Handwashing has many important health benefits in community settings: it can substantially cut rates of diarrhoeal and respiratory disease, two leading causes of death in children <5 years of age (Rabie & Curtis 2006; Ejemot et al. 2008; Black et al. 2010); reduce transmission of gastrointestinal illness in households (Sandora et al. 2005) and work environments (Hubner et al. 2010; Savolainen-Kopra et al. 2012); reduce school absenteeism (Bowen et al. 2007; Talaat et al. 2011); and may improve developmental outcomes among young children in low-resource settings (Bowen et al. 2012). However, little evidence exists to guide the design of interventions to produce sustained adoption of appropriate handwashing habits. A recent review found that maintenance of handwashing behaviours was assessed >1 year after the intervention in only four of 30 handwashing behaviour change studies (Vindigni et al. 2011). Thus, our understanding of effective and cost-effective strategies for achieving sustained, appropriate handwashing habit adoption is poor. A better understanding of how to improve and sustain handwashing habits globally could have broad public health impact.

In a study conducted in informal settlements in Karachi, Pakistan, in 2003, neighbourhoods were randomised to receive household-level handwashing promotion and free soap for 9 months or to serve as controls (Luby et al. 2006). Participants in intervention households reported 53% less diarrhoea than did those in control households (Luby et al. 2009a). Between August 2005 and September 2006, we reassessed diarrhoea burden and handwashing knowledge and practices among these households (Luby et al. 2009a). Children in the intervention groups appeared to experience less diarrhoea during the first 5 months of the study, although the difference across the entire study period was not statistically significant (Luby et al. 2009a). Intervention households were also 1.5 times more likely to have a place for handwashing that included soap and water and were more than twice as likely to demonstrate rigorous handwashing, suggesting that improved handwashing behaviour had become habitual and persisted more than 2 years after the intervention ended (Luby et al. 2009a). We returned to these households to re-evaluate handwashing knowledge and practices in 2009, more than 5 years after the original intervention.

We re-enrolled 160, 141 and 160 households in the control, handwashing and handwashing + water treatment groups, respectively, representing 84% of eligible households from the original study (Figure 2). Households who re-enrolled did not differ significantly on many key variables, including neighbourhood of residence and study group allocation, from those who declined re-enrolment or were lost to follow-up (Bowen et al. 2012). Reenrolled households were similar across study groups with respect to household size, parental education/literacy and socioeconomic indicators (Table 1).

Water sources were also similar across groups, but water treatment practices differed (Table 2). Most homes in all groups received drinking water from the municipal system and/or purchased it from delivery trucks, and storage of drinking water was nearly universal. Control households were significantly more likely to report not typically treating their drinking water, while households from both intervention groups were more likely to report usually boiling their drinking water. Across groups, fewer than 3% of households reported typically using filters, alum or bottled water, and no respondents reported using bleach, disinfectant tablets or PuR^® (Procter & Gamble) to treat their drinking water.

Bar hand soap was the most commonly used hand cleansing material among all groups; however, many also reported using water alone (Table 2). No households reported typically using liquid hand soap or ash to cleanse hands. During handwashing demonstration, mothers from intervention households were 14 times more likely both to rub hands at least three times with soap [95% confidence interval (CI), 5.5–35] and to lather hands for at least 10 s (95% CI, 5.5–38) than were mothers from control households.

Each study group generated an identical list of occasions to wash hands during the free-list exercise, and participants cited these occasions in very similar rank order across groups (Table 2). More than 93% of respondents in each group reported washing hands after using the toilet. Respondents from intervention households were significantly more likely to report washing hands before cooking [relative risk (RR) 1.2 (95% CI 1.0–1.4)] and before eating or feeding others [RR 1.7 (95% CI, 1.3–2.1)], although only about half of intervention households reported washing hands before meals. Intervention households generated significantly longer lists of occasions to wash hands (mean number of occasions cited per group: handwashing, 4.74; handwashing + water treatment, 4.78; and control, 3.87; P < 0.0001).

Intervention households were more likely than control households to have a handwashing station with both soap and water present during the study visit in 2009 (Table 3). In 2009, 8 (3%) intervention households kept soap intended for handwashing at a location other than the handwashing station, compared with 114 (72%) control households [RR 0.036 (95% CI 0.015, 0.088)]. All households in 2009 also reported purchasing hand soap during the past month, but significantly more bars of soap, and more bars of soap per capita, were purchased by intervention households (Table 3).

Changing – and then sustaining – personal health-related behaviours can be difficult. This may be particularly true for health behaviours that must be repeated many times daily throughout life, like using seat belts, eating healthful foods and washing hands. To habituate such behaviours, individuals must recurrently choose to perform them until the behaviours acquire a degree of unconscious automaticity (Ouellette & Wood 1998; Verplanken 2006). Interventions designed to change these types of behaviours are, therefore, typically intensive (Ouellette & Wood 1998; Verplanken & Wood 2006). Successful interventions to change the dietary habits and physical activity of people at risk for diabetes generally involve weekly coaching for several months, followed by intermittent contact for many additional months (Venditti & Kramer 2012). Interpersonal communication during frequent household visits over a period of months, as employed in this study, is also likely to be more effective than less intensive strategies for effecting the adoption of handwashing (Wilson & Chandler 1993; Cairncross & Shordt 2004; Cairncross et al. 2005). Improved environmental cues, such as a fully stocked functional handwashing station at the exit to the latrine, can also serve as a visual reminder and reduce the effort required to perform a behaviour (Verplanken & Wood 2006). As handwashing is repeated in this context, it may begin to require minimal thought about whether to wash hands with soap, how to perform the steps of handwashing or where to locate soap or fetch water; it becomes habitual. This process is crucial for successful handwashing promotion because habituation may be the strongest determinant of handwashing behaviour (Whitby et al. 2007; Aunger et al. 2010).

Although the intervention in this study was intensive, with handwashing reminders and soap provided to participants at no cost for 9 months, it was not designed to foster long-term changes in handwashing habits. Furthermore, no interventions were provided during the years between the trial and the present evaluation, although both intervention and control households were repeatedly questioned about soap purchases in 2005. Nonetheless, intervention households appear to have adopted improved handwashing behaviours and maintained them for more than 5 years after handwashing promotion ceased. We observed soap and water at the handwashing station in 139 (99%) handwashing and 154 (96%) handwashing + water treatment households, but in only 41 (26%) control households in 2009. Although most households in the study were able to produce a bar of soap when asked to demonstrate handwashing, control households were 24 times more likely than intervention households to keep the soap at a location other than the handwashing station. Intervention households were also significantly more likely to report washing hands during several of the occasions in which handwashing can prevent disease transmission, to list a larger number of occasions they believed handwashing was appropriate and to demonstrate appropriate handwashing technique. Additionally, intervention households reported purchasing approximately 50% more hand soap per household member than control households in 2009. Together, these findings suggest that at least some people in intervention households developed the habit of washing hands with soap and water.

In contrast, access to well-stocked handwashing stations appeared to worsen over time among control households (Table 3). Among control households, 104 (53%) were observed to have a handwashing station with soap and water upon re-enrolment in 2005, but only 41 (26%) did so in 2009. Although it is unclear why reported soap purchases among controls varied so widely over time, control households reported buying less soap, and less soap per household member, in 2009 than they did in 2005 (Table 3). Thus, they might have purchased insufficient quantities of soap in 2009 to keep the handwashing station stocked at all times. Alternatively, the decreased access to soap at control handwashing stations in 2009 may be related to the group’s tendency to store soap away from the handwashing station. Because hand soap is considered a luxury in many settings, sequestering it to keep it clean, or to prevent domestic animals from consuming it or children from squandering it, is not uncommon (Wilson & Chandler 1993; Curtis et al. 2009). It is possible that by 2009, study children had reached an age at which they were especially likely to squander soap, and control parents reacted differentially by removing it from children’s reach.

The presence of soap at the handwashing station is likely to affect handwashing behaviour in a number of ways. If soap is not immediately available, household members may find it inconvenient to retrieve the soap and therefore might develop the habit of washing hands with water alone. Indeed, women in Bangladesh were more than twice as likely to wash hands with soap after faecal contact if soap was available at the handwashing station (Luby et al. 2009b). Ease of access to and visibility of handwashing stations were strongly associated with observed handwashing behaviour in a Kenyan community (Schmidt et al. 2009) and among healthcare workers in the United States (Nevo et al. 2010). Storing soap where hands are washed, rather than in a distant location, may particularly benefit children, who might not otherwise know where soap is being stored or be able to access it independently and who are at greatest risk for diseases that can be transmitted by hands. Future handwashing promotion efforts may benefit by stressing the importance of storing soap at the handwashing station.

We believe that handwashing becomes habitual after the behaviour is repeated sufficiently and that additional assessments of handwashing practices after this period may be unnecessary. Previous evaluations of this study population suggested that handwashing behaviours improved markedly during the intervention and were at least partially sustained for more than two additional years in the handwashing promotion group (Luby et al. 2006, 2009a). During the present study, we also found evidence of increased handwashing with soap in this community 5 years after the intervention. A separate study in India reported improvements in handwashing indicators up to 9 years after community-based handwashing promotion (Cairncross et al. 2005). Other types of health habits also appear to be maintained once inculcated. Substantial improvements in physical activity, nutritional choices and seat belt use were measured after the first year of a workplace wellness campaign in the United States; these improvements were then maintained at similar levels for several additional years (Byrne et al. 2011). If this is also true for handwashing, the cost-effectiveness of handwashing interventions could be estimated more accurately and resources for measuring handwashing habituation could be conserved.

This analysis is constrained by several limitations. Some fieldworkers were not blinded to original group allocation, and study participants might have been inclined to provide answers they felt would please the investigators. While we measured sizable improvements in many handwashing indicators, we cannot be certain how these indicators correspond to handwashing behaviours. Most of the outcomes we report reflect handwashing knowledge, an important but frequently insufficient precursor to behaviour change (Cairncross et al. 2005; Curtis et al. 2009). Information about soap purchases was self-reported and therefore of unknown validity. We did not use other methods to assess handwashing behaviour because of resource limitations and because of challenges to the validity of measuring handwashing behaviour through self-report, direct observation, hand microbiology and other proxies (Biran et al. 2008; Danquah 2010; Ram et al. 2010a, 2011). Current efforts to validate and improve measurement of handwashing behaviour will be useful for future assessments (Ram et al. 2010b). Questions and observations differed somewhat across the three assessments; thus, we were unable to summarise handwashing infrastructure uniformly across time. Further, we did not investigate the level of motivation underlying handwashing behaviours and therefore cannot be certain that such behaviours had become automatic. Perhaps, most notably, the design of our evaluation did not permit us to explore which aspects of the original handwashing promotion programme were associated with changes to the household handwashing environment or with maintenance of handwashing knowledge and practices.

Nonetheless, greatly superior access to handwashing stations with soap and water in intervention households suggests that appropriate handwashing was habituated among at least some members of these households and that these habits persisted for more than 5 years. The prolonged persistence of handwashing behaviours in this study suggests that cost-benefit analyses of handwashing promotion could assume a long time horizon (the time horizon is long) for assessing benefit. Implementers of handwashing programmes and other stakeholders may wish to adopt and evaluate intensive but time-limited methods of handwashing promotion, as used in this trial, and to focus on improving access to functional handwashing facilities and soap.