Outcome measured by the proportion of women with employment

Outcome variable: ITN use

The
Rollback Malaria indicator for ITN use, defined as percentage of women who
slept under ITN the previous night before the survey was used as the operational
definition for the outcome 3

We Will Write a Custom Essay Specifically
For You For Only $13.90/page!


order now

Predictors

The
individual- level predictors included age of woman, parity, media exposure etc.
To examine community level factors, 3 community sociodemographics and four
community health factors were selected. The predictors are described in Table
1. 

 

Table
1 Description and measurement of independent variables

Community-level
factors

Description

Community
sociodemographics

 

%
of women with employment

Aggregated
values of community level women employment status measured by the proportion
of women with employment from individual level data categorized using
tertiles as low, middle and high

%
of women with any education

Aggregated
values of community level women education status measured by the proportion
of women with primary education and above from individual level data
categorized using tertiles as low, middle and high

%
of women who are rich

Aggregated
values of community level women wealth index measured by the proportion of
women in rich wealth index  from
individual level data categorized using tertiles as low, middle and high

Community
health factors

 

%
of women with autonomy on HC decisions

Aggregated
values of community level women’s decision status on health care measured by
the proportion of women who make self-decisions on health care from
individual level data categorized using tertiles as low, middle and high

%
of women in households with 2 or more ITNs

Aggregated
values of community level women households with 2 or more ITNs measured by
the proportion of women in household with 2 or more ITNs from individual
level data categorized using tertiles as low, middle and high

%
of women with perceived distance to HF a problem

Aggregated
values of community level women’s perceived distance to HF measured by the
proportion of women  who perceived
distance to HF to be a problem from individual level data categorized using
tertiles as low, middle and high

%
of women in households with IRS coverage

Aggregated
values of community level women in households with IRS coverage measured by
the proportion of women IRS coverage from individual level data categorized
as 0 and 1

Individual-level
Factors

Description

Age

15-24,
25-34, 35 and above

Media
exposure

Yes/No

Occupation

Employed/Unemployed

Education

No
formal education, Primary, Secondary and above

Wealth

Poor,
Middle, Rich

Parity

Primigravida,
Secundigravida, Multigravida

Residence

Rural,
Urban

Region

Northern,
Central and Southern

Number
of household members

<5, ?5 - Average number of household members in Malawi is 4.5 2 Number of Under-fives None, one, two or more IRS Yes, No Health behavior Constructed with a combination of two variables, boiling water and whether they have a hand washing facility at home. Categorized as no any of the two" low", one of the two "medium" and all of the two "high" Women autonomy on HC decisions Categorized as self, husband alone, with other Altitude >1600,
?1600

Number
of ITN in household

<2, ?2 Religion Catholics, Protestant, Muslims and others Distance to HF Yes/No   Ethics statement The survey protocol was reviewed and approved by the National Health Sciences Research Board of Malawi, the Institutional Review Board (IRB) of ICF Macro, and the Centers for Disease Control (CDC) in Atlanta. Informed consent was obtained at the beginning of each interview and permission was sought from the DHS program for the use of the data. Statistical analyses Descriptive analysis The distribution of participants' characteristics (both individual and community-level) according to ITN use were examined using chi-square tests. Logistic Model Approach and measures of association A two-level multilevel logistic regression analysis was applied to measure the association of Individual- and community-level factors on ITN use among women (level 1) nested within their communities (level 2). The association (fixed effects) was reported as Odds ratio and 95% CIs. In the multilevel logistic regression, four different models were fitted. First, a null model with outcome variable only was constructed to assess total variance between communities, and community effects on the outcome. This model was included to allow the decomposition of the variance that existed between the communities. Then model I included individual level factors only, followed by model II with community level factors only, and finally model III adjusted both individual and community-level factors. Measures of variation Furthermore, measures of variation were examined through area variance and 95% CI, Inter-cluster correlation (ICC), median odds ratio (MOR) and Proportional Change in Variance (PVC) 4, 5. Testing model fit To test for model fit, Akaike Information Criterion (AIC) was used. To test for multicollinearity, variance inflation factor (VIF) and tolerance were used 6. All variables were below 10 for VIF and more than 0.1 for tolerance hence no problem for multicollinearity. All analyses were done using Stata version 15.0. Results The study analyzed 16060 married women (level 1) nested within 850 communities (level 2). About 10.2% (1634) women were pregnant, 69.7% (11197) non-pregnant with under-5 children and 20.1% (3229) non-pregnant without under-5 children (Figure 1). All the women groups were mutually exclusive. The rate of ITN use was 45.9%, 46.9% and 39.1% for pregnant, non-pregnant with under-5 children and non-pregnant-women without under-5 children respectively. Distribution of participants according to ITN use Table 2 presents distribution of participants' individual- and community-level factors according to ITN use for pregnant women, non-pregnant women with under-5 children and non-pregnant women without under-5 children. In all study groups, significant (p < 0.05) differences were observed between those ITN users and non-users in terms of individual level factors (i.e. education, media exposure, distance to health facility, wealth, number of household members, number of ITN in household) and one community-level factor (i.e. percentage with two or more ITN). No significant difference was observed in all groups for IRS, Altitude, region, religion, and autonomy decision on health care (i.e. individual-level factors) and two community-level factors (i.e. percentage of women with autonomy on health care decisions and percentage with IRS). In both non pregnant women groups, the difference between ITN users and non-users was significant (p < 0.05) in terms of age, residence and health behavior. Furthermore, there was a significant (p < 0.05) difference between ITN users and non-users among pregnant women and non-pregnant women with under-5 in terms of parity, and percentage of rich women. No significant difference was observed for these factors in non-pregnant women without under-5 children. Finally, there was a significant difference between ITN users and non-users in terms of occupation in non-pregnant women without under-5 children, percentage with employment in pregnant women, percentage with any education and percentage of women perceiving distance to health facility a problem in non-pregnant women with under-5 children. Measures of Association (Fixed effects) Among pregnant women Table 3 showed the results of multivariable multilevel logistic regression analyses for pregnant women. Model 3 (Final model) which controls for individual- and community-level factors shows that compared to primigravida women, secundigravida and multigravida were more likely to use ITN (aOR = 1.79; 95% CI: 1.23 – 2.62 and aOR = 1.97; 95% CI: 1.28 - 3.28 respectively). Women who perceived distance to health facility as not a problem, with less than 5 household members, from households with two or more ITNs, central region, and from rural areas were more likely to use ITN compared to those who perceived distance to health facility as a problem, had 5 or more household members, from households with less than 2 ITNs, from northern region, and from urban areas  (aOR = 1.44; 95% CI: 1.09 – 1.89, aOR = 2.23; 95% CI: 1.74 – 3.11, aOR = 4.25; 95% CI: 3.19 – 5.64, aOR = 1.66; 95% CI: 1.2 – 2.45 and aOR = 1.63; 95% CI: 1.04 – 2.55). For community level factors, pregnant women in communities with a high percentage of households with 2 or more ITNs were more likely to use ITN compared to women in communities with low percentage of households with 2 or ITNs (aOR = 1.47; 95% CI: 1.05 – 2.06). Among non-pregnant women with under-5 children The results of multivariable multilevel logistic regression analyses for non-pregnant women with under-5 children are shown on Table 4. After controlling for both individual- and community-level factors, secundigravida and multigravida were more likely to use ITN (aOR = 4.97; 95% CI: 2.62 – 9.43 and aOR = 4.55; 95% CI: 2.40 – 8.61 respectively) compared to Primigravida (model 3). Women from middle wealth households, with 2 or more under-5 children, with less than 5 household members, from households with two or more ITNs, central and southern region, from rural areas and with good health behavior were more likely to use ITN compared to their counterparts (aOR = 1.13; 95% CI: 1.01 – 1.28, aOR = 1.12; 95% CI: 1.01 – 1.24, aOR = 1.85; 95% CI: 1.65 – 2.08, aOR = 6.89; 95% CI:6.19 – 7.65, aOR = 1.46; 95% CI:1.19 – 1.78, aOR = 1.40; 95% CI:1.16 – 1.70 , aOR = 1.48; 95% CI: 1.18 – 1.84 and aOR = 1.19; 95% CI:1.01 – 1.39). For community level factors, non-pregnant women in communities with middle and high percentage of households with 2 or more ITNs, and in communities with high percentage of women autonomy on health care decisions were more likely to use ITN compared to women in communities with low percentage of households with 2 or ITNs and low percentage of women with autonomy on health care decisions (aOR = 1.19; 95% CI: 1.02 – 1.38, aOR = 1.35; 95% CI:1.14 – 1.60, and aOR = 1.18; 95% CI:1.00 – 1.38 respectively). Among non-pregnant women without under-5 children For non-pregnant women without under-5 children, only those from households with less than 5 members, households with 2 or more ITNs, central and southern region, and rural areas were more likely to use ITN compared to their counterparts (aOR = 1.66; 95% CI: 1.37 – 2.02, aOR = 10.72; 95% CI: 8.59 – 13.37, aOR = 1.53; 95% CI: 1.12 – 2.09, aOR = 1.34, 95% CI:1.00 – 1.81, and aOR = 1.57; 95% CI:1.14 – 2.16) in model 3, table 5. For community level factors, non-pregnant women without under-5 children in communities with middle and high percentage of households with 2 or more ITNs were more likely to use ITN compared to women in communities with low percentage of households with 2 or more (aOR = 1.36; 95% CI: 1.06 – 1.75, and aOR = 1.49; 95% CI:1.15 – 1.96, respectively). Measures of variation Table 3 also shows the results of measure of variation. 11.6 %, 8.9% 6.6% and 7.3% of the variation to use ITN among pregnant women was attributable to unobserved community characteristics in the null model, model I, model II and model III respectively. The MOR showing the effects of community heterogeneity were 1.87, 1.71, 1.58 and 1.62 for null model, model 1, 2 and 3 respectively. On the other hand, the PCVs shows that 25.4%, 46.2%, and 39.8%, of the variance in ITN use among pregnant women across communities were explained by individual level factors, community level factors and both individual-level and community-level factors respectively. In non-pregnant women with under-5 children, 13.9 %, 10.9%, 8.3% and 10.4% of the variation to use ITN was attributable to unobserved community characteristics in the null model, model I, model II and model III respectively. The MOR for the communities were 2.01, 1.83, 1.68 and 1.80 for null model, model 1, 2 and 3 respectively (table 4). The PCVs showed that 24.7%, 44.4%, and 28.9%, of the variance in ITN use among non-pregnant women with under-5 children across communities were explained by individual level factors, community level factors and both individual-level and community-level factors respectively. Measure of variation indicated that 17.2 %, 9.6%, 8.2% and 8.2% of the variation to use ITN was attributable to unobserved community characteristics in the null model, model I, model II and model III respectively among non-pregnant women without under-5 children (table 5). The MOR were reduced from 2.20 in null model to 1.76, 1.68, 1.68 in model 1, 2 and 3 respectively (table 4). The PCVs showed that 49.2%, 57.1%, and 57.2%, of the variance in ITN use among non-pregnant women without under-5 children across communities were explained by individual level factors, community level factors and both individual-level and community-level factors respectively.