Data analysis

Quantitative data analysis (Papers I, III, IV)

Data analysis was done using STATA version 10 (I, III) and SPSS version 13 (IV).

Standard statistical methods were used for descriptive analysis. Estimated proportions were done with a 95% confidence interval. Chi-square test was performed to test the significant difference between proportions.

Multiple regression models were used to examine influencing factors to the S.

pneumoniae carriage prevalence (I), antibiotic resistance (I), main-caregivers’ overall-knowledge of antibiotics (III), antibiotic use among children (III) as well as HCPs’

knowledge and practical competence (IV). The independent variables were sex, age, education, occupation, geographical areas, household’s economic-status (I, III), present ARI symptoms (I), antibiotic use (I), healthcare-seeking places (III), duration of illness (III), and frequency of seeing childhood patients (IV).

In Paper I, interpretative non-meningitis breakpoints based on the CLSI standards were used to interpret the antibiotic susceptibility of S. pneumoniae isolates (CLSI, 2009). The inhibitory zone diameters for isolates to be considered resistant were: tetracycline

≤18mm, co-trimoxazole ≤15mm, erythromycin ≤15mm, ciprofloxacin: resistant ≤15mm, susceptible ≥30mm (SRGA, 2008b). MICs values for cefotaxime in the Etest were:

resistant ≥4mg/l, susceptible ≤1.0 mg/l. Modified CLSI breakpoints using EUCAST

breakpoints were used to categorize benzylpenicillin susceptibility as: susceptible MICs

≤0.5mg/l; intermediate 1.0mg/l≤ MICs ≤4.0mg/l; and resistant MICs ≥ 8.0 mg/l (CLSI, 2009; EUCAST, 2009b).

We defined the isolates as susceptible to amoxicillin and ampicillin using the same breakpoints as for benzylpenicillin (CLSI, 2009). Resistance to phenoxymethylpenicillin was derived from benzylpenicillin MICs >0.06 mg/l (EUCAST, 2009b). We defined MDR as isolates resistant to at least three of the six tested antibiotics.

In Paper III, multilevel logistic regression was used to adjust for intra-cluster correlation (ICC) of antibiotic use for children in the 28-day period in three levels: the mild ARI episode, the child and the cluster.

In Papers III and IV, the illness of children was classified based on the reported symptoms following the IMCI guidelines as: (i) mild ARIs, if the child presented any of following symptoms: cough, stuffy nose, runny nose, sore throat, without fast breathing or chest in-drawing; (ii) severe ARIs, if the child presented at least one of the pneumonia signs: fast breathing, chest in-drawing, or stridor; (iii) others, if the child had any other symptom such as watery faeces, bloody stools, vomit, ear ache, injury, abdominal pain, skin rash, or toothache.

Overall knowledge about antibiotic use for acute respiratory infections among caregivers and healthcare providers was considered as correct if the respondents provided answers to knowledge questions in accordance with the guidelines for all three stated ARI symptoms, i.e “No” for questions 1, 2, 4 and “Yes” for question 3 (Part II, appendix 1, 6).

Drugs used for the participating children or recommended by HCPs in the clinical scenarios were classified according to the Anatomical Therapeutic Chemical (ATC) classification system (WHO, 2008), with the help of VN-pharmacy software (Hanoi University of Pharmacy, 2004). Antibiotics that are classified as antibacterials for systemic use and aggregated at the level of the active ingredient were included (level 5 of the ATC class J01) (WHO, 2008).

Qualitative data analysis (Paper II)

After each FGD, the research team discussed and considered whether all information had been noted. The tapes and notes of the FGDs were transcribed in Vietnamese and then translated into English by two independent translators. The Vietnamese and English versions were used simultaneously side-by-side during the analysis to avoid misunderstanding of the real meaning of the texts.

Data from the transcripts was analyzed using qualitative content analysis (Graneheim &

Lundman, 2004) in order to describe and understand what people think, understand and do regarding drug use and seeking healthcare for children. My analysis consisted of a dynamic process between the descriptive and interpretive levels of the content. The descriptive level is referred to as the manifest content and described the visible and

obvious components in the text. The interpretative level is defined as the latent content of the text and deals with interpretation underlying meaning of the text (Graneheim &

Lundman, 2004). In practice, the distinction between the descriptive and interpretative level is vague. Graneheim and Lundman (2004) believed that both deal with interpretation, but that the interpretations vary in depth and level of abstraction.

From this point of view, I don’t see the manifest and latent analysis as mutually exclusive. In the analysis process, I was moving back and forth between various levels of abstraction in order to discern underlying meanings and interrelationships of the data.

First, I listened to the tapes and read through the text in order to become familiar with it and to gain an overall understanding of the content of the discussions. Meaning units apparently referring to the same content were identified and allotted to the tentative interview themes.

Condensing the meaning unit was done in two steps, descriptions close to the original text and interpretations of the underlying meanings. Reading and comparing the topics and meaning units established sub-themes (Table 5). Several emerging themes among each interview theme were found during analysis by comparing and sorting various sub-themes. In the final step, I compared the findings between the different FGDs, tackled research questions and decided on main themes. I have done this process manually in collaboration with my co-authors.

Table 5: Example of selecting meaning units and establishing sub-themes Condensed meaning units

Meaning units

Description close to the text

Interpretation of the underlying meaning

Sub-themes

‘It seemed to be amoxicillin. And there was a yellow medicine as well, that was

prednisolon.’

‘My child frequently has coughs and fever. I often take him to a private clinic. Once my child had a high temperature for a long period, the physician at private clinic said that he had a sore throat, and gave him Pamin,

amoxicillin and Dexa’

Antibiotic and corticoids are often recommended Cough and fever are frequent symptoms 1^st choice is private clinic

Long duration of fever is considered Antibiotics are used for sore throat Pamin, amoxicillin, and Dexa were often recommended

Children had cough and fever frequently, and antibiotic and corticoids are commonly used for childhood illness.

Common use of antibiotics and corticoids