The 1-minute prescription drug commercial is unlike any other commercial on TV: a battleground between happy and uplifting words and scenes (a woman shopping at a farmer’s market, a man golfing with his buddies,etc.) against words like “uncontrollable bleeding”, “fatal” and “may cause death”.
This dichotomy in advertising has put prescription drug commercials in a class of their own – well known to anyone who watches TV. To take a step beyond my own anecdotal perception of prescription commercials being loaded with side-effects, I decided to look at the transcript text of these commercials and see what a sentiment analysis would say.
So, how did I get the data? Assuming commercials with high sales would have readily accessible commercials, I looked up the top selling drugs by sales in quarter 4 of 2013. Looking for the transcripts for these prescription drug commercials, I used YouTube, searching for the most recent commercial of each drug. When I found a commercial I would select “… More” and then “Transcript”. Once the transcript was copied it needed cleaning, which required listening to the commercials and changing words as needed.
Once I had a clean transcript for each, I classified each drug into a group by disease treated and created a .csv file.
As a gauge for how positive or how negative these commercials are, I did a sentiment analysis using the open-source statistical software R. The algorithm I used scores words based on how negative or positive they are. The commercial transcripts then receive an average polarity score.
The upper graph shows the duration, or how many words per commercial. The color of each block indicates its polarity: the redder the block the more positive, the bluer the more negative.
The lower graph focuses on polarity of the commercials, grouped by illness or disease treated. Not surprisingly, multiple sclerosis, depression and neuropathic pain were the most negative in sentiment.
This last graph looks specifically at the drugs themselves, where we see that only 4 out of my top selling 35 drugs were overall positive in their messaging. That’s an 11% positive rate. Put differently, this means that 89% of the time when you see a prescription drug commercial you will feel a negative sentiment instead of a positive one. The effect of seeing a happy gardening couple or the convertible drive in the countryside haven’t been factored in in this case.
Most of the commercial transcripts scored below 0.0 on the polarity scale, indicating that, like our general sentiment about drug commercials, they are really advertisements of scary side effects.
We don’t always need a tidy data set to get a sense of trends, especially when it comes to people’s experiences with watching prescription drug commercials. Drug commercials have a stigma of barraging viewers with side-effects that often sound worse than the illness itself. In this case, data analysis has the ability to uncover a new layer of insight on these commercials, confirming visually what we originally thought.
List of top selling drugs I used: http://www.drugs.com/stats/top100/sales
QDAP polarity function found here: http://trinker.github.io/qdap/vignettes/qdap_vignette.html#polarity