Cross-national comparisons and correlates of harms from the drinking of people with whom you work

Abstract

Background:

While research in high-income countries (HIC) has established high costs associated with alcohol’s harm to others (AHTO) in the workplace, scant attention has been paid to AHTO in the workplace in lower- or middle-income countries (LMIC).

Aim:

To compare estimates and predictors of alcohol’s impacts upon co-workers among workers in 12 countries.

Methods:

Cross-sectional surveys from 9,693 men and 8,606 women employed in Switzerland, Australia, the United States (US), Ireland, New Zealand, Chile, Nigeria, Lao PDR, Thailand, Vietnam, India and Sri Lanka. Five questions were asked about harms in the past year because of co-workers’ drinking: had they (1) covered for another worker; (2) worked extra hours; (3) been involved in an accident or close call; or had their (4) own productivity been reduced; or (5) ability to do their job been affected? Logistic regression and meta-analyses were estimated with one or more harms (vs none) as the dependent variable, adjusting for age, sex, rurality of location and the respondent worker’s own drinking.

Results:

Between 1% (New Zealand) and 16% (Thailand) of workers reported that they had been adversely affected by a co-worker’s drinking in the previous year (with most countries in the 6–13% range). Smaller percentages (<1% to 12%) reported being in an accident or close call due to others’ drinking. Employed men were more likely to report harm from co-workers’ drinking then employed women in all countries apart from the US, New Zealand and Vietnam, and own drinking pattern was associated with increased harm in five countries. Harms were distributed fairly equally across age and geographic regions. Harm from co-workers’ drinking was less prevalent among men in HIC compared with LMIC.

Conclusions:

Workforce impairment because of drinking extends beyond the drinker in a range of countries and impacts productivity and economic development, particularly affecting men in LMIC.

Introduction

Drinking patterns among workers and at work

In high-income and a few other countries there have been studies of drinking patterns and problems in employed populations (Silva et al., 2003). In recent decades, alcohol consumption by full-time workers, although not necessarily at work, has been common. In one survey, 55–69% of US workers aged 18–49 years employed in a variety of specified occupations reported drinking in the past year; heavy drinking was reported by 5–14% of workers surveyed (Zhang and Snizek, 2003). Drinking varies by occupation, with heavier drinking reported, in Australia, for example, in the hospitality, agriculture, mining, retail, transport and financial services industries (Pidd et al., 2011, Pidd, 2005). Data gathered in the US indicated that lawyers, nurses and other health care professionals, and workers in the mining, construction, and hospitality and food service industries, as well as entertainers and managers, were at higher risk of heavy drinking or alcohol-related problems (American Addictions Centre, 2019).

While drinking among the workforce is common, drinking at work and being intoxicated whilst working are less common. In general in the last forty years workplace safety regulations have tightened and heavy-drinking business lunches are less widespread, with norms for drinking at such lunches, for example, having shown a downward trend in the US (Greenfield and Room, 1997). Workplace testing for alcohol and other drug use during, for instance in the US, prior to and after working has also become more common in high-risk occupations and transport industries (Cunradi et al., 2005). Nevertheless, a substantial proportion of Australians (8.7%) reported drinking alcohol at work in a 2007 survey (Pidd et al., 2011), and alcohol still plays a substantial role in work-related social events (Pidd, 2005).

Other factors may be correlated with drinking as well: for example, alcohol consumption in some countries (but not all) is higher in rural areas (Giang et al., 2013, Allan et al., 2012, Rehm et al., 2005). Consumption also varies by socio-economic factors within and between countries of differing affluence. There is some evidence that socio-economic differences operate differently in lower-and middle-income countries (LMIC) as compared with high-income countries (HIC), with lower-income respondents less able to afford to drink in LMIC (Huckle et al., 2018). Gender (male) and age (generally younger) are consistent predictors of heavy episodic drinking (Livingston et al., 2008) and potentially of associated injuries (Cherpitel et al., 2009). Particular occupational groups, as mentioned above, show patterns of heavier drinking (Hitz, 1973). These same factors may be associated with workplace injuries or impacts on work.

Alcohol’s impact on work-related illness, injury and productivity

Alcohol-related work injuries and incidents occur in the transport sector, factories, offices, retail and entertainment businesses, as well as in boardrooms and on farms in Australia (Roche et al., 2015). One in four industry accidents worldwide are estimated to be attributable to drugs and/or alcohol (Pidd, 2005). In Australia 16% of fatal workplace injuries tested for BAC levels involved workers who had BACs higher than zero (Hollo et al., 1993). In addition to workplace injuries, alcohol is responsible for substantial numbers of days missed due to heavy drinking. In the US, 38% of workers were absent for one or more days in the previous year. Of these days, 4.4% were attributed to the respondent’s own drinking, with a further 5.5% attributed to someone else’s drinking (Greenfield et al., 2016). In this study a clear association with drinking pattern was identified, with productivity losses greatest for the heaviest drinkers.

Impacts from co-workers’ drinking

The majority of research on alcohol-related workplace harms has focused on self-reported drinking by, and drinking-related problems of, workers; however, the drinking of co-workers, bosses and others at work can have a range of impacts on other workers. For example, accidents where workers are intoxicated or hung over can injure others, and absenteeism and inefficient work practices can mean co-workers ‘cover’ for others, creating a burden for those they work with, and diminishing overall productivity. Workers may also have to take time off work to take care of alcohol-related sickness of family members. A large number of people in the Australian Alcohol’s Harm to Others (AHTO) survey, when asked, reported being negatively affected by their co-workers’ drinking and working extra hours to cover for these drinkers (Dale and Livingston, 2010). These Australian findings indicated that the cost of co-workers’ drinking to the Australian economy was an estimated AUD$453 million annually. This was comparable to the cost estimated when drinkers were asked about their own absenteeism because of their drinking (Dale and Livingston, 2010).

Research so far

While we know that in general per capita consumption is higher and abstention rates are lower in HIC than in LMIC, for the same level of drinking the negative health consequences for people in less economically developed countries can be greater (World Health Organization, 2018). In HICs, a number of studies have focussed on alcohol-related work injuries, illnesses and absenteeism (Roman, 2002). However, in LMICs, very little attention has been paid to alcohol-related workplace harms in general, with even less focus turned to alcohol’s harm to others in the workplace. In LMICs, small studies have identified harms from others’ drinking in a range of occupational settings, including the trucking, hospitality (Obot et al., 2014) and sex work industries (Panchanadeswaran et al., 2008), and in farming families in rural India (a primary motivation of women’s “anti-arrack” movements) (Larsson, 2006). Given the vast majority of fatalities due to occupational injuries and illnesses are reported in Asia, including LMICs such as China, India and Vietnam (Wang et al., 2016), a large research knowledge gap is apparent. Moreover, given fewer resources in LMICs, there is less occupational injury-related research and related regulation and infrastructure support for increasing safety in LMIC workplaces (Sorensen et al., 2017). This research is intended to begin to fill the gap.

Aims

This study aims firstly to compare estimates and correlates of harms from co-workers’ drinking in 12 countries, and secondly focuses on differences by country-level income when these countries are grouped into LMIC and HIC. The specific objectives of the study are to estimate and describe co-worker related harms, and correlates of these harms, from others’ drinking for each country. The study seeks to provide evidence about which respondents are more likely to experience harms from co-workers’ drinking, specifically studying whether respondents who are male, younger, from rural areas and with heavier drinking patterns are at greater risk. The paper studies the correlates of harm from co-workers’ drinking in each country and uses meta-analysis to produce a pooled estimate of harm for men and women overall and in LMIC and HIC separately.

Materials and Methods

Data

Country-level analyses:

The study uses data from 12 countries involved in the Gender and Alcohol’s Harm to Others (GENAHTO) project (https://genahto.org/) (Wilsnack et al., 2018): Switzerland, Australia, the United States, Ireland, New Zealand, Chile, Thailand, Sri Lanka, Nigeria, Vietnam, Lao People’s Democratic Republic (Lao PDR) and India (listed in order of decreasing gross national income) (World Bank, 2015, WHO, 2014). All data are archived at the Centre for Alcohol Policy Research, La Trobe University. Only respondents aged 18–64 years who were employed are included in this analysis. Respondents who were unemployed, students and people primarily engaged in household duties were excluded. Recruited samples and the number of employed respondents are shown in Table 1.

Table 1:

Sample composition and prevalence of harm to respondents from co-workers’ drinking in ten societies.

Country	Year of data collection	Response rate	Sample	Number of respondents	Per cent not employed	Any harm from drinking of co-workers [1+ harms] in population % (95%CI)	Number of employed respondentsa	Any harm (1+) from drinking co- workers among those employed, % (95%CI)
Switzerland	2012/16	51.4/45.0	Men	1457	16.3	3.2 (2.1, 4.8)	1147	3.8 (2.5, 5.7)
			Women	1813	27.1	1.3 (0.7, 2.3)	1270	1.8 (1.0, 3.1)
			Total	3270	21.7	2.2 (1.6, 3.1)	2417	2.9 (2.0, 4.0)
Australia	2008	35.2	Men	889	5.6	5.4 (4.0, 7.2) c	852	5.7 (4.3, 7.7)b
			Women	1300	10.5	2.6 (1.8, 3.8) c	1181	2.9 (2.0, 4.3)b
			Total	2189	8.1	4.0 (3.1, 5.0) c	2033	4.3 (3.4, 5.5)b
US	2015	60.0 c	Men	801	24.2	3.5 (2.1, 5.7)	580	4.6 (2.8, 7.5)
			Women	1026	35.9	3.4 (2.0, 5.6)	623	5.3 (3.2, 8.7)
			Total	1827	30.8	3.4 (2.4, 4.9)	1203	4.9 (3.4, 7.0)
Ireland	2015		Men	782	27.1	7.9 (6.1, 10.3)	563	10.9 (8.4, 14.0)
			Women	826	35.9	4.1 (2.9, 5.9)	515	6.4 (4.5, 9.1)
			Total	1608	31.7	6.0 (4.8, 7.4)	1078	8.7 (7.1, 10.7)
New Zealand	2008/09	64.0	Men	950	14.7	1.2 (0.6, 2.3) b,d	823	1.4 (0.7, 2.7) b,d
			Women	1459	25.3	0.7 (0.4, 1.3) b,d	1065	1.0 (0.5, 1.8) b,d
			Total	2409	20.4	0.9 (0.6, 1.5) b,d	1888	1.2 (0.7, 1.9) b,d
Chile	2012/13	71.8	Men	653	14.2	17.8 (14.8, 21.3)	565	20.5 (17.1, 14.5)
			Women	741	36.2	7.5 (5.6, 9.9)	482	11.3 (8.5, 14.9)
			Total	1394	25.3	12.6 (10.8, 14.7)	1047	16.6 (14.2, 19.2)
Thailand	2012/13	94.2	Men	643	8.4	24.3 (20.8, 28.2)	593	26.4 (22.5, 30.6)
			Women	928	22.8	8.7 (6.7, 11.2)	731	11.0 (8.5, 14.1)
			Total	1571	15.8	16.3 (14.2, 18.6)	1324	19.1 (16.8, 21.8)
Sri Lanka	2013/14	93.0	Men	1091	15.1	22.1 (19.5, 25.0)	915	26.0 (22.9, 29.3)
			Women	1172	72.3	3.2 (2.3, 4.5)	329	11.5 (8.2, 15.9)
			Total	2263	44.9	12.3 (10.8, 13.9)	1244	22.2 (19.7, 24.9)
Nigeria	2012/13	99.0 e	Men	1355	10.3	2.8 (1.9, 4.3)	1227	3.2 (2.1, 4.8)
			Women	861	23.1	1.0 (0.5, 2.0)	705	1.3 (0.7, 2.6)
			Total	2216	16.7	1.9 (1.4, 2.7)	1932	2.3 (1.6, 3.3)
Vietnam	2012/13	99.2	Men	719	5.3	11.8 (9.5, 14.6)	683	21.5 (10.0, 15.4)
			Women	719	24.4	7.5 (5.7, 9.8)	551	9.5 (7.2, 12.5)
			Total	1438	15.0	9.6 (8.1, 11.4)	1234	11.1 (9.4, 13.2)
Lao PDR	2013	99.0	Men	504	13.4	8.9 (6.4, 12.2)	448	10.3 (7.4, 14.0)
			Women	708	21.1	3.3 (2.1, 5.2)	567	4.2 (2.6, 6.6)
			Total	1212	17.2	6.1 (4.7, 8.0)	1015	7.4 (5.6, 9.6)
India	2013/14	97.0	Men	1517	15.4	20.8 (18.8, 23.1)	1297	24.6 (22.2, 27.1)
			Women	1711	66.5	5.0 (4.0, 6.2)	587	12.3 (9.8, 15.5)
			Total	3228	40.6	13.1 (11.9, 14.4)	1884	21.2 (19.3, 23.2)
All countries			Men	11361	14.4	10.7 (6.8, 14.5) b,d,f [I2 = 98.3%]	9693	12.3 (7.8, 16.9) b,d,f [I2 = 98.4%]
			Women	13264	35.2	3.8 (2.6, 5.0) b,d,f [I2 = 93.8%]	8606	6.2 (4.3, 8.1) b,d,f [I2 = 94.8%]
			Total	24625	25.0	7.1 (5.5, 8.8) b,d,f [I2 = 97.7%]	18299	9.3 (7.1, 11.5) b,d,f [I2 = 97.7%]

^aexcluding unemployed, students and people engaged in house duties;

^bIn Australian and NZ the any harm from co-workers’ drinking variable was derived from three items for the Australian sample and four items for the New Zealand sample (compared to five items used for each of the remaining 10 countries)

^ca co-operation rate was published in the US;

^dNZ respondents were not asked about specific items unless they identified a co-worker unprompted;

^eIn Nigeria a response rate of 99% was reported among households where someone was home, but random selection was not followed within the household;

^foverall estimates are pooled across country-level estimates via the DerSimonian-Laird method of two-stage inverse-variance random-effects meta-analysis using individual participant data

The datasets used have been previously described in detail for Switzerland (Marmet and Gmel, 2017), Australia (Wilkinson et al., 2009), the US (Nayak et al., 2019), Ireland (Hope et al., 2015), New Zealand (Casswell et al., 2011) and for the remaining seven countries, as participants in a WHO-Thai Health collaborative project (Callinan et al., 2016). An overview of the methodology of the GENAHTO Project has also been published (Wilsnack et al., 2018). All studies were based on probability samples, with the majority national, and some from selected regions. Response rates varied from 35% in Australia to 99% in Vietnam and Lao PDR (Laslett et al., 2017). The cooperation rate for the US sample was 60% (Kaplan et al., 2017). The Australian study’s lower response rate reflected telephone sampling and interviewing; it slightly under-represented male, young and less educated persons in comparison with the Census (Wilkinson et al., 2009). The response rates in LMIC, on the other hand, were remarkably high, where survey burnout is uncommon, and in face-to-face surveys in general. This effect in LMIC may also reflect endorsement of participation in the study from trusted village health workers or heads of villages and adoption of a minimum of three call-back visits (Callinan et al., 2016).

HIC versus LMIC comparisons:

The present study produces estimates of harm from co-workers’ drinking stratified according to gross national income (GNI) per capita, determined by country-level indicators of 2015 per capita GNI (The World Bank, 2019a). High-income economies included countries whose 2015 GNI per capita was USD 12,476 or more (Chile: USD 14,340 – Switzerland: USD 85,780). LMIC economies (India: USD 1,600 -- Thailand: USD 5,690) were countries whose 2015 GNI per capita was less than USD12,476 (World Bank, 2017).

Dependent variables/Outcome measures

Respondents were asked the five questions:

a) Have you covered for workers because of their drinking; b) Has your productivity been reduced because of their drinking; c) Has your ability to do your job been negatively affected because of their drinking; d) Were you involved in an accident or close call at work because of others’ drinking; and e) Have you had to work extra hours because of others’ drinking? In Australia, only the third, fourth and fifth questions were asked, and in New Zealand the fourth was omitted. A summary dichotomous (0,1) outcome measure was produced for reporting one or more of these harms from the drinking of co-workers [any work harm]. Responding workers who reported experiencing any of the items were categorised as 1. Respondents who reported no to all items were categorised as 0. Respondents in all countries (apart from Australia) who reported no to all items but who failed to respond on one item (n=60) were coded 0. Twenty-seven respondents who failed to respond on more than one of the work harm items, yet may have answered no to other items, (apart from seven Australian respondents who had one or more missing items) were coded missing for the “any work harm” variable. (In Australia, missing even one response constituted a third of their responses.) Given that we are creating a dichotomous any harm versus no harm variable this procedure was considered adequate.

Independent variables

Respondents indicated whether they were male or female and their age at the time of interview. Ages were categorised into three groups: 18–29, 30–49 and 50–64 years. Respondents were classified as living in a rural or non-rural location. Rural areas were classified as “open country but not a farm”, “on a farm” or “in a small city or town”. A small city or town constituted fewer than 60,000 people in Sweden, fewer than 20,000 people in the US and fewer than 50,000 people in all other countries.

To categorise drinking pattern, respondents were classified as abstainers, moderate drinkers or heavy episodic drinkers (HED). HED were those who said they consumed five or more drinks (about 60 gm of ethanol) on an occasion at least monthly in the last 12 months. Moderate drinkers were those who drank in the last year, but drank 5 or more on an occasion less than monthly, or always drank less than that. 151 Nigerian respondents missing a response on the drinking pattern variable were categorized into a fourth group to avoid a large proportion of the Nigerian sample being coded as missing. 359 Chilean respondents known to drink who were incorrectly skipped and not asked about their frequency of drinking 60 g or more of alcohol on an occasion were also categorized as an additional group to prevent a large proportion of the Chilean sample being coded as missing for the drinking pattern variable.

Weighting:

All country data were weighted to adjust for participants’ probability of selection (based on the household’s number of adults). Thailand, Sri Lanka, Lao PDR, Vietnam, Chile and Nigeria were additionally weighted by gender (based on male-female population proportions). Country-specific data were weighted to improve the representativeness of the sample, e.g., derived pre- and post-stratification weights were applied for Australian (Callinan et al., 2016) and US (Kaplan et al., 2017) data.

Analytic approach

The following analyses were undertaken:

Country-level analyses

• aDescriptive statistics and confidence intervals using weighted data were generated for individual and combined measures of prevalence estimates of harm from co-workers’ drinking per country.
• bCorrelates of harm from co-workers’ drinking in each country were identified using logistic regression, including each independent variable one at a time, and in a multivariable model, adjusting for gender, age, rurality and drinking status.

HIC versus LMIC comparisons

• cCombining the data from all countries, pooled estimates of harm from co-workers’ drinking for men and women were generated, with these effect estimates unadjusted and adjusted for other factors to determine the size of this effect for men versus women.
• dCombining workers’ data from all countries, an overall effect estimate of the increased risk of harm from co-workers’ drinking associated with one’s own drinking was calculated, unadjusted and adjusted for gender, age and rurality.

Confidence intervals, statistical significance and statistical software:

Country prevalence and odds ratios are accompanied by 95% confidence intervals, with real country differences defined as non-overlapping confidence intervals (du Prel et al., 2009, Gardner and Altmann, 1986). All data analyses and construction of forest plots were completed using Stata version 14.0 (Stata Corp., 2015).

Rationale for meta-analysis:

Because the 12 surveys are similar but still heterogeneous studies from vastly different countries, with differences in sampling, methodology and sample compositions, random effects meta-analyses were conducted (Huedo-Medina et al., 2006, Borenstein et al., 2007). The pooled estimates resulting from these analyses are interpreted as the mean estimates of the true varying effects across all studies. In the first meta-analysis, the DerSimonian-Laird method of two-stage inverse-variance random-effects meta-analysis (via the ipdmetan command using Stata 14.0 (Fisher, 2015)) was used to estimate the pooled proportion of respondents who reported harms from co-workers’ drinking in the last 12 months, separately for men and women, pooled across 12 countries. Similarly, the likelihood of reporting harm from co-workers’ drinking is presented by respondent drinking pattern. Country-level and pooled effect estimates and accompanying I² statistics are presented as forest plots. The I² statistic indicates the total variability in effect sizes due to heterogeneity across the studies; I² values of 25%, 50%, and 75% indicate low, medium, and high heterogeneity.

Results

Country-level findings

Table 1 displays the unadjusted country-level findings. Employed men experienced significantly more harm from co-workers than employed women in Switzerland, Australia, Ireland, Chile, Thailand, Sri Lanka, Nigeria, Lao PDR and India, while in Switzerland, the US, Ireland, New Zealand and Vietnam differences were not significant. Overall, in the pooled analysis, employed men (12.3%; 95%CI 7.8%, 16.9%) reported more (but not statistically significantly more) harm from co-workers’ drinking than employed women (6.2%; 95%CI 4.3%, 8.1%). Seven per cent of respondents overall reported harms from co-workers’ drinking.

Table 2 provides more detail on the types of harms respondents reported experiencing from co-workers’ drinking in each country. The most commonly reported harm from co-workers’ drinking in six countries was having to cover for others because of their drinking, although Australia did not include this item. Reduced productivity and having to work extra hours because of co-workers’ drinking were also reported relatively commonly as the most or second-most common reason respondents were affected by co-workers. Being involved in an accident or close call at work because of others’ drinking was the least commonly reported impact of co-workers’ drinking in the majority of countries, although the percentage of respondents who reported this measure of harm was remarkably high in India (11.7%), but also relatively high in Sri Lanka and Thailand.

Table 2:

Percentage of employed respondents aged 18–64 that experienced harms from co-workers’ drinking in ten societies

Country	Number of employed respondentsa	Covered for workers because of their drinking, % (95%CI)	Has your productivity been reduced because of their drinking, % (95%CI)	Ability to do your job been negatively affected because of their drinking, % (95%CI)	Involved in an accident or close call at work because of others’ drinking, % (95%CI)	Have you had to work extra hours because of others’ drinking, % (95%CI)
Switzerland	2417	1.3 (0.8, 2.1)	1.6 (1.0, 2.5)	0.9 (0.5, 1.6)	0.4 (0.2, 1.0)	1.3 (0.8, 2.2)
Australia	2033	n/a	n/a	3.3 (2.5, 4.3)	0.5 (0.3, 0.9)	2.8 (2.1, 3.8)
US	1203	3.5 (2.2, 5.4)	2.9 (1.8, 4.7)	3.4 (2.2, 5.2)	0.9 (0.4, 2.1)	3.5 (2.2, 5.4)
Ireland	1078	7.2 (5.6, 9.0)	7.3 (5.8, 9.2)	5.9 (4.5, 7.6)	0.8 (0.4, 1.6)	5.3 (4.0, 6.9)
New Zealand	1888	0.7 (0.4, 1.3)	1.0 (0.6, 1.7)	n/a	0.0 (0.0, 0.3) b	0.6 (0.3, 1.1)
Chile	1047	11.3 (9.3, 13.6)	7.6 (6.0, 9.6)	6.8 (5.3, 8.7)	1.4 (0.8, 2.5)	9.0 (7.2, 11.1)
Thailand	1324	12.1 (10.2, 14.3)	11.0 (9.2, 13.1)	7.5 (6.0, 9.3)	3.9 (2.8, 5.3)	6.6 (5.2, 8.4)
Sri Lanka	1244	13.9 (11.8, 16.2)	11.8 (10.0, 14.0)	10.9 (9.1, 13.0)	4.9 (3.7, 6.4)	12.6 (10.6, 14.8)
Nigeria	1932	1.4 (0.9, 2.2)	0.9 (0.5, 1.5)	0.7 (0.4, 1.3)	0.4 (0.2, 1.0)	1.5 (1.0, 2.3)
Vietnam	1234	6.9 (5.5, 8.5)	6.5 (5.2, 8.1)	3.9 (2.9, 5.1)	2.2 (1.4, 3.3)	2.7 (1.9, 3.8)
Lao PDR	1015	5.1 (3.7, 7.0)	4.2 (2.9, 6.0)	4.1 (2.9, 5.9)	2.2 (1.3, 3.7)	2.0 (1.2, 3.2)
India	1884	12.5 (11.0, 14.2)	11.6 (10.2, 13.2)	14.9 (13.3, 16.7)	11.7 (10.2, 13.3)	14.0 (12.4, 15.7)

N = 18,299;

^aexcluding unemployed, students and people engaged in house duties; n/a: no equivalent item in country-level survey;

^bNZ respondents were not asked about specific items unless they identified a co-worker unprompted.

A higher percentage of men were employed than women in all countries, with quite diverse levels of reported employment among women. The percentage of women not reporting employment was low in Australia (10.5%), around 20–30% in Switzerland, NZ, Thailand, Nigeria, Vietnam, Lao PDR, and 31–40% in US, Ireland and Chile; over two-thirds of women reported not being employed in the paid workforce in Sri Lanka (72%) and India (67%) (See Table 1).

Unadjusted analyses

Table 3 compares odds ratios for demographic and drinking status subpopulations on the rate of occurrence of any of the harms. As noted earlier from Table 2, among those in the workforce, women were significantly less likely to report harm from co-workers’ drinking in the majority of countries.

Table 3:

Characteristics of employed respondents harmed by co-workers’ drinking: odds ratios by gender, age group, residence (rural/non-rural), respondent’s (R’s) drinking status

			Switzerland	Australia	US	Ireland	New Zealand	Chile	Thailand	Sri Lanka	Nigeria	Vietnam	Lao PDR	India
Bivariate	Gender (vs. Male)	Female	0.46*	0.50**	1.16	0.56*	0.67	0.49***	0.34***	0.37***	0.42*	0.74	0.38**	0.43***
	Age (vs. 18–29)	30–49	2.67*	1.26	0.82	1.15	1.25	1.53*	0.62*	0.61*	2.90*	0.69	0.52	1.51**
		50–64	1.17	0.50	0.70	0.58	0.35	1.19	0.35***	0.66	1.82	0.52*	0.41*	1.71**
	Residence (vs. Rural)	Not rural	n/a	0.78	0.34	0.92	2.30	1.51	0.92	0.64*	0.85	0.67	2.32*	0.95
	Drinking status (vs. Abstainer)	Moderate	0.82	0.82	1.01	1.09	1.51	2.21*	2.00**	2.03***	1.88	1.35	1.73	6.60***
		HED	1.28	1.40	0.82	1.28	1.48	2.03**	5.39***	5.03***	1.79	1.60	3.18*	6.39***
Multivariate	Gender (vs. Male)	Female	0.48*	0.54*	1.00	0.55*	0.67	0.54**	0.50**	0.72	0.45	0.88	0.40**	0.73
	Age (vs. 18–29)	30–49	2.77*	1.34	0.74	1.14	1.23	1.62*	0.68	0.54**	2.82	0.69	0.52	1.18
		50–65	1.27	0.54	0.62	0.57	0.34	1.60	0.42**	0.54*	1.64	0.53	0.36*	1.26
	Residence (vs. Rural)	Not rural	n/a	0.76	0.29	0.85	2.19	1.62	0.92	0.61*	0.90	0.73	2.27*	0.86
	Drinking status (vs. Abstainer)	Moderate	0.76	0.85	1.36	1.20	1.65	2.27*	1.58*	1.76*	1.43	1.26	1.34	6.19***
		HED	1.13	1.21	1.04	1.14	1.35	1.95*	3.40***	4.43***	1.13	1.39	1.79	5.88***

N = 18,299;

^*p<.05

^**p<.01

^***p<.001; n/a: no equivalent item in country-level survey

Respondents in the oldest age group were generally less likely to report harm from co-workers’ drinking than those in the youngest age group, although this was statistically significant only in Thailand, Vietnam, Lao PDR and India. Comparison of the youngest age group and the mid-aged group revealed inconsistent results across countries, with the younger group significantly less likely to report harm than 30–49 year olds in Switzerland, Chile, Nigeria and India, yet significantly more likely to report experiencing harm in Thailand and Sri Lanka. The findings regarding rurality were mixed, with respondents from rural areas significantly more likely to report harm in Sri Lanka and less likely to do so in Lao PDR, but with the results in other countries showing no significant rural-urban differences.

Respondents who were HED were more likely than abstainers to report harm from co-workers’ drinking in Chile, Thailand, Sri Lanka, Lao PDR and India. Moderate drinking was also associated with greater odds of harm from co-workers’ drinking compared to abstainers in Chile, Thailand, Sri Lanka and India.

Adjusted (multivariable) findings

In five countries -- Switzerland, Australia, New Zealand, the US and Ireland (all HIC) -- harms from co-workers’ drinking were fairly evenly spread in the employed population, with no adjusted odds ratios significant apart from gender in Australia and Ireland and age in Switzerland.

Other correlates aside from gender were identified for harms from co-workers’ drinking but few consistent patterns were identified. Respondents aged 50–65 in Thailand, Lao PDR and Sri Lanka, and 30–49 in Sri Lanka were less likely than those aged 18–29 to report harm from co-workers’ drinking. In Switzerland, Chile and Nigeria the mid-aged groups were more likely to report harm from co-workers’ drinking than the youngest age group. Rural workers in Sri Lanka were more likely than non-rural workers to and rural workers from Lao PDR less likely to report harm from co-workers’ drinking.

The adjusted results show that those who were themselves HED and moderate drinkers were much more likely than abstainers to have suffered harm from co-workers’ drinking in Chile, Thailand, Sri Lanka and particularly in India.

Analysing results across all countries

In pooled meta-analyses across nine countries (not shown here), adjusting for the factors in the regression models in Table 3, effect estimates for gender and HED were separately estimated. Men were more likely than women to report harms from co-workers’ drinking (OR = 0.62, 95%CI = 0.53, 0.72, I² = 0.0%), consistent with the unadjusted results presented in Table 1. Furthermore, HED were significantly more likely to report harms from co-workers’ drinking than abstainers (OR = 2.10, 95%CI = 1.36, 3.25, I² = 77.3%), as were moderate drinkers (OR = 1.69, 95%CI = 1.13, 2.50, I² = 74.0%).

Differences between high-income and low- and middle- income countries: a meta-analysis

In a second meta-analysis undertaken to formally test differences in harm from co-workers’ drinking between LMICs and HICs, although men from LMIC were generally more likely to report harm from co-workers’ drinking than men in HIC, this was not significant. Women in HIC were less likely to report harm than women in LMIC, although the difference was smaller and the rates less consistent within HICs and LMICs. With Chile and Nigeria as exceptions, Figure 1 breaks down into a split between LMIC and HIC countries: men report much more harm than women from others’ drinking in LMICs, while men and women report harm from co-workers more evenly and at lower levels in HICs.

Figure 1a.

Proportion of employed men in high-income countries and in low- and middle-income countries who experienced any harm from co-workers’ drinking in the last 12 months.

Any harm from co-workers’ drinking was derived from three items for the Australian sample and four items for the New Zealand sample (compared to five items used for each of the remaining 10 countries); Overall estimates are pooled across country-level estimates via the DerSimonian-Laird method of two-stage inverse-variance random-effects meta-analysis using individual participant data; Weights of the contribution of country-level estimates to pooled estimates are represented by the relative area of the corresponding grey square.

Discussion

The finding that 7.0% of employed respondents reported some harm from others’ drinking in the past 12 months, underscores the externalities of others’ drinking, in this case co-workers’ drinking, and suggests that harm from others’ drinking in the workplace may not be fully acknowledged in current statistics, which have tended to focus on absenteeism and harms reported only by the drinking worker. Where these harms have been estimated in the past (e.g., Australia), the costs associated with others’ drinking have been nearly equal to those already incurred by the respondent’s own drinking (Laslett et al., 2010), although there is probably some overlap between the two costs.

In general, high income-countries reported a much lower prevalence of harms from co-workers’ drinking, with the exception of Ireland and Chile. This is consistent with evidence that industrial injuries and illnesses are higher in LMICs (Wang et al., 2016) and that workplace policies and regulations are less likely to be in place and, if in place, enforced (Sorensen et al., 2017). Chile is the lowest income country in the HICs, with an income level more similar to Thailand than to New Zealand. It was only recently classified as a HIC by the World Bank in 2012. Inconsistent with the LMIC pattern were Nigeria and Lao PDR – respondents in these countries reported lower levels of harm from co-workers’ drinking among the employed population than respondents in other LMICs.

All countries reported relatively low rates of occupational “accidents” and close calls due to others’ drinking in the previous 12 months, with all countries except India reporting rates below 5%. However, even rates between 2 and 5% (reported in 5 LMICs) are substantially higher than rates under or around 1% found in all of the HICs in this study. The prevalence of accidents and near-misses in LMICs of between 2.2% and 4.9% (as high as 11.7% in India) is worrying. Including all negative effects from co-workers’ drinking measured in this study, these comparisons are even more stark, with 7–22% of respondents in most LMIC countries reporting one or more harms from co-workers’ drinking in the previous year. In South-East Asian countries like India, Thailand and Vietnam, the likely impact on LMIC co-workers is of particular concern, as these countries are viewed as key emerging markets by the alcohol industry, since increasing alcohol consumption is forecast (Babor et al., 2010).

Turning to factors that indicated working respondents were at higher odds of harm from co-workers’ drinking, gender (unadjusted) was a consistent marker of risk across almost all countries, with men generally at twice or more the odds of harm than women in the majority of countries. The meta-analysis of workers’ harms from co-workers’ drinking provides an important estimate of the harms experienced by employed men and women from a small number of HIC and LMIC countries. Overall, four percent of women and 11 per cent of men reported harms from co-workers’ drinking, with men significantly more likely to report these harms than women. Men’s greater likelihood of experiencing harms from co-workers’ drinking (possibly with the exception of men in the US) probably reflects at least in part that they are likely to associate more with co-workers who are men and who drink more.

This is consistent with the gendered drinking patterns identified previously (Wilsnack et al., 2009), and may be due to management decisions where there is some degree of gender separation in the workplace in LMICs: men may be more likely to be asked to cover for other men they work with (also true for women but less likely to occur because women drink less). Cultures in male-dominated workplaces (which can sometimes be more physical and more risky, e.g., construction and mining) have traditionally been associated with heavy drinking (Bacharach et al., 1994, Sonnenstuhl, 1996). Types of men’s and women’s work and the typical workplace culture may be more gendered across some countries than others (Dong et al., 2015, Bureau of Labour Statistics, 2017, Cui et al., 2015). Gender employment equity differs a lot across countries (The World Bank, 2019b) and might affect exposure to harm in general and in particular occupations. Indeed, because men comprise a larger proportion of the formal workforce, and tend to be the drinkers and heavy drinkers, they are likely to be at greater odds of co-worker harms, as in almost all countries in our study.

While employed persons of both genders are selectively targeted by advertising and marketing strategies, analyses of alcohol industry campaigns indicate that these campaigns often seek to reinforce traditional masculine drinking patterns, sometimes in association with workplace settings or work-related gatherings. These marketing campaigns may also seek to increase drinking in masculine work-related groups (Duff, 2003), potentially increasing the odds of harm in these settings for men. Currently it appears to be LMIC men who are at greater risk here, although women may also be at risk from these drinking men when they return home (Callinan et al., 2018). Women may be at more risk in the future if the alcohol industry continues to target underdeveloped markets.

Contrary to expectations, the youngest age group in this study was not always the group most likely to experience harm from co-workers’ drinking. Generally, the oldest group was less likely to experience harm from co-workers’ drinking than the youngest group, with less difference identified between the mid and youngest age groups. However, in Thailand, Vietnam, Lao PDR, and India younger workers did seem to be at greater risk of harm from coworkers’ drinking. This is consistent with evidence from Canada, for example, where it was found that young people may be more likely to be employed in riskier occupations, and that their inexperience may increase the risk they are exposed to (Breslin et al., 2007). However, we cannot determine from our data whether young people were employed in riskier occupations as that level of detail on occupation was not collected in the survey. It may also be the case that younger people work in occupations where there is a culture of heavier drinking (e.g., hospitality industry), and these occupations may not necessarily be more dangerous. It is important to remember that the harms to coworkers studied in this survey include harms related to covering for someone as well as risk of injury, so it is not only in more dangerous occupations where respondents are at greater risk. However, findings showing that middle- and older-aged males are at similar risk in some countries are consistent with the findings of Wilsnack and colleagues (Wilsnack et al., 2009), who found drinking patterns, particularly problematic drinking patterns, in some countries were not established until somewhat later in life.

HEDs, compared to abstainers, were more likely to report harm from co-workers’ drinking, and this was significant in the majority of LMICs, including Thailand, Sri Lanka and India (and for Lao PDR when the results are unadjusted). In HIC, apart from Chile (which has the lowest GNI of all the HICs in this study), there was no evidence of differences in reported harms from co-workers’ drinking by drinking pattern. Interpreting this fairly consistent pattern needs to keep in mind that in HICs the levels of harm were much lower and the prevalence of drinking much higher, diminishing the size of the baseline abstaining group and reducing the power to detect differences.

Limitations

Outcome variable:

The outcome – harm from a coworker’s drinking – is the subjective assessment of the respondent. This variable may under-report harms from a co-worker’s drinking, as the work colleagues may not know or be told why their co-worker is absent. However, often a colleague may be told about heavy drinking sessions, and also will know whether a co-worker is a heavy drinker in general (either through direct observation or conversations). When this person is absent -- perhaps especially on a Monday – reasonable assumptions may be made regarding the absence. While likely to produce conservative estimates, these questions are arguably no less likely to be valid than self-reports of individuals about their own drinking. This may also be the case for hung-over people who attend work and may cause an accident, or perhaps just go home ‘sick’ early.

A sensitivity analysis was undertaken on the outcome variable measuring one or more harms from co-workers. This analysis (presented in a supplementary appendix) compared an outcome variable that used the full five items with variables that used only four and three items. Although the outcome variable derived differently across countries was validated (See Supplemental Tables A1 and A2), the odds ratios determined in the logistic regression may be identifying slightly different conceptual links between predictors and outcomes. Nevertheless, identifying the correlates of harm from co-workers’ drinking in each country and studying outcomes in HIC and LMIC was useful and comparisons across countries should be interpreted with caution.

Underestimation:

The proportion of Australian and New Zealand respondents who experienced one or more harms from co-workers’ drinking has been underestimated because these countries’ surveys included three and four, respectively, of the five harm items included in this analysis. Moreover, in Australia only those respondents who identified that they were employed and who had answered affirmatively an earlier question about whether they had been negatively affected by heavy drinking co-workers were asked these questions. In New Zealand, respondents were asked to list people who had negatively affected them because of their drinking. Only if they listed (unprompted) a co-worker would they have been asked these further questions. In Switzerland, US and Ireland before completing specific harm items, respondents had to answer yes to “Any problems with colleagues/boss due to their drinking”, and in Nigeria, India, Sri Lanka, Thailand, Vietnam and Lao PDR respondents had to first respond that they worked with others to be asked questions about harms from their co-workers’ drinking. These procedures will have resulted in some underestimation of respondents who might have reported harms from co-workers’ drinking if they had been prompted by the specific items or if they had been harmed by co-workers who were less than heavy drinkers.

Attribution of harm to alcohol: In places like India where a minority of men drink and when they do, drink a lot, the abstaining others often have strong negative opinions about drinking. This anti-drinking bias may affect the tendency to find fault with immoderately (or even moderately) drinking co-workers.

Selection bias:

It is important that this paper is not interpreted as a definitive comparison of HIC versus LMIC as only a small number of countries were selected and this selection was not done randomly on the basis of specified income levels. Yet the findings do begin to highlight that harm from coworkers’ drinking is a problem common across many countries, and one that appears to be a somewhat greater problem in the LMICs versus the HICs in this study.

Additional questionnaire items, for example, on sexual harassment due to a co-worker’s drinking, should be added to future surveys. The absence of such questions may explain in part why women were less likely to report harms from co-workers’ drinking. Type of occupation, organisational size, part-time work, and casual and seasonal work are all likely to impact upon the exposure of respondents to potentially heavy-drinking coworkers. This study was part of a broader harm to others study and as such could not provide detailed information on all domains of life affected by others’ drinking. Future studies should more consistently measure part-time work and type of occupation. Nevertheless, this study suggests gender and country-level income differences in harms from co-workers’ drinking and begins to highlight the extent of the problem cross-nationally.

The I² values in Figures 1a and 1b showed a high level of heterogeneity and indicate between-studies variability (not the same as sampling error). While the confidence intervals around these estimates are relatively tight, these values should not be interpreted as measures of global prevalence. Finally, the data are cross-sectional and therefore cannot be used to attribute causality.

Figure 1b.

Proportion of employed women in high-income countries and in low- and middle-income countries who experienced any harm from co-workers’ drinking in the last 12 months.

Any harm from co-workers’ drinking was derived from three items for the Australian sample and four items for the New Zealand sample (compared to five items used for each of the remaining 10 countries); Overall estimates are pooled across country-level estimates via the DerSimonian-Laird method of two-stage inverse-variance random-effects meta-analysis using individual participant data; Weights of the contribution of country-level estimates to pooled estimates are represented by the relative area of the corresponding grey square.

Conclusion

This study estimates the prevalence of harms from co-workers’ drinking in each of the countries studied. Male workers were more at likely to experience harm than female workers overall. Although these results were based on only a small number of countries, workers in middle- and lower-income countries appeared to be more at risk than workers in higher-income countries in the study.

More stringent alcohol-related regulations, enforcing alcohol-free workplaces, providing supports and treatment (termed Employee Assistance Programs in the US) for workers who drink in heavy episodic ways (Carson and Balkin, 1992) and their colleagues within workplaces in LMIC may make workplaces safer for drinkers themselves and their co-workers. Strategies that address heavy drinking cultures associated with workplaces should also be considered. Whether prevention strategies are effective should be evaluated as programs to reduce harm from co-workers are introduced.