powered by NetLogo
Here is an “interactive” logic model, as I call it, inviting you to adjust processes to try to stop the spread of illness. It shows children and adults going from home to public places, illness spreading, and people taking protective measures to protect themselves. Each person has levels of risk-awareness and beliefs that change in response to education and public alerts. Some beliefs are contagious too.
You can try different strategies to try to stop the spread of illness, including education, vaccination and school closure as well as alerts. Agencies signal alerts and offer education.
You can vaccinate different segments of the population, and use alerts to ask people to stay home when sick. However, people will only comply if they have a certain set of beliefs. That is people will only engage in health protective action if they have high perceived risk and high perceived efficacy.
I call this an “interactive” logic model because this format allows anyone to modify parameters through a web browser. Logic modeling help articulate how we expect how processes affect outputs and outcomes, given inputs. The purpose of logic models is not to predict specific numbers of infections but rather to help program partners, citizens, policymakers and any “stakeholder” articulate what kinds of programs they expect to work and how they expect outcomes are produced by processes. I offer an “interactive” format so that stakeholders can produce better logic models.
I ground this logic model in health behavior theory. Each person has levels of risk-awareness, response-efficacy and self-efficacy that can increase in response to public alert and education. Those three variables correspond to the Protection Motivation Theory widely used in studies of response to Influenza and other threats (e.g. Zwart et als international survey, 2007). You can change the initial settings of these variables.
Self-efficacy is an “I can do it” belief. A person with low perceived self-efficacy has low confidence that they can engage in a behavior, or complete a task. One reason this is very important, is that self-efficacy is a necessary condition for action. So even when a person perceives a risk as high, they will not engage in risk protective behavior (staying home when sick and getting vaccinated when that’s available, in this case) if their perceived self-effiacy is low. For that reason, many theory-based studies of health behavior include measures of self-efficacy. Also for that reason, the APHA recommends interventions to increase self-efficacy in response to influenza.
In this model, alerts raise people’s risk awareness, but health education raises people’s self-efficacy. So when self-efficacy is low, we need to raise self-efficacy through health education. In fact, low self-efficacy may be often more of a problem than low perceived risk of influenza, at least based on the findings of an international survey relating to the Avian influenza (Zwart et al 2007). “Although in all countries an influenza pandemic is perceived as a real risk, the level of self-efficacy appears to be rather low.” Zwart et al also found that self-efficacy levels were lower for the youngest age group compared with older respondents. In very rough approximation of this, the elderly have a lower proportion with low self-efficacy, in the model.
Another type of perception that is important is response-efficacy, the belief that a response to influenza is effective. If a person’s perceived response-efficacy is very low, then they are likely not to engage in that protective behavior. Thus, if any of the three key variables–perceived risk, self-efficacy or response-efficacy–is very low, then the person will not engage in the risk behavior. That is the assumption in this very simplified model. If people have high levels of all three variables, they will stay home from public places when sick and they will get vaccinated, when the vaccine is available.
Alerts will raise the level of people’s risk-awareness, but will not raise their self-efficacy. By contrast, health education in this model raises people’s flu protective self-efficacy, as recommended by the APHA.
Watch what happens when health education is turned off: the total percentage of persons ready to take protective action is lower than those who simply intend action. The % actors limited by the number of people who remain at low self-efficacy levels. You can try turning on health education to improve people’s response to the outbreak.
For each alert that is “on” (red), and for each day, approximately 10% more people will have a high level of risk awareness (the probability of each person moving up to a high level of risk perception is 10% per day, per alert).
For each education that is “on” (blue), and for each day, approximately 10% more people will move up to a high level of self efficacy (the probability of each person moving up to a high level of self-efficacy is 10% per day, per agency giving flu health education). Response efficacy is not affected by education or by alert, in this model.
Alerts go off once a certain threshold of sick persons is reached. Once you turn on “flu education,” in this model, the education inteventions start at the same time the alerts go off. You may set the alerts at different levels, by scrolling down below the monitors (described below). For now, leave the alert thresholds alone and observe what happens.
If a person has a high perceived risk and a high perceived response efficacy, then they will have a high “intention” to perform an action. But they will not follow-through with action unless their self-efficacy is also high. So when some people have low self-efficacy you will see that a lower proportion of people are ready to take protective action, e.g. staying home when sick, as compared to the proportion of persons who merely “intend” to perform the action. Intentions are necessary but not sufficient for action.
Self-efficacy can be contagious. The default spread is 0%, but you can adjust the likelihood that self-efficacy spreads, the probability of spread per contact. This only spreads through adults and elderly because children beliefs about their own ability to get vaccinated and stay home when sick are not relevant in this model.
You may experiment with different vaccination and school closure interventions. If you scroll down below the monitor, you will see many buttons with which you can experiment with different interventions. You can set the total amount of vaccine available, and choose to vaccinate any of three demographic groups: children, adults, elderly or all three. You can set the proportion of adults and elderly that want to get vaccinated and parents that want their child to be vaccinated, the day that they go to get vaccinated, the effectiveness of the vaccine. Note that each of these three demographic groups has a different amount of contact with others (children have a high rate of contact, and elderly a low rate), and a different chance of becoming infected from any given contact.
Any of the agencies may have the authority to close school. We may authorize any or all of these agencies to close schools. So you can try turning the authority to close school (switch on or off). Once you switch that on, kids will no longer go to school, if the threshold of sick children is large enough. You can adjust that threshold, which is a certain number of sick children.
You can try out different combinations of vaccination, alerts, education and school closure.
Hit the “setup” inputs button. You will see homes, people, school and workplace, and agencies. Children, adults and elderly populate the homes.
You can change the initial levels of risk perception, self-efficacy etc. These indicate the proportion of persons with low levels (this is stochastic, so the percentages are actually the probability that each person will have a low level). If this is your first run, perhaps try leaving the settings where they are to see what happens.
Once you hit “go” you will see people move to public places: schools and work during the week. Children go to shool and have a high rate of contact. Adults go to work and have a slighly lower rate of contact.
Once we reach the weekend, people go to either a mall or a park. If the “vaccinate” is on, then people who want to get vaccinated go to the mall to get a vaccine on Saturday.
A window showing how many days have elapsed, and windows showing the spread of infection, and vaccines still available.
By default, one person is infected on day 2, but you can change that day of first infection with the slider. You also can initiate an infection whenever you want. Hit the “infect” button whenever you want to initiate an infection. That will cause one person, at random, to be infected. Watch the infection spread. You can use the sliders to change settings. You can change the infection rate and duration.
People will only engage in health protective action–staying home when sick and getting vaccinated when that’s available–only if they have high levels on all belief variables.
Watch what happens when health education is left turned off: the total percentage of persons ready to take protective action plateas at a level lower than those who simply intend action.
The percentage of “actors” is limited by the low self-efficacy. After you see a plateau in the the level of persons ready to take action, you can try turning on health education to improve people’s response to the outbreak. Now you will see the level of “actors” rise to the level of “intenders.”
There are other interventions you can try:
You can give the health department the authority to close the school by switching that setting “on.” You can add a delay to any school closing.
The model also allows you to experiment with vaccination, trying out different priorities, vaccinating only kids, only elderly etc. You will see that vaccinating only kids often can result in a lower overal case rate, and case mortality rate, as compared to vaccinating only the elderly. Although the elderly have a higher case mortality rate, children spread the disease more, in this model.
If you want to look at the effect of school closure without any vaccination, then turn the vaccination off by setting the proportion of adults and children getting vaccinated at 0% (the default is 0%).
If you want to see what happens with vaccination alone, turn the school closure off, and set some proportion of adults and children going to get vaccinated. You can adjust the proportion of adults and children that go to get vaccinated (more precisely the liklihood of going to get vaccinated), the day that they go to get vaccinated, the effectiveness of the vaccine and the amount of vaccine available. Note that the vaccine can run out if the proportion of vaccine per person is less than 100%.
You could set all beliefs to the highest levels, if you want to focus on the interventions alone. Then you also can have sick people stay home on the weekends, and see what effect that has on the spread of infection. You can have a certain proportion of sick adults decide to stay home from work. And you can try different combinations of all of the above, or compare different strategies and see which is most effective in mitigating the spread of infection. (Keep in mind that school closings and staying home from work are not have costs of their own).
If self-efficacy is low, then it is not enough to raise risk awareness. Thus alerts on their own may not be productive. We need to raise self-efficacy as well, through education or some type of intervention.
If one demographic group has particularly low self-efficacy, we may want to target that group with education. The behavioral variables determine readiness to take a vaccine, so in theory, one could waste limited vaccine effort on a group that will not get vaccinated. However, a better alternative is to raise the self-efficacy and perceived risk etc., among those who need it.
Self-efficacy should measure one’s perceived ability to carry out protective action against the flu despite barriers. Self-efficacy is measured through survey instruments, and the questions should “use the following semantic structure: ”I am certain that I can do xx, even if yy (barrier)" (Schwarzer, 2008). Of course, we also want to limit barriers and obstacles, e.g. make it easier to get vaccines, easier to stay home from work when sick. One common barrier is limited vaccine.
When there is a limited amount vaccine, you must choose which demographic groups to vaccinate. Note that the elderly may have a higher chance of becoming infected, but a lower contact rate. Note that children have a higher contact rate. Thus the children will be more effective spreaders of the disease even if the elderly are more susceptible. Under certain conditions, vaccinating children may be the most effective strategy to stop the spread of the disease. Thus even if the elderly are more susceptible and more likely to die from the disease everyone may benefit more if the disease is stopped.
You can set the total amount of vaccine available, and choose to vaccinate any of three demographic groups: children, adults, elderly or all three. You can set the proportion of adults and elderly that want to get vaccinated and parents that want their child to be vaccinated, the day that they go to get vaccinated, the effectiveness of the vaccine. Note that each of these three demographic groups has a different amount of contact with others (children have a high rate of contact, and elderly a low rate), and a different chance of becoming infected from any given contact.
Notice also that the school closure may be more complicated than we expect. Each of the three agencies (the department of education, and the emergency management agency, and the health department) might have the authority to close school. But they might have different criteria for closing down the schools. So here we have a separate slider for each of the three departments. So the user can set the department of education at a different threshold, a different number of children can get sick before the second agency sends an alert (red). In another version of this model I have a third agency, so that the three include the local health department, the department of education, and an emergency management agency, all of which can signal an alert). The main point here is to model how organizations can coordinate to close schools. After all, organizations “behave” too. (This is the topic of another of my models that focuses entirely on organizational coordination). It turns out that having more than one agency with authority is often an advantage).
Once a certain threshold of sick children is reached, the health department signals an alert. But the emergency management agency may have a different threshold. So with different thresholds we see the alerts go off successively. First the health department, then the department of education, then the emergency management agency alert.
Note that school only closes if the agency also has the authority to close the schools. Note that school closure along may have little effect if people still go to other public places. That is, even when kids don’t go to school interact in the parks and shops on the weekends, and parents still go to work. More interestingly, schools will not close in the first place, unless the agencies are coordinated.
See chapter 10 in my book, modeling behavior, for more detail (Keane 2013).
Authored by Christopher Keane The motion from homes to public places
was inspired by Burke, Epstein and colleagues’ model of smallpox control, 2006.