Define Variables

Define Variables

To define the variables on the DISCOVERY request form, consider the following:

  1. Define variable content
  • Include a basic working definition of your variable and units of measure;
  • Avoid acronyms;
  • Specify the range of values for that particular variable if possible;
  • Refer to the Top 50 Variables (see Section 2) and RedX to explore possible values;
  • Be aware that the variable may be stored in more than one location and may have more than one name; for example, “smoking” may appear as “smoking” or “tobacco use” in different parts of an individual’s health record;
  • Work with a clinician to help define the components in the patient record that meet your research question needs*;

* This will provide valuable insight into how these values are documented as part of a clinician’s workflow and what is being measured. This is especially useful for variables with more complex definitions (e.g. type II diabetes defined by past diagnosis, current insulin use, or HbA1c level).

  1. Define variable location/context
  • Include details on where the variable is stored in the database or patient record when possible
  • Refer to the Top 50 Variables List, WebCIS, RedX, or a collaborating clinician.
  • Work closely with your assigned programmer if location cannot be defined

Tip: If feasible, work with a clinician to find patient records that meet the criteria defined for the variable of interest. Be sure to compare what you obtained through DISCOVERY with the list of patient records generated from working with a clinician.

Example:
For the pancreatic surgery study, glucose lab values were needed. First, glucose values were defined by the type of lab result (e.g. blood glucose), and the appropriate unit (e.g. mg/dL).

Next, the location of the variable within the patient record is defined by referring to the Top 50 Variables list to determine which tables in the database contain the lab value of interest. If unable to specifically define where the lab values are stored, however, you may submit a description of where to expect the value to be stored in the patient record and work closely with the assigned programmer to determine its specific location.

In the case study insulin administered during the inpatient hospital stay is also a variable of interest. By exploring the data using RedX with “Insulin” as a keyword, you will find multiple options, listed below. After receiving the list, working closely with the clinician will help you determine the best variables to include.

  • Aspart Insulin Sliding Scale JC TEST
  • Insulin Aspart Inj
  • Insulin Aspart Inj (Novolog)
  • Insulin Aspart Prandial Scale
  • Insulin Aspart Protamine 70/30
  • Insulin Aspart Protamine 70/30 (Novolog)
  • Insulin Aspart Protamine70/Aspart30
  • Insulin Aspart. Sliding Scale (Novolog)
  • Insulin Aspart. Sliding Scale Inj
  • Insulin DRIP
  • Insulin Glargine Inj (Lantus)
  • Insulin Human NPH 70%/ Regular 30% +NF+
  • Insulin Human Ultra Lente Inj
  • Insulin Lente Human Inj
  • Insulin Lispro (Humalog) +NF+
  • Insulin Lispro SlidingScale(Humalog)+NF+
  • Insulin NPH Human Inj (Humulin N)
  • Insulin NPH Inj (HumuLIN N)
  • Insulin Pork Lente Inj +NF+
  • Insulin Pork NPH Inj +NF+ Insulin Pork Regular Inj +NF+
  • Insulin Reg Human Inj (Humulin R)
  • Insulin Reg Inj (HumuLIN R)
  • Insulin Reg. Slide Scale Inj
  • Insulin Reg. Slide Scale Inj (HumuLin R)

  • Find Columbia Doctors