Iteration 2 Architecture

This wiki page describes the architecture for Iteration 2 of the SCOOP project.

Context

The architecture context is basically the same as the previous iteration; however, we are looking at how to enrich the medication data in order to answer this iteration's question..

/iteration1/images/architecture/IT1-Context.png

This iteration is being deployed to a laboratory environment (non production) at UVic.

There is only one practice being simulated in this iteration (i.e. one endpoint).

Logical View

/iteration1/images/architecture/IT1-Systems.png

There are a number of architectural changes planned for this iteration of SCOOP.

1. We are continuing to enhance and refactor OSCAR to export the patient summary document, either for a single individual or by batch for all individuals with a record in the EMR

2. The middleware (Mirth) will play a much smaller role in this design. It will simply transfer individual summary documents between the EMR and the hQuery Gateway. It will not perform any aggregation or transformation.

3. hQuery Gateway will be enhanced to support the enrichment of patient summary data. Specifically, we will be adding the ability to classify medications by ATC code. This will be serving as a model for use of future mappings / classification / terminology services.

Constraints

Privacy Individual information never leaves the practice. Early feedback from discussions with the faculty physicians UBC highlighted their stewardship of patient information as a serious matter. There’s a genuine concern that anonymous data collected may be used for unintended research purposes in the future. Several are concerned by the risks of data aggregation, even when the data is de-identified.

Open Source As SCOOP development is sponsored with public funds it will be released as open source for any one to use. In addition, SCOOP will only rely on free, open source components so that its use is accessible to any individual/group.

OSCAR In addition to being open source, OSCAR is the EMR used by many UBC clinical faculty and physicians. A few of these faculty members will the initial SCOOP users. For this reason, OSCAR will be the first supported EMR.

Design View

Software design of the new export feature in OSCAR

We will attempt to stay true to the MVC design. Velocity based templating requires both a template and a data model. We have opted to make the Patient object draw on all the pre-existing DAOs and models and have Patient group up the relevant fields for easy velocity access. Most of the heavy code work will be found in E2EVelocityExport.java and DemographicExportAction4.java acts as the Event Handler and file management for the export function.

Software design for new hQuery enhancements

As that Iteration 1 already added E2E document support, adding support for ATC codes into the E2E specification only required verifying that the E2E specification supported it, updating the export template, and updating hQuery's importer to handle incoming ATC codes. The data structure within hQuery's database already supports multiple drug codes so no modification is required there.

Query Design

Since hQuery follows the map-reduce paradigm for gathering and returning results, we will also be mapping and reducing with design.

Map

The goal here is to emit every drug's class. To do that, it will check each patient's record for recorded medications, and emit the drug's class in ATC format. Below is the Javascript query for the map phase used in hQuery.

function map(patient) {
  var atcLevel = 4; // Level definition based on definition found on Wikipedia

  var drugClasses = findATCDrugClasses(patient.medications(), atcLevel);

  for(var i = 0; i < drugClasses.length; i++) {
    emit(drugClasses[i], 1);
  }
}

// Returns list of all ATC medication codes
function findATCDrugClasses(drugs, level) {
  var list = [];
  var cutoff;

  // Define ATC cutoff levels
  switch(level) {
    case 1:
      cutoff = 1;
      break;
    case 2:
      cutoff = 3;
      break;
    case 3:
      cutoff = 4;
      break;
    case 4:
      cutoff = 5;
      break;
/* Prevent data leakage - only want classes, not specific substance
    case 5:
      cutoff = 7;
      break;
*/
    default:
      return list;
  }

  for(var i = 0; i < drugs.length; i++) {
    // Get all represented codes for each drug
    var codes = drugs[i].medicationInformation().codedProduct();

    // Filter out only ATC codes
    for(var j = 0; j < codes.length; j++) {
      if(codes[j].codeSystemName() == "whoATC") {
        // Truncate to appropriate level length
        list.push(codes[j].code().substring(0, cutoff));
      }
    }
  }
  
  return list;
}

Reduce

The goal here is to gather all the emitted data from the map phase and sum them together in each respective category to yield a useful answer to the researcher. This is done with a straight-forward summation function. Below is the Javascript query for the reduce phase used in hQuery.

function reduce(key, values) {
  var result = 0;

  while (values.hasNext()) {
    result += values.next();
  }

  return result;
}

Data View

Introduction to E2E E2E, or EMR to EMR, is a CDA specification being developed by BC's PITO organization. Its main purpose is to create a standard in which EMR data can be transmitted to other Canadian EMRs in a standardized way.

Overview of E2E sections (including image)

Scope of content for Iteration 2 In order to sufficiently answer our question, we would require demographic information such as name and age of birth, as well as all current medication data. The medication data required would be when it was prescribed, how long its duration is, and what drug it is as well as its class. The only information being directly used in the query would be the drug's class information. All other data fields are not used explicitly to answer the query but are there to support previous iterations.

Technology Selection

OSCAR An Open Source EMR system used by many private physicians. This is way that individual information is captured - and the user interface for the physician practioner end user.

Mirth Mirth is an open source interfacing system. It acts as generic middleware between many healthcare software systems and connects them via HL7 messages (although several formats / protocols are supported).

Mirth Rest Adaptor This is an open source piece of middleware developeed by the SCOOP project. Mirth does not fully support restful web services. Specifically it does not allow naming of attachments to web service calls. For this reason, hQuery gateway cannot find the patient summary document when Mirth calls it directly. This component can be thought of as a restful web service relay that renames a single attachment (patient summary document) before forwarding the call.

hQuery Gateway The query gateway is a web based application that provides the back end for executing queries. The query gateway which exposes a query API, accepts queries, runs those queries against the patient data, and returns the results of the query back to the query composer.

hQuery Composer The query composer is a web based application that provides the front end for creating, managing, and executing queries. Those queries are executed against the query gateway which exposes a query API, accepts queries, runs those queries against the patient data, and returns the results of the query back to the query composer.

Scoop Endpoint refers to all the non-emr SCOOP software components that will reside at a practice in the future. This indludes hquery gateway, Mirth, and Mirth Rest adapter.

SCOOP Hub This is the application that the researcher uses to ask a "question" of the Research Network. This includes features for query management, policy enforcement, privacy management, and security. Currently, hQuery composer is the only software system in the hub - but more design and enhancements are needed.

• Libraries: template