User:Solstag/Spreading health information and behaviour through social media seeded by a national survey in Brazil

Here we propose using the coming National Crack Survey in Brazil to promote civic engagement in health information and behavior through the Internet, cellphones and other computer networks, while performing a controlled experiment to gather information on this process and improve the scientific knowledge on social contagion processes.

The formal project, which needs to be written in Portuguese for the funding agency, is located at pt.wikiversity.

New thoughts
Because of different appeal to subjects, although statistically uniform from the point of view of a mobilization campaign, the sample isn't statistically uniform from the point of view of social experiments. For those, it is going to be a convenience sample, although likely much more diverse than usual convenience samples.

In terms of mobilization, it might also help to focus on the household as our unit, so if respondents live with their relatives, we ask them to deliver the invitation to the person in their household most interested in the family's health.

Each city is its own experiment, and coordination with local health institutions is a requirement.

In each city, we can split our sample into 2 to 3 groups, under different experimental conditions.

Participants are invited to participate in an online social platform of "healthkeepers", within which they'll be in communication with health professionals.

They'll be invited by these professionals to participate in activities to improve their own health, their family's, or their neighborhood's.

Depending on the experimental condition, they may or may not be allowed to interact with other "keepers".

Online, they'll have acess to additional tools to keep tab of their family's health and map health issues in their neighborhood.

The idea is to monitor differentials in the process of participant engagement in three scenarios:
 * participants only communicate with health professionals
 * participants communicate with health professionals and receive information about other participants
 * participants communicate with both health professionals and other participants

Also, in the third condition, we will be able to monitor the process of social interaction and network formation in regards to participants' characteristics and engagement in activities.

The secondary objective of this study is to build a template for further studies, since this process can be repeated with every single sample by the government. Health professionals would therefore be provided with a refreshed sample of randomly sourced, even if not uniformly so, citizens whom they can interact with about the general situation of health care in their cities.

How do we preserve a statistically controlled sample?
If we randomly distribute seeds among different programs and treat the chance of them to not join as part of the process, sample seems random. We would however be restricted to conclusions about performance of programs, not so many conclusions about the general population.

Conclusions about average human behavior would require high likelihood of adoption and therefore a very good motivation for engagement with low barriers for important factors such as age and gender.

How do we engage people?
Values? Family care, improving the health system.

Incentive mechanism? Status and achievements, recruitment pyramids.

Social interaction? Matching mechanisms, collective power.

This is a municipal election year in Brasil, that may play a role.

How do we track influence?
Having the whole thing play out through SMS would be very good, phone companies could even provide anonymous aggregate data on geolocation and whatnot, but they are not known to be cooperative or even competent - I've heard there's about 5% of SMS loss overall.

If we're going to model diadic interpersonal influence in the process, there needs to be some kind of interface through which people interact. If they interact among themselves, either we provide a predefined network, or we have some kind of link addition process, which may or may not involve recruitment.

There are other kinds of influence we might want to use instead, or in addition, such as aggregate influence as in the music experiment.

What is the variable?
Is it in the network, in the individuals, or in their relationship?

One variable could be the program itself. People could be assigned to different programs whose effectiveness would then be compared. As noted before, this avoids some issues of statistical control, but only because it makes conclusions about the population harder, focusing on the programs instead.

Another variable could be the link aggregation process: whom to link with whom?

Related initiatives

 * Gripenet Portugal
 * Observatório Brasileiro de Informações Sobre Drogas
 * SISAP Idoso - indicadores de saúde
 * MonitorAIDS
 * Saútil (notícia)
 * Getfit MIT
 * Music Lab

Basic settings
The survey will run all of Brazil's state capitals, reaching a total of 25,000 people.

It should aim to deliver measurable health realted benefits at the same time it controls the environment for scientific outcomes.

Measurement of the outcomes of interest must be automatic.

The experimental design should be conceived from a network perspective.

Focus on family or community, general health issues
Not necessarily related to the issues approached by the survey, the game can invite participants to engage in many dimensions of health care, such as:


 * learn to access public health information and hold government accountable (Solstag has friends working on a platform to make accessible this kind of information)
 * learn about their consumer rights and make private providers respect them (Institutions like Idec and Procom could join here)
 * learn about the health risks of feeding children cheap industrialized high-fat food with low nutritional value
 * learn about the long term costs of sedentarism and promote the practice of sports through local groups
 * map public health issues, like drug abuse, open sewers, bursts of infectious diseases, dengue mosquitoes etc
 * track their own health and their family's, in particular for chronic diseases like diabetes, heart conditions, high cholesterol

Focus on crack users, crack use
Directly responding to the issues approached by the survey. This may be troublesome given the sensitivity of the issue.

A game where participants would seek crack users not in treatment and guide them to clinics by handing them coupons received from the interviewer. The coupons would have a code that would allow us to identify their source interviewee.

Invite people to collaboratively map crack zones, deploying something like Ushahidi, so health officials can work with more complete and up to date information. Phone apps could be developed to help this mapping, which can also be broader than crack use.

Engagement

 * Appeal to family care, to community care, to civic engagement
 * Facilitate collective organization, supporting groups that mobilize to fulfil tasks in their neighborhood
 * Have an incentive structure where engagement and achievements provide acess to more powerful tools, as an example: for community engagement
 * Involve public health schools at public and private universities around the country as "game shops" that recognize and validate volunteer effort
 * Make use of the game to promote and reinforce existing and already planned health related activities in each local context

Theoretical
Is it possible to go this far into the applied social realm and still hope to do some theoretical work?

Theoretical work should guide our strategic thinking, but can we directly model these processes?

Is it the case to, instead, structure the proccesses in a way that answers some theoretical question?

For example, employ different strategies to the targets and see how they perform?

In the case of reinforcing existing activity, as described in ==Engagement==, measure the gains and what else?

With a sample of 25000, though quite heterogeneous, can we use the sample size to test different approaches?

If so, should we pick randomly? Segment by location and answers in the questionnaire? Or distribute evenly among those with similar location and answers?