Saturday, April 19, 2014

Interoperabbable update

Report: Lack of Interoperability Limits Meaningful Use Program
April 17, 2014,

Meaningful use stages 1 and 2 fall short of implementing the interoperability among electronic health records that is necessary to facilitate information exchange and develop a robust health data infrastructure, according to a new report from a task force assembled by the MITRE Corporation, Health Data Management reports...

HHS released the report, which was developed by JASON, an independent task force that advises the federal government on issues pertaining to science and technology (DeSalvo, "Health IT Buzz," 4/16). The report was funded by the Agency for Healthcare Research and Quality.

In the report, the task force concluded that the criteria for meaningful use stages 1 and 2 "fall short of achieving meaningful use in any practical sense," adding that "large-scale interoperability amounts to little more than replacing fax machines with the electronic delivery of page-formatted medical records."

According to the task force, "most patients still cannot gain electronic access to their health information," and "rational access to EHRs for clinical care and biomedical research does not exist outside the boundaries of individual organizations."...
Link to the contractor's report here: "A Robust Health Data Infrastructure" (pdf)
1.4 Facing the Major Challenges
A meaningful exchange of information, electronic or otherwise, can take place between two parties only when the data are expressed in a mutually comprehensible format and include the information that both parties deem important. While these requirements are obvious, they have been major obstacles to the practical exchange of health care information.

With respect to data formats, the current lack of interoperability among the data resources for EHRs is a major impediment to the effective exchange of health information. These interoperability issues need to be solved going forward, or else the entire health data infrastructure will be crippled. One route to an interoperable solution is via the adoption of a common mark-up language for storing electronic health records, and this is already being undertaken by the HHS Office of the National Coordinator for Health IT (ONC) and other groups. However, simply moving to a common mark-up language will not suffice. It is equally necessary that there be published application program interfaces (APIs) that allow third-party programmers (and hence, users) to bridge from existing systems to a future software ecosystem that will be built on top of the stored data...
Open the pdf, hit Ctrl-F (PC) or Command-F (for us Mac snobs), search the document for keywords/phrases "dictionary," "data dictionary," "schema," or "RDBMS."

Negative. Zip. Zilch. Nada. Nein. Nyet.

1.5 A New Software Architecture
The various implementations of data formats, protocols, interfaces, and other elements of a HIT system should conform to an agreed-upon specification. Nonetheless, the software architecture that supports these systems must be robust in the face of reasonable deviations from the specification. The term “architecture” is used in this report to refer to the collective components of a software system that interact in specified ways and across specified interfaces to ensure specified functionality. This is not to be confused with the term “enterprise architecture,” referring to the way a particular enterprise’s business processes are organized. In this report, “architecture” is always used in the former sense...
...There would be opportunities to operate within the new software architecture even as it is starting to be implemented. The APIs provide portals to legacy HIT systems at four different levels within the architecture: medical records data, search and index functionality, semantic harmonization, and user interface applications...
"Semantic harmonization"? Lordy. Recall "Rigorability"?

They finish up:
7  Concluding Remarks
This report has expressed disappointment in current US progress towards the creation of a robust health data infrastructure, while praising ONC and HHS for their persistence in trying to tackle one of the most vexing problems of today’s society. JASON believes that the two overarching goals, improved health care and lower health care costs, can be achieved by moving to EHRs and the comprehensive electronic exchange of health information. JASON has provided a path toward realizing the promise of a robust health data infrastructure through the development of a unifying HIT software architecture that adheres to the following core principles, all embodying a focus on the patient:

  • Be agnostic as to the type, scale, platform, and storage location of the data
  • Use public APIs and open standards, interfaces, and protocols
  • Encrypt data at rest and in transit
  • Separate key management from data management
  • Include with the data the corresponding metadata, context, and provenance information
  • Represent the data as atomic data with associated metadata
  • Follow the robustness principle: be liberal in what you accept and conservative in what you send
  • Provide a migration pathway from legacy EHR systems.
Yeah, "interoperability is good." All that're needed are yet more irrelevancies ("encrypt data at rest and in transit"), broad, vague cliches ("robustness principle..." "atomic data...") and blinding glimpses of the obvious.

In fairness, the report contains much of substantive concern, e.g., noting that there is "a growing trend towards capturing large quantities of data associated with particular aspects of patient phenotype, analyzing those data, and reporting relevant information back to the patient. These come under the general heading of “omics” technologies, a designation derived from genomics, the first of such data types..."
  1. Genome sequence. The haploid human genome contains 3 x 109 base pairs of DNA. Humans are diploid, so each person has two copies of their genome, one maternal and one paternal. These two copies differ by approximately 0.1%, so it is necessary to sequence the DNA sufficiently deeply to capture all of the genetic variation of an individual in comparison to the reference human genome sequence. The current standard for individual genomes is to sequence to approximately 30-fold coverage, or approximately 1011 bases of sequence data. In the case of cancer, for which it is important to know the genotype of the tumor in comparison to that of normal tissue, a similar level of sequencing might be applied to a tumor sample, and this could include a sample of both the primary tumor and its metastases. Although these data can be compressed by denoting only the difference with respect to the reference human genome sequence, there is clearly a rapidly growing need to incorporate vast amounts of genome sequence information into individual EHRs.
  2. Transcriptome. The transcriptome is a quantitative description of the types and amounts of messenger RNA molecules transcribed from the genomic DNA. Most cells in the body have the same genome sequence, but differential expression of that genome allows cells to become differentiated. Differential expression also defines disease states; for example, breast cancers can be divided into subtypes based on gene expression patterns. The transcriptome can be assessed by microarray analysis or, increasingly, by “RNAseq,” in which DNA copies of the messenger RNAs are sequenced with high coverage. The amount of information generated in a transcriptomics experiment is typically similar to that of a genome sequence, although because every cell type is different and there are many possible variables of cell state, there is the potential for much larger datasets.
  3. Epigenome. The epigenome is a description of the modification states of the genomic DNA and the RNA and proteins that are physically associated with the DNA in the form of chromatin. These modifications are part of the basis for the differential expression of the genome that is manifested in the transcriptome. The epigenome is assessed by a variety of methods that allow for spatial resolution of particular modifications in the genome (e.g., “ChIP-Seq” for measuring modifications of histone proteins, bisulfite sequencing for determining sites of methylation along the DNA, and DNase I hypersensitivity analysis for assessing chromatin structure). Efforts are currently underway to establish reference epigenome information for all genes in all tissue types.
  4. Proteome. The proteome is a description of the types and amounts of proteins expressed from the genome; it is the protein analog of the transcriptome. The proteome is determined in part by the transcriptome from which it is derived, but also by the many subsequent processes that affect proteins, including their translation, transport, post-translational modification, and degradation. The proteome is usually assessed by mass spectrometry. Both the sensitivity of detection and the methods for determining the amount of each protein detected by mass spectrometry are improving rapidly.
  5. Microbiome. The human body contains approximately 10 times more microbial cells than human cells by cell number (although only about 1% by mass). The microbiome is the complete description of this microbial population, including commensal and symbiotic organisms as well as pathogens. The microbiome of an individual is a unique signature, changing with time and environment, and likely responsible for some elements of phenotype. Because many of the microorganisms living in and on humans cannot be cultured, the microbiome is usually assessed by deep sequencing of the genomic DNA of microbiome organisms. There is growing evidence that several pathogenic conditions are due to aberrant states of the microbiome, some of which can be corrected by altering or replacing an individual’s microbiome.
  6. Immunome. The immunome is a description of the state of the immune system of an individual, focusing on the diversity of immune responses based on past exposures. In a narrower sense, such information has long been a part of health records. For example, the Mantoux (or PPD) skin test, and its predecessor the tuberculin tine test, assess the immune response to Mycobacterium tuberculosis antigens as a measure of previous exposure to this pathogen. High-throughput methods now allow testing for reactivity to thousands of antigens at once, in combination with deep sequencing to characterize the genome rearrangements that occur in each immune cell and define its reactivity.
Indeed. But, if we permit haphazard foundational dictionary specifications of "omics" data, things are only going to get much worse.

I refer you to another of my posts:

To coin a technical term,


More to come...

No comments:

Post a Comment