3.2. Data Analysis

Three researchers performed a content analysis on the entire 6,713 posts. The analysis consists of the following steps (shown in Figure 1): First, we developed a codebook using 200 posts randomly sampled from the corpus. More specifically, the sampled posts were independently reviewed by two coders, C1 and C2, for relevance judgment. For posts that were considered relevant, we used an open coding technique to categorize posts and their chosen replies (also known as the “Best Answer”). As the quality of the post answers varies, the poster often chose a reply which fulfilled his/her information needs as the “Best Answer”. Therefore, examining the best answer responses could help us not only understand what types of support were provided but also confirm the poster’s information needs.

Data Collection and Analysis Steps.

Once an initial version of the codebook was established, the researchers met to discuss whether to keep, remove, merge, or replace any codes. After the researchers reached an agreement about every code in the codebook, the codebook was then utilized to standardize the coding procedure. The finalized codebook contained 14 themes, which were organized by three general stages of pregnancy: pre-pregnancy, during-pregnancy, and postpartum (shown in Table 2).

Summary of themes for information needs. Some posts fell into multiple themes, so percentages add up to more than 100%. Percentage equal to the number of occurrence divided by the totoal posts analyzed.

Next, the third coder (C3) coded another 100 randomly sampled posts to examine the exhaustiveness of the codebook. Through this process, no new codes merged, confirming that the codebook was comprehensive enough. Following that, C1 and C2 coded another random sample of 200 posts and applied Cohen’s Kappa Coefficient to test their inter-rater reliability (Figure 1, Step 3). Two coders achieved a substantial agreement (kappa’s value is 0.80). Once we resolved the disagreements, C1 and C2 coded the remaining posts to conclude the analysis (Figure 1, Step 4).