Hi everybody. This is Stefan Baral and today, I'll be presenting a module titled dissemination strategies and communications frameworks. Now, we've a series of different objectives for today and some of them include describing the methods for presentation of surveillance data, talking about the flow of surveillance data that are in place to meet legal reporting requirements, and also provide an overview of different privacy legislations that exists related to surveillance data. Then I'll spend some time talking about different communication frameworks and how they're important in terms of really communicating the sensitive data that are often included within surveillance systems. So I'll start by talking about the presentation of surveillance data. Now, some of this may seem basic but it's important to have a really organized and strategic approach to how you're going to organize your data. We'll start just by talking about the use of tables. Now, most people have generated a table and the format is pretty consistent. One arranges data in rows and columns and those data can take the form of counts, means, rates, or other analytic measures. We use tables to really present patterns within data and often tables are the source of data that are used for other graphics. But there are certain characteristics of tables to be mindful of. Some of those include that the title should really concisely describe what these data are, when they were collected, or what they refer to, and where they were collected. It's important that the tables title can stand on its own, that one could present that in a different setting and that people could really understand what those numbers refer to. Each row and each column should be clearly labeled. One should really clearly label also the units of measure being used, define any abbreviations and symbols, and importantly reference the data if it's not your own. Now, tables can either be single-variable tables or multi-variable tables. Here, you can see a single table that is titled estimated number, percentage, and rate of pool chemicals-associated injuries that were treated in emergency departments by selected characteristics in the United States in 2007. What one can see here, is the distribution of a total of 115 cases of pool chemical-associated injuries, by injury diagnosis, by the affected body part, by the patient disposition. Then they have a series of different analytic approaches including a weighted estimate which is to say, what is the estimate of all pool chemical associate injuries that likely took place across the United States, confidence intervals around those estimates, and then the percent attribution for each of the types of diagnosis or affected body part. In this table as well they provide an annualized rate as well as a conference interval for each of those. But again, a single variable table, is a frequency distribution for a single variable. Now, multi-variable tables intend to normally provide more detailed data on two different variables. Here, we're looking at a table of the estimated average annual number of a person with self-reported current asthma, by age, sex, race, ethnicity, region, and poverty level. Importantly, we can see down in our columns, are the characteristics, sex, race, ethnicity, region, and the ratio of family income to poverty threshold. Whereas along our different columns, we have the different age groups. So what this has allowed us to do is have both the characteristic as well as the distributions by age. As one could imagine, it's not really feasible to do this with much more than two different characteristics in the context of a table. But tables do represent a really important strategy to present large bodies of data in organized ways. Graphs are also commonly used by many to display data. The format again, is known to many, it is a visual display of different coordinates. We use graphs to visualize patterns, trends, aberrations, similarities, and differences in data. But there's also certain characteristics to really keep in mind when developing graft to present surveillance data. One, generally simple is best. If you have a simple and clear approach to developing a graph, ultimately it'll improve the clarity and the message will more clearly come across. It's important to differentiate variables with a legend or a key. It's also important to ensure that the scales are appropriate for the data that are presented. If there are scale breaks, ensure that you clearly indicate them in addition to the units of measure. Again, define the abbreviations and any symbols used. Again, if not your own data, reference it. There's lots of different types of graphs and I'll just present a few different examples commonly used. One is the arithmetic scale line graph. Now, the format is that there are equal distances across both the x and y axes that represent equal quantities. We normally use these to examine long series of data or to compare several data sets. In a different module, I presented historical rates of syphilis in the United States and we used a graph very similar to this where we looked at the data from 1989-2017. Here, we're looking at the number of salmonellosis events in livestock and chickens per year. We can see that slowly but surely, there has been a decline in the number of salmonellosis events from 1996-2007 in four primary different types of animals including sheeps, pigs, chickens, and cattle. Semi-logarithmic scale line graphs are also frequently used. Here, one axis is arithmetic and that's normally the x-axis. One axis is logarithmic, where we've transformed the data and that's normally the y-axis. We use these to examine the rates of change in surveillance data and it really helps us smooth the line, it allows us to show large differences in magnitude or where there are significant outliers in a single graph. The interpretation here is that, the slope of the line indicates either the rate of increase or decrease. The example that's provided here, is a semi-logarithmic scale line graph of injury rates by age from 2003-2005. On our x-axis, we have age and years, from under one all the way up to 85 and over. On our y-axis, we have events per 10,000 population, again, in a log scale. Thus the y-axis in a log scale, goes from 0.1 all the way up to 10,000. Now, as you can see you have three primary lines here, the lowest one are deaths which are most rare, hospital discharges, and then initial emergency department visits. We can see a clear relation between initial emergency department visits, hospital discharges. But importantly, we can show all three of those that are varying magnitude all on the same graph in a way that allows us to really compare the relationships across them. Finally, charts. The format is as well varied including bar charts, and pie charts, and histograms. We use charts to compare magnitude of events in categories. A bar chart is where the length of the bar is normally proportional to the frequency of event. We can use these for non-continuous variables including categorical or ordinal variables. Generally speaking, two-dimensional is best even if Excel allows you to do these in 3D, we recommend having spaces between the bars and you can have grouped or multi-unit bar charts. But generally, you would want to have less than three groups for effective presentation. So here's an example of live births by cesarean section, by plurality and year in the United States from 1996-2006. On the y-axis we have percentage from 0-100. On the x-axis we have two different elements. We have the plurality as well we have by year. So you can see we've grouped data from 2006, 2000, and 1996, and it's been grouped by plurality in terms of singleton, twin, and triplet and higher-order. Generally speaking, what we can see is that cesarean sections have been increasing from 1996 through to 2006. Although the greatest increases have really been in the context of singleton and twin pregnancies, whereas it was always common for triplet and higher-order pregnancies to be delivered by cesarean section. So again, this bar chart allows us to see a lot of different information and interpret it in the context of a single graph. Histograms are also commonly used within surveillance and in the section on outbreaks presented by Justin Leslie, you will have seen many of these histograms because they are commonly used in the context of outbreaks. But they are graphic representations of the frequency distribution of a continuous variable. Generally, the bases of rectangles lie on a linear scale representing different time intervals and the heights are proportional to the frequencies or the values within each of the intervals. Here, we're looking at an outbreak of influenza. On the x-axis, we can see dates. Those are really in days from early March through the early May. On the y-axis, we see numbers of cases with an axis ranging from 0-1200. We have two different colors represented here including suspected cases and confirmed cases. Then we can follow it along and see with the histogram where the outbreak really started both in terms of suspected cases as well as in terms of confirmed cases and also when the outbreak ended by the decrease in both of those.
