Intermittent screening and treatment with dihydroartemisinin-piperaquine for the prevention of malaria in pregnancy: implementation feasibility in a routine healthcare system setting in western Kenya

The study was conducted in 12 health facilities in four administrative units (subcounties) in Kisumu County in western Kenya: Muhoroni, Nyakach, Kisumu West and Seme. The study had two phases, an implementation phase which ran for ten months between September 2014 and June 2015, and an evaluation initiated mid-way from February to June 2015. The overall aims of this mixed methods study were to assess implementation feasibility using the framework constructs of Bowen et al. [24]: (1) acceptability, (2) demand, (3) implementation, (4) practicality, (5) integration, (6) adaptation and (7) expansion. The constructs of acceptability and demand are addressed in the qualitative study (Hoyt, pers. commun.), while implementation, practicality and integration are addressed directly by both quantitative and qualitative studies (Additional file 1: Table S1). The quantitative study addresses implementation through systems effectiveness analyses including both the overall implementation of each of the interventions, and each component of the interventions. Integration is addressed through exploration of the provision of RDT testing and provision of DP (to RDT test positive cases) by nurses in ANC, whereas in routine practice RDTs are performed by laboratory staff in laboratories, and artemisinin-based combination therapy for treatment of malaria given by clinical doctors. Other constructs are indirectly addressed in the quantitative study including practicality and adaptation and further explored in the qualitative study.

Methods used in the quantitative evaluation reported here comprised health facility audits, observations in ANC clinics, and exit interviews with pregnant women leaving ANC clinics. Health facilities were purposefully selected based on the feasibility of achieving the required sample size within a 3-month period using available ANC records. Three county or subcounty hospitals (level 4 facilities) and three health centres (level 3 facilities) were included in each study arm. The health facilities were first split into two similar groups using health management information systems (HMIS) data on factors likely to have an effect on study outcomes: (1) ratio of ANC registrants per professional staff in ANC, and (2) coverage of one and two doses of IPTp-SP in the previous 12 months. Six matched health facilities were then assigned to one of two arms, ISTp-DP (i.e. the intervention arm) or the current IPTp-SP policy (i.e. the control arm).

The implementation phase began with sensitization and training of Ministry of Health (MOH) staff in the study sites. Members of the County Health Management Team were first sensitized and provided with an overview of the study design. Health facility staff involved in delivering ANC services, including nurses, clinical officers, laboratory staff, data clerks, and pharmacists, were then trained at a central location in each subcounty between August and September 2014. Health staff received a refresher course on Kenyan malaria in pregnancy and related policies, including IPTp-SP, and staff in the ISTp-DP arm received additional training on the use of RDTs, quality control of drugs and RDTs, stock control and storage of RDTs and administration of DP. On completion of training, health staff in the ISTp-DP facilities were provided with standard operating procedures (SOPs) on managing malaria cases, performing RDTs, and administration of DP. DP was supplied to the facilities in the ISTp arm for the duration of the study, and buffer supplies of SP and RDTs of the same brands used by the MOH [namely histidine–rich protein-2 (HRP-2)/Plasmodium lactate dehydrogenase (pLDH) combination RDTs (First Response® Malaria pLDH/HRP2 Combo Test, Premier Medical Corporation Ltd, India)] provided in the event of stockouts in routine MOH supplies. Health facility stocks of anti-malarial drugs (DP and SP) and RDT kits were monitored and replenished when needed to avoid stock-outs, however gloves were not provided. The supply chain, therefore, does not reflect operational realities however other external interference with the study health facilities by study staff such as additional guidance or supervision was kept to a minimum. Respective County and subcounty Health Management Teams were encouraged to continue routine supervision of health facilities and ANC services throughout the study.

Pregnant women attending study health facilities for scheduled focused ANC visits during their second and third trimester received the standard ANC package of care, and either IPTp-SP or ISTp-DP depending on study arm. Pregnant women attending ANC in the IPTp-SP arm received three tablets of quality-assured SP (supplied by Durbin PLC, Middlesex, UK, 500 mg sulfadoxine/25 mg pyrimethamine /tablet) and those in the IPTp-DP arm received a standard 3-day course with DP (Eurartesim®, Sigma Tau, Pomezia, Italy, 40 mg dihydroartemisinin /320 mg piperaquine /tablet). According to national policy [25], women receiving IPTp-SP receive a single dose of three tablets of SP under directly observed therapy (DOT). Women infected with HIV on daily cotrimoxazole are not given IPTp-SP. Since 2010, all health facilities switched to low dose folic acid, with the recommendation for any high dose folic acid to be taken 14 days after a dose of SP [MOH, unpublished]. Women in the ISTp-DP arm who tested RDT positive were prescribed DP based on body weight (24 to 35.9 kg—2 tablets; 36 to 74.9 kg—3 tablets; > 75 kg—4 tablets). Health staff were trained to administer the first dose of DP under DOT and provide the remaining two doses and instructions for how to take them at home. Women with a known HIV positive status were not tested with an RDT unless they were symptomatic for malaria, and if test-positive they were treated following national guidelines for malaria treatment during pregnancy.

Data were double entered using Epi info. All analyses were adjusted for survey design and clustering of health facilities using STATA 12.1 (StataCorp). Two systems effectiveness analyses were performed for each study intervention and for RDT testing proficiency: (1) cumulative effectiveness and (2) intermediate effectiveness of each process in the delivery chain. For IPTp, intermediate processes assessed comprised: (1) 2nd or 3rd trimester visit; (2) given any SP; (3) given 3 tablets of SP; (4) given SP by DOT; or where was not given IPTp-SP by DOT (5) left the facility with 3 tablets of SP and (6) knows how they would be taken; and (7) a composite endpoint combining 4) and 6). IST intermediate processes were: (1) 2nd or 3rd trimester visit; (2) given an RDT test; (3) RDT negative and not given an anti-malarial; (4) RDT positive and given any DP; (5) given correct dose of DP; (6) given first dose of DP by DOT; (7) Correct number of tablets given; (8) Health worker describes how to take remaining doses; (9) Woman able to repeat instructions; and (10) a composite endpoint combining (3) and (9) [26]. Analyses were restricted to women attending ANC in the second and third trimester, and eligible for either intervention. Analysis was limited to asymptomatic HIV-uninfected participants determined as those who did not receive any antiretroviral during that visit, participants who did not receive any anti-malarials for treatment and for whom information on drugs prescribed during the visit was available on exit. RDT testing proficiency analysis was assessed for 16 steps laid out in the study SOP (listed in Table 6). Univariate analyses were conducted of potential predictors of receipt of: (i) IPTp-SP by DOT vs no DOT; and (ii) ISTp-DP full vs incomplete regimen. Logistic regression models adjusted for the survey design were used, and adjusted Wald tests were applied to test for associations between predictors and outcomes. Potential predictors significant at the 10% level (p ≤ 0.1) in the univariate analyses were included in multivariate logistic models to determine which factors were associated with the outcomes after accounting for other potential predictors, and predictors were considered significant at < 0.05 in the final model [26]. A list of potential predictors and the corresponding data source are summarized in Table S2. Principle components analysis (PCA) was used to construct a wealth index in order to assess the effect of socio-economic status (SES) [27].

The study was approved by the ethics committees of the Kenya Medical Research Institute’s (KEMRI) Scientific and Ethics Review Unit, Kenya; the Liverpool School of Tropical Medicine, UK; and the London School of Hygiene and Tropical Medicine, UK; the US Centers for Disease Control and Prevention, Atlanta, Georgia, USA relied on KEMRI. Health workers gave signed informed consent for structured questionnaires and for observations. Pregnant women gave signed informed consent for observations and exit interviews at ANC registration. In Kenya, pregnant women aged 15–17 years are considered emancipated minors and were consented directly. Permission from the County Health Management Teams and health facility in-charges was sought prior to initiating the study.

Using health management information systems (HMIS) data from 2014, health facilities in each study arm were reasonably well matched with respect to: total catchment population (94,112 and 88,679) and number of ANC visits (2938 and 2454) in the IPTp and ISTp arms, respectively. Health facilities in the ISTp arm had higher numbers of ANC staff (42 versus 21 staff); no data were reported for facility #3 (Table 1). The catchment population for county hospitals ranged from approximately 21,000–23,000, for subcounty hospitals from approximately 10,000–18,000, and health centres from approximately 9000–21,000. Client to staff ratios ranged from 46 to 543, with both extremes found at health centre level. All health facilities had functioning microscopes and all but one health centre had a resident microscopist. All health facilities in the ISTp arm reported using RDTs for malaria diagnosis in the 6 months before the survey with one reporting routine facility stockout of RDTs on the day of the facility audit which was replenished by the study team. One health facility in the ISTp arm reported a stockout of study DP in the 6 months before the survey.

A total of 112 health providers were interviewed; the mean age was 37 years and 68% were female (Table 2). Approximately one fifth lived in the local catchment area of the health facility where they worked (19%) and approximately one quarter were from the same subcounty as the health facility (28%). Approximately one third of health providers interviewed were registered nurses (35%), 13% were laboratory technicians, 10% were trained HIV counsellors and 8% were clinical officers. Years worked in the health facility ranged from 1 to 5 years (mean 3 years). Only some staff had received training relating to malaria in pregnancy either at a workshop or from a colleague within the last five years, ranging from 20–100% for malaria in pregnancy, 30–94% for case management, 27–50% for RDTs (94% of laboratory staff and 27% of nurses) and 18–78% for ISTp. In general, nurses had greater access to job aids and guidelines than laboratory staff.

A total of 777 pregnant women were observed during an ANC visit on the day of the survey and interviewed as they exited the health facility, 389 in the IPTp arm and 388 in the ISTp arm (Table 3). There were approximately equal numbers of women in the second and third trimester in both arms: 51% and 49%, respectively, in the IPTp arm (mean gestation 6.4 months), and 49% and 51%, respectively, in the ISTp arm (mean gestation 6.5 months). Similarly, there was an approximately equal distribution of women attending for their first, second, third, fourth or more (4+) ANC visit in both arms, with slightly more women attending for their first visit in the IPTp arm (29%) and for their second visit in the ISTp arm (27%) compared to the other visits (not statistically significant). Over 40% of women in each arm were multigravidae aged between 20 to 24 years. Most women were married, 78% and 82% in the IPTp and ISTp arms respectively, and of Luo ethnicity, 88% and 89%, respectively. Most women had primary level education (55% and 53%, respectively), and approximately one fifth had no education (30% and 24%, respectively). Over 70% of women had at least one child aged under 5 years, and 11–12% had had a child who died.

Of 336 women included in the analysis, 75% (239/319) received a dose of SP and all received the correct number of tablets, i.e. 3 tablets, however, only 76% (179/237) of these received the dose by DOT (Table 4). This translates into a cumulative effectiveness for receiving IPTp-SP by DOT of 57%. The cumulative effectiveness for the composite endpoint (doses given with or without DOT) was 53.6%.

Of 347 women included in the analysis, 90% (289/320) were tested with a malaria RDT of whom 11% (32/289) tested positive (Table 5). Of women who tested positive, 91% (29/32) were given DP of whom 86% (25/29) were given the correct dose based on the woman’s weight, with a cumulative effectiveness of 71%. Of the women receiving the correct dose, only 44% (11/25) received the initial dose of DP by DOT and only 18% (2/11) of these received instructions from the health provider on how to take the remaining doses. Neither of these women were able to repeat these dosing instructions on exit resulting in 0% cumulative effectiveness for women who tested positive. The overall cumulative effectiveness for the composite endpoint was 60.8%. The proportion of women screened with an RDT was relatively consistent across visit number (1–4+ visits), with slightly fewer women screened with an RDT at a first visit as these women have blood slides taken as part of ANC profile testing (20% [11.4, 33.9]) compared to 4+ visits (27% [17.7, 38.1]). More women attending their first ANC visit tested positive for malaria (4% [1.2, 9.6]) than women attending for 4+ visits (2% [0.5, 5.9]) (not significant).

Of the 289 women screened with an RDT in the ISTp arm, a higher proportion of tests were performed in ANC (95%, 274/289) than in the lab (Table 6). Cumulative effectiveness for adherence to each RDT proficiency step in ANC was 12% (35/289) if step #5, selecting the fourth finger to prick for blood (65%, 82/126), and step #7, labelling the RDT with the patient’s name (only 22%, 26/117 were labelled), are omitted. The omission of these two steps is based on the assumptions that: i) the finger used to collect blood does not affect the test result; ii) in this setting the woman is given the result during the consultation by the nurse during her ANC visit thereby removing the need to label the RDT. Two further steps with low adherence to the SOP in ANC were the wearing of gloves (47%, 129/274), which is important for health worker safety, and waiting time for the test result (30%, 35/117), which may affect accuracy of test results. Cumulative effectiveness for steps 1–15 of tests performed in the lab was higher than in ANC (36% vs 12% without steps 5 and 7, respectively), the steps with lowest adherence being wearing gloves (73%, 11/15), selecting the fourth finger (36%, 4/11), and waiting time for the test result (55%, 6/11). Steps where ANC nurses performed better than lab staff were selecting the fourth finger (percentage score difference of+ 29) and using guidelines (+ 11).

Predictors of women receiving SP by DOT in the univariate analysis were marital status, sick visit (as reported by or agreed with the woman), paid for a lab test, and being given drugs at the pharmacy (Table 7). In the final multivariate model, only payment for a lab test was independently associated with receipt of a dose of SP by DOT (AOR6.43, 95% CI 2.07,19.98).

Predictors of women being screened with an RDT in the univariate analysis were: marital status, gravidity, education, SES, urban or rural location, consultation time, consultation duration, sick visit (reported by the woman), the presence of a resident microscopist at the facility, and delivery of prevention-of-mother-to-child-(HIV) transmission (PMTCT) services (Table 8). In the final multivariate model, married women (AOR 0.77, 95% CI0.47,1.26), and women living in rural areas (AOR 0.78, 95% CI 0.02,0.27) were less likely to receive ISTp-DP while women making a sick visit (AOR 1.78, 95% CI 1.31,2.41) and being tested in a health facility with a resident microscopist (AOR 3.75, 95% CI 1.31,2.40) were more likely to receive ISTp-DP.