Effectiveness of a mobile antiretroviral pharmacy and HIV care intervention on the continuum of HIV care in rural Uganda

Introduction Adherence to antiretroviral therapy (ART) is critical in order to achieve viral suppression, one of three UNAIDS targets set for achievement before 2020. One of the main barriers to adherence is the long distance between patient residences and healthcare facilities. We designed an intervention, Mobile Antiretroviral Therapy and HIV care (MAP-HC) in rural southwestern Uganda aimed to reduce travel distance and hypothesized that MAP-HC would improve ART adherence and rates of viral load suppression. Methods The study was conducted at two sites, Kitagata and Itojo Hospitals, and these are public health facilities located in rural southwestern Uganda. Patients who lived >5km from the hospital were provided the option to participate. For each hospital, we identified 4 health centres in the catchment area to serve as site for the mobile pharmacy. Each site was visited once a month to provide ART refills, adherence counseling and treatment of other illnesses. We measured patient waiting time, adherence and viral load suppression before and after the intervention. Results We conducted baseline assessment among 292 patients at the two hospitals. The mean waiting time at Kitagata Hospital changed from 4.48 hours before the intervention but increased to 4.76 hours after the intervention (p=0.13). The proportion of patients who missed an ART dose in the last 30 days dropped from 20% at baseline to 8.5% at 12 months after the intervention (p=0.009). The proportion of patients with detectable viral load from 19.9% to 7.4% after the intervention (p=0.001). Conclusions Our study has showed that a mobile pharmacy intervention in rural Uganda is feasible and resulted in improvement in adherence and viral load suppression. Although it did not reduce patient waiting time at the clinic, we recommend a scale-up of this intervention in rural areas where patients face challenges of transportation to the clinic.


70
Adherence to antiretroviral therapy (ART) is critical in order to achieve viral suppression 71 [1] or the third 90, one of three UNAIDS targets set for achievement before 2020. In 72 order to achieve viral load suppression, patients should take at least 95% of their 73 medications [2][3][4]. Although adherence levels in sub Saharan Africa are on average 74 higher than estimates in North America [5], patients in countries such as Uganda face 75 structural barriers to achieve high level adherence. 76 77 One of the main barriers to adherence is the long distance between one's place of 78 residence and healthcare facilities. [6,7] Specifically, patients seeking HIV care and 79 treatment travel longer distances compared to their HIV negative counterparts. [8] 80 Several studies have examined geographical factors as barriers and systematic analysis 81 has shown that travel distance is a barrier across the continuum of HIV care from testing 82 to treatment and retention in care [9]. Patients narrate how they struggle to raise the 83 monthly transportation fee to the clinics to collect their medicines [10,11]. Some patients 84 devised innovative ways such as pooling resources in order to secure their monthly pills, 85 with some patients making two-day arduous journeys to an HIV clinic. There is now a need for interventions that not only reduce distance to HIV treatment 100 clinics, but also have potential to decongest them. These may result in better adherence to Study setting 120 The study was conducted at two sites, Kitagata and Itojo Hospitals, and these are public 121 health facilities located in rural southwestern Uganda. They were selected because they 122 were among the first district-level ART clinics serving a rural population.  143 We conducted a baseline survey at the hospital to measure patient demographics, place of 144 residence, distance travelled to the clinic, pre-intervention adherence to ART and viral 145 load suppression. We also defined the catchment area of the clinic and mapped patients' 146 places of residence by parish and village. We used this information to identify the 147 clusters or zones where majority of patients resided. The catchment area of residence was 148 then divided into 4 zones based on clustering of the patient population. Within each zone, 149 we identified a county-or sub-county-(Health Centre III) or parish-level (Health Centre 150 II) healthcare facilities, which was closer in distance than the district-level hospital 151 (Health Centre IV) to the majority of patients in that zone, to serve as a distribution point 152 to dispense antiretroviral therapy. 153 154 In the identification phase, clinicians at the hospital interacted with patients and informed 155 them about our proposed intervention. At Kitagata Hospital, where we first began to 156 work, patients were asked to choose whether they would wish to receive their refills at 157 our proposed ART dispensing sites or continue receiving their medications at the 158 hospital. Patients were enthusiastic for our proposed intervention, but expressed a desire 159 for healthcare services in addition to ART refills. We then worked with hospital 5 160 administration to determine the feasibility and logistics of adding healthcare delivery 161 elements to our intervention. Ultimately, we revised our intervention from being ART 162 dispensing alone (MAP) to one that also provides basic healthcare services (MAP-HC), 163 which is described further below.

165
Patients were offered MAP-HC services if they had been taking ART for at least 6 166 months, were considered stable on treatment by the physician, and lived further than five 167 kilometers from the district hospital, a distance determined by the Ministry of Health as 168 being excessive when traveling for one's healthcare. [17] In making their decision, 169 patients were asked to consider other issues such as privacy and whether the MAP-HC 170 site would be convenient to them. 171 172 In the implementation phase, we reviewed the appointment dates for all patients that The typical layout for the clinic at Kitagata Hospital is shown in Figure 1 below.
The circled numbers in Figure 1 represent the positions where the research assistants 224 were positioned to collect the waiting time data.

226
Adherence was measured using self-report. Blood samples were drawn from the patients 227 and sent to a regional testing laboratory in Mbarara for viral load testing. VL were 228 measured using Amplicor system® from Roche. Results were relayed back to the health 229 care providers 4 weeks after sample collection.

231
Sampling, sample size and data analysis 232 We used consecutive sampling to collect baseline data. We aimed to interview 300 study 233 participants at baseline to determine the proportion of participants that were living 234 outside of 5km from the hospital. At the follow up for evaluation, we did not calculate 235 sample size. The number was determined by the resources available to conduct viral load 236 testing, as this was the primary outcome for the evaluation. 237 We summarized the baseline characteristics using 5km as the cut off. This is because the 238 Ministry of Health recommends that persons should reside within a radius of 5km from a 239 health facility [18]. We compared continuous baseline variables using non-parametric 240 tests and categorical outcomes using Chi square tests. We compared the pre-and post-241 intervention proportions of patients who were adherence or had viral load suppression 242 using Chi square test and a t-test to compare waiting time. Paired tests were not used 243 since the sampling process did not necessarily include the same patients before and after 244 the intervention. Technology. Study procedures were explained to the clients at the HIV clinic. Individual 250 written informed consent was obtained and all study participants signed a consent form 251 once they understood study procedures and accepted to participate. Participants were 7 252 made aware they were free to decline participation in the MAP but continue to receive 253 their medications at the hospital. Confidentiality was maintained by using number 254 identifications for the patients on study materials and questionnaires.

257
We conducted baseline assessment interviews among 292 patients at both Kitagata and 258 Itojo Hospital. Almost two thirds were women, with a median age of 37 years. Majority 259 lived more than 5km away from the hospital and the results are shown in Table 1. As 260 expected, those who lived nearer were more likely to walk compared to those who lived 261 further. Participants who lived further than 5 km also had significantly lowed median 262 monthly incomes compared to those who lived near the health facility (p=0.021). Those 263 who lived further than 5km also spent more time and money in travel to the hospital.  Participants had comparable 3 and 30 day self-report adherence regardless of whether 275 they lived within or outside 5km from the hospital (p=0.577) and these results are shown 276 in Table 2 below. However, participants who lived further than 5km were more likely to 277 report having ever missed a pill compared to those who lived near (p=0.013). The 8 278 participants that lived further than 5km were also more likely to report distance as a 279 barrier to their travel to the clinic.  The mean waiting time at the hospital was 4.48 hours before the intervention and 295 increased to 4.76 hours after the intervention, however this change was not statistically 296 significant (p=0.13) and these results are shown in Table 3. The proportion of patients 297 who missed an ART dose in the last 30 days dropped from 20% at baseline to 8.5% at 12 298 months after the intervention and this drop was statistically significant (p=0.009). 299 Similarly, the proportion of patients with detectable viral load significantly dropped from 300 19.9% to 7.4% after the intervention (p=0.001).  Our intervention targeted participants that lived more than 5km from the hospital and our 320 data indicate this was the right target population. Not only did they take more time to 321 travel to the hospital, they also paid more to reach the facility and reported more 322 difficulty trying to reach it. Yet, these participants who lived further away were more 323 likely to earn less per month compared to those who lived nearer, placing more burden on 324 their meager resources to meet their transportation fee to the hospital per month. Those 325 who lived further than 5km were more likely to reside in more rural remote 326 establishments. Although majority of the districts are predominantly rural, the hospital 327 was located in a semi-urban setting. This more rural population needed the intervention. Our study has some limitations. In our sampling, some of the patients involved in the pre-376 intervention assessment were not necessarily involved in the post intervention VL and 377 adherence measurement. In the post-intervention assessment, we randomly sampled MAP 378 patients for viral load measurement, regardless of whether they had been involved in the 379 baseline assessment or not. Although this sampling approach partly engages components 380 of convenience in the sampling, it is unlikely to have influenced the results because all 381 patients receiving care at a MAP site lived more than 5 km from the hospital and reported 382 distance as a barrier to adherence. The sampling approach is consistent with other 383 pseudo-experimental study designs that involve a pre-and post-evaluation comparison. 384 Lastly, we did not conduct a cost-effectiveness analysis for this intervention and future 385 evaluations should incorporate data collection to support this analysis.

387 388
In conclusion, our study has showed that a mobile pharmacy intervention in rural Uganda 389 is feasible and resulted in improvement in adherence and viral load suppression.