Racial/Ethnic Disparities in the Risk of Injury Related to the Frequency of Heavy Drinking Occasions

Abstract

Aims

To estimate the risk of injury associated with the frequency of heavy drinking days overall and for black, white and Hispanic drinkers in a US sample.

Methods

Data are from the 2010 National Alcohol Survey and included 6506 respondents comprising the landline sample. Analyses utilize Cox proportional hazards models with age as the timescale in a retrospective cohort design. Life-course drinking is determined by age of onset and questions on heavy drinking by decade of life. The outcome measure is having had a serious injury at a certain age. Models estimate the risk of injury in relation to heavy drinking in each year controlling for demographics, risk taking and time varying measures of smoking and chronic disease.

Results

Results indicate that the risk of injury increases with the frequency of heavy drinking days to a hazard ratio of 2.14 (1.45–3.14) for daily heavy drinkers. Risks for white respondents were similar to the overall results but different risk relationships were found for black respondents among whom only daily heavy drinkers had increased risk of 4.09 (2.11–7.93), and for Hispanic respondents where elevated risk was seen among yearly heavy drinkers 2.71 (1.29–5.68), with a similar risk estimate for monthly heavy drinkers but lower and non-significant risks found for more frequent heavy drinking categories.

Conclusions

Different risk relationships were found across race/ethnicity groups suggesting elevated risk with less frequent heavy drinking among Hispanic respondents and very high risk from daily heavy drinking among black respondents.

INTRODUCTION

Although a well-documented literature from studies in hospital emergency rooms (ERs) has established the association of injury with drinking prior to the event (Romelsjö, 1995; Cherpitel, 2007), less is known about the risk at which various drinking patterns or quantities per occasion place the individual for injury or how representative these data are of the general population. ER studies have found that risk of injury increased with volume of drinking and episodic heavy drinking, however, controlling for drinking in the preceding 24 h, high volume drinkers were at a lower risk of injury than low volume drinkers (Gmel et al., 2006) suggesting that tolerance and behavior modification based on past intoxicated experiences may reduce injury risk for regular heavy drinkers (Cherpitel et al., 2006). Another analysis of ER data across 19 countries found usual volume strongly predicted an alcohol-related injury, and controlling for volume, both episodic heavy and frequent heavy drinking were also predictive of injury (Cherpitel et al., 2012). However, studies have found ER patients are more likely to be frequent heavy drinkers compared to those in the general population from which they come (Cherpitel, 1992) and compared to injured patients seeking other kinds of treatment or no treatment (Cherpitel, 1994), both of which raise the question as to the representativeness of data on risk of injury from drinking based on studies of ER patients.

Population studies of injury risk indicate increased risk for heavy and dependent drinkers (Vinson et al., 2003) and increased risk for heavy spirits drinkers (Watt et al., 2004). Risk function analyses of alcohol and injury in the US general population have found a ‘j-shaped curve’, with the lowest risk at lower levels of alcohol consumption compared to abstention, and with risk increasing more rapidly at higher levels of consumption before leveling off at 6–8 days of consuming five or more drinks (5+ days) during the last year (Cherpitel et al., 1995; Cherpitel and Ye, 2009), although risk for women continued to increase with the frequency of 5+ days (Cherpitel and Ye, 2009). However, an international review found a dose–response relationship suggesting that many injuries were associated with low and medium consumption levels (Kuendig et al., 2008).

Difference in risk of injury from drinking across racial and ethnic groups has received little attention in the literature, although the need for more research on injury risk disparities has been highlighted (Keyes et al., 2012). Previous findings regarding injury risk disparities have not been consistent with other alcohol-related problem disparities. For example, an ER sample of white, black and Hispanic injured patients found black respondents were less likely to have a positive BAC or to report heavy drinking or drunkenness either in the event or in the past year compared to whites and Hispanics (Cherpitel, 1998), although findings from the literature on alcohol-related problems suggest higher risks of injury might be expected among blacks (Zapolski et al., 2014) and at lower levels of drinking or 5+ frequency (Witbrodt et al., 2014). Similar to black men, Hispanic men have been shown to continue heavy drinking at older ages (Caetano and Kaskutas, 1995), and to have a greater risk for alcohol problems (Galvan and Caetano, 2003; Mulia et al., 2009), than white men; however, lower risks of problems have been found for Hispanic women (Galvan and Caetano, 2003). Recent findings from the US general population found blacks and Hispanics were significantly more likely to report drinking prior to injury (RR = 2.59 and 2.68, respectively) while risk of an alcohol-related injury was not significant for whites (RR = 0.89) (Cherpitel and Ye, 2014). Disparities in injury risk from drinking may differ from those in social problems in that behaviors may be problematized at lower levels in black and Hispanic cultures with higher levels of abstainers and may become problems more quickly for those with fewer social and economic resources (Zapolski et al., 2014), while injury risk may be more consistently defined across groups and cultures.

Analysis here of the relationship between life-course heavy drinking and ever having had a serious injury, utilizing a retrospective cohort design (Hudson et al., 2005), will add a new dimension to injury risk studies by considering, in a general population sample, risk of injury from heavy drinking during earlier life decades where this type of drinking is most common, and examine racial/ethnic differences in this relationship. This life-course perspective on injury risk differs from event-focused studies in assessing relationships between individual's general drinking pattern and injury risk. This perspective allows for less proximate influences such as the effects of hangovers on work-related or other injuries the following day and indirect effects through other long-term health problems or impaired decision making. We hypothesize that the frequency of 5+ days will be positively related to the risk of serious injury in a given year and will address the question of whether risk levels off at a low frequency or continues to rise with more heavy drinking occasions. We also hypothesize that black and Hispanic drinkers will have a greater risk of injury across levels of heavy drinking frequency.

METHODS

Data

The 2010 US National Alcohol Survey was a Computer Assisted Telephone Interview (CATI) household survey of the US adult population aged 18 or older, with a sampling frame including all 50 states and the District of Columbia conducted for the Alcohol Research Group by ICF Macro between June 2009 and March 2010. A Dual-Frame design, including both Landline and Cellular Phone cases, was implemented. The landline sample included a base sample and ethnic minority oversamples for Hispanic and African American populations where an adult in the surveyed household at a private residence was randomly selected for interview using the Kish Grid Method (Kish, 1965). The average interview time was 55 min for landline completed interviews. Questions on demographic, alcohol use and health problem measures comprise the early sections of the instrument. Cell phone respondents were asked a limited set of questions that did not include the injury outcome measure and were therefore excluded from these analyses. The cooperation rate was 49.9% for the landline samples (N = 6855). Of these, 140 respondents did not answer the injury question, and among those reporting injury, 59 failed to report a valid age at injury. Another 150 respondents were both missing on age and reported no injury during their lifetime. As the relevant information is needed for survival analysis, these individuals were excluded, resulting in an analytic sample of 6506. Although low, the cooperation rate is consistent with those from recent telephone surveys (Curtin et al., 2005). The cooperation rate is also of reduced concern in this study as our interest is in the association between lifetime drinking and injury occurrence, rather than any population prevalence estimates.

Measures

Lifetime injury occurrence was assessed by the question ‘have you ever been told by a doctor or other health professional that you had injuries from a serious accident?’ Those who answered yes were then asked ‘at what age were you first told?’ These questions were embedded in a series of question on the respondents' lifetime health conditions including heart disease, diabetes, etc. (described below).

Lifetime heavy drinking was derived from a series of questions on the frequency of heavy drinking beginning with the respondents' age of drinking onset. Four items elicited the frequency of drinking five or more (5+) drinks in a day during specific life decades, i.e. the teens, 20s, 30s and 40s, using five response options: ‘every day or nearly every day’, ‘at least once a week’, ‘at least once a month’, ‘at least once a year’ and ‘never’ (Greenfield et al., 2014). For those aged 50 and older, 5+ frequency during the last 12 months prior to the interview date, based on the graduated frequency series items on 12 or more, 8–11 and 5–7 drinks (Greenfield, 2000), was used. Drinks were defined as the US standard 14 g of ethanol examples for beer, wine and liquor pours were provided. These retrospective lifetime drinking questions are combined to create the measure of annual heavy drinking at each age, assuming a constant heavy drinking pattern for each decade and, for those aged 50 and more, from age 50 to the year before the interview.

Time-invariant control variables included demographics such as gender, race/ethnicity, family income, education, employment status and marital status, as well as risk taking disposition. The risk taking/impulsivity and sensation seeking scale (Greenfield et al., 2011) is the mean of seven items including ‘I often act on the spur of moment without stopping to think’ and ‘I like to try new things just for excitement’. Note while all of these measures except gender and age could potentially vary during a respondent's life history, because only the status at the time of interview was recorded, these measures were necessarily treated as time-invariant in the analyses.

Time-varying control variables included the age of first diagnosis for measured diseases and conditions that could potentially influence injury risk including hypertension, heart disease, diabetes, stroke and cancer, using questions in the same format as the serious injury variable described above. These health condition measures were coded as negative for the years before the onset of each disease and for those with no lifetime incidence and positive for the years after the age of onset for each disease or condition. The other time-varying control variable was smoking status in each year of life, coded from questions on the age of smoking onset and age of last use.

Data analysis

Cox proportional hazards models (Cox, 1972) were used to estimate the hazard of injury occurrence from both the time-varying and time-invariant predictors, assuming a relationship of h(t) = h0(t) exp(β₁X₁ + β₂X_2t), in which h0(t) is the baseline hazard function, X₁ is the vector of time-invariant variables, X_2t is the vector of time-varying variables at time t, and exp(β) is the estimated hazard ratio. The dataset was structured with each respondent aligned by age, right-censored at either the age of injury occurrence or, for those had no major injury during their lifetime, the age of interview. This allows for the control of age by design (Korn et al., 1997). The data were also left-censored at age 14, as injury before that is presumably highly unlikely to be related to drinking. All analyses were performed in Stata 11 (Stata Corp., 2009) utilizing sampling weights accounting for probability of selection, non-response and US population distributions. The proportionality of the hazards assumption was tested using Schoenfeld residuals based on the correlation between the scaled Schoenfeld residuals and rank of survival time. The tests have both covariate specific form and a global form for all covariates combined (Grambsch and Therneau, 1994; Keele, 2010). Covariates highly significant in the Schoenfeld residuals tests were stratified in the final models. Baseline hazards in the race/ethnicity group-specific models were stratified for variables highly significant in the Schoenfeld residuals tests thus violating the proportionality assumption. These variables were marital status and income in the white group model, marital status and education in the black group model and employment and income in the Hispanic group model.

RESULTS

Table 1 shows the numbers and weighted percentages of injury occurrence for the overall sample and for white, black and Hispanic samples separately. Also shown are the weighted percentages of 5+ drinks in a day frequencies during respondents teens, 20s, 30s and 40s as well as mean values or category percentages for demographic and other control measures. Rates of lifetime serious injury are highest for white and lowest for Hispanic respondents. Differences in 5+ frequencies are seen for each decade with white respondents having the highest prevalence for more frequent categories in their 20s while Hispanic respondents are generally highest in the other decades. Black drinkers generally report lower frequencies of 5+ days but differences are smaller for the 30s and 40s age categories.

Table 1.

Sample descriptive statistics including injury occurrence, heavy drinking during decades, basic demographics and control variables^a

	Total (n = 6506)	White (n = 3641)	Black (n = 1353)	Hispanic (n = 1299)
Any lifetime injury: n %	1061	674	190	161
	18.0	19.5	14.8	11.9
5+ during teens (%)
Nearly everyday	2.7	2.3	3.0	2.9
At least weekly	12.5	13.0	6.7	15.2
At least monthly	14.8	17.3	6.4	11.3
At least yearly	10.2	11.4	7.7	7.8
5+ during 20s (%)b
Nearly everyday	6.7	6.7	8.1	4.1
At least weekly	19.7	21.1	13.4	18.8
At least monthly	17.2	18.6	12.2	16.5
At least yearly	12.4	13.7	7.5	10.0
5+ during 30s (%)b
Nearly everyday	3.9	3.6	3.6	4.5
At least weekly	10.1	9.5	10.4	12.6
At least monthly	14.8	15.5	11.1	15.5
At least yearly	16.7	18.7	7.9	13.4
5+ during 40s (%)b
Nearly everyday	2.2	2.3	2.6	1.0
At least weekly	6.7	6.2	7.7	9.7
At least monthly	10.2	11.0	7.1	11.0
At least yearly	15.8	16.8	10.3	15.6
Gender male (%)	49.5	49.0	47.7	52.9
Mean age	46.2	48.2	43.1	39.2
Mean age at first injuryc	30.2	30.7	30.7	27.2
Race/ethnicity (%)
White	68.9
Black	11.2
Hispanics	13.1
Other	6.8
Family income (%)
≤$10,000	11.1	7.5	23.9	16.6
$10,001–20,000	15.2	12.4	22.4	25.5
$20,001–30,000	11.2	10.5	11.3	15.3
$30,001–40,000	10.5	10.7	11.5	11.2
$40,001–60,000	14.6	16.1	11.4	11.4
$60,001–100,000	22.3	25.3	13.2	11.4
>$100,000	15.1	17.5	6.4	8.5
Education (%)
Less than HS grad	15.0	10.2	20.2	35.7
HS grad	30.3	30.5	36.7	28.6
Some college	28.8	29.8	23.1	25.5
College grad	25.9	29.5	20.0	10.2
Marital status (%)
Married/cohabiting	64.2	68.1	42.1	63.7
Separated/divorced	8.2	7.7	12.5	7.1
Widowed	6.0	6.5	6.6	2.7
Never married	21.7	17.7	38.9	26.5
Employment status (%)
Employed full time	42.9	44.9	39.6	36.5
Employed part time	12.4	12.0	9.0	15.9
Unemployed	9.3	7.6	17.8	13.0
Others	35.4	35.6	33.6	34.6
Mean risk taking (0–3)	0.81	0.84	0.69	0.71
Lifetime disease (%)
Hypertension	31.1	31.3	39.5	24.5
Heart disease	12.2	13.2	11.4	9.0
Diabetes	11.0	10.5	12.2	10.4
Stroke	2.7	2.5	5.4	1.4
Cancer	7.4	8.8	4.1	3.0
Smoking lifetime (%)	45.1	48.3	42.0	30.4

^aReported means and percentages are all weighted, but numbers (n) are unweighted.

^bOnly for those aged at least 20, 30 or 40 at interview.

^cOnly for those who reported experiencing a severe injury.

Table 2 shows the hazard ratios (HR) for heavy drinking (5+) frequencies from yearly to daily, with the broad group reporting no 5+ drinking days as the reference, estimated using the Cox proportional hazards models with age as the timescale. The analysis was first performed for the overall sample, then for white, black and Hispanic samples separately. We first fit raw models that controlled only for gender and, for the overall sample, race/ethnicity. The final models further controlled for income, education, marital status, employment and impulsivity and risk taking, all time-invariant measures, as well as other time-varying control measures including lifecourse histories of hypertension, heart disease, diabetes, stroke, cancer and smoking status. Compared to the raw model, adding control variables affected estimates of heavy drinking on injury particularly for whites. For example, the HR for those reporting 5+ drinks daily was reduced from 4.24 to 2.11.

Table 2.

Hazard ratios (HRs) and 95% confidence intervals (CIs) from Cox proportional hazard models predicting injury occurrence from lifecourse yearly heavy drinking (5+ drinks in a day) frequencies with a reference group of those reporting no 5+ drinking days at each age

	Total sample	White	Black	Hispanic
Raw modela
Yearly	1.50 (1.09, 2.07)*	1.42 (0.99, 2.04)	2.13 (0.67, 6.77)	3.07 (1.25, 7.57)*
Monthly	1.53 (1.07, 2.17)*	1.54 (1.02, 2.32)*	0.54 (0.20, 1.44)	2.73 (1.29, 5.78)**
Weekly	2.56 (1.88, 3.48)**	2.66 (1.88, 3.77)**	1.84 (0.73, 4.66)	2.45 (1.08, 5.59)*
Daily	3.86 (2.50, 5.96)**	4.24 (2.54, 7.08)**	5.24 (2.36, 11.62)**	2.78 (0.48, 16.14)
Full modelb
Yearly	1.32 (0.95, 1.84)	1.32 (0.92, 1.89)	1.92 (0.79, 4.71)	2.71 (1.29, 5.68)**
Monthly	1.38 (0.96, 1.98)	1.29 (0.87, 1.93)	0.56 (0.21, 1.48)	2.68 (1.34, 5.31)**
Weekly	1.79 (1.34, 2.39)**	1.86 (1.33, 2.60)**	1.90 (0.98, 3.67)	1.88 (0.91, 3.87)
Daily	2.14 (1.45, 3.14)**	2.11 (1.30, 3.42)**	4.09 (2.11, 7.93)**	1.98 (0.52, 7.47)

^aControlling for gender and, for total sample, race/ethnicity.

^bFurther controlling for income, education, employment status, marital status, risk taking and impulsivity and lifecourse history of hypertension, heart disease, diabetes, stroke and cancer and smoking.

*P < 0.05, **P < 0.01.

Focusing on the final model results, for the overall and white sample the HRs of heavy drinking on injury were about 1.3–1.4 and not significant for the 5+ yearly and monthly groups, increasing to 1.8–1.9 for those reporting 5+ weekly and 2.1 for the 5+ daily group (P < 0.01). For the black sample, the HRs across heavy drinking frequency groups fluctuated and were significant only at the daily level (HR = 4.1, P < 0.01). In contrast, for the Hispanic sample the HRs of the yearly and monthly 5+ groups were both significant (HR = 2.7, P < 0.01), dropping to elevated but non-significant HRs of 1.9–2.0 at the weekly and daily levels.

Results for the control variables (not shown in the tables) include an association of male gender with injury for all three ethnic groups (HR = 1.33, P < 0.05 for whites and 2.00–2.10, P < 0.01 for black/Hispanics). The impulsivity and risk taking scale was strongly predictive of injury in the white sample (HR = 1.91, P < 0.01), but not for the black or Hispanic samples. Higher educational attainment was found to be protective of injury in the white sample, but increased risk of injury for the Hispanic sample. Compared to less than high school graduate, HRs of high school graduate, some college and college graduate were 0.59 (P < 0.01), 0.70 and 0.45 (P < 0.01) for whites, but for Hispanics 2.86 (P < 0.01), 2.76 (P < 0.01) and 2.17, respectively. For the black sample, education was stratified in the final Cox model as it was highly significant in Schoenfeld residual tests, as described in the Methods section. For the black sample, higher income was found to be protective, with an HR of 0.48 (P < 0.05) for those with a family income of $30–60k and an HR of 0.61 for incomes >60k as compared to less than $30k. (Note income was stratified for both whites and Hispanics.)

DISCUSSION

Our results indicate significant risk of serious injury for weekly and daily 5+ drinkers at about double the rate for those with no 5+ days. While this pattern reflects results for white heavy drinkers, different risk patterns were seen for the minority groups. Black daily heavy drinkers had a risk that was nearly four times that of those with no 5+ days. Hispanic heavy drinkers did not evidence a similar gradient of serious injury risk, rather, risk was found to be non-significant for the weekly and daily groups, but was significantly elevated to an HR of about 2.7 for the lower frequency yearly and monthly heavy drinkers. These results differ from some previous studies with overall risks suggesting that risk rises with the frequency of heavy drinking rather than leveling off (Cherpitel et al., 1995; Cherpitel and Ye, 2009). Findings of differential risk profiles across the three race/ethnicity groups are generally as hypothesized with some higher risks among black and Hispanic drinkers at particular frequency levels but with some surprising aspects, particularly the results for Hispanics indicating greater risks at lower 5+ drinking frequencies. Potential explanations for this include higher tolerance and/or risk mitigating behaviors such as less engagement in activities among the frequent heavy drinkers in this group or that the less frequent Hispanic heavy drinkers only engage in very heavy events in contexts with greater risk of aggression and injury such as bars and parties (Chartier et al., 2014). Alternatively, any engagement in heavy drinking may be associated with some unmeasured injury risk characteristic, lifestyle or behavior such as work in higher injury risk occupations, in which case the results for alcohol risk would be spurious. Findings of an elevated risk disparity among daily black heavy drinkers may indicate a greater intensity of heavy occasions, potentially reaching higher BAC levels, consistent with a relative preference for spirits and malt liquor and with drinking in response to stressors such as racial discrimination, or greater vulnerability to injury through intoxication in contexts with more injury risk such as outdoor drinking and activity in disordered neighborhoods with poor infrastructure maintenance (Karriker-Jaffe et al., 2012; Jones-Webb and Karriker-Jaffe, 2013; Zapolski et al., 2014).

The interpretation of these analyses should consider several limitations. The disparities in injury risk seen for the most frequent black heavy drinkers and for less frequent Hispanic heavy drinkers could reflect a correlation with related, but unmeasured variables, such as drug use, propensity to engage in risky activities, or occupation-related injury risks, although we do control for important predictors of this, particularly the impulsivity and risk taking scale. Other unmeasured injury risk factors related to the respondent's environment such as gun ownership prevalence, high crime rate and lack or poor maintenance of infrastructure such as sidewalks and roads could also have potentially affected results. Related to such factors, findings may have differed if the cause of injury were taken into account; however, the relatively limited number of respondents precluded such analyses and should be the focus of future studies on life-course heavy drinking and the risk of injury.

Data on both injury and heavy drinking were based on self-report and retrospective recall. Test-retest analyses of the decades 5+ measures were conducted with the 2005 NAS and a follow-up 2–3 years later. Results indicated moderate consistency for each decade (ρ = 0.63–0.68); however, black and Hispanic respondents were less consistent (ρ = 0.56) (Greenfield et al., 2014). Only the frequency of 5+ days averaged by the respondent over each decade of life was assessed. Frequencies of lower or higher amounts, potentially relevant to injury risk, are not known. Only the earliest serious injury was recorded. Individuals who may have had multiple serious injuries were removed from the analysis after their first, potentially resulting in conservative estimates here. While models controlled for respondent's income, educational attainment and employment status at the time of the interview, information on these variables in the year of injury was not collected, potentially affecting study findings.

Disparities in injury risks related to the frequency of heavy drinking episodes suggest the possibility of greater risk of serious injury at lower frequencies of heavy drinking among Hispanic respondents and for the most frequent heavy drinkers among black respondents compared to relationships seen for non-Hispanic whites. Findings emphasize the importance of considering race/ethnicity in studies of injury risks related to alcohol use and heavy drinking patterns. This diversity of findings regarding life-course risks of heavy episodic drinking frequencies is, in part, consistent with findings regarding higher rates of alcohol-related problems among black and Hispanic drinkers, particularly at lower levels of drinking (Herd, 1994; Jones-Webb et al., 1997; Galvan and Caetano, 2003; Mulia et al., 2008, 2009). However, the elevated risk among black heavy drinkers was seen only in the most frequent category here and diverges from findings for other alcohol problems (Witbrodt et al., 2014; Zapolski et al., 2014). Studies with designs and additional measures addressing the limitations we have noted are needed to confirm and potentially explain these differences. This study employed a retrospective cohort design to evaluate life-course risk of serious injury related to heavy drinking frequency. This framing differs from most studies of injury risk where drinking in the event is the focus, allowing consideration of longer-term consequences of heavy drinking including hangover and impaired decision making. Overall results differ from some drinking-in-the-event studies indicating increasing risk with frequency rather than a leveling out of risk. However, findings for Hispanic and African American respondents show an increased risk for heavy drinking but with very different patternings across frequency categories.

FUNDING

This work was supported by the National Institute on Alcohol Abuse and Alcoholism (P50AA005595). Study protocols were approved by the Public Health Institute Institutional Review Board, IRB# I11-019.

CONFLICT OF INTEREST STATEMENT

The authors declare no financial conflicts of interest.