Now we will look at how the sixth step in Barrier Analysis is carried out when you have used option #2 - individual interviews. We will use the example of ORS.

1. Develop a coding guide similar to the one below to hold the data. You will tabulate the proportions of Doers and Non-Doers answering each question.
   
   Let’s say that you were looking at the positive attributes of the behavior, “using ORS when your child has diarrhea.” One question that you may have in your questionnaire to examine that determinant is, “What do you see as the advantages or good things that (would) happen if/when you used ORS when your child has diarrhea?”
   
   Let’s say that the responses seen for that question are: “Can prepare quickly, low cost of packet, easy to make, child’s older sister can make it when I’m not home”. As you go through each questionnaire, make tic marks to register the responses on each questionnaire. For this type of questionnaire, you may have more than one response per questionnaire. Your table would look like this.
   
   For closed (yes/no) questions (e.g., “When a person exclusively breastfeeds a child for the first six months of life, does that help to avoid diarrhea in the child?”), you can make up a coding sheet using the responses included in the questionnaire (e.g., Yes, No, Don’t Know). As you go through each questionnaire, make tick marks to register the responses on each questionnaire. Your table would look something like this.
   
   For another example, refer to The “Exercise” Exercise page, which provides a complete questionnaire and its accompanying coding guide.

2. Divide the questionnaires into two stacks: people who reported THEY DID DO THE BEHAVIOR (e.g., used ORS) versus those who reported THEY DID NOT DO THE BEHAVIOR (e.g., did not use ORS).

3. Mark each questionnaire: For the stack of questionnaires from those who reported YES, mark each sheet of the questionnaire with a “D” for “Doer.” For the stack from respondents who reported NO, mark each sheet with “ND” for “Non-Doer.”

4. Keep the stacks separate and divide each stack among your staff who are tabulating the responses.
   
   The tabulator should look at each participant’s responses and try to find the same or a very similar response on the coding sheet illustrated above. If the tabulator finds a genuinely different response, write the response on the “Other” line and add a tick mark in the appropriate column of the coding guide
   
   As each response is coded, the tabulator should place a tick mark next to that response in either the “Doer” or “Non-Doer” column of the coding guide, depending on the stack from which it came (“D” or “ND”). At the same time, the tabulator should place a check on the questionnaire beside that question to indicate that the response has now been coded. Tabulators should register a tick mark for each different response, even if some seem similar.

5. Once all questionnaires have been tabulated, calculate percentages for each possible response. To do that, first write down in each cell the total number of tick marks in that cell. Then calculate percentages by using the total number of “D” questionnaires as the denominator for the “Doers” column. Use the total number of “ND” questionnaires as the denominator for the “Non-Doers” column.

6. Now look for five or six of the biggest differences in percentages between the Doers and Non-Doers responses, or responses where there was surprisingly little difference between Doers and Non-Doers.
   
   Keep in mind the following:

1. When Doers and Non-Doers report similar percentages for any item, that item is not a likely determinant of the behavior for this target group.

2. When Doers’ responses are radically different from Non-Doers’ responses, that item is very likely an important determinant of the behavior for this target group.

3. This rapid survey technique is not a rigorous statistical analysis of your findings. Therefore, when we speak of “differences” between responses of Doers and Non-Doers, it is important to look for relatively “big” differences, that is, differences of more than a few percentage points. If you calculate confidence intervals on each proportion, you will be looking for differences where the confidence intervals do not overlap. If all overlap, it means your sample size should be larger. In that case, you can still look for those with the smallest amount of overlap; these differences will be the ones that are more likely to be important.If you have a larger number of people in your sample (e.g., 300), smaller differences may be significant. For small samples (e.g., 60 people), only larger differences (>30 percentage points) are generally meaningful.

4. Knowledge about the health benefits of the behavior will often be very similar among Doers and Non-Doers and therefore often not a practical focus for an intervention.

5. Doers’ responses may include ideas for strategies on how to make the behavior easier or more appealing, and could provide clues for messages to Non-Doers. Examine these carefully.

6. Sometimes more Doers list a particular disadvantage of the behavior than do Non-Doers. This may simply indicate that the Doers are more familiar with the behavior. Despite familiarity with the disadvantage, they have overcome it to be Doers. Program planners will need to consider whether a difference between Doers and Non-Doers, in this case, indicates an item that the intervention should address. They may need to talk further with Doers and Non-Doers to determine what to do with such data.

7. Looking at differences between Doers and Non-Doers regarding those who approve or disapprove of the behavior may provide important information on who to target for your intervention. If differences are noted, this implies that you may need to work with a different target group than you had originally intended. You may first have to work with the “influentials” to change their attitudes towards the behavior (e.g., convincing mothers-in-law that ORS is good for their grandchildren).

7. To summarize your results for program planning, list your selected findings in a table like the Sample Individual Interview Results Table (61kb). (An actual Barrier Analysis results table would have more rows since it would be summarizing more questions.) In the Research Findings column, list the findings for each determinant (summarizing the questions) and then report the percentage of Doers and Non-Doers for those findings in columns 2 and 3. Leave the “Implications” and “Program Focus” columns blank for the moment.

8. Now you should discuss the results from the Barrier Analysis and how it should affect your program planning. Make notes (on newsprint) about the implications of the results and to what degree your program or intervention should focus on that determinant. In the “Implications” column, mention whether there is a significant difference between Doers and Non-Doers, whether the intervention should target each determinant analyzed, and whether an intervention is likely to change the situation. Add to your table the implications and to what degree the program or intervention should focus on the determinant/barrier.Your summary could look something like: Research findings table