I had the opportunity to lecture at the Foundations of Personalized Health Care course offered at the University of Utah, of which I also TA. The next two weeks are on informatics so I provided an introductory overview of biomedical informatics (@2:00) and the different fields therein. I then discussed why health informatics will be necessary to achieve the promises of genetics and personalized health care (@15:30). I invite you to view the lecture entitled ‘Introduction to Biomedical Informatics for Personalized Health Care‘. I lecture from minute 2:00-29:30.

Following my lecture, my PhD adviser Kensaku Kawamoto, MD, PhD lectures on the role CDS plays in personalized health care. His lecture entitled ‘Clinical Decision Support for Genetically Guided Personalized Health Care‘ summarizes work he and we did in this area. His lecture starts at minute 30:00 and goes through 1:01:30. I also encourage you to listen to his lecture as well.

 

Clinical decision support for genetically guided personalized medicine: a systematic review

My PhD adviser, Kensaku Kawamoto,MD,PhD, and I recently had a systematic review of ours published in JAMIA. Our manuscript entitled “Clinical decision support for genetically guided personalized medicine: a systematic review” received the Editor’s choice designation and was selected for presentation at an AMIA journal club in February.

In summary we only found 38 manuscripts related to the design and/or evaluation of CDS for personalized health care. Most focused on family history, however in recent years there has been an increase in the number of papers with larger focus on CDS using genetic information.

I will share more in the future.

Welch, BM, Kawamoto, K. Clinical decision support for genetically guided personalized medicine: a systematic review. J Am Med Inform Assoc 2013;20:388–400

The World of Informatics

The World of Informatics

Click image to view and interact.

Informatics is a very large field with many different applications, as such I often need to clarify I fit into the overall world of informatics. I’ve created this diagram to help people understand the different fields within informatics. Enjoy!

FYI- I mostly fit into ‘health informatics’ and ‘consumer health informatics’

Hello, World

I’ve created this blog as a way to share interesting thoughts, work, and information from myself and others related to the use of health IT and informatics to realize personalized health care and genetics in health care. There are so mach interesting knowledge and information out there that I feel the need to organize it for my sake, as well as others.

I’m currently a PhD student in Biomedical Informatics at the University of Utah. I have a master’s degree in Human Genetics from Tulane University. I recommend both programs for anyone interested in going into this field. You can learn more of my background and experiences on my LinkedIn profile. I also have a Twitter account which I will fire up at conferences and other important times to broadcast my thoughts.

Thanks.

bmw