Association Between Organophosphate-based IRS and Reduction of Malaria in Manicaland, Eastern Zimbabwe Ashley Thomas 7th MIM Conference April 17th Dakar, Senegal
Introduction: IRS and Manicaland PMI-led IRS in 4 districts in Manicaland Province Chimanimani Mutare Mutasa Nyagatare Bullet 1: Briefly describe IRS and its application in sub-Saharan Africa Bullet 2: six decades of spraying by GoZ, nearly 6 years of partnership with PMI through TA, direct IRS implementation began in Manicaland in 2013 Bullet 3: Selected for malaria burden
Known mosquito resistance IRS Insecticides District Insecticide used Known mosquito resistance 2011- 2013 2014 2015 Buhera PY DDT Unknown Chimanimani* OP Chipinge Makoni Mixed Mutare* PY, carbamate resistance (2013) Mutasa* Nyanga* PMI Districts: note that this was as the result of insec susceptibility testing and in consultation with the NMCP Makoni district started spraying with DDT, but moved to PY when the DDT was fully used. Following the completion of the PY, OP was used to finish spraying the district. The decision to switch insecticides was made in an effort to deplete all remaining stocks of insecticide. *PMI-supported districts
Objective To understand the association between the change to OPs for IRS and parasitologically- confirmed malaria cases by health facility catchment area in the four PMI-targeted districts and three comparison districts in Manicaland Province. However, all seven districts in the province changed insecticides to OP, DDT or a “Mixed” approach (OP, DDT, and PY) during the period under study. The study thus assesses the effects of changing insecticides without restricting the change to OPs only. The study team collected the total number of parasitologically-confirmed malaria cases per month from each health facility from July 2013 through June 2016 in all districts in Manicaland Province and weather variables by estimated health facility catchment area for the same period. Text above Figure 7 says “Note that data from July 2013 through September 2013 are excluded from the analysis as weather data for the three months preceding these data (needed for the final model which includes lagged weather data) were not available.”
Methodology Parameters Manicaland province only (7 districts) Health facilities for which GPS could be determined: 1,252 278 Health facilities had GPS + DHIS2 data available 2013 – 2016 Data Collected Insecticide used DHIS2 HF-level case data Meteorological data B1: 3 PMI districts + 4 GOZ (w/ GF funding) districts = 7 districts B2: GPS locations determined or given for each and catchment areas estimated using Voronoi-Thiessen polygonal method (1,219 (originally provided by MOHCC) – 3 (no longer active) + 61 (health facilities with DHIS2 data, but without geolocations) – 15 (health facilities whose geolocations could not be identified due to security issues) – 10 (health facilities that were initially thought to have DHIS2 data but no geolocation, but were, in fact, duplicates of previously provided health facilities) = 1,252) B3: V-T Analysis done on all HFs to restrict catchment area appropriately, but only 278 actually had data + GPS and were included in analyses B5:The number of people tested (all ages) and number of children under 5 years of age tested for malaria per health facility per month. The total number of people testing positive (all ages) and total number of children under 5 years of age testing positive for malaria per health facility per month. The percentage of those tested with malaria positive results for all ages and for children under 5 years of age, on average. B6: Humidity, Rainfall, Temp We compute these numbers per health facility per month rather than in total because some health facilities did not report data for all months, and thus reporting in total would potentially misrepresent changes over time due to missing data. We then assess the ratio of ‘the percentage of those tested with malaria positive results’ in the years 2015 and 2016 to 2014. This latter figure is intended to show the magnitude of change in ‘the percentage of those tested with malaria positive results’ over time (as compared to 2014 for the time period in question). We also calculated the average rainfall, temperature, and humidity per month for each of the treatment categories, and assessed the number of health facilities that did or did not report data for each month, by the data point that was or was not reported.
Methodology (Contd.) 2 Transmission Seasons: 4 treatment categories: Pyrethroid to Mixed (Makoni) Pyrethroid to DDT (Buhera) DDT to Organophosphate (Chipinge) Pyrethroid to Organophosphate (Chimanimani, Mutare, Mutasa, Nyanga) 2 Transmission Seasons: High transmission (January – June) Peak transmission (February - April In Makoni district, PYs were used until the 2015 campaign. During the 2015 campaign, the district used three different classes of insecticide: DDT, PY, and OP. Makoni serves as the main before/after comparator area for the other districts because this district retained PYs use the longest, and did not fully move away from PYs in 2015. The study team considered partial treatment in 2015 for Makoni, based on the percentage of non-pyrethroid insecticide used. The 2015 IRS campaign in Makoni district took place from October through 1December 2015. In Buhera district, PYs were used previously prior to the switch to DDT in 2014. This treatment category is distinct in order to separate the effectiveness of DDT from OPs. The 2015 IRS campaign in Buhera district took place from October through December 2015. In Chipinge district, since insecticide was switched from DDT to OPs in 2015, this treatment group separates the effectiveness of DDT from PYs. The 2015 IRS campaign in Chipinge district took place from October through December 2015. IRS treatment switched from PY to OP in 2014. This treatment group serves as the main intervention of interest for this analysis. The 2015 IRS campaign in Chimanimani, Mutare, Mutasa, and Nyanga took place from October through November 2015. All data were compiled for each year and each district
Methodology: Analytical Approach Case numbers computed per health facility per month Assess the ratio of ‘the percentage of those tested with malaria positive results’ in the years 2015 and 2016 to 2014 Difference-in-difference to compare changes over time across treatment groups B1: rather than in total because some health facilities did not report data for all months, and thus reporting in total would potentially misrepresent changes over time due to missing data. B2: This latter figure is intended to show the magnitude of change in ‘the percentage of those tested with malaria positive results’ over time (as compared to 2014 for the time period in question). B4: The study team then used a difference-in-difference approach (which compares data from treatment and control groups both before and after the treatment to determine the treatment effect) to assess if changes over time in areas that switched insecticides are different than the changes over time in areas that either did not switch or switched to a different class of insecticide. The primary outcome of these analyses is the effect of the change in insecticide class on the number of parasitologically-confirmed malaria cases presenting at health facilities for people of all ages, and the secondary outcome looks specifically at the number of parasitologically-confirmed malaria cases presenting at health facilities in children under the age of five years. The number of parasitologically-confirmed malaria cases is considered on a per health facility per month basis in the regression model. The study team used a fixed effect negative binomial regression to assess the difference-in-difference, with the fixed effect at the health facility level and the health facility specific catchment population as the exposure variable
Results Average Monthly Humidity Average Monthly Temperature Average Monthly Rainfall B1: Zimbabwe’s rainy season typically lasts from November through March. During the study period, average monthly humidity (Figure 4) followed a similar pattern to average monthly temperature (Figure 5), while rainfall patterns were slightly less regular (Figure 6). Temperatures began increasing slightly before the rainy season, while humidity increased in conjunction with rainfall levels. Of note were high average rainfall levels in December 2014. While differences in individual weather indicators were observed among all districts in Manicaland Province, the patterns of weather change were similar among the districts. B2: Over the three-year period, an average of 9% (25) of the 278 health facilities included in this investigation did not have data for number of cases tested using RDT (Figure 7). Over the same period, an average of 12% (33) of these same facilities did not have data for number of parasitologically-confirmed malaria cases. In August of 2013, nearly 50% (139) of health facilities were missing data on the number of parasitologically-confirmed malaria cases, with high levels of missing data also seen in July of 2013; levels of missing data decreased after that time.
high transmission period peak transmission period Results (Contd.) Proportion of confirmed cases out of suspected cases in health facilities (All Ages) January to June: high transmission period February to April: peak transmission period Treatment Category (District[s]) 2014* 2015 2016 PY to Mixed (Makoni) 53% 46% 23% 48% 40% 22% PY to DDT (Buhera) 37% 12% 34% 11% 10% DDT to OP (Chipinge) 44% 42% 41% 36% PY to OP (Chimanimani, Mutare, Mutasa, Nyanga) 60% 39% 35% 58% 32% *baseline comparison year
high transmission period peak transmission period Results (Contd.) Proportion of confirmed cases out of suspected cases in health facilities (Children under 5) January to June: high transmission period February to April: peak transmission period Treatment Category (District[s]) 2014* 2015 2016 PY to Mixed (Makoni) 34% 28% 13% 40% 33% PY to DDT (Buhera) 20% 4% 4 % 21% DDT to OP (Chipinge) 27% 36% 31% PY to OP (Chimanimani, Mutare, Mutasa, Nyanga) 50% 25% 52% *baseline comparison year
Discussion PYs to OPs: substantial reduction in proportion of cases in PMI-supported districts in high-transmission season: from 60% in 2014 to 39% in 2015 HFs in OP districts had more cases prior to switching than Makoni, but also a greater decrease in cases post-switch PYs to DDT: reduced cases in Buhera in high- transmission seasons: from 37% in 2014 to 12% in 2015 B2: Because Makoni does not serve as a ‘true’ comparison (i.e., no change in treatment over the timeframe), the results are based on a modeled counterfactual as if there had been no switch from PYs. Makoni only serves partially as a comparator for this counterfactual; in this sense, these analyses are not true ‘difference-in-difference’ estimates, but reflect a before-after assessment with only partial assessment of difference-in-difference based on the partial spray in Makoni. B3: Previous studies demonstrated that changing from PYs to DDT substantially reduced the number of malaria cases in South Africa during the 2000-2001 transmission season (Coetzee et al, 2013). This was primarily due to PY resistance in the An. funestus population which is susceptible to DDT. Anopheles funestus is generally resistant to PY insecticides in Southern Africa, but susceptible to DDT in parts of Manicaland Province in eastern Zimbabwe (Sande et al, 2015).
Limitations & Recommendations Missing data IRS/ITN coverage data Entomological data Gross population factors Disaggregation B1: missing data will always confound data – either inflating or deflating effect, and missing data was a big problem B2: no IRS/ITN coverage data included, but should be because low coverage rates will impact the herd immunity conferred + confounding factor of IRS & ITN in same location for compound effect B3: scarcity of mosquitoes; updated susceptibility testing necessary especially in non-PMI regions B4: socioeconomic status, water access, living conditions, migration B5: lumping districts into the same categories can confound effects, especially in a place like Makoni where there was a complex change in insec
Acknowledgements GOZ Ministry of Health and Child Care GOZ National Malaria Control Program PMI/Washington: Allison Belemvire, Laura Norris PMI/Zimbabwe: Christie Billingsley, Regis Magauzi, Gail Stennies PMI AIRS: Brad Lucas, Dereje Dengela, Aklilu Seyoum, Ben Johns, Pearl Zhang, Godfrey Tinarwo, Shadreck Sande, Hieronymo Masendu
Thank you!