We planned to include items measuring felt stigma and enacted stigma at both the family and community level to encompass stigma found in day-to-day interactions and in wider social participation. We also aimed to include both those affected by podoconiosis and unaffected community members.
Potential items to be included in the scales were identified by analysing the findings of studies investigating community knowledge, attitudes and practices towards individuals with podoconiosis [
8‐
11]. Other potential items were identified from a literature review on measuring stigma towards leprosy [
12], another highly stigmatised NTD with symptoms that are hard to conceal, like podoconiosis. Since individuals with podoconiosis face similar exclusionary social treatment to those with leprosy, commonly used items in leprosy stigma assessments were also included in this study.
Potential items were then grouped according to the WHO International Classification of Functioning, Disability and Health (ICF) [
13]. Of the nine domains, leprosy stigma had greatest impact on mobility, domestic life, interpersonal interactions and relationships, major life areas, and community, social and civic life [
12]. Again, existing qualitative research [
10,
11] has demonstrated the effect podoconiosis stigma has on these areas, so these formed the basis through which stigma was assessed (Table
1).
Table 1
Indicators of podoconiosis stigma in three domains
• Buying items at market | • Employment | • Leadership and decision making |
• Eating/living separately | • Marriage | • Participation in community affairs, public events and social organisations |
• Interactions with family, friends, neighbours and health professionals | |
• Isolation from others |
• Shame/embarrassment | | |
Two structured 50-item pilot scales were produced based on the indicators in Table
1, one for podoconiosis patients and one for unaffected community members. The scales generated were first evaluated through a Delphi process in which six people with expertise in the area of podoconiosis were asked to rate the relevance of items included in the preliminary scales. After several discussions, an exclusion consensus was reached and items considered irrelevant were rejected. The inclusion criteria decided by the research team were that each item had to be universal (applicable in different cultural contexts and across different groups such as age, marital status and social class); had to measure stigma (rather than confidence, or self-esteem) and could be clearly categorised as felt or enacted stigma. On the basis of these criteria, 18 items were excluded from the preliminary podoconiosis patient scales.
The scales agreed for testing were formatted as questionnaires containing both positively and negatively framed items to address response set bias, and translated into both the Ethiopian national language (Amharic) and the local language (Wolaitigna) and back translated into English to check for consistency.
Drawing from responses used in scales to measure stigma towards onchocerciasis [
14,
15] participants were asked to respond with either ‘yes’, ‘possibly’, ‘uncertain’ or ‘no’ to each item on the questionnaire. A score of 3 was given for a ‘yes’ response to an item on the scale, 2 for ‘possibly’, 1 for ‘uncertain’ and 0 for ‘no’. A mean value for each item on the scales is then calculated from the participant responses ranging from 0 indicating low stigma to 3 indicating high stigma. This scoring system, and the scales themselves, was developed to aid comparisons rather than establish criteria of stigma or no stigma with a clear numerical cut off point.
Participant selection
Study area and population
The scales were piloted in Wolaita zone, Southern Ethiopia, which has a population of around 1.53 million, 88.3% living in rural areas, the majority working as subsistence farmers [
16]. Wolaita zone was chosen because of the high prevalence of podoconiosis (5.5%) [
17], and the community’s acceptance and responsiveness to previous podoconiosis research. The Mossy Foot Treatment and Prevention Association (MFTPA, a non-government organisation specialising in podoconiosis prevention and care) also provides a contact base for over 35,000 podoconiosis patients [
18].
Sample size
Assumptions used in the sample size estimate were drawn from Brieger’s study measuring onchocerciasis (another NTD) stigma in Western Nigeria (Briefer et al., 1998), and the desired size of reduction in stigma level towards podoconiosis patients were an intervention to be introduced.
We used a mean stigma score of 17 and standard deviation of 10 (Brieger’s 13 item, 39 point scale), and the following equation:
(where u = 1.28 for power of 90%; v = 1.96 for 2-sided significance level of 5%; SD1 = 10, SD2 = 10) (Kirkwood 1988).
A series of sample sizes were calculated based on a range of stigma reductions anticipated. A sample size of 150 podoconiosis patients was calculated to give 90% power to detect a stigma reduction of 20–25%. Less dramatic reduction in stigma level was anticipated in the non-affected community, and 500 community controls were calculated to give 90% power to detect a reduction of 10–15%. A sample of 500 unaffected community members and 150 podoconiosis patients is also adequate to enable factor analysis and thereby validate the scales, based on item to case ratio recommendations for factory analyses. Costello and colleagues recommend an optimal sample of 20 individuals to one item [
19].
Sampling technique
A two stage sampling was used to identify 150 podoconiosis patients in Wolaita Zone. First, five of fourteen active clinic sites of the MFTPA were randomly selected. Systematic sampling was then employed using the patient list at each clinic site to select 30 patients. Any participants who might have been experiencing stigma for other health afflictions such as leprosy or other visible skin diseases were excluded.
Unaffected community members were identified through multistage stratified random sampling. Firstly four weredas (districts) were randomly identified and within each wereda, two kebeles (smallest administrative unit) or Peasant Associations (PAs) were randomly selected. Within each of the weredas, 125 adult (≥ 18 years) household heads were systematically selected from the list obtained from the two kebele administrative offices to make a total sample size of 500 community members. If any of those individuals identified themselves as having podoconiosis, they were excluded from the sample.
Data collection
Due to the low literacy rate of participants the questionnaire was interview-based. Six data collectors were fully briefed on the purpose of study, the eligibility criteria for participants and given training on how to request consent and administer the questionnaires.
Data were collected in May 2011. Podoconiosis patients were interviewed in a private area of the clinic they were approached in. Non-affected community members were interviewed in a private area of their home.
All participants were asked to give their responses according to a 12 month time frame (‘since the start of the last rainy season…’) in order to measure recent accounts and perceptions of stigma.When administering the questionnaire, the participant’s age, gender and education level were recorded, and for podoconiosis patients their disease stage was also noted for further analysis.
Ethical considerations and regulatory approval
Special attention was given to dealing with the sensitive topic of stigma when obtaining informed consent from participants. Following guidance from early studies of the ethics of approaching this community [
20], podoconiosis patients were first asked to participate by MFTPA staff with whom they were familiar, before official consent was obtained by data collectors.
Information sheets and consent forms were inapplicable in this particular setting due to the low-literacy level of participants. Instead, the purpose of the study was explained orally, in the style of a conversation rather than reading out an information sheet, and potential participants gave their consent verbally in front of a witness as confirmation. Ethical approval for this study was given by the IRB of Wolaita Sodo University in Ethiopia.
Data analysis
Before analysis, we decided that since felt stigma and enacted stigma represent two distinct dimensions, they should be assessed through two independent scales. Four scales were therefore analysed in total: a podoconiosis patient felt stigma scale, a community felt stigma scale, a podoconiosis patient enacted stigma scale and a community enacted stigma scale.
The scales were evaluated through reliability assessment, content and construct validity analysis of the items, factor analysis and internal consistency analysis.
Reliability assessment
Reliability was assessed through consistency analysis using Cronbach’s alpha. An alpha of 0.7 to 0.9 was considered good consistency [
21]. The impact of deleting an item on overall consistency was also assessed, with the aim of discarding an item whenever its deletion led to a noticeable improvement in the overall consistency of the rest of the items.
Content and construct validity
Content validity was examined using Spearman’s correlation coefficients. Construct validity was analysed using exploratory factor analysis (FA) with principal components analysis (PCA). Separate factor analyses were performed for felt and enacted stigma for both podoconiosis patients and the community. The first round of un-rotated factor analyses yielded Scree Plots to determine the number of factors underlying the respective stigma dimensions. To find the best fit to the data, orthogonal and non-orthogonal analyses were also conducted. Factor loadings of more than 0.4 were considered satisfactory for the patient and community questionnaires [
19]. Findings of the PCA were further validated by split-half reliability analysis, in which the two random halves of the data were expected to provide similar factor solutions. Following recommendations [
22,
23], factors generated by the PCA were extracted as valid if at least two of the following criteria were met: (1) eigenvalues were equal to or greater than the randomly generated factors from Horn’s parallel analysis; (2) the scree (Cattel’s) test was passed and; (3) eigenvalues of equal or more than unity [
24,
25].
For analysis on any given variable, cases were excluded if data on that variable were missing. Total scores were only calculated if a case had responded to more than 50% of the items. SPSS 15.0 was used for reliability and validity analysis, while Monte-Carlo PA software was used for parallel analysis. An alpha of 0.05 or less was considered significant.