text4baby Evalulation Info!

I'm excited to share some new information about evaluation of text4baby. A small (n=160) survey-based study was done in San Diego with some pretty neat results. There's a great pdf showing it all but it was emailed to me in the text4baby newsletter, instead of posted online, so I can't link to it, boo.

I think my favorite stat is that "63.1% reported that text4baby helped them remember an appointment of immunization that they or their child needed." Missed appointments are a big deal, and this is a pretty direct text->action connection.

I also liked "38.5% ... of participants reported that they called a service or phone number that they received from a  text4baby message. These messages range from perinatal and infant care services such as breastfeeding and post-partum support lines, teratogen information, low-cost health services to product safety, poison control, and infant immunization information lines." This kind of integration is a strength of mHealth. You're looking at a phone number on your phone so you just go ahead and dial. Also neat is that this percentage jumped to 58.3% for those participants without health insurance.

The study also looked at cultural and linguistic appropriateness for the Spanish version of text4baby. Oddly, a third of recipients reported incorrect Spanish in the messages. Were they not checked by native speakers? Fortunately, a large majority said that the messages were understandable anyway.

More general information in a whitehouse.gov post here, and if you're interested in reading details about other upcoming larger scale evaluations, let me know and I can forward the email! (Some of it is available at the links here.)

Recent text4baby thoughts

I subscribe to the weekly text4baby Tuesday newsletter - sign up at the link here if you're interested - and have long been vaguely annoyed at the paucity of data they publish. For a while, (I've been receiving the newsletter over a year, since about 2 months after the text4baby launch) they showed only number of subscribers per state in a bar chart (not weighted by population). It didn't used to have a scale, either, so it was like, hm, about 5 mm of people in Vermont and wow! about a decimeter of people in California. (Woops, I just checked and they also gave % English/% Spanish, and % pregnant/% first year (there are texts through the baby's first birthday) and % would recommend to a friend (consistently 93-97%).

Anyway, today's issue included a better graph, labeled, weighted and all, so I thought I'd share. Sorry for the craptastic graphic quality:

from text4baby Tuesday newsletter May 10, 2011

I am really eager to see some evaluation from text4baby (spoken like a true public health nerd). I know Professor Doug Evans at George Washington University is working on something formal. Information is hard to find.

Regarding the extremely ambitious text4baby goal of one million subscribers by the end of 2012. Here is my homegrown crappy data (from the same newsletters since March 2010) about their growth rate:

text4baby subscribers. the cute little february 2011 bump is this

At this rate, by my rather haphazard calculations (approx 369 new subscribers a day, beautifully linear, and taking nothing extra into account), at the end of 2012*, they'll be about 608,145 people short.

But then, I feel like a million subscribers is a dubious goal in the first place without any (any!) evaluation data.

Someone else has similar things to say.

I would be critical for some of the same and some new reasons now, a year after I wrote this paper and having taken many more public health courses.


*601 days from today!

Populations and Individuals

One of the hallmarks of public health is that unlike medicine, you're thinking at the level of a population, not an individual.  This is one of the first things I learned about public health, but I feel like I still don't have a sophisticated understanding of how to think about the relationship between the two.


There are two aspects of this split that I'm thinking about: 1. how to intervene, how to avoid just telling collections of individuals the same thing a doctor would tell an individual (e.g. "eat less salt"); and 2. how to understand population level outcome measurements, how to make those statistics more meaningful than "you have an x% chance (based on data about other people like you and what happened to them) of getting this disease."

The NYT ran an article this week about salt reduction.  The inspiration for the article is a NEJM article which I have not read yet on the same topic, specifically with reference to the reduction of cardiovascular disease.
This sentence from the NYT article caught my attention:

"If everyone consumed half a teaspoon less salt per day, there would be between 54,000 and 99,000 fewer heart attacks each year and between 44,000 and 92,000 fewer deaths..."

This made me laugh because it reminded me of some sort of motivational group effort, like, if only we each did our part, now Get It Together, People!.  But half a teaspoon less salt per day is a LOT less salt.  Depending on gender, that would be a 30-40% reduction for the average American adult.  Small wonder that you'd see such big results.  I can totally believe that for a given individual, that reduction in salt could have dramatic effects on his or her health.


Now suppose it claimed, instead (not suggesting this is equivalent or true! thought experiment in progress): "If everyone consumed one pinch less salt per day, there would be between 540 and 990 fewer heart attacks each year and between 440 and 920 fewer deaths."


In that scenario, it seems less plausible that the good benefits are because any given individual who was eating x amount of salt is now eating (x - 1 pinch) of salt.  That is, it seems harder to interpret the sentence as being strictly true.  Put another way, if we had a laboratory of humans who were all fed exactly x grams of salt per day, and then we cut their salt intake to (x - 1 pinch) per day, would you see the results above?


Or: does the sentence really mean the following: some number of people (perhaps those who were eating way too much salt) suddenly dropped their salt intake dramatically (and the rest of us didn't make any changes), so that the average came out to each person eating 1 pinch less per day.  Then, those people who dramatically reduced their intake made up the group who didn't get heart attacks or die after all.


There are probably things I should already know that address this question.  For example, more than the averages are important, and standard deviations will address the question of variance in individual levels.  Furthermore, the basic split between longitudinal and cross-sectional studies addresses these concerns (hey, I learned something in my speech-language pathology class!).  I know there are aspects of cross-sectional study methodologies that try to overcome the effect of all the averages (average of day-to-day salt intake per individual, average of individuals' salt intake to give population level estimates).  Maybe after I read the actual study I'll have a better idea of how it works, and whether the NYT quote above is a misrepresentation or not.


Final notes on the NYT article.  Kudos for:


1. acknowledging that action beyond the level of the individual might be appropriate
2. quoting someone on the idea that when we legislate these kinds of things, we must be very wary of unintended consequences
3. mentioning some subpopulations that would particularly benefit (would have gotten extra points for mentioning health disparities more broadly)
4. quoting the study author on the population-individual issue


Boos for:


1. being wishy-washy about where the bulk of the sodium intake is from. As far as I know, the research clearly points to processed and restaurant foods
2. having a kind of pointless quotation about how instead we should be working on anti-smoking efforts because they'd make a bigger difference.  Who said we should put all our energies toward one single problem? This just seemed irrelevant, like the reporter simply wanted one more dissenting opinion