According to Haiti’s 2011 health facility census, its health care system includes approximately 800 health facilities distributed among 10 departments [11]. About half of these facilities are dispensaires (outpatient clinics typically staffed by those with a nursing degree equivalent), and about one quarter are centres sans lits (outpatient clinics often with limited beds for short observation and staffed by at least one physician). The remainder are centres avec lits (facilities with inpatient services and one or more physician) or hospitals (facilities with inpatient wards, and specialized services and staff). All types of facilities offer outpatient services. While most hospitals have laboratories on-site, laboratory capacity varies among other types of health facilities.

This nationally representative, cross-sectional survey of febrile outpatients used a stratified cluster-sampling design where the primary sampling units, or clusters, were facilities in each department (or stratum). From each stratum, a sample of health facilities was randomly selected in proportion to the number of total facilities in each department. The sampling frame of health facilities for this evaluation included 833 functional facilities with outpatient services, identified in the MSPP national health facility census report of 2011 [11]. All eligible febrile patients presenting for outpatient consultations at the sampled health facility during regular working hours on the survey day were eligible for enrolment. Reliable patient volume data was not available when the survey was planned; therefore, facilities were not selected in proportion to their utilization.

Criteria for patient enrolment in the study were: (1) experiencing fever, defined as a measured temperature ≥37.5 °C, or history of fever at any time during the previous 2 weeks, (2) being aged 18 years or older, or having a guardian present who was 18 years or older, and (3) providing informed consent. Patients were excluded from participation if they had signs of severe disease, specifically: impaired consciousness, prostration, intractable vomiting, convulsions, respiratory distress, shock, jaundice, or spontaneous bleeding [12].

A sample size of 533 enrolled subjects was calculated to detect a malaria prevalence of 50 % with a precision of ±5 %, taking into account a type 1 error of 0.05, a design effect of 1.25, and a non-participation rate of 10 %. The malaria prevalence estimate was chosen to provide the most conservative estimate of sample size, and was informed by a post-earthquake survey which demonstrated a malaria prevalence of 20.3 % among febrile patients [6], much higher than had been observed in previous health facility surveys [3–5] which raised uncertainty about the true burden of malaria in this population. Thirty facilities were sampled across the 10 departments in Haiti (Fig. 2) to ensure an appropriate minimum number of units to account for clustering at the facility-level. Fig. 2

Map of sampled health facilities in Haiti

The field work for the survey was conducted between December 7 and 14, 2012. Each health facility visit was unannounced and took place over two consecutive days. On the first day, the survey teams conducted a quantitative inventory of the facility’s physical and human resources, and performed interviews with all available and consenting health care providers in the outpatient department. For the facility inventory, the survey teams utilized a standardized data collection instrument to assess resources in outpatient clinics, laboratories, and pharmacies. A second day was dedicated to an outpatient survey where patients presenting to the facility’s outpatient clinic were screened for fever. Eligible, consenting, febrile patients were enrolled and finger-prick blood samples were collected to prepare thick and thin blood smears from each subject. These blood smears were reserved for later interpretation at the LNSP, and were considered the gold standard. In addition, several drops of blood from each participant were collected onto Whatman 903 protein saver cards for PCR to assess sub-microscopic parasitaemia. Each patient was evaluated and treated per facility standard of care by the provider and laboratory. After the clinical encounter, the provider was asked to complete a short form on each study participant describing the diagnosis, tests ordered, and treatments prescribed for febrile illness, including anti-malarial drugs. Occasionally, if the provider was unable to fill out the form, it was completed by the survey team based on the clinical note completed by the provider. Enrolled patients were administered a post-consultation questionnaire to assess illness history, mosquito net ownership, and malaria knowledge.

Laboratory diagnosis of malaria was performed at two levels: (1) by the facility laboratory using microscopy if operational, or by RDTs if available; and (2) by the LNSP reference laboratory, which analysed survey-prepared blood smears from enrolled patients as the “gold standard”. Facility laboratory results were documented by the survey team if the laboratory completed and recorded the results by the end of the clinic day. Gold-standard microscopy was performed according to a standard protocol [13]. Briefly, after fixation of the thin smear in methanol, thin and thick smears were stained with 10 % Giemsa for 5–10 min. The survey-prepared blood slides, were double read by two expert, microscopists at the reference laboratory who were blinded to the facility laboratory results. The examination of slides began in January 2013, approximately 1 month after preparation in the field. In order to declare a microscopy specimen free of Plasmodium falciparum infection, 300–500 thick-smear fields were examined.

Dried blood spots, on Whatman 903 protein saver cards, were analysed in duplicate by LNSP in June 2014 by polymerase chain reaction using photo-induced electron transfer fluorogenic genus-specific primers (PET-PCR). Sample preparation, storage, extraction, and assays were performed using protocols described previously [10, 14], but modified to account for World Health Organization (WHO) Evidence Review Group recommendations made in March 2014 for PCR analysis in low transmission settings [15]. Briefly, the amplification of Plasmodium genus (5′–3′, forward primer: GCTCTTTCTTGATTTCTTGGATG; reverse primer: FAM-aggcgcatagcgcctgg AGCAGGTTAAGATCTCGTTCG) was performed in a 30 µL reaction containing 2X TaqMan Environmental buffer 2.0 (Applied BioSystems, Grand Island, NY, USA), 400 nm each of forward and reverse primers. For each sample, duplicate PET-PCR reactions were run with 6 µL of DNA template used in the PCR reaction with the following cycling parameters: initialization at 95 °C for 15 min, followed by 45 cycles of denaturation at 95 °C for 15 s, annealing at 63 °C for 60 s. The correct fluorescence channel was selected for FAM dye and the cycle threshold (CT) values recorded at the end of annealing step. All assays were performed using Applied Biosystems ABI 7500 or StepOne thermocyclers (Life Technologies, Carlsbad, CA, USA). Cycle threshold values, which are continuous, semi-quantitative measurements of parasite load, were obtained for all specimens. Those with CT values of less than 40.0 were considered positive for malaria.

Positive samples were subjected to additional testing by nested 18S rRNA PCR (nPCR) to confirm the Plasmodium species, utilizing a method previously described by Singh et al. [16]. Briefly, both primary and secondary PCR reactions were performed using 2 µL of DNA template in 25 µL total volume containing 1X buffer, 2.5 mM MgCl₂, 200 µM dNTPs, 200 nM primers, and 1.25 units of Taq Polymerase (New England Biolabs, Ipswich, MA, USA). The products were analysed for appropriate band size on a 2 % agarose gel with a positive showing a single visible band.

Data were entered using Epi Info™ 7 (7.1.1.14) (CDC, Atlanta, GA, USA) and Microsoft Excel for Windows 7 (Microsoft, Redmond, WA, USA). Analysis was performed with SAS 9.3 (SAS Institute, Cary, NC, USA) and involved simple counts and percentages. National estimates and 95 % confidence intervals (CI) were calculated using patient-level or health facility-level weights as appropriate using the SAS PROC SURVEYFREQ procedure in order to account for the complex sampling design. Sensitivity and specificity analyses were conducted comparing facility diagnostic results to the survey-performed gold-standard microscopy. The map was created using ARCGIS 10.2.2 (Esri, Redlands, CA, USA) using spatial data of health facilities courtesy of the Haitian Institute of Statistics and Informatics (Institut Haïtien de Statistiques et d’Informatiques) and National Center of Geo-spatial Information Systems (Centre National de l’Information Géo-Spatiale).

For the purposes of this survey, health facilities were defined as having the capacity to diagnose malaria if they fulfilled the following criteria: (1) a facility laboratory completed a malaria test on the survey day for at least one enrolled patient using either an RDT or a microscopy examination of a blood smear, or (2) the survey inventory results determined that there was on-site capacity to perform either microscopy (functional microscope, reliable electricity, staining reagents, and a trained laboratory technician) or RDTs (current stock of any one of the three MSPP-approved RDT brands and at least one employee trained in performing RDTs). Reliable electricity was defined as having electricity for four or more hours a day, 5 days a week or having at least one generator in the hospital.

“Treatment according to guidelines” referred to directives set in the MSPP 2012 malaria treatment policy and was defined as: (1) prescription of the correct dose of chloroquine and primaquine to those with a positive malaria test result, and (2) no prescription of anti-malarials to those with a negative test result [1]. For treatment doses, the total adult regimen is 1500 mg of chloroquine base (administered over 3 days, with four tablets containing 150 mg each given on day one, and three tablets given on days two and three) and a single dose of 0.75 mg/kg primaquine (maximum adult dosage is 45 mg) on the first day of treatment. Paediatric regimens are adjusted according to age and weight [1]. The MSPP/PNCM supplies tablets containing 150 mg of chloroquine base and 7.5 mg tablets of primaquine.

The survey protocol and questionnaire were reviewed and approved by the human subjects review boards of the MSPP (reference number 1112-30) and the US Centers for Disease Control and Prevention (Atlanta, GA, USA). Written consent was obtained from each subject, or from guardians accompanying subjects younger than 18 years; children 7–17 years old provided written assent to participate.

Figure 2 shows a map of the thirty health facilities sampled proportionally across all ten Departments; of these 19 offered outpatient services only (14 were dispensaires, and five were centres sans lits); 11 facilities offered both outpatient and inpatient services (five were centres avec lits, and six were hôpitaux). Figure 3 shows the flow chart for participant enrolment. Four hundred and fifty-nine patients presenting to the sampled outpatient clinics on the day of the survey were screened for inclusion. Two hundred and fifty-seven (56 %) patients had either an axillary temperature ≥37.5 °C at evaluation or a history of fever. Of these, 64 were ineligible because they were below the age of consent and did not have a guardian present (n = 9, 4 %) or because they had symptoms consistent with severe illness (n = 55, 21 %). Of the 193 eligible patients, 40 (16 %) patients refused, resulting in 153 (79 %) enrolled patients.Fig. 3

Flow chart of patient enrolment

One hundred and fifteen providers were interviewed, of whom 102 provided direct patient care and were included in our analysis. Providers reported a mean 4.3 years of medical training (range 0–12, median of 3), and 61 (60 %) providers were female. Fifty (49 %) providers had received in-service training in the case management of malaria or severe malaria, and 16 (14 %) had this training in 2012, after the malaria treatment guidelines had been revised.

Malaria diagnostic capacity of health facilities is shown in Table 2. Eleven (37 %) facilities surveyed had diagnostic capacity on the basis of having completed malaria testing of at least one patient. According to inventory results, eight (27 %) and two (7 %) facilities had the materials and personnel necessary to conduct either microscopy or RDT, respectively. Eighteen (60 %) facilities reported having electricity at least 4 h a day, 5 days a week; six (20 %) facilities had neither electricity nor a generator. Adequate supplies and equipment for microscopy were present in 11 (37 %) facilities. Three (10 %) facilities had at least one brand of RDT in stock (≥1 test) on the day of the survey. Of these, two had an approved brand and two had non-approved brands; one facility had both an approved and a non-approved brand in stock. Five (19 %) facilities previously had RDTs available, but were out of stock on the day of the survey. Among these facilities, three (60 %) used an approved brand, one used a non-approved brand, and one could not specify the brand. Thirteen facilities (43 %) had at least one provider or laboratory technician trained to perform RDTs. Altogether, 16 (53 %) of 30 facilities had the ability to diagnose malaria, and 91 (59 %) enrolled patients were seen at facilities with this capacity.Table 2

Facilities with malaria diagnostic capacity

Training on malaria microscopy, RDTs and current MSPP malaria guidelines was reported by the facilities, according to health worker cadre (health care provider or laboratory technician). Trained microscopy laboratory technicians were staffed at 16 (53 %) facilities (Table 2). For RDTs, a total of 13 (43 %) facilities had at least one staff (of any cadre) trained; seven facilities had at least one laboratory technician trained, nine facilities had at least one clinical health provider trained, and three facilities had both laboratory technicians and health providers trained on RDTs. Fifteen facilities (50 %) reported that personnel (of any cadre) were trained on the malaria treatment guidelines. Of these, 14 facilities had at least one clinical health provider trained; seven facilities had both providers and laboratory technicians trained on the new guidelines, and one facility had only a laboratory technician trained on the new guidelines.

Data on the ordering and completion of malaria tests are shown in Fig. 4. Malaria test ordering practices were assessed using the post-consultation form completed by health care providers. Among the 91 febrile patients seen in 16 facilities with malaria diagnostic capacity, 53 (58 %) had a malaria test ordered by a provider (45 [50 %] microscopy, and 16 [18 %] RDT). At 14 facilities without diagnostic capacity, nine (15 %) of 62 patients had an order written for a test to be done at an outside laboratory: all nine included an order for microscopy, and eight included an RDT order.Fig. 4

Flow chart for malaria testing

The reference laboratory examined 153 gold-standard blood smears prepared by the survey team, of which 13 could not be read due to poor staining or fixation, or due to degradation of the stain in the interval period. Of the remaining 140, none were positive for any Plasmodium species. Specimens from 33 patients had results available from both the gold-standard survey microscopy and the facility laboratory. However, since no ‘true positives’ were identified by the survey microscopy, it was not possible to calculate the sensitivity for facility-read blood smears. Specificity of the facility results was 70 % (95 % CI 51–84 %).

PET-PCR was performed at LNSP on dried blood spots from each of the 153 enrolled patients. One sample was positive with duplicate CT values of 39.0 and 39.1, which are just under the CT cutoff threshold for positivity of ≤40. Calibration assessments using the PET-PCR protocol described herein have determined that a CT value of 34 corresponds to approximately 100 parasites per microlitre, and therefore specimens with CT values between 34 and 40 are generally considered below the threshold of detection by traditional microscopy and most commercial RDTs (unpublished observations, E. Rogier, JW Barnwell, V. Udhayakumar). Nested PCR was conducted at the CDC-Atlanta malaria laboratory and confirmed that this specimen contained a single-species infection with P.falciparum. The patient from whom the sample had been obtained was a 28-year old male seen in a facility in Port-au-Prince, who had previously sought care elsewhere and may have had prior treatment. On the survey day, a blood smear and RDT were ordered; the RDT was positive, but the facility microscopy result was not available. The patient was treated with chloroquine alone on the survey day. The gold standard blood smear was negative.

Table 3 shows the national estimates calculated for key indicators. The national estimate of the prevalence of malaria among patients seeking care at health facilities, detected by PCR, was 0.5 % (95 % CI 0.0–1.7). Two criteria were used to classify patients as being managed by the health facilities according to 2012 national guidelines: (1) those with positive malaria test results (from the facility) and prescribed the correct dose of chloroquine and primaquine (n = 1), and (2) those with negative test results not prescribed any anti-malarial (n = 27). The national estimate for febrile patients being treated according to the 2012 guidelines is 16 % (95 % CI 0.0–39). Patients with malaria results available for clinical decision-making was estimated at 17 % (95 % CI 0.0–40).Table 3

National estimates of key indicators

Health facilities with diagnostic capacity for malaria were estimated at 56 % (95 % CI 36–77), and 45 % (95 % CI 26–65) of health facilities were estimated to have a provider trained to use RDTs.

Twenty-three percent (95 % CI 11–36) of health providers were trained on RDTs, and 16 % (95 % CI 9–23) of them were trained on malaria treatment guidelines in 2012.