Extent of inappropriate prescription of artemisinin and anti-malarial injections to febrile outpatients, a cross-sectional analytic survey in the Greater Accra region, Ghana

This is a health facility-based, cross-sectional analytic study conducted in the Greater Accra region (GAR). There are 16 administrative units in the region. Six health facilities in three municipalities (Ga South, La Dade-Kotopon, La-Nkwantanang Madina) of the region were involved. The three municipalities were selected from a list of 9 administrative units with the proportion of confirmed malaria cases not greater than 24.9% per GAR’s Health Management Information System (HMIS) between January and December, 2014.

For each municipality, the two facilities that contributed the most to total OPD visits in the public health delivery sector were selected. These facilities had no recorded stock-outs of all three recommended ACT nor inability to perform malaria tests within 6 months prior to data collection. They included Kekele Polyclinic and Pentecost Hospital from the La-Nkwantanang Madina municipality; La General Hospital and Police Hospital from the La Dade-Kotopon Municipality and Ga South Municipal Hospital and Ngleshie Amamfrom Health Centre from the Ga South Municipality.

Following the recruitment of field staff and pre-testing of data collection tools at the Ashaiman Polyclinic, located in Ashaiman Municipality, a health facility survey within the selected study municipalities was conducted between October and December 2015. Excluding the records of review patients, afebrile patients, pregnant women and febrile patients with danger signs, all other records of febrile OPD patients who had visited these health facilities were obtained from the records department. OPD folders of patients who had axillary temperature (T), T ≥ 37.5 °C or fever recorded in clinical notes were classified as febrile.

The improved data extraction tools (following the pre-test) were used to retrieve information on patient age, gender and ownership of a health insurance card, presenting symptoms, laboratory investigations conducted, laboratory results, diagnoses, and prescribed drugs. Malaria diagnoses documented in OPD records of patients encompassed both confirmed and suspected malaria cases. Other diagnoses may not have been based strictly on standard International Classification of Diseases (ICD) codes.

Interviews were held with the prescribers who saw these patients in order to obtain information on prescriber demographic characteristics, professional category, work experience, exposure to in-service training on Integrated Management of Neonatal and Childhood Illnesses (IMNCI), Malaria Case Management (MCM), access to MCM guidelines and wall-charts and supervision. Assessment tools were employed to determine availability of microscopy, RDT services and anti-malarials in laboratories and pharmacies of each health facility.

Since the study mostly used routinely collected data, potential inconsistency in the measurement of temperature due to differences in thermometers used and inter-operator variability, use of different brands of RDT, differences in the efficiency of different microscopes used at the different hospitals, and between prescriber variabilities that were not measured, may have had an impact on the findings of this study.

Epidata version 3.1 and Stata version 13 SE (College Station, TX, USA) were used to manage and analyse the data. The data were checked to ensure internal consistency and completeness. The analyses were restricted to the initial visits of febrile patients who were at least 6 months or older who did not present with any danger signs or severe disease. The data obtained from prescribers, patient records, facility pharmacies, and laboratories were merged into one dataset using identifiers for municipalities, facilities and prescribers. The number of febrile cases who were attended by a prescriber in a facility on a survey day constituted a cluster. On the basis of this, the data was “svyset” and analyses were conducted, adjusting for clustering. The characteristics of the prescribers and the patients seen were described using simple frequencies. For continuous variables such as age and prescriber number of years of work, means and standard deviations were used to summarize them after log transformations of continuous variables that were skewed. For categorical variables, Chi square tests were used to determine differences.

The primary outcome investigated was inappropriate prescription of ACT, a composite binary indicator which combined the proportion of febrile patients who were either presumptively treated or over-treated. Febrile patients with negative malaria test results who were prescribed ACT were considered to have been over-treated, whilst febrile patients who were not tested but prescribed ACT were considered to have been treated presumptively. Combining these two proportions provided a good measure of the fraction of ACT prescribed inappropriately. Explanatory variables included patient level factors (age-group and ownership of a health insurance card), prescriber level factors and health facility level factors (availability of microscopy services, RDTs and anti-malarials). The crude association between the outcome variable and all other predictor variables were examined using Odds Ratios. A logistic regression analysis was performed to determine the predictors of the primary outcome at 95% CI (α-value of 0.05). The model comprised the outcome, a variance covariate estimator (vce) with the clustering variable and all independent variables that were predictors of the outcome at p < 0.1 in the crude analysis.

Overall, the records of 2519 independent febrile OPD patients were analysed. Of this number of patients, 45.6% (n = 1149) were younger than 5 years and 1.2% (n = 30) had no record of age nor date of birth on their folders. Their ages ranged from 6 months to 99 years, with a mean age of 6.8 years and a standard deviation of 4.4 years. Females constituted 52.7% (1327) of patients. Additionally, 57.6% of these patients were registrants of the National Health Insurance Scheme (NHIS) or other health insurance schemes. These patients were seen by 82 prescribers, whose mean age was 32.2 years with a standard deviation (SD) of ± 1.28, and the distribution showed that 80.2% of them were below the age of 40 years. Females constituted 61.0% of the prescribers. Medical officers and physician assistants constituted 59.8 and 29.3% of the prescribers, respectively. The mean number of years of work experience for the prescribers was 3.4 years (SD ± 2.6).

A little over half of all the patients (51.6%) were attended by physician assistants. Medical officers saw 1065 (42.3%) patients whilst the remaining 153 were seen by nurses and student prescribers. During these interactions, as many as 1512 (60.0%) of the patients were not referred for malaria tests and the rest [40.0% (95% CI 37.1–43.0%)] were tested by microscopy or RDTs.

Overall, as high as 98.0% of the patients were prescribed drugs, with a mean number of drugs per prescription of 3.2 (SD ± 1.2). Recommended oral ACT was prescribed for 1036 (41.1%) patients, whilst 234 patients were prescribed artemisinin monotherapy injections. However, 19.4% of the patients (n = 201) who were prescribed oral ACT were also prescribed artemisinin monotherapy injections concurrently. On the average, out of every 5 febrile patients prescribed ACT, 4 were also prescribed another drug in addition to the ACT; 83.4% were prescribed antipyretics, 61.4% were prescribed antibiotics and 10.4% (108 of 1036) were prescribed haematinics. None of the OPD folders sampled contained prescriptions of non-artemisinin monotherapy, including quinine and amodiaquine. Figure 1 provides detail of the malaria case management practices of the prescribers at each municipality.

In respect of the relationship between failure to test and prescription, almost half (42.7%: 95% CI 38.8–46.7) of the 1512 patients who were not tested were prescribed ACT. Among the 1007 patients who were tested, 158 tested positive for malaria, of which 147 (93.0%; 95% CI 87.6–96.2%) were prescribed recommended ACT. The remaining 11 patients with confirmed malaria were neither prescribed recommended nor non-recommended anti-malarials. Seven out of these 11 patients were however prescribed antibiotics. On the other hand, of the 679 patients who had negative malaria tests results, 145 (21.4%; 95% CI 17.9–25.3) were prescribed ACT.

In all, the proportion of patients who were prescribed ACT inappropriately was 76.4% (n = 791 of 1036), (95% CI 73.6–78.9). Of these 791 patients, 141 (17.8%) patients were prescribed artemisinin monotherapy injections in addition. The fractions of ACT prescriptions based on malaria test results are highlighted in Fig. 2.

The association of patient characteristics and treatment showed that under-fives were significantly less likely (χ² = 7.0388, p < 0.04) to be tested for malaria 37.2% (95% CI 33.5–41.0%), compared to those who were 5 years and older 42.4% (95% CI 39.0–45.9%). Additionally, patients younger than 5 years were significantly less likely (χ² = 6.5045, p < 0.05) to be treated presumptively (39.2%, 95% CI 33.9–44.8) compared to patients who were 5 years and older (45.7%, 95% CI 41.3–50.2). Significantly fewer (χ² = 6.7472, p < 0.05) under-fives were overtreated (17.2%: 95% CI 13.1–22.4%) compared to those above 5 years, 25.5% (95% CI 20.6–31.0%), Over a third (36.8%) of the under-fives were diagnosed with malaria compared to 44.1% of patients above 5 years of age, (χ² = 13.6163, p < 0.01). There was no significant difference (χ² = 0.1827, p > 0.05) in the prescription of ACT to patients with positive test results by age-group, 94.1% of under-fives compared to 92.2% of patients above 5 years. The top ten symptoms and diagnoses of the patients are presented in Table 1.

In a crude analysis (Tables 2 and 3), a number of variables were significantly associated with inappropriate ACT prescription, and these included, age-group of patients, treatment chart availability, access to reference material on Malaria Case Management (MCM) and training in IMNCI or MCM in the past 12 months. However, after adjusting for covariates, prescriber factors, such as number of training programmes in MCM or MCM attendance, number of supervisory visits on MCM, access to ACT treatment charts or wall charts on MCM, work experience, professional category and RDT stock-outs became significantly associated with inappropriate ACT prescription.

Specifically, febrile patients who were seen in health facilities with RDTs in stock were less likely to be prescribed ACT inappropriately, (AOR: 0.04, 95% CI 0.01–0.14, p < 0.001) compared to those who were seen in health facilities with RDT stock-outs. Patients who were seen by medical officers were more likely to be prescribed ACT inappropriately than the patients seen by physician assistants (AOR: 5.8, 95% CI 2.5–13.7, p < 0.001). The odds of patients being prescribed ACT inappropriately by student and nurse prescribers was reduced by 90% compared to those seen by physician assistants (AOR: 0.1, 95% CI 0.03–0.27, p < 0.001). Patients who were seen by prescribers who had been supervised once within the past 6 months were 8 times more likely to be prescribed ACT inappropriately compared those seen by prescribers who had not benefited from any supervision on MCM within the past 6 months. The odds of patients being prescribed ACT inappropriately by prescribers who had access to MCM wall charts was reduced by more than half (88%) compared to the prescribers who did not have the MCM wall chart (AOR: 0.12, 95% CI 0.06–0.24, p < 0.001). However, the odds of patients being prescribed ACT inappropriately by prescribers who had access to the ACT treatment chart was increased by 800% compared to the prescribers who did not have the ACT treatment chart (AOR: 9.0, 95% CI 3.7–22.2, p < 0.001).

Prescribers who had benefited from in-service training intervention on malaria case management at least once within the past year were 4 times more likely to prescribe ACT inappropriately compared to those who had not benefited from any MCM training intervention in the past year (AOR: 4.1; 95% CI (1.5–11.6); p < 0.01). However, the odds of patients being prescribed ACT inappropriately by prescribers who had been trained at least once in IMNCI within the past 12 months was reduced by 93% compared to the prescribers who had not been trained IMNCI the past year (AOR: 0.07, 95% CI (0.03–0.15), p < 0.001). In this study, over 70% (1774) of patients were prescribed antibiotics including 45.6% (72) of febrile patients with confirmed malaria and 70.4% (478) of patients with negative malaria test results.

To obtain data on febrile outpatients in this study, the investigators relied on secondary data alone without a gold standard, implying that any potential misclassifications could not be accounted for. Findings of the study should therefore be interpreted with caution. The data used to arrive at these conclusions is cross-sectional in nature and the observed temporal associations between exposure and outcome variables may not be causal.