Participant Demographics
In April 2019, one-hundred and forty-one respondents were recruited and randomised into one of eight arms. Figure 2 presents the CONSORT flow chart. There were no withdrawals, and all respondents who wished to take part were eligible. Table 2 reports the summary statistics for the study cohort. The average age of the respondents was 28·9 years – with most (92·2%) having completed primary school or above. Respondents were predominantly female (97·9%), and most (51·1%) had a household income below 50,000TZS (21·75USD) a month. The average household size was 5·5 people.
Table 2: Summary statistics of demographics and quantitative study outcomes for all participants, by study arm
|
|
A
|
B
|
C
|
D
|
E
|
F
|
G
|
H
|
ALL
|
Total, n (%)
|
|
17 (12·1)
|
18 (12·8)
|
18 (12·8)
|
18 (12·8)
|
18 (12·8)
|
18 (12·8)
|
18 (12·8)
|
16 (11·3)
|
141
|
Age, Mean (SD)
|
|
28·9 (6·2)
|
28·4 (6·8)
|
27·6 (5·2)
|
30·4 (9·2)
|
30·3 (6·7)
|
29·3 (6·7)
|
29·9 (8·1)
|
27·9 (7·2)
|
29·1 (7·0)
|
Education, n (%)
|
None
|
0
|
0
|
0
|
1 (5·6)
|
0
|
0
|
2 (11·1)
|
1 (6·3)
|
4 (2·8)
|
Some Primary
|
0
|
0
|
0
|
1 (5·6)
|
1 (5·6)
|
1 (5·6)
|
3 (16·7)
|
1 (6·3)
|
7 (5·0)
|
Finished Primary
|
10 (58·9)
|
10 (55·6)
|
13 (72·2)
|
11 (61·1)
|
9 (50·0)
|
12 (66·6)
|
8 (44·4)
|
11 (68·8)
|
84 (59·6)
|
Some Secondary
|
1 (5·9)
|
4 (22·2)
|
0
|
2 (11·1)
|
0
|
4 (22·2)
|
3 (16·7)
|
2 (12·5)
|
16 (11·3)
|
Finished Secondary
|
6 (35·3)
|
4 (22·2)
|
3 (16·7)
|
2 (11·1)
|
8 (44·4)
|
1 (5·6)
|
2 (11·1)
|
0
|
26 (18·4)
|
Some Tertiary
|
0
|
0
|
2 (11·1)
|
0
|
0
|
0
|
0
|
1 (6·3)
|
3 (2·1)
|
Finished Tertiary
|
0
|
0
|
0
|
1 (5·6)
|
0
|
0
|
0
|
0
|
1 (0·7)
|
Sex, n (%)
|
Male
|
0
|
0
|
0
|
1 (5·6)
|
2 (11·1)
|
0
|
0
|
0
|
3 (2·1)
|
Female
|
17 (100%)
|
18 (100%)
|
18 (100%)
|
17 (94·4)
|
16 (88·9)
|
18 (100%)
|
18 (100%)
|
16 (100%)
|
138 (97·9)
|
Income, n (%)
|
>50,000
|
8 (47·1)
|
10 (55·6)
|
8 (44·4)
|
12
|
8 (44·4)
|
9
|
8 (44·4)
|
9 (56·3)
|
72 (51·1)
|
50,000-100,000
|
5 (29·4)
|
4 (22·2)
|
7 (38·9)
|
6 (35·3)
|
7 (38·9)
|
4 (22·2)
|
4 (22·2)
|
7 (43·8)
|
44 (31·2)
|
100,000-500,000
|
2 (11·8)
|
3 (16·7)
|
2 (11·1)
|
0
|
3 (16·7)
|
3 (16·7)
|
4 (22·2)
|
0
|
17 (12·1)
|
500,000+
|
0
|
0
|
0
|
0
|
0
|
1 (5·6)
|
1 (5·6)
|
0
|
2 (1·4)
|
Decline to say
|
2 (11·8)
|
1 (5·6)
|
1 (5·6)
|
0
|
0
|
1 (5·6)
|
1 (5·6)
|
0
|
6 (4·3)
|
Area, n (%)
|
Unguja
|
6 (35·3)
|
6 (35·3)
|
6 (35·3)
|
6 (35·3)
|
6 (35·3)
|
6 (35·3)
|
7 (38·9)
|
7 (43·8)
|
50 (35·5)
|
Igogo
|
6 (35·3)
|
7 (38·9)
|
7 (38·9)
|
6 (35·3)
|
6 (35·3)
|
6 (35·3)
|
6 (35·3)
|
5 (31·25)
|
49 (34·8)
|
Kilmahaewa
|
5 (29·4)
|
5 (27·8)
|
5 (27·8)
|
6 (35·3)
|
6 (35·3)
|
6 (35·3)
|
5 (27·8)
|
4 (25·9)
|
42 (29·8)
|
Household Size, Mean (SD)
|
|
5·6 (2·2)
|
4·9 (1·7)
|
5·7 (2·2)
|
5·4 (3·2)
|
5·1 (1·6)
|
5·9 (2·6)
|
6·8 (3·0)
|
4·8 (2·1)
|
5·53 (2·4)
|
14-day Diarrhoea Prevalence, %
|
|
34·67
|
27·12
|
38·26
|
42·86
|
40·62
|
30·36
|
24·00
|
52·38
|
36·35
|
Complete Response Rate, %
|
|
55·15
|
42·75
|
79·86
|
48·61
|
44·44
|
38·89
|
34·72
|
32·81
|
47·33
|
Survey Response Rates
Over the course of the study, between April and September 2019, 8,215 surveys were distributed: 7,655 daily texts and 560 fortnightly texts, with an even split between the high question load (3-question) and low question load (1-question) surveys, and incentive and no incentive. These can be broken down into 1,122 child-rounds of observation (each round lasting two weeks). The trial concluded in September 2019 when all arms progressed through all treatment combinations.
The mean response rate was 47%. Daily questioning had a similar mean response rate to fortnightly questioning (46·6 % vs 48·0%); the 3-question survey was lower than the 1-question (43·8% vs 51·0%); and the incentivsed surveys was higher than the surveys without incentive (50·6% vs 44·0%) (Fig. 3). When examining mean response rates by interactions between treatments, there was little evidence of any interaction between treatments, other than response rates being lower when daily questioning and the 3-question survey were combined (Fig. 4). Response rates increased as the study progressed (Fig. 3).
Table 3 reports the results from the adjusted model-based analysis. Daily questioning was associated with a non-significant reduction in the response rate by a 1·2 percentage point difference (ppd) (95%CI[-4·9,2·5]), compared to fortnightly questioning. The 3-question survey was associated with a significant reduction of response rates by 7·0ppd (95%CI[-10·8,-3·1]) compared to the 1-question survey. Incentivisation was associated with a significant increase in response rates by 6·5ppd (95%CI[2·6,10·2]) compared to no incentive.
There was also evidence that respondent age affected response rates, with each additional year of age being associated with an increased in response rate by 1·1ppd (95%CI[0·2, 2·1]), as did time, with a 0·9ppd (95%CI[0·0,1·7]) increase per round (Table 3). Having education beyond the primary stage was associated with an increase in response rates by 11·7ppd (95%CI[-1·6,25·1]) when compared to having primary education or lower. Having a low income (below 50,000TZS) was associated with a decrease in response rate by 3·8ppd (95%CI[-16·2,8·6]) when compared to middle or high income (Table 3).
Table 3: Estimated Adjusted Treatment Effects and Effects of Demographic Factors on Response Rate
|
Adjusted Treatment Effect (percentage point difference, (95%CI))
|
Daily Recall vs 14 day Recall
|
-1·2 (-4·9,2·5)
|
3-question Survey vs 1-question Survey
|
-7·0 (-10·8,-3·1)
|
Incentive vs No Incentive
|
6·4 (2·6,10·2)
|
Age of Respondent (continuous in years)
|
1·1 (0·2,2·1)
|
Beyond Primary Education vs Primary Education or Lower
|
11·7 (-1·6,25·1)
|
Low Income (below 50,000TZS) vs Middle or High Income
|
-3·8 (-16·2,8·6)
|
Study Round (continuous)
|
0·9 (0·0,1·7)
|
Kilimahewa vs Igogo
|
-3·4 (-18·6,11·2)
|
Unguja vs Igogo
|
-6·0 (-21·2,9·1)
|
Estimated Diarrhoea Rates
Overall, 36·4% of the 14-day child-rounds reported diarrhoea. When broken down by treatment, daily questioning had an estimated diarrhoea rate of 51·2% (compared to 21·9% for fortnightly questioning); the 3-question survey had a 36·3% estimated diarrhoea rate (compared to 36·4% for the 1-question survey); and the incentivesed surveys had a 38·7% estimated diarrhoea rate (compared to 33·6% for surveys without incentivisation) (Fig. 3). When looking at the impact of interactions between interventions on diarrhoea rate, we see a similar trend, with all treatment combinations that included the fortnightly survey having a similar lower estimated rate, regardless of interaction (Fig. 4). The estimated diarrhoea rate appeared to decrease as the study progressed (Fig. 3)
Table 4 reports the results from the model-based analysis. Compared to fortnightly questioning, daily questioning was associated with a significant increase in the estimated diarrhoea rate, with an adjusted treatment effect of 29·9ppd (95%CI[22·8,36·9]). There was no evidence to suggest that the 3-question survey had a significant impact on the estimated diarrhoea rate, with an adjusted treatment effect of 0·0ppd (95%CI[-6·0,5·9]). There was little evidence indicating that financial incentivisation had a significant impact on the estimated diarrhoea rate, with the incentive raising the estimated diarrhoea rate by 3·0ppd (95%CI[-3·1,9·0]).
Evidence showed an impact by respondent age, with each additional year in age associated with a decrease in the estimated diarrhoea rate by 1·2ppd (95%CI[-2·2,-0·2]), but not by other demographics (Table 3). Evidence also indicated a decrease in the estimated diarrhoea rate over the course of the study by 2·9 ppd per round (95%CI[-4·3,-1·5]) .
Table 4: Estimated Adjusted Treatment Effects and Effects of Demographic Factors on Estimated Diarrhoea Rate
|
Adjusted Treatment Effect (percentage point difference, (95%CI))
|
Daily Recall vs 14-day Recall
|
29·9 (22·8,36·9)
|
3-question Survey vs 1-question Survey
|
-0·0 (-6·0,5·9)
|
Incentive vs No Incentive
|
3·0 (-3·1,9·0)
|
Age of Respondent (continuous in years)
|
-1·2 (-2·2,-0·2)
|
Beyond Primary Education vs Primary Education or lower
|
3·7 (-10·2,17·6)
|
Low Income vs Middle or High Income
|
-0·3 (-12·9,12·3)
|
Study Round (continuous)
|
-2·9 (-4·3,-1·5)
|
Kilimahewa vs Igogo
|
4·5 (-10·5,19·6)
|
Unguja vs Igogo
|
-1·6 (-16·7,13·6)
|
Qualitative Findings
A high degree of acceptance with the SMS surveying system was observed during the analysis of the qualitative interviews. Participants were accustomed to using mobile phones, as they use them in daily life for work, communicating with friends and family, and studying. Participants reported that the messages were not perceived as intrusive and that late morning receipt of messages was convenient. Participants further reported appreciation of the reminders, as they were sometimes busy when the first message came. Participants did, however, prefer infrequent questioning, stating that they believed that they were able to recall diarrhoea over 14-days.
Participants generally stated that while the incentive was appreciated, it did not factor into their decision whether or not to take part in the survey. Participants reported appreciation of being able to feedback on their child’s health and that the messages encouraged the carers to pay more attention to their child’s health.
Participants were mixed regarding preference towards face to face surveys vs SMS surveys, but for the most part, appreciated the ease and privacy of SMS surveying. Participants suggested that in the future SMS and face to face be integrated, with more emphasis on education rather than purely surveying.