screen-shot-2016-11-15-at-6-59-11-am

How much do you know about what words do in your organization?  …or about what they’re doing to your organization? …or about how you can harness them to work better for your organization?

  • Are they typed hastily into an electronic health record to capture an emergency department patient’s chief complaint?
  • Are they texted and tweeted as real-time disaster event updates or unsolicited product reviews?
  • Are they recorded on countless hours of support calls?
  • Or are the most important words deliberately absent, subtly leaving their marks in, say, the order that information is presented in a report, or the cues that team members give as they vie for leadership roles in a collaborative effort?

As a starting point, here I offer some insights on uncovering language and communication patterns through the use of keywords.

The Goldilocks effect

One of the most common ways that organizations think to use language is to look at the frequency that a word or sentiment appears in a given context, whether they’re delivered to your customer service department or to your hospital emergency department (I’ll give you examples from each).  While this sounds straightforward enough, there’s a trick to it:  your list needs to be sensitive enough capture as many cases as possible that define a given category, and also specific enough that it excludes the non-relevant cases.  Not too small and not too big.

Playing telephone

On one of my call center studies I was making my way around the floor, talking to agents about how they used the drop-down lists to code their calls.  These are the lists that inform management that 9% of January’s calls were about equipment issues, say, while 36% were about service.

One agent volunteered the following:

There are too many codes to choose from.  I just pick one that’s close.

Ten minutes later, someone else at the other end of the room told said,

There aren’t enough codes to choose from.  I just pick one that’s close.

Same list, same job, same calls.

That was enough for me to put ‘redesign call codes’ at the top of my to-do list for that center.  Since the codes had no value to the agents (coding calls was a chore that added unwanted time to the call handling length they were evaluated on), a lot of agents tended to stick to a half-dozen they considered “good enough.”  As a result, a number of call issues–and the creative solutions that agents developed to solve them–were flying under the radar because they weren’t adequately listed in call logs.  This in turn meant that management was unaware of these issues, unable to make informed decisions about how to handle them, and unable to recognize and harness the ingenuity of their top agents.

Chief complaints

Another area where keywords play a critical role is in public health syndromic surveillance.  In a nutshell, syndromic surveillance involves daily searches of hospital emergency department (ED) records to see if there are any blips in the number of flu cases or drug overdoses or other conditions across a region that might require the mobilization of a public health response.

It’s a fabulous idea, except that the information available in a 24-hour window isn’t a reliable ICD code or standardized diagnosis, but what’s called a “chief complaint.”  This is generally a free-text, often messy, description of what appears to be wrong with the patient when they walk through the door.  This can include everything from

HEROINE OD

(make sure you pick up that typo), to

HIT ON HEAD WITH FROZEN TURKEY

(no, I didn’t make that up, but more importantly, how are you going to classify it?).

Developing keyword lists and trolling through tens of thousands of ED records every day is as much an art as a science, even with an automated system.  Outlining the process is a bit more than I can cover here, so I’ll direct you to a recent methods paper of mine if you’d like to find out more.  Although it was published in a disaster preparedness journal, the core technique can be applied to other fields and used in conjunction with more sophisticated tools, such as natural language processing.  (I love NLP!  But not every organization has the resources or time to invest in new technology infrastructure.)  And for those of you in public health and syndromic surveillance, I’m happy to share more information on this particular project.

You can access the paper here:

Window of Opportunity for New Disease Surveillance: Developing Keyword Lists for Monitoring Mental Health and Injury Through Syndromic Surveillance

Whether your focus is call centers or technology adoption or leadership training or disaster preparedness, you can always contact me if you have questions or would like additional guidance on sorting through the words that are steering your organization.