Study area and setting
The study was conducted in Luuka district which is one of the 10 districts making up the Busoga sub-region in East Central Uganda.The east central region where Luuka district is located has an under five mortality ranging between 73 to 90 per 1000 live births[26]. The district is made up of 7 sub-counties(Ikumbya,Bukooma,Bulongo,Irongo,Nawampiti,Waibuga and Bukanga) and 1 town council(Luuka town council) as shown in Figure 1.
Presently, the districthas no hospital, has 1 Health Centre(HC) IV, 6 HC IIIs and 16 HC IIs. Only 61 out of the 340 villages (18%) have community health workers (CHWs) locally known as village health teams (VHTs). Up to 49% of the total population do not live within the recommended 5 km of a health facility[26]. However, all villages have drug shops where preliminary self-treatment can be sought before ahealth worker is seen. In Luuka district,drug shops are under a drug shop association which has a committee headed by a chairman. Drug sellers meet every month under the auspices of the drug shop association to deliberate on matters concerning licences and working conditions affecting their daily operations.
The Intervention
Lived experienceson peer supervision were captured from peer supervisors and drug sellers after the peer supervision model was tested for effectiveness on appropriate treatment among drug sellers in Luuka district. In Luuka district, the peer supervisors were chosen by registered drug sellersthrough a democratic process by show of handsusing a criteria based on age, experience and academic qualifications.Among the registered drug sellers in every sub-county,a peer supervisor was chosen. In total, eight peer supervisors were selected and instructed to report to the District Drug Inspector (DDI)who derives the inspection mandate from the District Health Officer (DHO). The DDI continued the traditional inspection role as per statutory mandate during the course of the intervention.
The peer supervisors underwent a three day training where emphasis on adhering to standard treatment guidelines was stressed[27]. Other topics handled during the training sessions included: adhering to ethical standards of supervision; reporting drug sellers who do not adhere to treatment guidelines to the DDI or DHO;and mediating disputes such as those that may arise between drug sellers, peer supervisors and district or central government inspectors. Peer supervisors were given an allowance of 80,000 Uganda shillings equivalent to twenty two united states dollars ($ 22) at the end of each month for a period of one year to cater for lunch, transport and other incidentals during the intervention period. The assumption was that each peer supervisor would visit all drug sellers within their designated sub-county once every month, and that supervision visits would not exceed one day.
Peer supervisors were equiped with supervision checklistswhose purpose was to make monthly summaries of appropriate treatment of children under five years from drug shop sick child registers.Peer supervisors also checked whether sick child registers were being filled by drug sellers. This way, they were able to assess whether drug sellers were adhering to the standard treatment guidelines. The peer supervisors were instructed to adhere to the highest form of privacy, professionalism, integrity, continuous learning and empathy. In the peer supervision model, we worked with an active district drug shop association where many drug sellers met every month particularly to attend continuous medical education-related seminars.The peer supervision model therefore strengthened self-supervision that is currently prescribed by policy guidelines.
Study design,participants and sampling procedures
We conducted a qualitative study based on in-depth interviews(IDIs) and focus group discussions(FGDs).Participants were selected based on: 1) being a statutorily licensed drug seller; and 2) being a democratically elected peer supervisor involved in the supervision process. IDIs were conducted with peer supervisors while FGDs were conducted with drug sellers.Each FGD was composed of nine nursing assistants and either a nurse or midwife.This is because the number of nurses and midwives was too small to make a group of 8 to 12people which is desirable for a FGD[28].As such, the five members (nurses and midwives) were placed in each of the five focus groups with nursing assistants which was aimed at creating homogeneity across groups.The IDIs and FGDs were conducted from the sub-county headquarters in the seven sub-counties. Interviews held within the town council were conducted from the town council main hall. All places were devoid of noise.Registered drug sellers and peer supervisors who were involved in the peer supervision exercise were purposively sampled and mobilised through the DDI. The DDI got official communication concerning the interviews from the DHO who was informed in writing by the lead researcher (AB).
Data collection
We developed an interview guide for IDIs and FGDs with the aim of capturing experiencess of peer supervisors and drug sellers on the peer supervision intervention. Included in the interview guides were key questions aimed at understanding how drug sellers felt about being supervised by peers and how peers felt supervising colleagues.The interview guide for IDIs explored how peer supervisors felt supervising fellow drug sellers, what they wanted improved in peer supervision ,how best peer supervision and the drug shop association could be merged and which community members can be added to and benefit from peer supervision. The interview guide for FGDs assessed how drug sellers interacted with peer supervisors, to what extent the drug shop association amalgamated with peer supervision and what drug sellers wanted added to peer supervision to improve its smooth running.During indepth interviews, at individual level, peer supervisors were probed until the team lead(AB) felt that additional data collected was redundant of data already collected[29].At group level, in both the IDIs and FGDs, saturation was achieved when additional interviews yielded no new information[30]. Saturation occurred with the sixth IDI and fourth FGD. However, two additional IDIs and one additional FGD were done to ensure that all information was captured and no new information was left out.Whereas the interview guide was in English, during interviews, moderators experienced in conducting IDIs and FGDs conversant with English and Lusoga-the most widely spoken language in Luuka district were used as interviewers. Interview questions were asked and recorded in the local dialect(Lusoga).Total time taken for each IDI ranged between 45 to 55 minutes, while time taken for each FGD ranged between 50 to 70 minutes. A note taker assisted the moderator in taking notes and digital audio recording conversations.All peer supervisors and registered drug sellers present at the time interviews were held participated.
Data management and analysis
All audio recorded interviews and group discussions were translated from Lusoga to English as they were transcribed verbatim into Microsoft word documents. All transcripts were verified to be a true reflection of what transpired in the IDIs and FGDs before analysis.The translated scripts were then uploaded into Atlas ti.7 qualitative data management software (ATLAS.ti GmbH, Berlin) for analysis. After four IDIs and two FGDs were independently coded by two researchers(AB and MM),a discussion with intention to agree on how to resolve differences as well as improve validity and reliability of developed codes ensued until consensus was reached. The agreed upon codes were then used to code the rest of the transcripts. A similar procedure of agreeing upon sub-categories,categories and themes was followedto address issues of rigour. In both instances,we made sure that we achieved inductive thematic saturation, that is to say,new codes and themes were redundant of codes and themes already constructed. Graneheim and Lundman’s framework for capturing both latent and manifest content in transcripts was used for this content thematic analysis approach[31].During analysis, sub-categories were derived from codes. Codes were derived from unifying meaning units abstracted from condensed meaning units.Abstract level categories were then formed out of sub-categories with similar meaning[31]. In the final analysis,a larger narrative of themes from condensed categorieswere developed. To address rigour and trustworthiness of the qualitative data collected, a dissemination meeting was held with participants after the interviews were done to confirm whether it was reflective of what was discussed.