CTSI Annual Pilot Awards to Improve the Conduct of Research

An Open Proposal Opportunity

IT Improvements Needed to Make APEX useful for Clinical Research

Type: 
Proposal Status: 

Rationale: The APEX system has been promoted to the faculty as providing a new and extremely valuable means of doing clinical research, especially for discovering previously unappreciated clinical associations such as the relationship of kidney stone and myocardial infarction, etc..  Such associations would be very valuable in discovering underlying mechanisms of disease.  However, the APEX system falls very short of fulfilling this promise, because it has been primarily designed for maintaining information for medical practice, and for billing.  The problem is that it forces the physician to use vague descriptors of disease that are so imprecise as to be of no value whatever in discovery of associations.  For example in our Lipid Clinic we have discovered a form of dyslipidemia, cholesterol 7 alpha hydroxylase deficiency, that we believe afflicts about 100,000 Americans.  The APEX system refuses to allow us to enter this diagnosis for search purposes, insisting on “cholesterol problem” as a substitute.  Because there are probably fifty individual diagnoses that could fall under this heading, all value to translational research is lost.

Also there are data from emerging biomarkers, etc. that it will not accept for search processes. An example is the biomarker, prebeta-1 HDL, also discovered by our group, which is emerging as a robust indicator of risk of myocardial infarction. We have measured this biomarker on nearly four thousand patients.  We should be able to enter it  with  the restriction that it cannot be used for clinical decision making, but allowing faculty to access its value for clinical research.

An academician from the Cleveland Clinic, lecturing here recently, stated they also found APEX nearly useless for research, leading them to re-engineer the IT to overcome these issues, making detailed information searchable.  This was done carefully within the envelope of APEX without altering its function for clinical practice, billing, etc.

Plan: We would bring in expert IT personnel to make the changes in APEX, fitting it to our particular needs and system characteristics at UCSF.  We would support part-time effort of a clinical faculty member to supervise the entry of appropriate descriptors, following the experience at the Cleveland Clinic.

Criteria and Metric:

We can test the system for association of descriptors in widely diverse information fields.

Approximate cost:

Expert IT consultation and engineering $40K

Faculty time for descriptors: $ 15K

Collaborators:  Faculty from diverse disciplines would be welcomed to bring precision to the descriptors in their respective fields.

Comments

Great idea to leverage APEX for research. It would be helpful if you could articulate in more detail, how you plan to bring in faculty from diverse disciplines to participate in this effort. Another barrier in conducting clinic-based interventions is that APEX does not allow integration of external data (e.g. those generated from patient-facing technologies/apps) into the system to facilitate/improve health care. Could this be addressed as well?

I think this would be an opportunity for any UCSF contributor to participate. Surely individuals in diverse fields will have much more refined concepts of clinical syndromes and perhaps newly defined disorders. The more participation, the better this will be able to serve the translational investigation community. The problem with data mining approaches is that the ICD-9 codes and other restricted lists are just that. They do not keep up with the progress of discovery. There has to be a means of putting searchable descriptors that have scientific meaning into the system if it is going to be of use in translational research at an institution like ours.I think the collective wisdom of investigators in the diverse fields at UCSF can make this a valuable resource for discovering many relationships among disease states and their clinical and laboratory manifestations that have escaped discovery to date.

I could not agree more with the limitations of Apex in terms of coding of diseases and phenotypes. I share the same concern expressed by Kathleen, however, that it will take an enormous effort to modify the system in a way that solves this across all diseases. Are you sure that $45K would do this?

Sounds like a great idea, but I am on the same page as Dan. For such a large upgrade to the system, from a software perspective, it doesn't seem like $45K is enough money to get it done. In terms of coding versus mining text fields, there are no other controlled vocabularies to use in such an effort? While this too would be daunting, you get so much more from using an established ontology - from synonyms to related terms.

John and Mini, This project sounds like a precursor to making APEX amenable to more detailed diagnoses that will be available with ICD-10 coding. If this is the case, one potential method for improving data capture from APeX might be to instead focus on the hospital and physician billing data, which is much more uniform in format than unstructured notes. I am very interested in participating in the implementation of MyChart. This is the patient portal where it is envisioned physicians will be able to have patients fill in surveys prior to their visit, and together, review these data in the clinic. In the past, the distribution of surveys to obtain SF12, pain and disability scores were considered "research". The goal is to routinely distributed these surveys to patients. Integrate the results into their charts, and use these data to facilitate patient/physician discussion regarding the course of care

Great idea! You have articulated a couple of concerns with APEX and possible solutions. Are there any other concerns (in addition to the ones stated above) with using APEX for research and would re-engineering the system tackle all of the issues?

I heard a comment at the Dean's retreat from the keynote speaker that (as I heard it) said that getting research data from the form entry in EMR's was notoriously unreliable and advocated for mining the text fiels as a better alternative. As the VA EMR is mostly text, this was encouraging. The last thing I'd do is to add yet more terms on top of the thousands already in Apex!

Agree this proposal represents an important vision for how the Apex project should be keeping research efforts in mind while being implemented. While the priority of this issue is high, the scope of the project described is extremely large. Even if narrowed substantially, it will be hard to carry forward in the next 6-9 months given priority placed on implementing Apex in a clinically safe and efficient manner. Whether large or small in scale, this work will require integration with Apex analysts, the IDR working group, and would probably would require more than the small amount of funds requested.

Adapting EMRs to improve their utility for research is a really key area not only for UCSF but across other academic medical centers. Couldn't agree more, and I'm glad you're highlighting this. Since reading the submission, I've been learning more about status of APEX here at UCSF, and it seems like there's actually good momentum in integrating the research mission with the medctr rollout. In particular, APEX does apparently have a richer vocabulary for use than what may be viewable right now - its just going to take time to roll these pieces out carefully. Of course, I'm sure any progress will seem too slow for those involved in research, but I fear there may not be much room for quick fixes here. Here's a related proposal - how about we propose bringing in academics who've seen good examples of EMRs adapted for research (like the colleague you mention) - exposing ourselves to their progress may help accelerate our own interventions.

One advantage of electronic medical records is "easy" access to data for research. I think it’s important to make that access a priority and a reality at UCSF. Departments across campus will have similar needs to this one. Is it possible that IT modules from EPIC already exist that can help capture data for research purposes? What can we learn from other research institutions that have adopted EPIC?

The comments on this post reflect the intense interest in access to APEX that we have seen on the mHealth side as well. The need to leverage APEX for clinical research purposes outlined by this proposal is well thought out and very worthy of consideration. It also highlights a bigger issue. How does UCSF best leverage the tremendous investment UCSF Med Ctr has made and the tremendous wealth of information and information flow that APEX provides? This is the start of a continuous flow of ideas all competing for finite resources on how to best use APEX. At Apple the yearly planning cycle would include dozens upon dozens of proposals for each major system. All of them were very good ideas. The proposal of the good idea itself was not enough for the proposal to be approved. It also needed to include the value and the return on investment (yes the dreaded ROI) of each proposal. I think all APEX proposals will need to provide the epxected value derived for UCSF. They need not all be monetary - though lets face it in this climate those should get looked at first.

This is an incredibly important project; Apex was implemented as a means to optimize patient care and billing, not for research. In an academic institution such as ours, Apex needs to be able to keep up to date with the discoveries. This is a problem that will be easier to tackle at these early stages, as Apex is being implemented. From here on, we can move forward and build this tool for the future. This proposal addresses this fundamental need at this early stage. For sure the academic community will welcome this initiative - it will help our institution provide a valuable tool to the research community. Apex has the potential to become a valuable source of information, that is able to provide accurate diagnosis for our patients as new discoveries are made. It will not only improve research, but patient care as a whole. We should be able to learn from other insitutions, such as the Cleveland Clinic and implement our own valuable, up-to-date database. This project can become large, but early intevention in the process, even at a small scale, is a priority. From then on, we can scale it up to fit the needs of the research community and patient care of the whole institution.

It's clear that, in an institution like UCSF, there are many clinical insights that could provide the basis for important new discoveries in the interrelationships of clinical observation and laboratory findings. However, there must be a means of getting these academic observations of emerging syndromes and sophisticated new biomarkers into the clinical database in a searchable format. Clearly, the Apex system was designed for efficient clinic operation and billing and, in its present state, is totally unserviceable with respect to supporting translational discovery. The experience at the Cleveland Clinic indicates that a sophisticated, searchable database can be created within the envelope of Apex, leaving its utility to medical practice and billing intact. We should draw upon their experience and that of other academic centers but develop a system adapted to our own clinical practices. This is an opportunity for all UCSF academicians to contribute their individual insights to this endeavor. An example of the kind of discovery this could engender is the recent discovery of a relationship between certain aspects of HDL metabolism and ovarian cancer. There's no way to make a cost/benefit analysis of this proposal as one would in a clinical practice. How would one evaluate the monetary value of a new insight in the etiology of cancer? Surely, in the long run, this would bring in research grant support, etc. but the value to medicine and humanity is paramount. This is an endeavor in keeping with the caliber of clinical investigation at UCSF.

This is a very interesting proposal as the need is shared amongst anyone at UCSF doing clinical research. APEX or EPIC was designed to serve a particular set of use cases, clinical EMR related for instance. As such instead of trying to make that system research friendly, it may be worth considering getting the data from APEX and merging it with other sources within a system designed for research use cases. An example would be what is being done here: https://i2b2.cchmc.org/faq The research data-warehouse approach does not address the immediate need to support better annotation of diagnostic information within APEX, a need specified in this proposal. However it would be a more scalable solution for generalized research data storage and mining. As a disclaimer, I am not an expert on either APEX or i2b2 (it is installed at UCSF, per).

I read with interest the comments about how APeX was implemented to make billing easier, though it is not clear how to promote access to billing and APeX/UCare data. Though many colleagues remain skeptical that diagnosis and procedure coding is sufficiently accurate for informing patient-centered research, most UCSF researchers lack sufficient data access to assess this notion. There are 3 sources of claims that should be considered: Hospital (including purchasing data with itemized bills for every consumable), campus (many departments submit bills for professional services - CPT codes), and the admit-transfer-discharge (ADT) used to distinguish and communicate "present on admission" and emergent conditions. I believe cohort discovery is using ADT data. These data are in standard formats. Barrier to gain access remains because of the perceived distinction between the medical center "quality of care" and campus "research" focus. It will be difficult to build clinical research platforms by pulling data from APeX. But not many researchers have considered the existing administrative claims data streams that are ready to be mined.

APEX as now constituted is very unfriendly to the entrance of the precise diagnosis necessary for support of clinical research. The proposed effort promises to remedy this deficiency. We need a module that with concordance will accept new descriptors.

This is a proposal that has my full support. I have been using the limited abilities of STOR for twenty years in my research and am in the process of transitioning to APEX. Now is the time for us to make sure that we have a much better system established for the future at UCSF. I think this proposal is very timely, and what is suggested a necessary undertaking. I believe that the comments about mining billing data and the current clinical descriptors in APEX misses the point of this proposal. The problem is that the informative descriptors i.e. newly emerging syndromes, biomarkers etc do not exist in APEX and cannot be inserted in a searchable format. It will be possible to insert an internal system for recoding these valuable data in a searchable format within APEX based on the Cleveland Clinic experience.

Is there an open source APeX community? How can we be sure that in the future, we and our colleagues do not have to license from EPIC, technologies we develop to make APeX data available for research?

Commenting is closed.