Data Mining Patient Experience

Data Mining Patient Experience


The CDC has been gathering and publishing patient surveys since 1973. In recent years, the CDC National Ambulatory Medical Care Survey (NAMCS) has become a valuable resource for understanding what medications and therapies physicians are prescribing in US emergency rooms and clinics. The CDC publishes the NAMCS dataset with 603 fields of data densely packed as numeric codes. It translates into reasons for patient visits, patient demographics, hospital services, physician services, and then services, therapies, treatments, and medications prescribed. The CDC published a public version of the study about 3-years after collection year.

Starting in 2010, the NAMCS added capabilities to understand the prescription of medications. The NAMCS records up to 10 medication prescriptions per patient record. Three features for each medication: a code for the medications brand name, a generic coding for the type of medication, and a three-level hierarchical therapeutic categorization of the medication. The challenge is that the raw data (see below) is difficult to use.

0260524-072-7-7020423401000000001-800000444-9055521050-0009-000902021010011499-02-9 -7-9 -7114990-00009-0000900-7-71000000000000010-91145-0009.0010984111307520000000000000000000000000-70000000-70000000-70000000000000000000000000-9 -9 -9 -9 -9 -9 -9 -9 -9 -009-009-009-009-009-009-009-009-0090031108585-9 -9 -9 -9 -9 -9 -9 -9 -9 01-7-7-7-7-7-7-7-7-7010100010000000300010002-09-90002-09-90002-09-900-999-99902-09-90002-9.0-90002-09-900001000001011228012153120203020200000000000101-8020102010101010101010101010101010101010201-8-8-8-80402-7-7010102010101-8-8-8-801010402010101010101020102-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7-7010103010103010101030101020101010401010101-8-8-801010201020100000000010000000020149951d00626161138 133 136 138 1291310211122820131003725.6427000000.00001

Foxtrot Code recently published a new suite of algorithms that provides a foundation for utilizing the 2013 CDC NAMCS Survey. The algorithms unpack the 603 fields, and then translates the numeric codes in the fields associated with brand name medications, generic coding, therapies into medication names (brand and generic), and the three-level hierarchy details. We constructed these algorithms with lookup tables from the CDC documentation. The algorithms used Foxtrot Code lookup functions to enhance the data with brand names and generic drug codes using the Cerner Multnum standard.

The therapeutic categorization is valuable, and requires some extra work to reduce the data to make it useful. The CDC data is generally more statistically relevant when there are more than 30-records in a result. The therapeutic categorization makes it easier to roundup 30-results by reducing data based on a category of medications. The Foxtrot Code system creates algorithms that search the data. Being able to search based on words is a lot simpler and straight forward than searching on codes.

With this capability in place you can quickly deliver the top 10 medications categorized at the first-level of the hierarchy as: “central nervous system agents”,

Screen Shot 2017-02-22 at 6.29.57 PM

and then further reduce down the search to a third-level of the hierarchy such as:  “central nervous system agents > anxiolytics, sedatives, and hypnotics > benzodiazepines”

Screen Shot 2017-02-22 at 6.35.24 PM

Included in the foundational algorithms are a suite of medication analytics that sort the Top 10 medications by Age Group, Patient Visit Information, and Therapeutic Category. Keep in mind that there are up to 10 medications recorded per customer. These analyses use only the first medication recorded DRGNM1, so there are 9 more DRGNM1-DRGNM10 that you can build a search around for a fuller analysis.

We created an example of this more complete analysis under the “By Therapeutic ALL Entry” within the “Top Meds” folder where all these analyses are stored. “By Therapeutic ALL Entry” searches for “psychotherapeutic agents” in the 4 therapeutic categories for each of the 10 generic drugs. This yields a significantly larger number of results than searching on only the DRGNM1 field as in the other analyses.

Screen Shot 2017-02-25 at 5.26.57 PM

To understand the breadth of the possibilities for an analysis take a quick look through the documentation on the dataset here.

The foundational algorithms in the 2013 CDC NAMCS Survey marketplace listing are available for FREE on Foxtrot Code for use in non-commercial as well as commercial activities. If you don’t have a Foxtrot Code subscription, you can START FREE,


Leave a Reply


Log in with your credentials


Forgot your details?

Create Account