At baseline, the age of the women in the control and intervention groups ranged between 18 and 45 years with a mean age of 27. Very few women in control and intervention groups (16 and 13%, respectively) had completed primary school education and most were uneducated. More than 60% of the women were married at or earlier than 18 years of age. More than half of the women in both groups had three to four children. Most women lived in mud houses consisting of three to four rooms (>0.05), did not own a mobile phone, and did not have an improved source of drinking water, latrine and sewage drainage system in their households. Approximately 35 and 20% women in control and intervention areas, respectively, experienced stillbirths. Similarly, about a quarter of the women said that they had at least one newborn death in the past. Only about a quarter of the women from both groups said that they were counselled by an antenatal care provider about protection against malaria and the commonest topic on which the women in the control and intervention groups received any instruction was indoor residual spraying, 27 and 19%, respectively. Some 32% of women in the control and 21% in the intervention group identified health workers as the most common source of information about malaria. Most women in both groups were aware that malaria was caused by mosquitoes (Table 1).
Table 1 Baseline characteristics and information about malaria among control and intervention groups in the study
Variables
|
Control group (n=100)
|
Intervention group (n=100)
|
p-value
|
Age (years)
|
≤25
|
22 (22%)
|
39 (39%)
|
0.029*
|
26-30
|
53 (53%)
|
39 (39%)
|
31 and above
|
25 (25%)
|
22 (22%)
|
Education
|
Uneducated
|
62 (62%)
|
83 (83%)
|
0.001*
|
Primary
|
16 (16%)
|
13 (13%)
|
Any other type of education
|
22 (22%)
|
4 (4%)
|
Number of living children
|
1-2
|
17 (17%)
|
10 (10%)
|
0.111
|
3-4
|
54 (54%)
|
63 (63%)
|
5-6
|
9 (9%)
|
15 (15%)
|
7-9
|
20 (20%)
|
12 (12%)
|
Age at marriage (years)
|
15, 16
|
12 (12%)
|
11 (11%)
|
0.073
|
17
|
18 (18%)
|
34 (34%)
|
18
|
43 (43%)
|
36 (36%)
|
19-22
|
27 (27%)
|
19 (19%)
|
Type of household
|
Mud house
|
79 (79%)
|
82 (82%)
|
0.592
|
Brick house
|
21 (21%)
|
18 (18%)
|
Number of rooms
|
1-2
|
30 (30%)
|
21 (21%)
|
0.144
|
3-4
|
70 (70%)
|
79 (79%)
|
Owns a mobile phone
|
Yes
|
54 (54%)
|
21 (21%)
|
0.001*
|
No
|
46 (46%)
|
79 (79%)
|
Source of drinking water
|
Well outside home
|
71 (71%)
|
49 (49%)
|
0.001*
|
Well inside home
|
29 (29%)
|
51 (51%)
|
Type of latrine
|
Open
|
51 (51%)
|
26 (26%)
|
0.001*
|
Pit latrine
|
49 (49%)
|
74 (74%)
|
Mode of sewage drainage in house
|
Open sewers
|
64 (64%)
|
33 (33%)
|
0.001*
|
Underground sewers
|
2 (2%)
|
25 (25%)
|
Open pond
|
34 (34%)
|
42 (42%)
|
Income (PKR)
|
5,000
|
37 (37%)
|
76 (76%)
|
0.001*
|
6,000-10,000
|
48 (48%)
|
22 (22%)
|
11,000-15,000
|
15 (15%)
|
2 (2%)
|
Previous stillbirth (28 weeks)
|
Yes
|
35 (35%)
|
20 (20%)
|
0.018*
|
No
|
65 (65%)
|
80 (80%)
|
Previous newborn death
|
Yes
|
26 (26%)
|
18 (18%)
|
0.172
|
No
|
74 (74%)
|
82 (82%)
|
Antenatal counselling on malaria
|
Yes
|
35 (35%)
|
24 (24%)
|
0.088
|
No
|
65 (65%)
|
76 (76%)
|
Counselling topics
|
Use LLINs
|
4 (4%)
|
3 (3%)
|
0.925
|
Indoor spray
|
27 (27%)
|
19 (19%)
|
Take preventive medicine
|
4 (4%)
|
2 (2%)
|
Malaria during previous pregnancies
|
Yes
|
15 (15%)
|
29 (29%)
|
0.017*
|
No
|
85 (85%)
|
71 (71%)
|
Malaria during current pregnancy
|
Yes
|
2 (2%)
|
1 (1%)
|
0.561
|
No
|
98 (98%)
|
99 (99%)
|
Ever heard about malaria
|
Yes
|
34 (34%)
|
23 (23%)
|
0.085
|
No
|
66 (66%)
|
77 (77%)
|
Source of information about malaria
|
Health worker
|
32 (32%)
|
21 (21%)
|
0.683
|
Other
|
2 (2%)
|
2 2(%)
|
*Significant (p=<0.05)
Descriptive statistics were calculated for each knowledge and use of LLINs item on both intervention and control groups before the intervention to explore trends regarding use of LLINs between the two groups before and after intervention (Table 2). The women’s perception that malaria was a harmful disease increased from 75 to 97% after the intervention, whereas in the control arm this change was only 10% points from 30 to 40%. The use of LLINs increased in the intervention group at post-intervention assessment due to health education, from 10-30%, while in the control group only 1-3% improved. Most of the study population increased their knowledge on mode of transmission from 75-97% in the intervention group. Most of the participants heard about mosquito nets after the intervention (58-100%) and their ownership regarding mosquito nets increased up to 76% after intervention (<0.05). The LHWs were the most prominent source of obtaining malaria information and it significantly increased in intervention group (<0.05). LLIN use is important in prevention of malaria both in mothers and children and it increased substantially after the intervention (<0.05). Generally, community knowledge regarding malaria prevention was high. Mosquito net use during pregnancy (practices), importance of LLINs during pregnancy (knowledge) and importance (attitude) increased in both groups; however, a significant change (<0.05) was observed in intervention group (31-73%). Participants have shown a positive improvement while using indoor mosquito spray after intervention (<0.05). Most of the respondents reported that they used LLINs the previous night while sleeping; the frequency of LLIN users also increased after intervention (<0.05). However, no significant differences were found between the groups at baseline except source of information, type of LLINs available and their importance (p >0.05).
Table 2 Comparison of change in the use of long-lasting insecticide-treated bed nets and other malaria preventive measures (before and after, and between control and intervention groups)
Variables
|
Control (n=100)
|
Intervention (n=100)
|
Before
|
After
|
p value
|
Before
|
After
|
p value
|
|
Knowledge
|
|
Transmission
|
Insect bite
|
21 (21%)
|
17 (17%)
|
0.106
|
25 (25%)
|
3 (3%)
|
0.765
|
|
Mosquito bite
|
79 (79%)
|
83 (83%)
|
75 (75%)
|
97 (97%)
|
|
Symptoms
|
Fever
|
69 (69%)
|
66 (66%)
|
0.314
|
61 (61%)
|
77 (77%)
|
0.286
|
|
Headache
|
31 (31%)
|
34 (34%)
|
26 (26%)
|
23 (23%)
|
|
Severity
|
Yes
|
30 (30%)
|
40 (40%)
|
0.210
|
75 (75%)
|
97 (97%)
|
0.002*
|
|
No
|
70 (70%)
|
60 (60%)
|
25 (25%)
|
3 (3%)
|
|
Complications
|
Fever
|
69 (69%)
|
66 (66%)
|
0.314
|
61 (61%)
|
77 (77%)
|
0.286
|
|
Vomiting
|
31 (31%)
|
34 (34%)
|
39 (39%)
|
23 (23%)
|
|
Prevention
|
By using LLINs
|
46 (46%)
|
44 (44%)
|
0.255
|
31 (31%)
|
87 (87%)
|
0.893
|
|
Spray
|
54 (54%)
|
46 (46%)
|
69 (69%)
|
13 (13%)
|
|
Heard about LLINs
|
Yes
|
61 (61%)
|
48 (48%)
|
0.065
|
58 (58%)
|
100 (100%)
|
<0.001*
|
|
No
|
39 (39%)
|
52 (52%)
|
42 (42%)
|
0
|
|
LLINs use
|
Protects newborn from malaria
|
46 (46%)
|
50 (50%)
|
0.093
|
28 (28%)
|
6 (6%)
|
0.053*
|
|
Protects mother from malaria
|
54 (54%)
|
50 (50%)
|
72 (72%)
|
94 (94%)
|
|
Use LLINs prevents malaria
|
Yes
|
35 (35%)
|
40 (40%)
|
0.465
|
100 (100%)
|
100 (100%)
|
<0.001*
|
|
No
|
65 (65%)
|
60 (60%)
|
0
|
0
|
|
Use of LLINs
|
|
LLIN present
|
Yes
|
81 (81%)
|
70 (70%)
|
0.091
|
85 (85%)
|
97 (97%)
|
0.003*
|
|
No
|
19 (19%)
|
30 (30%)
|
15 (15%)
|
3 (3%)
|
|
Use LLINs in pregnancy
|
Yes
|
35 (35%)
|
39 (39%)
|
0.474
|
31 (31%)
|
73 (73%)
|
<0.001*
|
|
No
|
65 (65%)
|
61 (61%)
|
69 (69%)
|
27 (27%)
|
|
LLINs use important in pregnancy
|
Yes
|
35 (35%)
|
39 (39%)
|
0.474
|
31 (31%)
|
73 (73%)
|
<0.001*
|
|
No
|
65 (65%)
|
61 (61%)
|
69 (69%)
|
27 (27%)
|
|
Use LLINs previous night
|
Yes
|
35 (35%)
|
36 (36%)
|
0.883
|
31 (31%)
|
96 (96%)
|
<0.001*
|
|
No
|
65 (65%)
|
64 (64%)
|
69 (69%)
|
4 (4%)
|
|
Use spray in last 3 months
|
Yes
|
35 (35%)
|
33 (33%)
|
0.765
|
31 (31%)
|
81 (81%)
|
<0.001*
|
|
No
|
65 (65%)
|
67 (67%)
|
69 (69%)
|
19 (19%)
|
|
*Significant (p=<0.05)
DID is usually implemented as an interaction term between time and treatment group dummy variables in a regression model. The coefficient of the treatment variable (intervention), is the estimated mean difference in outcome (knowledge and use of LLINs) between the treatment and control groups prior to the intervention; it represents whatever baseline differences existed between the groups before the intervention was applied to the control group. At baseline, the coefficient of intervention is 0.43 (knowledge) and -0.02 (use of LLINs), however it is not significant (Table 3).
The time trend (post intervention) is the expected mean change in outcome from before to after the onset of the intervention among the control group. It reflects the pure effect of the passage of time in the absence of the actual intervention. The coefficient of time trend (post-intervention) is 0.57 (knowledge) and it is significant at 5% level of significance. Similarly, the coefficient of time trend (post-intervention) for use of LLINs is 0.57, which is significant at 5% level of significance (Table 3).
Table 3 Difference in differences (DID) with time trend and interaction term
Variables
|
Model-1a
|
Model-2b
|
Model-3c
|
Model-4d
|
Treatment
|
0.430
|
-0.020
|
0.106
|
-0.238
|
|
(0.276)
|
(0.221)
|
(0.364)
|
(0.277)
|
Time (Post-intervention)
|
0.570**
|
0.320
|
0.570**
|
0.320
|
|
(0.276)
|
(0.221)
|
(0.284)
|
(0.229)
|
Interaction term
|
4.170***
|
3.360***
|
4.170***
|
3.360***
|
|
(0.390)
|
(0.312)
|
(0.382)
|
(0.309)
|
Confounders
|
No
|
No
|
Yes
|
Yes
|
R square
|
0.540
|
0.500
|
0.570
|
0.530
|
Prob > F
|
0.000
|
0.000
|
0.000
|
0.000
|
Confounders: age, education, income, etc
Standard errors in parentheses; * p<0.1, ** p<0.05, *** p<0.01
a = Model 1: Estimation of knowledge without confounders
b = Model 2: Estimation of use of LLINs without confounders
c = Model 3: Estimation of knowledge with confounders
d = Model 4: Estimation of use of LLINs with confounders
The interaction term by itself is the difference in differences estimator. The coefficient for interaction term is the differences-in-differences estimator. The effect is significant at 1% level of significance with the treatment having a positive effect.
Results of DID regression estimation with an interaction between time (baseline vs post-intervention) and groups (control and intervention) were used to the estimate magnitude of the effect of intervention (knowledge and use of LLINs change due to intervention by controlling time) and its statistical significance.
The results (Table 4) showed that there were no significant differences between the mean knowledge and use of LLINs scores of the groups at baseline. However, the mean knowledge and use of LLINs scores increased significantly in the intervention group as compared to the control group after intervention. Therefore, it can be inferred that the estimated DID value of 4.170 (p <0.01) represents an increase in scores of knowledge in the intervention group compared with control after intervention. Similarly, it is estimated that increase in use of LLINs score after intervention is represented by DID value of 3.360 (p <0.01). The analysis was also done by controlling for confounders and it showed similar results. Thus, it can be concluded that the scores of knowledge and use of LLINs improved in the intervention group, which is due to the effect of intervention.
Table 4 Difference in Differences (DID) Estimation with and without confounders
Difference in Differences (DID) without confounders
|
|
Baseline (BL)
|
Post-intervention (PI)
|
|
Outcome variables
|
Control
|
Intervention
|
Diff (BL)
|
Control
|
Intervention
|
Diff (EL)
|
Diff-In-Diff
|
|
Knowledge
|
5.250
|
5.680
|
0.430
|
5.820
|
10.420
|
4.600
|
4.170
|
|
SE**
|
|
|
0.312
|
|
|
0.235
|
0.390
|
|
t value
|
|
|
1.380
|
|
|
19.57
|
10.68
|
|
p value
|
|
|
0.169
|
|
|
0.000***
|
0.000***
|
|
R square
|
0.54
|
|
|
|
|
|
|
|
Use of LLINs
|
4.530
|
4.510
|
-0.020
|
4.850
|
8.190
|
3.340
|
3.360
|
|
SE**
|
|
|
0.245
|
|
|
0.194
|
0.313
|
|
t value
|
|
|
-0.08
|
|
|
17.18
|
10.75
|
|
p value
|
|
|
0.935
|
|
|
0.000***
|
0.000***
|
|
R square
|
0.59
|
|
|
|
|
|
|
|
Difference in Differences (DID) with confounders
|
Knowledge
|
5.605
|
5.712
|
0.107
|
6.175
|
10.452
|
4.277
|
4.170
|
|
SE
|
|
|
0.309
|
|
|
0.309
|
0.383 |
|
|
t value
|
|
|
0.35 0
|
|
|
13.82
|
10.90
|
|
p value
|
|
|
0.730
|
|
|
0.000***
|
0.000***
|
|
R square
|
0.58
|
|
|
|
|
|
|
|
Confounders
|
Yes
|
|
|
|
|
|
|
|
Use of LLINs
|
2.785
|
2.546
|
-0.239
|
3.105
|
6.226
|
3.121
|
3.360
|
|
SE
|
|
|
0.250
|
|
|
0.250 |
|
0.309
|
|
t value
|
|
|
-0.95
|
|
|
12.48
|
10.87
|
|
p value
|
|
|
0.341
|
|
|
0.000***
|
0.000***
|
|
R square
|
0.53
|
|
|
|
|
|
|
|
Confounders
|
Yes
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
Confounders: age, education, income, etc
*Means and SE are estimated by linear regression
**Robust Std. Errors
*** p<0.01