Effects of Lowering the Minimum Alcohol Purchasing Age on Weekend Assaults Resulting in Hospitalization in New Zealand

Abstract

Objectives. We estimated the effects on assault rates of lowering the minimum alcohol purchasing age in New Zealand from 20 to 18 years. We hypothesized that the law change would increase assaults among young people aged 18 to 19 years (the target group) and those aged 15 to 17 years via illegal sales or alcohol supplied by older friends or family members.

Methods. Using Poisson regression, we examined weekend assaults resulting in hospitalization from 1995 to 2011. Outcomes were assessed separately by gender among young people aged 15 to 17 years and those aged 18 to 19 years, with those aged 20 and 21 years included as a control group.

Results. Relative to young men aged 20 to 21 years, assaults increased significantly among young men aged 18 to 19 years between 1995 and 1999 (the period before the law change), as well as the postchange periods 2003 to 2007 (incidence rate ratio [IRR] = 1.21; 95% confidence interval [CI] = 1.05, 1.39) and 2008 to 2011 (IRR = 1.20; 95% CI = 1.05, 1.37). Among boys aged 15 to 17 years, assaults increased during the postchange periods 1999 to 2003 (IRR = 1.28; 95% CI = 1.10, 1.49) and 2004 to 2007 (IRR = 1.25; 95% CI = 1.08, 1.45). There were no statistically significant effects among girls and young women.

Conclusions. Lowering the minimum alcohol purchasing age increased weekend assaults resulting in hospitalization among young males 15 to 19 years of age.

Hazardous consumption of alcohol is a leading contributor to the global burden of disease, causing 3.8% of all deaths and 4.6% of disability-adjusted life-years (DALYs).¹ In New Zealand, it causes 5.4% of deaths and 6.5% of DALYs.² The burden is borne disproportionately by men (8.8% of DALYs vs 4.3% of DALYs among women), the young (mortality rates peak among those aged 15–29 years), and Māori (the indigenous people of New Zealand, whose alcohol-related mortality rate is 2.5 times that of the non-Māori population).² The economic burden of hazardous drinking is estimated to be 4% of the country’s gross domestic product,³ placing it among the most costly modifiable risk factors in New Zealand.

There are various evidence-based strategies for reducing alcohol-related harm, prominent among them restricting the availability of alcohol to young people via a minimum drinking or purchasing age.⁴ In 1999, contrary to this evidence, the New Zealand Parliament lowered the minimum alcohol purchasing age from 20 to 18 years. Four studies of the short-term effects of that law change on young people have been published in the scientific literature; they showed increases in emergency department admissions for intoxication,⁵ increases in disorder offenses and drunk driving,⁶ and higher traffic crash injury rates than would have been expected in the absence of the change.^7,8

Injuries stemming from assault have rarely been studied as an outcome of a change in the minimum alcohol consumption or purchasing age.^9,10 Etiologic fractions estimated in the Global Burden of Disease project suggest that between one third and one half of assaults resulting in hospitalizations are attributable to alcohol consumption on the part of either the assailant or the victim.² Efforts to evaluate the short-term effects of New Zealand’s law change on assaults have been thwarted by data limitations and insufficient statistical power.¹¹ Data limitations have included the lack of a routinely used indicator of alcohol involvement in hospital records.¹² Analyzing all assaults—irrespective of alcohol involvement—is likely to increase the risk of type II error because many are not alcohol related and their inclusion adds noise to the data.

Lack of knowledge about the impact of this law on assault rates is a shortcoming with respect to public health policy because, in contrast to the situation with traffic crash injuries, there are no evidence-based targeted interventions (e.g., roadside breath testing or sobriety checkpoints) to ameliorate the effects of the increased availability of alcohol brought about by lowering the minimum drinking or purchasing age.

Our conceptual model of the basic mechanism of the law change is shown in Figure 1. Reducing the minimum purchasing age from 20 to 18 years increased legal access to alcohol among young people aged 18 to 19 years. It is also likely to have increased informal access through those aged 18 to 19 years supplying alcohol to their friends and siblings younger than 18 years, which was legal in the context of a “private party” under New Zealand law until December 2013. Research conducted in 2000 showed that underage purchases of alcohol were relatively common.¹³ Although it is likely that prescriptive norms concerning alcohol consumption among those aged 18 to 19 years became more permissive after the law change, data are lacking on this presumption.

FIGURE 1—

Conceptual model of the effects of lowering the minimum alcohol purchasing age on assault rates.

All of these factors are likely to increase alcohol consumption and the exposure of young people aged 18 to 19 years to licensed premises where they encounter other young people impaired by alcohol. Greater alcohol consumption and increased exposure to other alcohol-impaired people in licensed premises would be expected to independently and multiplicatively increase the probability of involvement in an assault.

With the additional person-years of exposure accumulating since the 1999 law change, we sought to evaluate the effects on weekend assaults (Friday through Sunday), incidents in which there is a high probability that alcohol consumption is a contributing factor, thus reducing the degree of type II error. There is evidence from other countries that assaults are much more common during weekends than earlier in the week^14,15 and that the proportion of assaults attributable to alcohol is also greater on weekends.¹⁵

Accordingly, we tested a pair of hypotheses using young people aged 20 to 21 years as a control age group. First, we predicted that weekend assaults would increase among those aged 18 to 19 years, the age group whose legal access to alcohol increased with the law change. Second, we predicted that assaults would increase among those aged 15 to 17 years, a group with an already high prevalence of hazardous drinking¹⁶ in whom informal access to alcohol probably increased after the law change.

METHODS

We used a controlled before-and-after design with 3 age groups: the target group (18–19 years), those potentially affected by “trickle down” exposure (15–17 years), and a control age group (20–21 years). We used 16 years of data, with a prechange period (December 1, 1995–November 30, 1999) and 3 postchange periods each spanning 4 years (December 1999–November 2003 [postchange period 1], December 2003–November 2007 [postchange period 2], and December 2007–November 2011 [postchange period 3]).

Outcomes

The outcomes of interest were the counts of all individuals in each gender and age group who were admitted to hospitals in New Zealand between 12:01 am on Friday and midnight on Sunday with injuries caused by an assault (irrespective of alcohol impairment). In New Zealand, the Ministry of Health Information Directorate is the custodian of the National Minimum Data Set (NMDS), which contains information on all publicly funded inpatient treatment of injuries in New Zealand hospitals. Emergency department admissions were included if the treatment period was more than 3 hours, and readmissions were excluded according to the procedure described by Davie et al.¹⁷ This case definition was used because it is applied systematically for the entire country.

In the NMDS, the International Classification of Diseases, Ninth Revision (ICD-9),¹⁸ was used to code injuries up to 1999; subsequently, the Australian modification of the 10th revision (ICD-10-AM) was used.¹⁹ Assaults were identified as discharges with a first-listed external cause of injury code in the range of E9600 to E9690 (ICD-9) or X85 to Y09 and Y871 (ICD-10-AM). Hospitalized assaults occurring at times when alcohol is likely to be involved (e.g., 10 pm to 6 am) could not be used as outcomes because times of injury and details of the event are not routinely recorded.

We examined rates of events occurring over the study period among the 18 to 19-year-old group, the 15 to 17-year-old group, and 20 to 21-year-old control group. Data on estimated resident populations obtained from Statistics New Zealand were used to produce the denominators for incidence rates.

Data Analysis

The assumption underlying the inclusion of those aged 20 to 21 years as a control group is that they will have been exposed to the same economic conditions, police enforcement levels, and other alcohol availability variables (e.g., density and proximity of outlets²⁰) influencing assault rates in the younger age groups. Also, the rate of events in this age group would not be affected by the law change.

We used Poisson regression to model changes (from the prechange period to each of the 3 postchange periods) in incidence rates among young people aged 18 to 19 years and 15 to 17 years relative to those aged 20 to 21 years. The exponents of the fitted coefficients are equivalent to incidence rate ratios (IRRs), with the prechange or postchange by age group interaction terms providing prechange and postchange incidence rate ratios relative to the comparison group.

We considered the merits of conducting a separate analysis involving the same data and case definitions in which monthly, quarterly, or annual counts were used to compare trends before and after the law change in the 3 age groups. Of most interest in such an analysis is the 3-way interaction of age, period, and time, which indicates whether changes in trends in the number of people injured per year within each age group differ between the periods before and after the law change. This approach has been used to examine the effects of changes in pub trading hours on assault rates.¹⁰ We rejected the approach because counts were relatively low and the results highly sensitive to volatility in prechange trends, which were also shorter than desired because of changes in the ICD coding system and hospital coding practices.

The approach we selected averaged estimates over each of the 4-year prechange and postchange periods and is relatively easy to interpret. We used 4-year periods to reduce volatility in our rate estimates, particularly those among young women, and to ensure consistency with our previous work.⁸ Given previous evidence of male–female differences in the effects of this change and similar legislative changes,⁹ all of the analyses were undertaken separately by gender.

Changes in service delivery variables, such as hospital coding practices, can create trends in injury rate estimates that do not reflect changes in the true incidence of injury.²¹ In the present case, it should be noted that in 1999 emergency department cases requiring more than 3 hours of treatment began to be recorded in the NMDS in an increasing number of geographic areas, and in 2009 this practice became mandatory.²² The effect on the overall data is a spurious increase that reflects, at least in part, this change in recording. We included all events recorded in the NMDS on the assumption that there would be no systematic variations in the timing and completeness of the transition to inclusion of emergency department cases according to age, particularly in contiguous age groups (i.e., 15–17 years, 18–19 years, and 20–21 years).

RESULTS

Figure 2 presents assault rates (as recorded in administrative hospital discharge records) in each of the 3 age groups. Rates increased in each group over the entire period studied; however, as noted, the increase beginning in 1999 was attributable in part to the inclusion of emergency department events requiring more than 3 hours of treatment. Accordingly, differences in the change in rates over time between the age groups are the basis of inference from these data regarding possible effects of the law change.

FIGURE 2—

Changes in the incidence of assaults, by age group, among 15- to 21-year-old (a) boys and young men and (b) girls and young women: New Zealand, 1995–2011.

^aThe rate is per 10 000 population years.

The results of the Poisson regression analysis are presented in Table 1. Relative to the prechange period (1995–1999), there were significantly greater increases in weekend assaults involving hospitalizations among young men aged 18 to 19 years than among young men aged 20 to 21 years between 2004 and 2007 (IRR = 1.21; 95% confidence interval [CI] = 1.05, 1.39) and between 2008 and 2011 (IRR = 1.20; 95% CI = 1.05, 1.37). Among boys aged 15 to 17 years, assaults increased during 1999 to 2003 (IRR = 1.28; 95% CI = 1.10, 1.49) and 2004 to 2007 (IRR = 1.25; 95% CI = 1.08, 1.45). For periods in which effect estimates were nonsignificant (1999–2003 for those aged 18–19 years and 2008–2011 for those aged 15–17 years), the estimates were in the hypothesized direction (i.e., incidence rate ratios < 1).

TABLE 1—

Weekend Assaults Involving Hospitalizations, by Age and Time Period, Among Young People 15–21 Years of Age in New Zealand: 1995–2011

Age Group and Period	Mean No. of Assaults per Year	Population (per Year)	Rate (per 10 000 per Year)	Within-Age-Group Postchange/Prechange IRR (95% CI)	IRR Ratio (Effect Estimate) (95% CI)a
Males
15–17 y
Prechange	133	83 453	15.9	1.00	1.00
Postchange 1	199	87 531	22.8	1.43 (1.28, 1.60)	1.28* (1.10, 1.49)
Postchange 2	234	97 036	24.1	1.52 (1.36, 1.69)	1.25* (1.08, 1.45)
Postchange 3	214	96 858	22.0	1.39 (1.24, 1.54)	1.04 (0.90, 1.21)
18–19 y
Prechange	166	54 726	30.3	1.00	1.00
Postchange 1	211	57 422	36.5	1.20 (1.09, 1.33)	1.08 (0.93, 1.24)
Postchange 2	274	61 698	44.4	1.46 (1.33, 1.61)	1.21* (1.05, 1.39)
Postchange 3	324	67 319	48.2	1.59 (1.45, 1.74)	1.20* (1.05, 1.37)
20–21 y
Prechange	170	53 735	31.5	1.00	1.00
Postchange 1	200	56 734	35.2	1.12 (1.01, 1.24)	1.00
Postchange 2	229	60 008	38.1	1.21 (1.09, 1.33)	1.00
Postchange 3	281	67 196	41.9	1.33 (1.21, 1.46)	1.00
Females
15–17 y
Prechange	29.8	79 658	3.7	1.00	1.00
Postchange 1	38.0	84 211	4.5	1.21 (0.95, 1.54)	0.82 (0.58, 1.15)
Postchange 2	51.3	93 529	5.5	1.47 (1.17, 1.84)	0.96 (0.69, 1.33)
Postchange 3	56.0	92 071	6.1	1.63 (1.30, 2.03)	0.79 (0.58, 1.09)
18–19 y
Prechange	26.0	53 142	4.9	1.00	1.00
Postchange 1	37.3	55 951	6.7	1.36 (1.06, 1.75)	0.92 (0.65, 1.30)
Postchange 2	43.0	59 847	7.2	1.47 (1.15, 1.87)	0.96 (0.68, 1.35)
Postchange 3	69.3	63 970	10.8	2.21 (1.78, 2.77)	1.08 (0.78, 1.48)
20–21 y
Prechange	27.0	53 055	5.1	1.00	1.00
Postchange 1	41.8	55 355	7.5	1.48 (1.16, 1.89)	1.00
Postchange 2	46.0	59 032	7.8	1.53 (1.21, 1.94)	1.00
Postchange 3	66.5	63 684	10.4	2.05 (1.64, 2.57)	1.00

Note. CI = confidence interval; IRR = incidence rate ratio. The prechange period was 1995–1999; postchange 1 was 1999–2003, postchange 2 was 2003–2007, and postchange 3 was 2007–2011. Comparisons within and between age groups are IRRs from Poisson regression models.

^aRatio for the 15–17-year-old and 18–19-year-old groups relative to the 20–21-year-old group.

*P < .05.

Rates varied similarly in all 3 age groups of females. The Poisson regression analysis showed no significant effects after the law change relative to the preceding 4 years for either girls aged 15 to 17 years or young women aged 18 to 19 years (compared with those aged 20–21 years).

DISCUSSION

Increases in weekend assaults involving hospitalizations among young men in the age group (18–19 years) affected by New Zealand’s change in the minimum alcohol purchasing age were greater 5 to 12 years after the change than the increases seen among young men aged 20 to 21 years, which is consistent with our primary hypothesis. Results were similar for boys aged 15 to 17 years, consistent with a trickle-down effect. We found no increases in the risk of weekend assaults involving hospitalizations among girls and young women that could be attributed to the lowering of the minimum purchasing age.

In the first postchange period, there was a 28% increase in weekend assaults involving hospitalizations among boys aged 15 to 17 years (relative to young men aged 20–21 years), and this increase was largely sustained in the second postchange period. In contrast, young men aged 18 to 19 years exhibited a small increase in the first postchange period that strengthened over time. It is plausible that some of the delayed effect observed among young men aged 18 to 19 years resulted from their having been exposed to both more drinking and more violence when they were 15 to 17 years of age; however, this age group may also have been responsible for some of the increase seen in weekend assaults in the 15- to 17-years group.

In considering the pattern of findings, one needs to bear in mind that the age and gender of an individual hospitalized after an assault may not be the same as the age and gender of the perpetrator of the assault and that it is the degree of intoxication of either or both parties that is hypothesized to increase assault risk. Differences in age and gender distributions between perpetrators and hospitalized victims are likely to be larger among injured young women than among injured young men.²³ If this is the case, and given that alcohol-related assaults among young women in these age groups commonly involve older men,²³ it would not be expected that changes in hospitalizations among young women would as closely reflect their own increases in alcohol access as is the case among young men. It is, however, surprising that no effect was seen among young females aged 15 to 19 years, and there would be value in investigating this gender difference in future studies focusing on the minimum alcohol purchasing or drinking age.

Threats to the validity of the inferences drawn here include the possibility that young people aged 20 to 21 years are not a suitable group against which to compare assault rates in younger groups. We chose this group because it was likely to be affected most similarly to the younger groups by other influences on alcohol consumption and violence over time (e.g., reductions in disposable income caused by the global financial crisis and other changes in the availability of alcohol, such as the introduction of beer sales in supermarkets at the same time the minimum alcohol purchasing age was reduced). It seems unlikely that these extraneous factors could explain the divergent trends in assault rates between these age groups that coincided with the law change.

Caution should be exercised in interpreting trends in these assault data. The inclusion of emergency department events requiring more than 3 hours of treatment in the hospitalization record beginning in 1999 led to an apparent increase in assault hospitalizations. Accordingly, our inferences concerning effects of the law change rely not on change per se but, rather, on differences between changes among young people aged 15 to 17 years and those aged 18 to 19 years, and changes among those aged 20 to 21 years. We have no reason to expect that the coding of emergency department events or hospital admissions varied between the age groups over time; if there were variations, however, our effect estimates could be biased toward or away from the null.

Another limitation of our study is the use of a proxy for alcohol-related assaults, namely assaults occurring on Fridays, Saturdays, or Sundays that resulted in hospitalization. Some of the events included in the data set were not related to alcohol. Assuming that these events were not systematically distributed between the groups or over time in a way that could bias our effect estimates, they would merely have increased the width of the confidence intervals surrounding those estimates. Accordingly, we judge that this limitation biased the findings toward rather than away from the null. Similarly, it is plausible that young people aged 20 to 21 years were involved in assault incidents with those aged 18 to 19 years (as either perpetrators or victims) to a greater extent after the law change than beforehand as a consequence of the latter gaining access to licensed premises and coming into greater contact with young people aged 20 to 21 years in a high-risk setting. Such changes would also bias our effect estimates toward the null.

A recent study in Canada did not show an association between whether young people were above or below the minimum legal drinking age and the incidence of assault hospitalizations in which they were involved.²⁴ To our knowledge, no previous research has examined assaults in relation to changes in the drinking or purchasing age. Laboratory studies suggest that the effect of alcohol consumption on aggression is weaker among women than among men.^25,26 It should be noted that, in our study, assault rates were 4 to 6 times higher among males aged 15–19 years (Table 1), indicating that the deleterious effects of this law change include increasing the already large gender gap in morbidity resulting from adolescent injuries.²⁷

We conclude that lowering the minimum alcohol purchasing age probably increased the incidence of assaults inflicted on young men aged 15 to 19 years. The mechanism of this effect is uncertain but plausibly includes a combination of factors that need further investigation. One factor is increased access to alcohol among those aged 15 to 17 years via illegal sales. There is evidence that those aged 15 to 17 years were commonly purchasing alcohol in the years after the law change,¹³ but we know of no studies that have examined changes in access from before to after the law change.

A second possibility is increased provision of alcohol by older friends, siblings, parents, and other adults. There is evidence that such provision of alcohol—legal in private settings, where most drinking occurs—is commonplace in New Zealand,²⁸ and many parents view providing their 15- to 17-year-old children alcohol in some circumstances as potentially protective against later risky drinking.²⁹ A final consideration is the possibility of increased contact with young people aged 18 to 19 years who are impaired by alcohol. Risk factors for assault include being young (18–25 years), male, unemployed, and a frequent or heavy drinker.³⁰

Lowering the minimum alcohol purchasing age to 18 years has the effect of introducing a large number of legal purchasers into the school system given that, on average, at least half of the students in the final year of high school turn 18 during that year. It would be surprising if members of this group did not supply their 17-year-old friends with alcohol in private settings. In December 2013, it became illegal in New Zealand to supply alcohol to individuals younger than 18 years without the “express consent” of their parent or guardian. It will be important to investigate the degree of compliance with this restriction and determine the effects on drinking and related harms.

Our findings add to a growing body of evidence showing deleterious effects of reducing the age at which young people can access alcohol. This evidence base is particularly strong with respect to traffic crash injuries³¹ but is more sparse in relation to other outcomes. In contrast to decreasing trends in traffic injuries, rates of serious assaults are increasing in New Zealand, particularly among young people, Māori, and people living in deprived areas.³² Increasing the minimum alcohol purchasing age should be considered as a countermeasure.