Designing a behavioral intervention using the COM-B model and the theoretical domains framework to promote gas stove use in rural Guatemala: a formative research study

Forty-one percent of the world’s population cook with solid fuels (e.g. wood, coal, crop residue, and animal dung) [1, 2]. Inefficient solid fuel stoves produce large amounts of particulate matter (PM) and carbon monoxide (CO), as well as a range of toxic, mutagenic or carcinogenic by-products [3‐7]. In 2016, household air pollution (HAP) was the 8th leading risk factor for death globally, attributable to between 2.8 million [8] and 4 million deaths annually [9, 10]. HAP is associated with many chronic diseases such as ischemic heart disease, stroke, chronic obstructive pulmonary disease, cataracts and lung cancer [10, 11]. For children, the damaging effects of HAP are often severe, ranging from pneumonia, low birth weight, stillbirth and preterm birth in children [12‐14]. The global burden of disease from HAP is concentrated in the poorest households in low-income countries. However, given scant evidence of relationship between HAP and low birth weight and preterm birth, these two health outcomes are not included in HAP-related global burden of disease estimates [15], despite their contribution to a large proportion of infant mortality [16]. It is therefore important to conduct studies that assess household air pollution reductions among pregnant women, and subsequent birth outcomes.

Chimney cookstoves that vent indoor air pollution outdoors or clean fuel stoves that use electricity, solar or liquefied petroleum gas (LPG) have the potential to improve health. However, little attention has been paid to how household members participate in decisions around cooking practices. Shifting households from using traditional stoves to a new LPG stove may improve indoor air quality only if the stove meets the demands of the household and is used consistently. Studies need to examine current stove practices, including the “entrenched practices” that make it difficult to “de-implement” [17], or abandon, the traditional stove. These “entrenched practices” may go beyond intervening with the primary cook, and may slip into the realm of other important decision-makers, such as  the cook’s mother-in-law or spouse. Resistance to de-implementation may also be influenced by poverty, lack of access to clean fuels, or discriminatory practices against the poor, including denying loans for new stove purchases. Similarly to de-implementation studies in the field of medicine [18], even the cleanest stove will not be used if social and structural forces are not acknowledged.

A well-known problem for health practitioners and researchers is addressing the “know-do gap” that lies between knowing something needs to be changed, and having the ability to make changes [19, 20]. This same gap exists in poor households relying on solid cooking fuels. Educating women about health risks from smoke exposure will do little to change their daily need for cooking under high exposure conditions. In these settings, women are constrained by fuel supply and cost, the need to prepare food in a certain way, or other household members who prefer to use the traditional stove instead of the gas stove [21].

To date, three randomized stove intervention trials in Malawi [22], Mexico [23, 24] and Guatemala [25] have shown that even the most fuel efficient, well-maintained solid fuel stoves have not succeeded in lowering HAP levels sufficiently to reduce infant pneumonia [26]. The reasons for this lack of translation into practice are numerous including:

Implementation science has developed tools and methods to understand and intervene upon the resistance to the successful adoption and use of a clean fuel stove and the pervasive exposure to additional sources of HAP. However, few studies in this field have used an implementation science framework to systematically assess and address the underlying barriers and enablers of this household risk factor that affects three billion people globally [31].

Since 2000, we have conducted studies to reduce HAP in rural Guatemala where over 90% of rural households cook with wood [1]. We recently began a study using an implementation science approach to develop an intervention that promotes the sustained use of an introduced LPG stove. We will first discuss the evolving stove research before presenting the current study.

During the RESPIRE randomized stove intervention study (2002–04), a well-maintained wood-fired chimney stove (referred to as the plancha) reduced kitchen CO levels by 90%, but only reduced maternal personal exposure by 61% and children’s personal exposure by 52%, compared to control homes with open cooking fires [32]. During the NACER study conducted in the same region (2012–13), 48-h personal PM_2.5 exposures among women who used the plancha ranged from 130 μg/m³ (standard deviation, SD 65) during pregnancy to 63.8 μg/m³ (SD 17.3) during the neonatal period [33]. However, the chimney stove was not able to reduce absolute PM_2.5 to levels needed to protect human health (World Health Organization interim target of 35 μg/m³ averaged over a 24-h period) [34]. Based on these and similar findings around the world, there has been a recent call for an acceleration of research that go beyond locally accepted “cleaner” solid fuel stoves [31, 35]. However, full adoption in low-resources settings of the “cleanest” cookstoves (e.g. LPG, biogas, ethanol, electricity or solar) has yet to be determined.

Theory-driven behavioral change interventions that are grounded in an understanding of the enablers and barriers to change, and that address these factors head-on, can begin to fill in the “know-do” or implementation gap. A review of 55 stove interventions in 20 countries found that most behavior change techniques were employed during the adoption of a clean cooking stove, when accepting to use a new stove becomes the essence of program success or failure [36]. To our knowledge, no studies to date have applied behavior change theory to influence the sustained use of clean cookstoves among pregnant women exposed to HAP. The COM-B Model is a comprehensive “behavior system” that provides the structure for assessing capabilities, opportunities and motivations, the fundamental conditions that must be understood at multiple levels, prior to the development of effective interventions. Capabilities are demonstrated by necessary skills and knowledge. Opportunities are the social or environmental factors that are external to an individual and allow change to occur. Motivation is guided by processes, either reflective or emotional, that direct behavior. The COM-B Model forms the hub of the Behavior Change Wheel (BCW) framework and nine intervention functions surround this hub. The nine intervention functions address the gaps in capabilities, opportunities and motivations that, when implemented correctly and consistently, will lead to sustained behavior changes [37, 38]. Related to the COM-B Model, the Theoretical Domains Framework (TDF) is a synthesis of 84 theoretical constructs of behavior change organized into 14 domains [39] . Each of the COM-B components maps onto several unique TDF domains [38].

The COM-B Model/BCW has been used to develop interventions for low-income cancer patients [40], stroke victims [41], preventative health checklists for children [42] and diabetes prevention for Latina women with gestational diabetes [43]. One recent study in Uganda used an opportunity-ability-motivation model [44] to identify the behavioral determinants that would lead to the purchase and use of a locally-produced gasifier stove [45]. Studies like this one, which characterizes formative research within a behavior change theory and an application framework similar to the COM-B/BCW, allow researchers to anticipate the bottlenecks before the introduction of the stove into the local market [45].

The overall purpose of the present study, referred to as NACER II (nacer means “to be born” in Spanish), is to implement and evaluate an introduced LPG stove coupled with a phased behavioral intervention to encourage use of the gas stove over traditional stoves among a group of 50 pregnant women in rural Guatemala. We focused on pregnant women because these young women may be the most amenable to behavior change messaging (to protect their unborn infants from harm) and because the fetus is susceptible to air pollutants, perhaps even more so in utero than during infancy [46, 47]. In this Guatemalan population, we have found that women do not receive preferential treatment that would reduce their exposure to cooking smoke during pregnancy. However, during the immediate post-partum period (and up to three weeks after birth when they observe the dieta, or are in repose), other women are tasked with cooking. Thus, our messages during pregnancy included other family members who cooked in the home. Here we describe the formative research to design and implement a behavioral intervention to promote the sustained and exclusive use of the gas stove using the COM-B framework, Theoretical Domains Framework (TDF) and BCW. This study is innovative in its approach to reducing exposure to HAP, which includes provision of a gas stove complemented by a tailored behavioral intervention and implemented by peer educators from local communities. We aim to assess the feasibility and acceptability of a tailored intervention that targets the entire family to be used in a future randomized, controlled trial.

This prospective cohort study began in May 2016 and concluded in June 2017. Participants were screened for eligibility by fieldworkers during their prenatal visits at the Guatemalan Ministry of Health Clinic in the Western Highlands of Guatemala. Fifty women who met eligibility criteria, and who were < 20 weeks gestation confirmed by physician-administered ultrasound, were enrolled in the study. Inclusion criteria were: 18–45 years old; non-smoking; responsible for cooking meals in the home; use wood-fired open fire or deteriorated chimney-stove (e.g. 2 or more defects in stove leading to observable smoke seepage into kitchen); purchase wood fuel (if they gathered free wood, more likely to continue wood stove use); no plans to migrate in the next year; and have a cell phone that we can contact (86% of households in the area have a cell phone). Exclusion criteria were: use gas stove as their primary stove in the home; multigestational pregnancy, associated with preterm birth and low birth weight; and, because women wore personal air pollution monitoring equipment during pregnancy, we excluded women with a child < 2 years routinely carried on mother’s back. After women met initial eligibility criteria, we visited them in their homes to assess the condition and types of cookstoves, to assure that their biomass stove was in poor condition and that they did not have a gas stove in the home. At that visit, women who agreed to participate signed a written consent. All women were visited at < 20 weeks, 24–28 weeks, 32–36 weeks, within 48 h of birth and when infant reached 1 month age. At three months, we visited the households to conduct a final survey about stove use at which time they were disenrolled from the study.

This study was approved by the Committee for Human Research at the University of California, San Francisco and the Ethics Committee at Universidad del Valle de Guatemala. All women provided written consent to participate after consents were verbally read to them in their preferred language, either Mam or Spanish. If women could not sign, they provided a thumbprint and the consent was witnessed by a third party.

Focus group findings were summarized into 28 themes organized into 11 categories (Fig. 1) each with a critical role needed to support LPG stove use. Emerging themes were then categorized into whether they would be a barrier or would enable LPG stove use (Table 1). The first theme described was environmental influences, which included 1) wood fuel quality/accessibility, and 2) the social context for undertaking a cooking activity considered outside the community norm. For the first, wood can either be gathered freely on owned land, or can be purchased from delivery trucks that pass by homes. Participants viewed wood supply as abundant and accessible. However, wet or “green” (uncured) wood is very smoky and “bad to use”; during the rainy season, some women indicated that they have no choice but to use wet wood. For the second, participants described not wanting to appear different from their neighbors, to avoid envy or shame. Acquiring household goods, like a new gas stove, is desired, but discussing these purchases with neighbors was viewed negatively. One woman said, “We are all the same”.

The second category was fuel and stove preferences. All women who owned a plancha stated that they prefer the plancha because “you can cook everything on it”. Women in the open-fire focus group were younger and newly married. They aspired to own a clean stove, but were living in their mother-in-laws’ homes, and could not yet afford to buy a stove. Women cook together in the home, and stated that older women are accustomed to using the open fire or the plancha and that it would be hard for them to learn to use an LPG stove. Only five participants with plancha stoves also had LPG stoves; they discussed the difficulty with LPG tank availability. LPG fuel tanks were delivered to households, but more commonly were picked up from a distributor in town. Women did not know whether wood or LPG gas was more costly. In these high-altitude communities of Guatemala, women felt that the plancha kept the house warm while the gas stove provided no additional warmth.

Proper stove use and maintenance was the third category. Having a chimney in good repair was important, but women stated that repairing or replacing the chimney required cooperation from their husbands, who would need to provide money for repairs. Women mentioned that smoldering fires produced smoke. One woman discussed blowing through a plastic hose to stoke the fire in the plancha; this kept her from leaning into the fire. In response, another woman joked, “we don’t do that, even the baby learns to blow on the fire”. The fourth category was general knowledge about HAP. This concerned issues of air pollution and climate change. One woman stated, “Burning trash, like plastic, heats up our environment and makes the sun hotter”. Another stated “we have heard about this on television”.

Child health and safety included many themes, but the most predominant themes were danger from burns when a child is near an open fire or danger from explosions when gas stoves or fuel tanks are improperly used. None of the women had experienced this, but had heard stories from other women about these potential dangers. To keep children safe, women mentioned keeping children outside the kitchen or asking other family members to care for children. Several younger women said they didn’t have support from others, and had to carry their children on their backs while they cooked. Health related to smoke exposure was the sixth category. One woman mentioned that a doctor told her that her child’s pneumonia was “caused” by smoke; another stated “doctors and nurses tell us smoke is bad for babies.” Women stated that breathing smoke “affects the lungs and eyes, especially in older women”.

The seventh category, hygiene related to smoke exposure, included a discussion of kitchens with clean instead of sooty walls, pots that were shiny and not blackened, and clothes that did not smell smoky, which made some women feel shameful. Ventilation, the eighth category, was possible if doors and windows were kept open, especially during the start-up of the fire when smoke was thick. Some women stated that elder family members preferred to keep the windows shut during cold weather; women did not feel that they could question those decisions. Stove placement was important. If the open fire was placed in a well-ventilated area, near a window, smoke would be minimized. Others stated that too much wind circulating in the kitchen made it difficult to control the gas flame. The ninth category was construction. Most women stated that men’s decisions around spending household income limited them from making changes they preferred, like adding a window to the kitchen. The women in the open fire group were thinking about new kitchen construction, but were unable to do so because they had not yet saved enough money to purchase materials and still lived with their mother in law. Modernization was the tenth category. Women spoke of the “old days” when women used clay pots on open fires, and about their current desire to use metal pots that stayed shiny because they use clean stoves. The final category was refuse disposal. Outdoor trash fires were burned to dispose of garbage, such as rubber shoes and plastic containers, because there was no garbage service. Open fires were used to cook animal food, and women mentioned burning plastic in these fires because they had no other way of disposing of ubiquitous plastic bags. Several women composted organic material and sold plastic bottles, but most were resigned to burning the inorganic solids that accumulate in their homes.

During NACER II, 238 women were approached at the antenatal clinic by trained, Mam-speaking fieldworkers. Of those, 73 met the eligibility criteria and were < 20 weeks confirmed by ultrasound. Of the 73, two did not wish to participate in the study, two had miscarriages before they consented to participate in the study, six had chimney stoves in good repair and three already had a gas stove at the stove inspection visit, and ten could not be located for the stove inspection visit.

Among the 50 women who participated in the study, the mean gestational age at baseline was 13.1 weeks (range, 6.7–19.7 weeks) and for 14% of the women this was their first pregnancy (Table 2). Twenty-seven (56%) and 23 (46%) were recruited in the first and second trimester, respectively. The majority (86%) self-identified as Mam-indigenous. None of the households had a stove in the same room where they sleep; 88% percent had a stove in a separate structure from the main house. Ten (20%) used an open fire and 40 (80%) used a plancha stove without a chimney or in a deteriorated condition as their principal stove. Additional sources of smoke were as follows: 36 (72%) used a wood-fired steam bath (a temascal, or chuj) for bathing [53], 33 (66%) burned garbage near the home, 23 (46%) cooked animal food outdoors over an open fire and 7 (14%) were exposed to second-hand smoke (median of 1 cigarette/week smoked at home).

Twelve to thirteen women attended each class in the 3-part series. Attendance rate was 93% for Phase 1 and 92% for Phase 2. Classes were based on focus group themes (Table 3). For example, we discussed feasible ways to reduce exposure to HAP since households often continued to use wood stoves. Nine behaviors were discussed, based on the evidence for HAP sources, and from the focus group findings: 1) cook in well-ventilated areas where air can circulate freely; 2) take turns cooking with other women in the family; 3) have other people care for the baby or children; let older children play outside during cooking; 4) find ways to prepare food that require less cooking time (while still thoroughly cooking food); 5) use dry wood, not wet wood, for cooking; 6) DO NOT burn plastic or garbage inside the kitchen; 7) make your fire at a distance no greater than 1.5 m from the closest window so that smoke can exit through the window; 8) keep windows and door open while the fire is burning; 8) use a fan or blow through a tube to fuel the fire; and 9) use LPG stoves that produce less smoke. These behaviors were reviewed at each class. Because we heard from women that stove safety was a concern, we reviewed gas stove and fuel tank safety, including how to detect and respond to a gas leak. We conducted one cooking event at each session: 1) a porridge drink (atol) was cooked at the first session to demonstrate ease of use, and how to cook a liquefied with a regulated flame to keep it from boiling over; 2) a “blind taste test” of beans cooked on an LPG stove and a plancha was conducted at the second session, with discussion about preference for one or another; and 3) a “cook off” of atol on an LPG stove and over an open fire was done at the final session, with a discussion of time savings, taste and appearance of the drink, and what women would do with time freed up from gathering wood, building fire and cooking for longer periods of time. Finally, at the last class, participants broke into three groups and acted out a socio-drama. They negotiated with “family members” (other participants) about refilling the LPG tank. This activity helped them anticipate discussions they would have at home after the study ended and fuel refills were no longer free.

During home visits, the stove technician provided LPG tank refills and stove maintenance and reinforced messages around checking for gas leaks and keeping the stove clean and in good repair. We provided all homes with a calendar that illustrated safety messages about the gas stove. We used the calendar to keep track of delivered LPG tanks. At each visit, we discussed usage patterns, especially if the interval between tank visits was longer or shorter than expected.

In Phase 2, we applied the COM-B model [37, 38] and the 14 related domains in the TDF [39]. Of the 84 TDF constructs, we mapped 35 constructs in 14 domains to questions we developed about behaviors related to LPG stove use. The remaining components were not relevant to the behavioral diagnosis. For example, professional roles and identity were not relevant for women who work at home. Punishment or sanctions, part of reinforcement, were not socially acceptable.

We attempted to be as specific as possible, to frame questions around women’s behaviors and, for the most part, to use the household as the target location (Table 4). We asked questions during home visits when LPG tank refills were delivered, usually every 2–3 weeks. At each visit a stove technician assessed the Capability component questions, and a bilingual Mam-speaking female fieldworker asked Motivation and Opportunity component questions. All members of the team assessed the household environment and family engagement at each visit. Because the list of target questions was considerable, the field team assessed up to four constructs at each visit until all domains had been assessed in the 25 homes. Home visits averaged 20 min.

We addressed constructs in the Capability component of COM-B (specifically skills, skills development and procedural knowledge), such as showing household members how to regulate the gas flame so that heat could instantly speed up or slow down cooking. However, the construct of fear (in the Automatic Motivation COM-B component) was related to skill development, because women had “heard stories” from family and neighbors that made them hesitant to use gas stoves. Reflective Motivation constructs, such as identity, empowerment and intentions were challenging to elicit and often involved ongoing interactions with female fieldworkers, who could speak with women about their own experiences with gas stoves.

Young women, often living in households belonging to their in-laws or parents, expressed their inability to make household changes, despite their intentions to use the LPG stove; this affected many of the motivational and opportunity domains. One woman stated, “If my mother-in-law tells me to light the plancha because she is cold, I have to obey”. One the other hand, one woman said, “Now even my dad cooks with the gas stove”. Women who did not live with their in-laws were more likely to state positive intentions to use the gas stove. One woman stated that she was proud to own her stove, and when she moved to her own home, she would take the gas stove with her. In terms of group norms, a component of the Social Opportunity construct, women didn’t know other women who had prior experience with LPG stoves. One woman stated, “I am a good example to my neighbors. I am happy because my house is free of smoke and that keeps me healthy. But I feel bad because some women may think I am too full of pride now that I own a gas stove”.

We developed several intervention functions from the BCW based on focus group results, group interactions during Phase 1 classes and the diagnostic questions applied in 25 homes [38]. We tailored household-level interventions to extend the messages from the class setting to the home. We addressed issues that related to the capability factors in that environment. A set of proposed intervention functions guided each tank refill visit (Table 5). To illustrate, in Phase 1, we trained women to use the stove, but we found that mother-in-laws were resistant. During the post-partum period, when women were in repose during the first three weeks, older women cooked for the family and returned to using their biomass stoves. Therefore, in Phase 2, we trained other cooks in the home to light and regulate the flame at the tank refill visits (Training function). Because women’s spouses make decisions about large purchases, including stove repairs, tank refills and home construction, in Phase 2 we involved spouses in discussion of what can be done to add a window to the kitchen where the biomass stove is located to ventilate the smoke (Environmental Restructuring function). Despite apparent enthusiasm for the LPG stove, lack of money was the primary reason for inability to make home modifications or repair stoves.

In our future intervention work, video vignettes presented on an electronic tablet in the home will be used to stimulate household-level conversation about the proposed changes will be developed. An example of wood stove dis-implementation could show a baby on a mother’s back while she is building the fire, and a sister-in-law who offers to carry the baby on her back outside of the kitchen while the woman is building the fire. An example of gas stove implementation would be a video showing beans cooked in traditional clay pots over a gas stove. These modelling exercises could stimulate a conversation among all household members about the possibility of doing this in their home in a non-judgmental way.

Wood stove use dropped upon introduction of the gas stove from 6.4 h cooking duration, or 1.8 events at baseline to 1.9 h cooking duration, or 1.0 event at 24–28 weeks gestation (Fig. 2). The number of cooking hours decreased significantly across all rounds compared to baseline (Wilcoxon rank sum; p < 0.001). However, wood stove use increased gradually over the course of the study, with a concomitant decrease in gas stove use from 2.6 h, or 3.8 events at 24–28 weeks gestation to 1.1 h, or 2.1 events at 1 month post-partum (Wilcoxon rank sum; p < 0.001). We saw a marked decrease in the fraction of cooking represented by gas stove use from nearly 60% after introduction (24–28 weeks gestation), to 54% (32–36 weeks gestation), to just over 30% (at 1 month post-partum).

There were several limitations to this study. First, this was a formative research study with 50 participants. We have provided here several intervention functions that we piloted, and which may serve as examples for others. Second, this study was located in the Highlands of Western Guatemala where several high pollution exposure patterns predominate: the use of the plancha indoors when weather is cold; the use of the wood-fired sauna bath (temascal) for bathing; and weekly use of an outdoor open fire to cook animal food for many hours. Thus, even if women use the gas stove consistently, these other contributors to HAP will persist.