Analysis of Instagram® Posts Referring to Cleft Lip

Background: Social media has become a source of medical information. Cleft lip and palate is a visible congenital anomaly. The aim of the study was to analyze Instagram® posts on the topic of cleft lip. Methods: Instagram® posts with “#cleftlip” from March 2014–March 2017 were accessed. Separate lists of expressions (hashtags, meaningful words, words with emojis or emojis alone) were prepared for primary posts and for replies. Thirty expressions statistically most frequent in primary versus secondary posts and 30 in secondary versus primary posts were identified (Group 1) as well as 30 English words or hashtags (Group 2), non-English words or hashtags (Group 3) and emojis (Group 4). The frequencies of expressions were compared (Z-test for the difference of two population proportions). Results: There were 34,129 posts, (5427 primary posts and 28,702 replies), containing 62,163 expressions, (35,004 in primary posts). The occurrence of all expressions was 454,162, (225,418 in primary posts and 228,744 in replies). Posts with positive expressions such as “beautiful”, “love”, “cute”, “great”, “awesome” occurred more often than these with negative ones. In replies all emojis were positive. Conclusions: Numerous Instagram® posts referring to cleft lip are published and do provoke discussion. People express their solidarity and sympathize with persons affected by cleft.


Introduction
Instagram ® (Facebook, Inc., Menlo Park, CA, USA) is an online photo-sharing application and service, where users may share pictures or videos. It was launched in October 2010 as a free mobile application. On 21 December 2016 it was announced that its community has grown to more than 600 million users and the last 100 million of Instagrammers joined in the past six months [1]. On 26 April 2017 it was announced on Instagram ® website that its popularity has grown to 700 million Instagrammers [1].
Since 2012, when 13% people used this service, a significant increase of usage is observed. According to a national survey carried out between 7 March and 4 April 2016 on 1520 adults, about 32% of online adults (e.g., American people who currently use the Internet, according to the same survey it was 86% of the population) or 28% of all adult Americans report using Instagram ® -roughly the same share as in 2015, when 27% of online adults used the application. Instagram ® use was especially high among younger adults; 59% Instagrammers were people between [18][19][20][21][22][23][24][25][26][27][28][29] year, whereas 8% only were older than 65. According to Pew Research Center women were more likely to use this service (38%) than men (26%). Half of Instagram ® users access the platform daily, 35% of them several times a day [2].
Instagram ® reflects major social trends, especially among young population. People use Instagram ® for communication and entertainment or to express thoughts, moods and feelings. Nowadays, social media play an important role in searching for information by medical caregivers or patients who might look for answers to their questions online [3][4][5][6]. Several papers pertaining to the use of social media in the context of a medical problem (diabetes, Zika virus, arthroplasty) could be found. By interacting with other users, people with medical problems may provide or gain support and share information.
Numerous social media network analyzers are available online. One of them is Netlytic [7]. It allows to automatically summarize and gather data from online conversations found on social media sites [7]. A few questionnaire studies on the use of social media in the context of cleft lip has been published in the recent years [8,9]. No studies based on social media surveillance for cleft lip could be found.
The aim of the study was to analyze the frequency of individual meaningful words, emojis, emoticons or hashtags and to compare their frequency in Instagram ® posts and their replies.

Instagram ® Surveillance
The analysis of the content of Instagram ® posts was initiated by querying the hashtag #cleftlip using the Netlytic service (netylic.org, Toronto, ON, Canada), an open-sourced software. All tagged messages with the #cleftlip hashtag on Instagram ® were downloaded and exported to a spreadsheet. The posts collected were divided into primary posts and secondary posts (replies). Primary posts consisted of pictures and their descriptions posted by Instagram ® users. Secondary posts were responses to primary posts written by other users or by the author of primary post. Two separate lists of all expressions (hashtags, meaningful words, words combined with emojis or emojis alone) present in the posts were prepared for primary posts and for secondary ones. For the purpose of this article both emojis and emoticons were later stated as emojis. A spreadsheet macro was written to assign each expression the number of occurrences in all primary and secondary posts. Capitalization of letters was ignored as well as dots, commas, exclamation marks and question marks. Then, both lists were grouped into one, containing expressions sorted in descending order by the number of occurrences in primary posts (x 1 ), and each expression was also assigned the number of occurrences in secondary posts (x 2 ). The total list consisted of 62,163 expressions.
The proportionp 1 of the number of occurrences of each expression in the total number of occurrences of all expressions in primary posts (n 1 ) was calculated. A similar procedure was applied to secondary posts in order to calculate proportionp 2 from x 2 and the total number of occurrences of all expressions in secondary posts (n 2 ):p On the basis of the calculated absolute value of the test statistic z (see the section "Statistical analysis"), 30 expressions statistically most frequent in the primary versus secondary posts were identified and tabularized. Similarly, 30 expressions statistically most frequent in the secondary versus primary posts were identified. All together the 60 expressions have been placed in Table 1 and designated as Group 1 of expressions. The same procedure has been applied to English words or hashtags (Group 2, Table 2), non-English words or hashtags (Group 3, Table 3) and emojis (Group 4, Table 4). There were no emojis statistically more frequent found in primary versus secondary posts, thus Table 4 consists of 30 emojis from secondary posts only. On the basis of the calculated absolute value of the test statistic z (see the section "Statistical analysis"), 30 expressions statistically most frequent in the primary versus secondary posts were identified and tabularized. Similarly, 30 expressions statistically most frequent in the secondary versus primary posts were identified. All together the 60 expressions have been placed in Table 1 and designated as Group 1 of expressions. The same procedure has been applied to English words or hashtags (Group 2, Table 2), non-English words or hashtags (Group 3, Table 3) and emojis (Group 4, Table 4). There were no emojis statistically more frequent found in primary versus secondary posts, thus Table 4 consists of 30 emojis from secondary posts only. On the basis of the calculated absolute value of the test statistic z (see the section "Statistical analysis"), 30 expressions statistically most frequent in the primary versus secondary posts were identified and tabularized. Similarly, 30 expressions statistically most frequent in the secondary versus primary posts were identified. All together the 60 expressions have been placed in Table 1 and designated as Group 1 of expressions. The same procedure has been applied to English words or hashtags (Group 2, Table 2), non-English words or hashtags (Group 3, Table 3) and emojis (Group 4, Table 4). There were no emojis statistically more frequent found in primary versus secondary posts, thus Table 4 consists of 30 emojis from secondary posts only. On the basis of the calculated absolute value of the test statistic z (see the section "Statistical analysis"), 30 expressions statistically most frequent in the primary versus secondary posts were identified and tabularized. Similarly, 30 expressions statistically most frequent in the secondary versus primary posts were identified. All together the 60 expressions have been placed in Table 1 and designated as Group 1 of expressions. The same procedure has been applied to English words or hashtags (Group 2, Table 2), non-English words or hashtags (Group 3, Table 3) and emojis (Group 4, Table 4). There were no emojis statistically more frequent found in primary versus secondary posts, thus Table 4 consists of 30 emojis from secondary posts only. On the basis of the calculated absolute value of the test statistic z (see the section "Statistical analysis"), 30 expressions statistically most frequent in the primary versus secondary posts were identified and tabularized. Similarly, 30 expressions statistically most frequent in the secondary versus primary posts were identified. All together the 60 expressions have been placed in Table 1 and designated as Group 1 of expressions. The same procedure has been applied to English words or hashtags (Group 2, Table 2), non-English words or hashtags (Group 3, Table 3) and emojis (Group 4, Table 4). There were no emojis statistically more frequent found in primary versus secondary posts, thus Table 4 consists of 30 emojis from secondary posts only. On the basis of the calculated absolute value of the test statistic z (see the section "Statistical analysis"), 30 expressions statistically most frequent in the primary versus secondary posts were identified and tabularized. Similarly, 30 expressions statistically most frequent in the secondary versus primary posts were identified. All together the 60 expressions have been placed in Table 1 and designated as Group 1 of expressions. The same procedure has been applied to English words or hashtags (Group 2, Table 2), non-English words or hashtags (Group 3, Table 3) and emojis (Group 4, Table 4). There were no emojis statistically more frequent found in primary versus secondary posts, thus Table 4 consists of 30 emojis from secondary posts only. analysis"), 30 expressions statistically most frequent in the primary versus secondary posts were identified and tabularized. Similarly, 30 expressions statistically most frequent in the secondary versus primary posts were identified. All together the 60 expressions have been placed in Table 1 and designated as Group 1 of expressions. The same procedure has been applied to English words or hashtags (Group 2, Table 2), non-English words or hashtags (Group 3, Table 3) and emojis (Group 4, Table 4). There were no emojis statistically more frequent found in primary versus secondary posts, thus Table 4 consists of 30 emojis from secondary posts only.                 Names, surnames, question marks, conjunctions, colons, pronouns and dashes were ignored as well as non-meaningful expressions. Emojis occurring one by one were considered as an independent type of emoji. For example, three hearts that occurred one by one were treated independently, as a more expressive form. Non-English expressions were recognized using available online dictionaries and translators (in most cases by Google Translator). If a non-English expression consisted of several words (e.g., #pormaissorrisossemvergonha), they were separated (#por mais sorrisos sem vergonha) and translated with Google Translator (#for more shameless smile) and then concatenated (#formoreshamelesssmile).

Statistical Analysis
The frequencies of selected expressions in the posts and their replies were compared using a ztest for the difference of two population proportions. The test statistic were applied for expressions meeting the following conditions: min( ̂ , (1 − ̂ )) ≥ 5 and min( ̂ , (1 − ̂ )) ≥ 5.
The null hypothesis : = (the proportions in both statistical populations are equal) and the following alternative hypotheses were checked: • two-tailed : ≠ (the proportions in both populations differ); • one-tailed : > or : < (proportion among the expressions occurring in primary Names, surnames, question marks, conjunctions, colons, pronouns and dashes were ignored as well as non-meaningful expressions. Emojis occurring one by one were considered as an independent type of emoji. For example, three hearts that occurred one by one were treated independently, as a more expressive form. Non-English expressions were recognized using available online dictionaries and translators (in most cases by Google Translator). If a non-English expression consisted of several words (e.g., #pormaissorrisossemvergonha), they were separated (#por mais sorrisos sem vergonha) and translated with Google Translator (#for more shameless smile) and then concatenated (#formoreshamelesssmile).

Statistical Analysis
The frequencies of selected expressions in the posts and their replies were compared using a ztest for the difference of two population proportions. The test statistic were applied for expressions meeting the following conditions: min( ̂ , (1 − ̂ )) ≥ 5 and min( ̂ , (1 − ̂ )) ≥ 5.
The null hypothesis : = (the proportions in both statistical populations are equal) and the following alternative hypotheses were checked: • two-tailed : ≠ (the proportions in both populations differ); • one-tailed : > or : < (proportion among the expressions occurring in primary Names, surnames, question marks, conjunctions, colons, pronouns and dashes were ignored as well as non-meaningful expressions. Emojis occurring one by one were considered as an independent type of emoji. For example, three hearts that occurred one by one were treated independently, as a more expressive form. Non-English expressions were recognized using available online dictionaries and translators (in most cases by Google Translator). If a non-English expression consisted of several words (e.g., #pormaissorrisossemvergonha), they were separated (#por mais sorrisos sem vergonha) and translated with Google Translator (#for more shameless smile) and then concatenated (#formoreshamelesssmile).

Statistical Analysis
The frequencies of selected expressions in the posts and their replies were compared using a ztest for the difference of two population proportions. The test statistic were applied for expressions meeting the following conditions: min( ̂ , (1 − ̂ )) ≥ 5 and min( ̂ , (1 − ̂ )) ≥ 5.
The null hypothesis : = (the proportions in both statistical populations are equal) and the following alternative hypotheses were checked: • two-tailed : ≠ (the proportions in both populations differ); • one-tailed : > or : < (proportion among the expressions occurring in primary Names, surnames, question marks, conjunctions, colons, pronouns and dashes were ignored as well as non-meaningful expressions. Emojis occurring one by one were considered as an independent type of emoji. For example, three hearts that occurred one by one were treated independently, as a more expressive form. Non-English expressions were recognized using available online dictionaries and translators (in most cases by Google Translator). If a non-English expression consisted of several words (e.g., #pormaissorrisossemvergonha), they were separated (#por mais sorrisos sem vergonha) and translated with Google Translator (#for more shameless smile) and then concatenated (#formoreshamelesssmile).

Statistical Analysis
The frequencies of selected expressions in the posts and their replies were compared using a ztest for the difference of two population proportions. The test statistic were applied for expressions meeting the following conditions: min( ̂ , (1 − ̂ )) ≥ 5 and min( ̂ , (1 − ̂ )) ≥ 5.
The null hypothesis : = (the proportions in both statistical populations are equal) and the following alternative hypotheses were checked: • two-tailed : ≠ (the proportions in both populations differ); • one-tailed : > or : < (proportion among the expressions occurring in primary Names, surnames, question marks, conjunctions, colons, pronouns and dashes were ignored as well as non-meaningful expressions. Emojis occurring one by one were considered as an independent type of emoji. For example, three hearts that occurred one by one were treated independently, as a more expressive form. Non-English expressions were recognized using available online dictionaries and translators (in most cases by Google Translator). If a non-English expression consisted of several words (e.g., #pormaissorrisossemvergonha), they were separated (#por mais sorrisos sem vergonha) and translated with Google Translator (#for more shameless smile) and then concatenated (#formoreshamelesssmile).

Statistical Analysis
The frequencies of selected expressions in the posts and their replies were compared using a ztest for the difference of two population proportions. The test statistic were applied for expressions meeting the following conditions: min( ̂ , (1 − ̂ )) ≥ 5 and min( ̂ , (1 − ̂ )) ≥ 5.
The null hypothesis : = (the proportions in both statistical populations are equal) and the following alternative hypotheses were checked: • two-tailed : ≠ (the proportions in both populations differ); • one-tailed : > or : < (proportion among the expressions occurring in primary Names, surnames, question marks, conjunctions, colons, pronouns and dashes were ignored as well as non-meaningful expressions. Emojis occurring one by one were considered as an independent type of emoji. For example, three hearts that occurred one by one were treated independently, as a more expressive form. Non-English expressions were recognized using available online dictionaries and translators (in most cases by Google Translator). If a non-English expression consisted of several words (e.g., #pormaissorrisossemvergonha), they were separated (#por mais sorrisos sem vergonha) and translated with Google Translator (#for more shameless smile) and then concatenated (#formoreshamelesssmile).

Statistical Analysis
The frequencies of selected expressions in the posts and their replies were compared using a z-test for the difference of two population proportions. The test statistic were applied for expressions meeting the following conditions: min(n 1p1 , n 1 (1 −p 1 )) ≥ 5 and min(n 2p2 , n 2 (1 −p 2 )) ≥ 5.
In the present study, hashtags dominated in primary versus secondary posts. It may be explained by the fact that use of hashtags helps the author of a message to link post with the group of desired subjects.
Pew Research stated that Instagram ® is very popular among non-white users. According to this demographic statistics in 2014 Hispanic origin people represent 34% of Instagram ® online adult users in the United States of America [20]. This is visible in our study among non-English words or hashtags group (Table 4) represented most often by Spanish and Portuguese languages.
The statistically significant difference in the occurrence of emoji in secondary versus primary posts indicates that replies gave positive responses or comments. This means that many individuals expressed their solidarity and sympathized with persons affected by cleft. The authors find this aspect as very optimistic.
What was remarkable during the review was the fact that there were posts pertained to animals (dogs, cats and a post on a squirrel). This shows that owners of animals with cleft problem also post on Instagram ® .
It is worth noticing that the number of replies (28,702) was much higher than of the primary posts (5427). This indicates a great interest in the posts concerning cleft lip. No comparison of primary versus secondary posts that might be used for comparison could be found in other studies analyzing social media.
From the fact that for primary posts, the total occurrence of all expressions was 225,418 (in 5427 posts) and for secondary posts 228,744 (in 28,702 posts), we may assume that the replies were shorter. The possible explanation could be that the primary posts contained detailed descriptions and the replies were spontaneous (often short and emotional) reactions to them.
Personalized medicine may create optimal treatment for the group of patients with cleft lip. Each cleft is unique in terms of its morphology. Every individual affected may suffer from different medical and psychological problems.
It is evident that people affected by a cleft interact with one another via Instagram ® . As social media become frequently used as a source of medical information, professionals should be aware of content available through Instagram ® and consider using it as a means to provide health education. In the future, a more detailed surveillance of Instagram ® posts with #cleftlip hashtag may help us better understand motivation, experiences and expectations of patients with clefts. In this way we may provide more accurate interdisciplinary and holistic treatment for affected persons.

Conclusions
(1) Numerous Instagram ® posts referring to cleft lip are published and provoke discussion.
(2) In Instagram ® posts two groups of meaningful expressions can be identified: one that appear more frequently in primary posts than in secondary posts and the other appearing more often in replies than in primary posts. Funding: This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.