A prospective longitudinal survey was designed which used a mixed methods approach to collect both quantitative and qualitative data on the CDD contribution to the Ugandan NTD Programme. The CDDs were selected across four randomly selected districts. Districts were stratified according to the number of drug delivery rounds a CDD would have to carry out in their community over an annual mass treatment campaign. These were either one, two or three delivery rounds depending on the underlying endemicity of NTDs and required drug package within the district. Kamuli district represented three delivery rounds ivermectin (IVM) and albendazole (ALB) for lymphatic filariasis, praziquantel (PZQ) for schistosomiasis, azithromycin (ZIT)) for trachoma, Mayuge district (PZQ + ALB, ZIT) and Yumbe (IVM + ALB, PZQ), where IVM was also for treatment for onchocerciasis, were representative of two rounds and Pallisa (IVM + ALB) was selected from the stratum with one delivery round.
Across the four districts a sample of 64 CDDs was sufficient for both the depth of information from repeat engagement with the participants (using PDs), as well as breath of information across variables such as number of deliveries, distribution method and number of households served to allow for generalizability of the findings  and was in-line with other similar studies [22–24]. Each village has two CDDs which it elects to take part in the NTD Programme (A. Onapa, RTI International pers comm.), thus 32 villages were selected. Decisions on methods for distribution of tablets tend to be made, and so similarities occur, at the sub-county level or parish level and so for each of the four districts, multi-stage randomisation was used to select two sub-counties (8 sub-counties) and from these, two parishes (16 parishes) and finally from each parish, two villages were selected (32 villages). The total population of the study areas was an estimated 34,615 with approximately 5,724 households, thus 89 households would be served by each CDD (Ugandan Bureau of Statistics, www.ubos.org). The CDDs were enrolled into the study at least two weeks prior to the beginning of NTD Programme activities in October 2008, until June 2009 when programme activities for which they were involved, had been completed.
CDD role in the NTD programme
Distribution mechanisms for IPC through the Ugandan NTD Programme are community-based and school-based. CDDs are elected by their communities or requested by their local councillors to participate in the programme. After receiving training from sub-district health personnel, the CDDs ‘volunteer’ their time to sensitise and mobilise communities about treatment for the NTDs and subsequent health benefits. Following this they distribute drugs to the eligible target population either from a focal point such as a church, health clinic, trading centre or by the CDD moving from door to door. Following distribution CDDs write coverage reports based on their treatment registers and submit the results and any remaining drugs to the nearest health unit.
Data collection was carried out in teams all trained by the study investigators and an experienced social scientist from Makerere University, Kampala. All interviews and FGD were held in the relevant local languages for the study and pre-testing districts: Lusoga for Kamuli and Mayuge districts, Lugwere in Pallisa district, Luganda in Mukono district (for pre-test) and Aringa Lugbara in Yumbe district. Data collection comprised of four components described herewith.
The first component determined how CDDs conducted their days, in terms of daily routines and during NTD treatment campaigns. Two FGDs were held and CDDs were asked to recount their daily activities from when they awoke to when they went to bed until no new activities were mentioned. The CDDs were then asked to describe what symbols or pictures they thought might best represent these daily and routine activities. Meanwhile, two Ugandan artists were sketching these symbols and pictures. Subsequently the CDDs were requested to review the symbols and pictures developed by the artists and identify which activities they represented. Where the pictures were identified incorrectly or were not clear the CDDs were asked to describe how they could be improved to be more recognizable.
The pictures were further refined and FGD were conducted across three districts to draw out any daily routine and NTD activities not previously mentioned, highlight any seasonal variations in routines and to review the set of illustrations to suggest any improvements for clarity. CDDs were also asked to describe how they measured their time, for instance, what tools or features they might use to identify the time of day and how they split the day. The most common methods of telling the time used by the CDDs were radio programmes, especially the news; mobile phones; clocks in their houses; Muslim call to prayers; and for those who lived near schools, the school bell (the banging of a wheel) at different times during the day. In each district the CDDs split the day into morning, afternoon, evening and night with only minor differences at what time they began and ended. All CDDs reported that they used ‘Swahili’ time throughout the day, where 7 a.m. is 1 o’clock, 8 a.m. is 2 o’clock and so on. Finally, during the FGD, the CDDs were asked about the feasibility of filling a diary of their activities on a daily basis.
Subsequently, two pictorial diary formats were developed. The first broke each day into hours e.g. 1–2, 2–3, 3–4 and would require the CDD to identify what activities they were involved in hour by hour and tick in the relevant boxes. This was called the ‘hour’ PD. The other format known as the ‘time’ diary, had only one column for each day and would require them to mark down the number of minutes spent doing an activity each time they carried it out in that day. Both formats were pre-tested in Mukono district chosen for its proximity to Kampala and because it was still a rural environment with CDDs involved in the NTD Programme. Four villages were chosen and seven, out of a potential eight, CDDs were at their homes. The CDDs were then randomly given the ‘hour’ or the ‘time’ PD and were talked through each of the pictures and asked to interpret what routine or NTD activity they saw. An explanation of how the PD works and the purpose of it were given and the CDD asked if they consented to record their time in the diary for one week. On follow-up the ‘hour’ PD (Additional file 1: Figure S1) was revealed to be easier to use, with fewer inconsistencies, than the ‘time’ PD, with a few minor amendments to the pictures highlighted. CDDs indicated they would be happy to fill the PD from between two months to one year, and all felt that the time it took to complete on a daily basis was not a burden.
Final adaptations were made to the ‘hour’ PD and an aide mémoire was developed (see Additional file 1). In the four selected study districts, the study teams set up the PDs with the 64 CDDs at their homes. During the initial visit CDDs were given an explanation to the purpose of the diary and how to complete it, including a practice for the previous day’s activities. The CDDs were followed-up one week later to check progress on PD entries by reviewing the activities recorded hour by hour for each day. Issues encountered, for example, not remembering exactly where a certain activity be marked, were discussed in detail and all the pictures reviewed to ensure the CDD fully understood each one. The CDDs were visited again two weeks after the previous visit where accuracy in PD completion was reviewed and a repetition of the pictures and how to fill the PD were given if there was insufficient clarity. CDDs were subsequently visited every two weeks, during which PDs were reviewed and semi-structured interviews were held after each NTD activity had been completed.
CDD semi-structured interviews and focus group discussions
The second and third components explored the CDDs participation in the NTD Programme through semi-structured interviews and FGDs. The interviews explored a series of background questions pertaining to socio-demographic characteristics, past and present involvement in CDD activities for NTDs and involvement in other health interventions. A set of pre-defined closed and open-ended questions were asked after each NTD Programme activity e.g. training, collecting drugs and mass drug administration (MDA) of PC drugs. These interviews focused on the CDDs role in and experience of the activity. Once all the NTD Programme activities were complete self-reported performance, motivation and an attitude scale were assessed through a final interview. Motivation in addition to overall experiences of and perceived responsibilities within an NTD Programme were explored further in an FGD held in each sub-county with between six and eight CDD participants.
Interview schedules for the semi-structured interviews and topic guides for FGD were developed. The tools were translated and back-translated, then piloted and adapted accordingly. Interviews and FGD were conducted in local language by the trained, local research assistants. The interviews were held at each CDD residence alongside the PDs. The FGD were conducted in convenient public places identified by the CDDs. All sessions were recorded, transcribed, translated, and back translated to ensure accuracy and quality.
Post-MDA drug coverage survey
Effectiveness of NTD Programmes is measured by performance indicators, the main being coverage of at-risk and eligible populations [9, 25]. Reported treatment coverage originates from each CDDs treatment register. CDD treatment registers were retrospectively assessed for accuracy of total treated for each drug against two recounts of CDD entered numbers, using a calculator by the study investigators. In all cases the total numbers recorded by the CDDs were inaccurate by more than 12 % with the majority being overestimations. Validated treatment coverage was obtained by including the study villages in a concurrently run independent post-MDA survey. The methodology for this fourth component, a multistage cluster sample coverage survey, is described elsewhere .
Data management and analysis
CDD time data from the PDs were double-entered into a customised Microsoft Excel database (Microsoft Corp. Seattle, WA, USA). Mean number of minutes were calculated for all routine daily activities and for each NTD Programme activity. Minutes were then converted into hours. Working days were based on an eight-hour day and working weeks were based on five working days i.e. 40 h were in a working week. To calculate the annual proportion of time spent on NTD Programme activities 246 working days was used.
As part of the national NTD Programme all CDDs received, regardless of whether they were part of the study, a financial stipend of 2,000 to 4,000 Ugandan shillings (USh) (US$0.89–1.79) when they attended training. No other remuneration was provided. From a societal perspective, CDDs incur an opportunity cost for participating in the NTD Programme as they are unable to perform their normal activities. Opportunity costs include the value forgone by the CDDs time not working in their shamba (gardens), doing casual labour or carrying out retail business. CDDs volunteer time was valued with a base case of 6,000 USh per day (US$2.70) which was the value of local casual labour wages . This was equivalent to an hourly rate of 750 USh (> US$0.34) based on an eight hour working day. The minimum wage on the Government of Uganda salary scale (4,193 USh or US$1.95 per day) and GNI per capita (7,931 USh or US$3.70 per day) were also used. All prices were adjusted for inflation over time using the GDP implicit price deflator and expressed in US$ 2010 prices (IMF, 2008 http://www.imf.org/external/pubs/ft/weo/2008/01/weodata/index.aspx).
Data were tested and approved for normality and all statistical analyses were carried out using STATA 11.2 (StatCorp LP, TX, USA). Time variables which were not non-normally distributed were transformed to the logarithmic scale. PD time data were analysed using a paired t-test to compare means of daily routine activities between days with, and without, NTD activities, during the same period. One-way analysis of variance (ANOVA) and simple linear regression were used to test for differences in the mean times between different levels of the independent variables, such as, administration level, distribution method, population and number of households served, length of tenure as a CDD, the number of drug deliveries (one delivery against two deliveries, one delivery against three deliveries, two deliveries against three deliveries), and socio-demographic variables. Mixed model linear regression was performed to determine if the total hours spent by the CDDs on the NTD Programme, whilst controlling for confounding and clustering at the parish level, were still significantly associated with those independent variables which had been identified during simple linear regression.
Treatment coverage data were entered into EPI Info (Version 6.04, USA CDC, Atlanta, GA). Mean coverage were calculated using the survey function in STATA which takes into account the clustered sampling design. Simple linear regression was also used to look at which variables were associated with treatment coverage. For both time and coverage, multiple linear regression included all variables that were found to be statistically associated with the outcome variable to adjust for the effects observed by these variables.
For the semi-structured and FGD quantitative and qualitative data, a coded scheme was developed by pre-defined topics together with themes emerging from the data using the qualitative data analyses software NVIVO (Version 9. QSR International, Doncaster, Australia).