Predictors of the use of interventions to prevent malaria in pregnancy in Cameroon

Surveys were collected in Cameroon between January and August of 2011 as part of the cross-sectional National Demographic Health Survey (DHS) [22]. These recurring household level surveys have been carried out using previously described methods and survey data was weighted to make it nationally representative [22‐24]. Data was collected from 15,426 eligible women who were interviewed using two-stage stratified sampling techniques. This study captures responses from the subset of 7647 women who reported a birth in the previous 5 years. Surveys included detailed questions about socio-demographics, pregnancy history, access to antenatal care services, site of delivery and the use of interventions to prevent malaria during the most recent pregnancy. Blood was collected from a subset of women for rapid HIV testing. DHS data has been cleaned, is devoid of personal identifiers and datasets are publicly available to researchers upon request.

Descriptive statistics were used for women surveyed who reported a pregnancy in the previous 5 years. Logistic regression models were fit to evaluate associations between independent variables and the six dichotomous outcomes of interest (bed net ownership, bed net usage, use of any medication for malaria prophylaxis, and 3 separate SP dosing measures (at least 1 vs >1 dose, at least 2 vs >2 doses, at least 3 vs >3 doses). Separate univariate (UV) and multi-variable (MV) models were fit for each outcome. All variables in Table 1 were considered for the MV models and variables were selected based on statistical significance in the univariate models (p < 0.05), review of existing literature and collinearity considerations. The same set of independent variables was used in all MV models. Sensitivity analyses were performed by fitting various MV models with and without the variables that showed collinearity. Since region of residence and urban/rural residence were highly collinear with the wealth variable, they were excluded from the MV models. The timing of the initial ANC visit, provider type and facility type were also excluded from MV models due to collinearity with one another. Missing data is presented in the tables but data points were generally complete. UV and MV odds ratios with 95% confidence intervals are presented in the tables and UV and MV odds ratios are shown in a figure. All analyses were performed with SAS 9.4 (Cary, NC) and results were adjusted for weighting, clustering and stratification using the SAS/STAT^® “SURVEY” procedures.

Table 2 shows the results of the multivariable analysis. In terms of bed net ownership, the univariate analysis showed an association with maternal primary school education (compared to no education), owning a method of transportation, the presence of children under age 5 in the home and measures of improved access to healthcare. Women who were not married were less likely to own a bed net. In the adjusted models, bed net ownership was associated with the presence of children under 5 in the home (aOR 2.3, 95% CI 1.7–3.1 for 3 or more children), more frequent ANC visits (aOR 1.6, 95% CI 1.3–1.9 for 4 or more visits), maternal literacy (aOR 1.2, 95% CI 1.02–1.5) and ownership of transportation (aOR 1.2, 95% CI 1.1–1.4). There was very little regional variation in bed net ownership.

Factors in the unadjusted models associated with bed net usage (among those who owned a bed net) included older maternal age, higher educational level (woman and partner), literacy, higher wealth quintile, use of contraception and having children under age 5 in the home. Decreased bed net usage was associated with polygamy and owning a means of transportation. Regional variation was noted in bed net usage: women in more populated regions (such as urban Douala and Adamaoua) had higher usage than women in the more sparsely populated Extreme North region. After multivariable adjustment, usage was associated with a higher number of children under age 5 in the home (aOR 2, CI 1.3–3 for 3+ children), number of ANC visits (aOR 2, CI 1.4–2.8 for 1+ visit), maternal primary school education (aOR 1.5, CI 1.1–2.1), maternal literacy (aOR 1.4, CI 1.1–1.8) and having a facility delivery (aOR 1.3, CI 1.1–1.7). Women who were widowed, divorced or separated were less likely to use a bed net compared to women who were married (aOR 0.6, CI 0.4–0.8). Wealth quintile and maternal age did not predict any of the outcomes in the adjusted models. When the model was run with urban/rural residence replacing wealth index (two collinear variables), it was not associated with the outcome.

In the univariate models for usage of any malaria prophylaxis medication during the most recent pregnancy, an association was seen with literacy, education, higher wealth index, urban residence, HIV-positive status, number of ANC visits and facility delivery. Polygamy, having an older partner, parity >5, ANC care at a health centre (compared to a hospital clinic) and later presentation for ANC care were negatively associated (data not shown). In the adjusted models, an independent association was seen for more ANC visits (aOR 17.8, 95% CI 13–24.5 for 4+ visits), maternal education (aOR 1.5, 95% CI 1.1–2.1 for secondary compared to none), maternal literacy (aOR 1.4, 95% CI 1.1–1.7), having visited a health facility in the past year (aOR 1.3, 95% CI 1.1–1.5) and higher parity (aOR 1.2, 95% CI 1.02–1.4 for 3–4 compared to 1–2).

Factors associated with the use of higher doses of SP (three separate outcomes) are displayed in Table 3. The factor most strongly associated with increased SP dosing was a higher number of ANC visits (aOR 25.4, CI 12.3–52.5 for the association of 4+ visits with 3+ SP doses). Maternal primary school education and having visited a health facility in the past year were associated with the receipt of at least one dose of SP and literacy was associated with the receipt of 1–2 doses of SP. Wealth quintile was not associated with the number of SP doses received. Parity greater than 2 had a negative association with receiving at least 2 doses of SP and a relationship status of “living with partner” or “never in union” was negatively associated with receipt of higher doses of SP compared to married women (aOR 0.6, 95% CI 0.5–0.8 for never in union).

Figure 1 shows the differences in odds ratios and 95% CIs between the UV and MV models by outcome. It highlights the importance of the MV model in adjusting for covariates and the primacy of ANC care in terms of the medication-related outcomes.

Only one in three women with a recent pregnancy owned a bed net. There was a wide range of bed net ownership (3–80%) documented by Singh et al. among pregnant women in sub-Saharan Africa [13]. The strongest predictors of net ownership in our study were maternal literacy, the number of young children in the house, and antenatal care. The association with maternal education is consistently seen, but the current analysis did not confirm the previous association with wealth index, parity or urban residence [13]. Also, bed net ownership rates were similar in all regions of the country. These positive findings point to equity in net ownership which may be attributed to public health campaigns that aim to disseminate bed nets widely at no cost [25]. The strong association with the presence of young children at home is likely due to patterns of prioritized net distribution for these households in the past [13, 23].

In this study, bed net usage rates were low at 16.9%. The rate of net usage during pregnancy was similarly low in Nigeria (26%) and 17% in sub-Saharan Africa overall [13, 14]. The clear discrepancy between net ownership and usage in pregnancy has been documented and some of the cited explanations include: discomfort, torn nets and a perception that nets are unnecessary [13, 26]. Maternal education, literacy, wealth, urban residence and older age have been variably associated with bed net usage in earlier studies but only maternal literacy and education were associated in this adjusted model [13, 27]. One explanation for this difference may be that public health messaging in support of bed nets in Cameroon has reached a broader population, including younger women, those with a lower socioeconomic status or women in rural areas but the major message is that usage rates are too low. This study, like others, supports the need for long-term investment in the education of girls.

Only 6 in 10 women were prescribed medication for malaria prophylaxis during pregnancy and about one in four (27.2%) received the recommended 2 or more doses of SP. This is consistent with the recent World Malaria Report, where 57% of pregnant women in sub-Saharan Africa received SP, 43% received 2 doses and 17% received 3 or more doses [21]. Nonetheless, prophylaxis rates have improved significantly in sub-Saharan Africa (SP coverage was 19% in 2007) [14]. The strong association between the number of ANC visits and use of prophylactic medication is expected since these medications are most often administered from ANC clinic. There is an obvious clinical benefit to 2 or 3 doses of SP compared to single dose and longer time (>10 weeks) between doses has been associated with suboptimal outcomes [28]. Wealth was associated with access to and use of malaria prophylaxis medications in prior studies but was not seen in this study [21, 29]. Access to malaria prophylaxis among pregnant women may have become more equitable in Cameroon due to affordable essential drug schemes which lower costs for all.

The strengths of this study include its size and the survey methods involving weighted data from national household surveys designed to be representative. The comprehensive data collection allowed for exploration of several individual level factors as well as measures of healthcare access. Limitations of this study include the single country design, use of data from 2011 and self-reported measures. A few additional limitations inherent to the survey methods include recall bias (pregnancy up to 5 years before the survey) and social desirability bias (the desire to give the “correct” answer to the person administering the survey). Analysis based on region, ANC provider type or facility type was not performed because of collinearity. Most of the women queried about bed nets were not pregnant at the time of the survey but were of childbearing age with a recent pregnancy and this was used as a proxy for the behavior of women during pregnancy. Bed net usage was defined by the standard indicator of having slept under a net on the night prior to the survey. This may not be best indicator of actual usage although it is commonly used and it is the outcome for the “Roll Back Malaria Initiative”. Pregnancy outcomes related to malaria infection were not available. Finally, because of multiple comparisons, some of the associations noted may have occurred by chance alone.