Introduction

Zoonotic infections can kill people, spread globally and have a devastating social and economic impact on affected regions (Dawood et al. 2012; WHO 2016). Intervention strategies aimed at interrupting spillover of zoonotic infections could reduce the risk of disease occurrence and its consequences.

Nipah virus (NiV) infection, a zoonotic disease transmitted from infected Pteropus bats to humans, is often fatal and can cause neurological sequelae among survivors (Chua et al. 1999; Reynes et al. 2005; Wacharapluesadee et al. 2005; Sejvar et al. 2007). Since 2001, NiV outbreaks have been identified almost every year in Bangladesh (Luby 2013). People in Bangladesh most commonly acquire NiV from drinking bat-contaminated raw date palm sap (Luby et al. 2006; Rahman et al. 2012; Hegde et al. 2016). Once a person is infected with NiV, they can transmit it to other people (Gurley et al. 2007; Homaira et al. 2010; Sazzad et al. 2013).

In Bangladesh, date palm sap harvesters, locally called gachhis, collect raw sap during cold months, from November to March, by shaving the bark of the tree and hanging a pot to collect sap overnight (Nahar et al. 2010). In 2009, the Government of Bangladesh began discouraging people from drinking raw sap to prevent NiV.

Traditionally, gachhis occasionally used skirt-like barriers called banas to cover the shaved area, the sap flow and the collection pot to stop bats, rodents, birds and insects from accessing the sap (Nahar et al. 2010). Banas can interrupt bats’ access to the sap, potentially preventing NiV spillover, and were acceptable to gachhis when promoted (Khan et al. 2012; Nahar et al. 2013, 2014). Some people reported stopping drinking raw sap after they learned about NiV (Sultana et al. 2013). To reduce the risk of NiV transmission, we implemented a behavior change intervention in two areas, promoting not drinking raw sap in one area and encouraging drinking only bana-protected sap in the second area. This study assessed the impact of these interventions.

Methods

Study Sites

We selected three NiV-affected districts—Rajbari and Faridpur for the interventions and Kushtia as a control. These three neighboring districts have similar population density, household size and literacy rate (Table 1) (BBS 2012). Date palm sap is harvested and consumed raw in these districts. From each district, we selected two adjacent sub-districts that did not share borders with the other intervention and control districts (Nahar et al. 2015). From Faridpur, we excluded 145 hard to reach villages located in river islands because they had no gachhis and no electricity, and could not run TV public service announcements, an important component of our intervention. Our study sub-districts included 342 villages in Rajbari, 381 in Faridpur and 276 in Kushtia. The approximate population of these sub-districts were 361,000 in Rajbari, 335,000 in Faridpur and 530,000 in Kushtia (BBS 2012).

Table 1 Demographic Characteristics of the “No Raw Sap” and the “Only Safe Sap” Intervention and Control Areas
Full size table

Study Design

We assessed two different community-based behavior change communication messages to improve knowledge about NiV and ultimately change behavior. In Rajbari district, we disseminated a “no raw sap” message for two seasons, discouraging community residents from drinking raw date palm sap (Fig. 1). Because we were delayed in securing Government of Bangladesh approval, in Faridpur district we disseminated an “only safe sap” message for only the second season. This message targeted community residents and gachhis, discouraging drinking raw sap but offering the option of drinking bana-protected sap. In both areas, local non-governmental organizations (NGOs) implemented the intervention by convening meetings and placing posters with key messages in public places with heavy traffic of people (Table 2). We broadcast televised public service announcements on closed-circuit local television, reaching both intervention areas exclusively, about five times daily, during the intervention period. Residents in both intervention areas had similar access to television and other communication channels used to disseminate the campaign messages (Nahar et al. 2017).

Figure 1
figure 1

Study activities in the “no raw sap” “only safe sap” and control areas between 2012 and 2014 in Bangladesh

Full size image
Table 2 Audience, Key Messages and Targeted Behaviors for the “No Raw Sap” and “Only Safe Sap” Intervention Areas, Implemented During 2012–2014 in Bangladesh
Full size table

To assess the effect of the intervention, we conducted baseline and endline surveys using pretested standardized questionnaires for face-to-face-interviews and observed gachhis collecting sap and sap selling points (Fig. 1).

During the 2012–2013 sap harvesting season, the evaluation team collected baseline survey data which asked about consumption in the prior year from community residents in Rajbari, the “no raw sap” area, and Kushtia, the control area. After the survey, the intervention consisted of meetings with opinion leaders and community residents, posters and TV public service announcements, from December 26, 2012, to March 29, 2013. During the 2013–2014 sap harvesting season, we continued with a limited intervention, only broadcasting the public service announcement from mid-November 2013 to January 2014 reminding residents of the risk of drinking raw sap. The evaluation team observed gachhis collecting and selling raw sap in the “no raw sap” intervention and control areas, between December and February during the 2012–2013 and 2013–2014 seasons. We did not conduct a baseline with gachhis in the “no raw sap” area because the message did not target them.

Before the 2013–2014 sap harvesting season, the evaluation team collected baseline data among community residents and gachhis in Faridpur, the “only safe sap” area, and Kushtia, the control area (Fig. 1). After the survey, we implemented the intervention from October 2013 to January 2014 including opinion leaders’ meetings, community residents’ meetings, posters and public service announcements. NGO workers also trained gachhis on making banas, encouraging their use on trees which sap was collected for raw consumption, provided stickers that gachhis could use to identify bana-protected sap pots, and offered sweatshirts as an incentive to use banas. Between December 2013 and February 2014, the evaluation team visited gachhis to observe them collecting and selling sap and to assess bana usage.

After the interventions ended in both areas, from March 31 to April 21, 2014, the evaluation team conducted endline surveys among community residents and gachhis from the intervention and control areas (Fig. 1).

During baseline and endline surveys, the evaluation team interviewed separate samples of adult men and women (Nahar et al. 2015). They collected data on respondents’ sap consumption behavior during the previous sap collection season, their exposure to the intervention communication channels and message recall from elements of the intervention. They interviewed gachhis on their NiV knowledge, number of trees harvested, use of banas and raw sap selling practices.

Sample Size Calculation

Based on our previous studies (Sultana et al. 2013) (Stephen P Luby, unpublished data), our intervention aimed for a 15% absolute reduction in proportion of people reporting raw sap consumption in the intervention areas, compared to the control area. For a difference in difference analysis, we calculated the desired sample size as 372 men and 372 women from 75 villages per area, per survey. We followed a probability proportionate to size sampling procedure, using 2011 census information on total population per village, to select villages from each area. From each village, we selected independent samples of men from six households and women from six other households, targeting 450 households per gender per area, allowing for a 15% refusal rate. We followed the same procedure for baseline and endline sampling.

To measure changes in gachhi behavior, targeting a 15% absolute increase in reported bana usage between the intervention and control groups in a difference in difference analysis, we calculated the desired sample size as 53 gachhis (from 27 villages) in the intervention and control areas. Since the evaluation team would have already visited 75 villages during the community surveys, we increased the sample to 150, two gachhis per village, per area per year.

The evaluation team asked community respondents to identify gachhis in their village and listed their names, phone numbers and household locations. The evaluation team used a Kish grid (Kish 1949), to select gachhis at random for interviews.

Observational Data Collection

Every two weeks, the evaluation team conducted observations in at least five different sites in the intervention and control areas, randomly selecting 10 gachhis per area, from the initial gachhi list. They observed their activity between 5:25 and 9:45 am, when they collected and sold raw sap, stopping when gachhis sold all their sap or started making molasses. They observed and recorded the number of trees harvested, trees covered by banas, amount of raw sap collected and consumed and the number of persons who consumed or purchased raw sap to take home. They estimated the amount of sap collected and consumed by observing the size of the pots and consulting with the gachhis.

Data Analysis

We compared respondents’ knowledge about NiV and sap harvesting and consumption practices between baseline and endline surveys. We assessed the relationship between intervention exposure and behavioral outcomes using a logistic regression model calculating odds ratios (OR). We adjusted for clustering within villages when calculating 95% confidence limits and P values. We used difference in difference analysis to compare changes from baseline to endline between the control and intervention groups.

Using the observation data, we calculated the median amount of raw sap collected and consumed; the mean number of people who consumed and purchased raw sap; and the proportion of observations during which at least one person consumed raw sap. We used difference in difference analysis to understand the changes in observed behavior among the control and the “no raw sap” intervention groups, from the 2012–2013 to the 2013–2014 sap season. We calculated the proportion of gachhis using banas and selling bana-protected sap for raw consumption, to understand the effect of the “only safe sap” intervention on bana usage.

Results

The evaluation team interviewed a total of 6220 community residents and 665 gachhis during baseline and endline surveys.

Knowledge About the Disease

Post-intervention 40% more respondents in both intervention areas reported knowing about a disease contracted through raw sap consumption compared with control. This increase was markedly higher in the intervention areas than in the control area. A larger proportion of gachhis reported knowing about NiV than community residents, both at baseline and endline (Table 3).

Table 3 Reported Community’s and gachhis Knowledge of Nipah, at Baseline and Endline, from the “No Raw Sap” and “Only Safe Sap” Intervention and Control Areas, Bangladesh 2012–2014
Full size table

Community Practices in the “No Raw Sap” Area

Reported raw sap consumption decreased markedly between the 2012 baseline and 2014 endline, from 43 to 18%, among residents of the “no raw sap” area, and from 57 to 40%, among residents of the control area (Table 4). The reduction in the intervention area was not significantly larger than in the control (−7%, 95% CI −15, 1%, P = 0.07). Reported raw sap purchasing from local gachhis and tree owners declined from 34 to 18% in the intervention area, a decline that was not significantly larger than in the control (−11%, 95% CI −22, 1%, P = 0.07). Reported collection of raw sap from one’s own trees or purchasing from market or mobile vendors for raw consumption did not change significantly among any group.

Table 4 Reported Community Raw Sap Consumption Behavior, at Baseline and Endline, from the “No Raw Sap” and “Only Safe Sap” Intervention and Control Areas, Bangladesh 2012–2014
Full size table

Community Practices in the “Only Safe Sap” Area

Reported overall raw sap consumption decreased markedly, between the 2013 baseline and the 2014 endline, from 60 to 44% among residents of the intervention area, and from 49 to 40%, in the control area. This reduction was not significantly larger than the control (−7%, CI −14, 2%, P = 0.12). Reported consumption of unprotected sap declined in the intervention area (59 to 26%) and the control (36 to 29%). This decline was significantly larger in the “only safe sap” area than in the control (−26%, 95% CI −33, −18%, P < 0.001) (Table 4).

Restricting the analysis to residents who consumed sap, reported consumption of bana-protected sap significantly increased in the “only safe sap” area (3 to 43%), while it hardly changed in the control (26 to 27%). In addition, reported inquiries about bana use before purchasing sap significantly increased in the intervention area but decreased in the control. Reported raw sap purchased from local gachhis significantly declined in the “only safe sap” intervention (39 to 30%) and control (38 to 25%) areas. However, reported buying sap from local gachhis increased among residents who reported consuming bana-protected sap (13 to 31%). This increase was significantly larger in the intervention area than in the control (27%, 95% CI 3, 50%, P < 0.05) (Table 4).

Gachhi-Reported Date Palm Tree Harvesting and Bana Use

In the “only safe sap” area, between the 2013 baseline and 2014 endline, the total number of trees harvested and those harvested for raw sap consumption significantly declined. In the control, there were fewer harvested trees than in the intervention area, but the number remained constant throughout the study. The 2014 endline data suggest a higher number of harvested trees in the “no raw sap” area than the control (Table 5).

Table 5 Reporting of gachhis on Their Harvested Trees and bana Usage During Baseline and Endline Data Collection from the “No Raw Sap,” “Only Safe Sap” and Control Areas, Bangladesh 2013–2014
Full size table

In the “only safe sap” area, gachhi-reported bana usage increased from 11 to 90% between the 2013 baseline and 2014 endline. In the control, an unexpectedly high proportion of gachhis reported bana use during the 2013 baseline, though it decreased during the 2014 endline (66 to 57%). The increase in the intervention area was significantly higher than in the control (Table 5).

Observation of Protected and Unprotected Raw Sap Consumption and Bana Usage

Comparing the 2012–2013 and the 2013–2014 sap harvesting seasons, in the “no raw sap” area, the observed amount of raw sap collected increased (median 36 vs. 50 L), while it decreased in the control (median 48 vs. 14 L). The proportion of observations of at least one person consuming raw sap at the gachhi’s place declined more in the “no raw sap” area (46 to 22%) than in the control (61 to 53%; difference in difference 95% CI −45, 14% and P = 0.30). During the 2013–2014 sap harvesting season, the evaluation team observed a higher number of harvested trees in the intervention area than the control. They observed a higher percentage of gachhis using banas in the control than in intervention villages (Table 6).

Table 6 Observation of Date Palm Sap Harvesting, Consumption and Selling at gachhis’ Households, from 5.25 to 9.45 am, Until gachhis Finished the Raw Sap Selling and/or Start Making Molasses, During the 2012–2013 and 2013–2014 Sap Harvesting Seasons
Full size table

Association of Exposure to the Intervention with Raw Sap Consumption

During the 2014 endline, more respondents from the “only safe sap” area than from the “no raw sap” area reported direct exposure to any element of the intervention (41 vs. 30%) as well as more indirect exposure by learning from others (36 vs. 28%) (“Appendix” of Table 11).

In the “only safe sap” area, respondents with direct exposure to at least one intervention element were more likely to report consuming raw sap from a protected source than those with no exposure (25 vs. 15%, OR 2.0, 95% CI 1.5–2.6, P < 0.001) (Table 7). Similarly, there were noticeable differences among respondents exposed to an individual element and those with no exposure. Respondents who attended community meetings were more likely to report consuming raw sap (54 vs. 43% OR 1.5, 95% CI 1.0–2.3, P = 0.02) than those who did not. They were also more likely to report consuming raw sap from a protected source than those who did not attend community meetings (38 vs. 16%, OR 3.1, 95% CI 2.1–4.6, P < 0.001). In addition, respondents who saw a poster were more likely to report consuming raw sap from a protected source than those who did not (26 vs. 16%, OR 1.8, 95% CI 1.3–2.5, P < 0.001). Respondents who watched the public service announcements were more likely to report consuming raw sap from a protected source than those who did not (27 vs. 18%, OR 1.7, 95% CI 1.0–2.7, P < 0.05) (Table 7). By contrast, exposure to individual elements of the “no raw sap” intervention was not associated with reported avoidance of raw sap consumption.

Table 7 Association of Exposure to Interventions with Reported Consumption of Raw Date Palm Sap During Previous Sap Harvesting Season in the Intervention Areas—Endline 2014
Full size table

Discussion

Respondents’ knowledge of NiV transmission in both intervention areas markedly improved, while there was no significant change among controls. Reported raw sap consumption declined in both intervention and control areas. Reported bana usage and consumption of bana-protected sap increased in the “only safe sap” area. Direct exposure to the intervention was significantly associated with drinking sap from a protected source in the “only safe sap” area.

The primary objective of the “no raw sap” intervention area was to reduce raw sap consumption. We expected a 15% absolute reduction in the proportion of people reporting raw sap consumption in the intervention area and observed a 25% absolute reduction, but we also observed an unexpected 17% reduction in the control; thus, the reduction in the intervention area was not significantly different than in the control. Our observational data found no change in the mean number of persons consuming raw sap at the gachhis place nor in the amount consumed, suggesting that the decrease in the number of people reporting sap consumption may have been due to social desirability bias (Wood et al. 2008). We also did not find any association between exposure to the intervention and decline in raw sap consumption. Overall, the evaluation does not provide compelling evidence that the “no raw sap” message markedly reduced raw sap consumption.

In the “only safe sap” area, there was no reported or observed reduction in raw sap consumption. However, reported consumption of bana-protected sap, inquiring about the use of bana prior to sap purchasing, and drinking bana-protected sap from the local gachhi significantly increased in comparison with the control. Direct exposure to the intervention was associated with increased consumption of raw sap, increased consumption of protected sap and reduced consumption of unprotected sap.

In the “only safe sap” area, reported bana usage significantly increased and exceeded the 15% absolute increase projection, though observed bana usage was much lower than that reported by gachhis. This higher reporting might be due to social desirability bias. However, in the “only safe sap” area, the observed proportion of gachhis using banas during the intervention period was about three times higher than the reported bana usage at baseline, suggesting that bana use increased because of the intervention.

The “no raw sap” message did not achieve the expected outcome. Exposure to community meetings and posters occurred one year prior to conducting the survey, probably affecting recall. Although we re-broadcast the public service announcement the second year, exposure to it, during two seasons, was not related to behavior change. This might be because a two-season intervention was not enough to eliminate an existing food behavior such as drinking raw sap. People acquire eating behaviors over a lifetime, and changing them requires alterations in habits with long-term interventions (Nestle et al. 1998). Although many respondents reported ceasing drinking raw sap, continuous intervention efforts may be required to ultimately modify raw sap consumption behavior. When raw sap is available, it might be difficult for people to abstain from drinking it (Nahar et al. 2015). Indeed, drinking raw sap from one’s own household trees remained constant among intervention and control groups over time.

Even though the “only safe sap” message was disseminated for a single season, reported community and gachhi behaviors were in line with the expected outcomes. The intervention offered the option of continuing an existing food behavior rather than completely eliminating a preferred food item. The use of bana and drinking bana-protected sap was an already existing occasional practice among some gachhis (Nahar et al. 2010). The cultural environment and social interaction exerts a strong influence on perceptions about food, food choices eating behavior (Nestle et al. 1998; Shepherd and Shepherd 2002). Thus, having the option to drink safe sap might be more acceptable compared with completely stopping consumption.

The “only safe sap” message is a harm reduction approach that recognizes abstinence as an ideal outcome but accepts alternatives that reduce harm (Marlatt 1996). These approaches have proven useful for reducing other public health risks, such as HIV transmission through needle sharing among drug users (Aspinall et al. 2014). Harm reduction studies demonstrate that modest changes to behaviors are easier to achieve than more substantive changes (Luby et al. 2005; Plautz and Meekers 2007; Kirby 2008). We know that some people drink raw sap even after learning about the risk of NiV (Nahar et al. 2015). Asking them to drink only safe sap is less demanding than asking them to stop drinking sap altogether. It also targeted gachhis who are the source of sap and could be held responsible for NiV transmission, giving them a safe option to provide sap. Thus, disseminating an “only safe sap” message may be a more pragmatic strategy to reduce the risk of NiV transmission.

Our study has limitations. Neither the community nor our evaluators were blinded to the intervention. Our primary outcome was measured through reported behavior. Although social desirability bias may have induced respondents to underreport their sap consumption practice and over-report bana usage (Wood et al. 2008), there is little reason to expect more social desirability bias in the “only safe sap” than in the “no raw sap” area. To interpret our reported data, we looked at changes in knowledge and observation data, though the number of observed outlets was small and the presence of our observer might have altered some behavior. Nevertheless, the association between exposure to specific elements of the intervention and reported safe sap behaviors suggests behavior change resulted from the “only safe sap” message but not from the “no raw sap” message.

Our intervention and control areas were not comparable in terms of number of date palm trees harvested and preexisting behaviors related to bana usage. The constraints of a mass media intervention trial requiring large units of intervention make balance more difficult to achieve than a trial with many smaller units of intervention that permits random assignment. Without any intervention, raw sap consumption decreased among community respondents in the control area, and this decline remains unexplained. There might be other characteristics specific to the districts that contributed to different practices in different years, rather than the intervention. Further research to better explicate year-to-year variability in sap consumption practices, which may be related to weather, sap harvest productivity, competing employment opportunities for gachhis or other factors, may help and guide future interventions and evaluations. Nevertheless, the changes in the “only safe sap” area likely resulted from the intervention, because of the association between exposure and behaviors and the significant changes identified in the difference in difference analysis between intervention and control groups on a number of outcomes. The control districts provided a useful counterfactual illustrating underlying variability and supporting a more conservative interpretation than a design that would have only looked at baseline and endline.

We did not fully understand why a high proportion of gachhis from the control area used bana at baseline and endline. Perhaps, since gachhis from the control area harvested a small number of trees, they have less work related to tree harvesting and more time to make banas. In our earlier work, even without intervention, some gachhis reported to occasionally using banas to collect more, cleaner sap (Nahar et al. 2010, 2014). Other gachhis recalled their previous experience seeing colleagues using banas more frequently when harvesting fewer trees (Nahar et al. 2010, 2014). In the “only safe sap” area, the proportion of gachhis that used banas increased, thus increasing their workload, which may have resulted in the significant decline in the number of harvested trees after the intervention. Better understanding of control area gachhis’ motivation to use banas without any intervention might provide useful insights to support the expansion of this intervention strategy.

Our measurements were unable to confirm if people reporting drinking protected sap, actually consumed protected sap. Their ability to reliably assess whether the sap was bana-protected or not likely depends on how well they know the gachhi. Future studies could investigate the effect of the message to ask for bana-protected sap from local gachhis (Table 2) and to observe actual use of banas.

The “only safe sap” message resulted in changes in reported behavior that may reduce the risk of NiV spillovers; thus, this intervention could be further promoted and evaluated. Prospective efforts to track raw sap consumption practices, and explore year-to-year variation in Bangladesh, would be particularly useful to guide government policy.