Trends in alcohol-attributable morbidity and mortality for Victoria

Journal of Public Health | Vol. 36, No. 3, pp. 399 –407 | doi:10.1093/pubmed/fdt063 | Advance Access Publication 30 June 2013

Trends in alcohol-attributable morbidity and mortality for Victoria, Australia from 2000/01 to 2009/10 Harindra Jayasekara1,2, Jason Ferris3,4, Sharon Matthews2,5, Michael Livingston6,7, Belinda Lloyd2,5 1

Centre for Molecular, Environmental, Genetic and Analytic Epidemiology, University of Melbourne, Melbourne, VIC, Australia Turning Point Alcohol and Drug Centre, Melbourne, VIC, Australia 3 The Institute for Social Science Research, The University of Queensland, Building 31B, Room 104, St Lucia, QLD, Australia 4 ARC Centre for Excellence in Policing and Security, The University of Queensland, St Lucia, QLD, Australia 5 Eastern Health Clinical School, Faculty of Medicine, Nursing and Health Sciences, Monash University, VIC, Australia 6 Drug Policy Modelling Program, National Drug and Alcohol Research Centre, University of New South Wales, Sydney, NSW, Australia 7 Centre for Alcohol Policy Research, Turning Point Alcohol and Drug Centre, Melbourne, VIC, Australia Address correspondence to Jason Ferris, E-mail: [email protected] 2

Background To examine trends in alcohol-attributable morbidity (AAMorb) (2000/01– 2009/10) and mortality (AAMort) (2000–07) by age, sex and region. Methods Time-series analyses of population data for Victoria, Australia. We used joinpoint regression to quantify trends by estimating quarterly percent change (QPC) for rates of morbidity and mortality. We present the average QPC (AQPC) as a weighted average of QPCs. A test of parallelism was used to examine pairwise differences. Results AAMorb increased significantly over time for Victoria (AQPC ¼ 1.0%, 95% confidence interval 0.8– 1.2). While females (1.6, 1.1–2.0), age groups 25 –44 (1.0, 0.9–1.1) and 45– 64 (1.2, 0.2– 2.2), and metropolitan population (1.2, 0.5–1.9) were broad subgroups more at risk, multivariate analysis detected specific increases for metropolitan females aged 15 –44 (1.8, 1.0– 2.6) and 45þ (1.6, 0.2–3.0). Relatively greater increases in morbidity among metropolitan subgroups were widespread. AAMort remained stable for Victoria and for most subgroups, although significant declines in mortality were specifically experienced by metropolitan 15– 24 (22.0, 22.9 to 21.0) and 25 –44 (21.0, 21.7 to 20.3) age groups, and by regional males aged 45þ (20.8, 21.3 to 20.3). Metropolitan males aged 45þ were a special high-risk population. Discussion Our study has identified overlooked subgroups as being at increasing risk for alcohol-attributable chronic harm necessitating their inclusion in future policies for harm reduction. Keywords alcohol consumption, joinpoint regression, morbidity, mortality, trends

Background Alcohol is a major determinant of global burden of disease, accounting for nearly 4% of deaths and 4.6% of DALYs.1 It is second only to tobacco as a preventable cause of drug-related hospitalization and death in Australia2 where the cost of alcohol-related social problems to the community was estimated to be around $15.3 billion in 2004/05.3 The increasingly heavy burden of alcohol-related harm on the healthcare system is more evident in Victoria, the second most populous state in Australia, than in any other state or territory where alcohol-attributable hospitalizations have seen the biggest increases in recent times. Livingston et al. 4 estimated

that alcohol-attributable hospitalization rates in Victoria had increased by 50% between 1999/2000 and 2007/08, whereas Pascal et al.5 observed a 75% increase between 1995/ 96 and 2004/05. This is in spite of alcohol-attributable deaths in Victoria remaining stable, which may reflect improvements

Harindra Jayasekara, Research Fellow Jason Ferris, Senior Research Fellow Sharon Matthews, Research Fellow Michael Livingston, Postdoctoral Research Fellow Belinda Lloyd, Senior Research Fellow

# The Author 2013. Published by Oxford University Press on behalf of Faculty of Public Health. All rights reserved. For permissions, please e-mail: [email protected].

399

Downloaded from http://jpubhealth.oxfordjournals.org/ by guest on October 30, 2015

A B S T R AC T

400

J O U RN A L O F P U B LI C H E A LT H

Methods The authors conducted time-series analyses to examine trends over time for AAMorb (2000/01 to 2009/10) and mortality (2000 – 07 excluding data for the fourth quarter of 2007 due to administrative processes related to data reporting across the 2007/08 period) using Victorian population data. Differences in trends by age, sex and geographical area (metropolitan versus regional) were explored. AAMorb data have been obtained from the Victorian Admitted Episodes Data Set of the Department of Health, Victoria with details of all acute hospital separations. The hospital separations related to alcohol were extracted based on the ICD-10-AM8 codes of the principal diagnosis associated with the separation using the method of alcohol-attributable fractions (aetiological fractions) derived from previous reviews of the literature by English and Holman2 and Ridolfo and Stevenson.9 Alcohol-attributable mortality (AAMort) data are based on the Unit Record Mortality data collected by the Australian Bureau of Statistics (ABS) for the respective calendar years. Deaths were attributed to alcohol based on the primary cause of death using the same method as described for morbidity.

Alcohol-related deaths reported here include deaths due to alcohol-specific diagnoses (e.g. alcoholic liver cirrhosis) as well as proportions of other diagnoses to which alcohol contributes (e.g. motor vehicle crashes, stroke and some cancers). Statistical analysis

Rates of AAMorb ( per 10 000 residents) and AAMort ( per 100 000 residents) were calculated per quarter (3 months). Using annual estimated residential populations (ERPs) produced by the ABS, quarterly rates were extrapolated by interpolating growth between each pair of annual ERPs. The quarterly rates were used (instead of annual rates) to derive more precise estimates of trends, to identify more exact time points where trends changed in a particular year and to enable the statistical software to generate the maximum number of points where the trends have changed. We used the exact method based on the Poisson distribution to calculate the standard error for rates.10 We used Joinpoint Regression Program11 to quantify temporal trends by estimating the quarterly percent change (QPC) for the rates. Joinpoint (or piecewise) regression is a statistical method for identifying sudden changes in long-term epoch trends for which rates are relatively stable and avoids the need to arbitrarily select a base for estimating the direction and magnitude of the slope. The approach uses statistical criteria to determine when and how often the QPC varies by fitting rates using joined log-linear segments. In the present analysis we specified the model to test with the maximum number of five joinpoints: each segment characterized by a QPC.12 The associated 95% confidence interval is indicative of the adequacy of the final model and the degree of random variation inherent in the underlying rates. The technique uses a Monte Carlo permutation (resampling across 5000 iterations) to test if an apparent change in trend is statistically significant. In order to derive a summary measure of the trends over the entire period, we calculated the average QPC (AQPC) for AAMorb and mortality. This is computed as a weighted average of the QPCs from the joinpoint model, with weights equal to the length of the linear segments. Furthermore, we tested for pairwise differences between subgroups of the population (e.g. males versus females) using a test of parallelism to see whether trends over time between groups are statistically different.13 All statistical analyses were performed using Joinpoint Regression Program and graphical outputs were created using Stata.14

Results An initial inspection of AAMorb rates comparing the first year 2001 to the last year 2009 revealed an overall increase in rates


in healthcare and perhaps enhanced access to services. Alcohol is the most widely used drug in the state with nearly half of all Victorians aged 15 years and above drinking daily or weekly,6 although the level of alcohol consumption has remained relatively stable over time.4 Traditionally, population subgroups such as young people, rural and regional populations, people with a mental illness and culturally diverse communities are considered to be particularly vulnerable to effects of harmful drinking.7 However, there is a scarcity of useful data around alcohol-related harms specifically focusing on age, sex and metropolitan-regional disparities including their trends over time even though it has been postulated that this may help explain disparities between harms and consumption.4 In other words, most research has focused on trends in overall rates of harm, with no attempt to determine which population subgroups are driving these trends. Understanding which components of the population are experiencing increasing rates of harm from alcohol is critical in targeting appropriate policy responses. Thus, in the present study, we examined trends in alcohol-attributable morbidity (AAMorb) and mortality, comparing the differences in trends across subgroups defined by age, sex and region. This study also extends previous analyses of trends in alcohol-related harm by modelling rates in order to derive linear trends and their magnitude, and identify actual points in time where the trends have significantly changed.

A L CO H O L - AT T R I BU TA B L E M O R B ID I T Y A N D M O R TA L I T Y

across all population subgroups. The overall rate of AAMorb for Victoria increased from 38.5 (95% confidence interval (CI) 37.9–39.0) per 10 000 residents in 2001 to 52.6 (52.0–53.2) per 10 000 residents in 2010. In contrast, the AAMort rate comparing the first year 2000 to the last year 2007 revealed an overall decrease in rates across all population subgroups. The overall rate for Victoria decreased from 15.1 (14.0–16.2) per 100 000 residents in 2001 to 13.4 (12.4–14.4) per 100 000 residents in 2007 (see Supplementary data, Table S1 for a complete summary of population demographics). Morbidity

by 3.1% (1.5 –4.7) per quarter from 2000q3 to 2002q2 followed by a much smaller increase by 0.7% per quarter (0.6 – 0.8) thereafter. Metropolitan females aged 15 –44 and 45þ showed the greatest overall increases in morbidity trends for the entire period under study (AQPC ¼ 1.8, 1.0 – 2.6 and AQPC ¼ 1.6, 0.2 – 3.0, respectively), although both groups showed a downward trend that overlapped (between 2005q2 and 2006q4 for metropolitan females aged 15– 44 and between 2005q4 and 2006q3 metropolitan females aged 45þ). Pairwise comparisons of trends between all sex-age subgroups by region (e.g. metropolitan males aged 15 –44 versus regional males 15 –44) rejected parallelism indicating the existence of diverging trends between these pairs. Mortality

As shown in Table 1 and Fig. 2, mortality remained stable for Victoria between 2000 and 07 (AQPC ¼ 0.0, 20.2 to 0.2). The rates of mortality also remained stable across subgroups of sex and region with no significant differences between subgroups (P ¼ 0.07 for test of parallelism for sex; P ¼ 0.27 for region). However, mortality decreased significantly and in parallel for the two younger age groups 15– 24 and 25– 44 (P ¼ 0.15 for test of parallelism) but not for the two older age groups where trends have remained stable. When mortality trends were examined separately for metropolitan and regional Victoria (see Table 2), the only significant observable trends were for decreasing mortality for regional males (AQPC ¼ 20.4, 20.8 to 20.1) and for people aged 15 –24 (22.0, 22.9 to 21.0) and 25– 44 (21.0, 21.7 to 20.3) in metropolitan Victoria. For all other sub-populations based on age and sex, the mortality trends were seen to be stable while none of the subgroups had evidenced any joinpoints. The only pairwise difference of significance across comparable sub-populations by region was seen for those aged 15 –24 (P , 0.05 for test of parallelism). In multivariate analysis of trends by age (using 15–44 and 45þ), sex and region, rates of mortality converged for regional and metropolitan males aged 45þ, with a consistent, significant decrease for regional males aged 45þ (AQPC ¼ 20.8, 21.3 to 20.3) and a consistent, significant increase for metropolitan males aged 45þ (0.5, 0.1–0.9). For metropolitan males aged 15–44, rates of mortality increased by 5.1% (20.2–10.7) per quarter from 2000q1 till 2001q4 but continued to decline significantly thereafter by 2.7% (23.6 to 21.9) per quarter. Mortality trends for all other sub-populations were stable (see Supplementary data, Fig. S2). Pairwise comparisons by region (metropolitan versus regional) between males aged 45þ and males aged 15–44 rejected parallelism, indicating the existence of diverging trends between these pairs.


Table 1 and Fig. 1 show the results for morbidity trends for Victoria overall and by age, sex and region. AAMorb had increased steadily for Victoria by 1.4% per quarter from third quarter of 2000 (2000q3) to 2005q1 (meaning that in the modelled values, 9.18 per 10 000 for 2000q3 had increased to 9.18 1.014 ¼ 9.31 per 10 000 for 2000q4). Thereafter, we saw a much smaller, but still significant, increase in morbidity (QPC ¼ 0.6, 0.4–0.8). The rate of increase for the whole period is denoted by the AQPC of 1.0% (0.8–1.2) per quarter, significantly different from an increase of zero. The overall rate of increase in morbidity for females was twice that for males, although the rates have slightly declined between 2005q3 and 2008q4, albeit non-significantly, for females. A test for parallelism indicated that differences in trends between males and females were statistically significant (P , 0.001). Largest increases in morbidity by age were seen for those aged 25–44 (consistent rise with no joinpoints) and 45–64 (with a decline between 2005q4 and 2006q3). Those aged 15–24 and 65þ showed parallel lines (P ¼ 0.22) with a steady rise followed by a recent downward trend. Metropolitan trends showed a significantly larger overall increase compared with regional trends. When metropolitan and regional data were modelled separately (see Table 2), morbidity rates of metropolitan males and metropolitan females had increased markedly compared with their regional counterparts (in spite of a decline from 2005q4 to 2006q3 for metropolitan females). By age, metropolitan populations aged 45–64 and 25–44 have experienced significantly larger increases in morbidity than their regional counterparts (P , 0.01 and P , 0.001, respectively, for test of parallelism). Morbidity rates had also increased significantly for people aged 65þ in metropolitan Victoria (with a dip after 2007q4 till 2010q2) than for the corresponding age group in regional Victoria (P , 0.001 for parallel trends). Multivariate analysis of trends by age (using 15 – 44 and 45þ categories), sex and region are shown in Supplementary data, Fig. S1. Metropolitan males aged 45þ had the highest rates of morbidity out of all subgroups, with a sharp increase

401

402


Table 1 QPCs and estimated AQPCs using joinpoint regression for alcohol-attributable hospital admissions (2000/01 –2009/10) and deaths (2000 – 07) by age group, sex, region and all Victoria Segment

QPC (95% CI)

AQPC (95% CI)

Alcohol-attributable hospital admissions Sex Male Female

2000q3– 2010q2

0.8a (0.7, 0.9) a

2000q3– 2005q3

2.3 (1.9, 2.7)

2005q3– 2008q4

20.3 (21.0, 0.5)

2008q4– 2010q2

3.4a (1.3, 5.5)

15– 24

2000q3– 2009q1

0.8a (0.6, 1.0)

2009q1– 2010q2

22.4 (25.4, 0.8)

25– 44

2000q3– 2010q2

1.0a (0.9, 1.1)

0.8a (0.7, 0.9) 1.6a (1.1, 2.0)

Age (years)

2000q3– 2005q4 2005q4– 2006q3

65þ

2.0 (1.6, 2.3)

1.0a (0.9, 1.1) 1.2a (0.2, 2.2)

24.4 (215.6, 8.3)

2006q3– 2010q2

1.3a (0.8, 1.9)

2000q3– 2007q4

0.9a (0.8, 1.1)

2007q4– 2010q2

20.2 (20.8, 0.4)

2000q3– 2005q4

1.8a (1.5, 2.1)

0.7a (0.5, 0.8)

Region Metropolitan

2005q4– 2006q3 Regional

1.2a (0.5, 1.9)

22.4 (210.7, 6.6)

2006q3– 2010q2

1.1a (0.8, 1.5)

2000q3– 2010q2

0.5a (0.5, 0.6)

0.5a (0.5, 0.6)

2000q3– 2005q1

1.4a (1.1 –1.7)

1.0a (0.8, 1.2)

All Victoria a

2005q1– 2010q2

0.6 (0.4, 0.8)

Male

2000q1– 2007q3

20.1 (20.4, 0.1)

20.1 (20.4, 0.1)

Female

2000q1– 2007q3

0.1 (20.2, 0.4)

0.1 (20.2, 0.4)

2000q1– 2007q3

21.6a (22.5, 20.7)

21.6a (22.5, 20.7)

Alcohol-attributable deaths Sex

Age (years) 15– 24

a

25– 44

2000q1– 2007q3

20.8 (21.3, 20.3)

20.8a (21.3, 20.3)

45– 64

2000q1– 2007q3

0.0 (20.4, 0.4)

0.0 (20.4, 0.4)

65þ

2000q1– 2007q3

0.1 (20.3, 0.5)

0.1 (20.3, 0.5)

Region Metropolitan Regional

2000q1– 2007q3

0.1 (20.2, 0.3)

0.1 (20.2, 0.3)

2000q1– 2007q3

20.2 (20.5, 0.1)

20.2 (20.5, 0.1)

2000q1– 2007q3

0.0 (20.2, 0.2)

0.0 (20.2, 0.2)

All Victoria

CI, confidence interval. Based on alcohol-attributable hospital admissions per 10 000 residents in Victoria per quarter of a year and alcohol-attributable deaths per 100 000 residents in Victoria per quarter of a year. a

Indicates that the QPC or AQPC is significantly different from zero at a ¼ 0.05.

Discussion In the present study, we conducted time-series analyses of routinely collected population data on AAMorb and AAMort for

Victoria, Australia. We modelled quarterly rates using joinpoint regression in order to quantify temporal trends and to identify sudden changes in log-linear trends while also comparing differences in trends between subgroups of the population.


45– 64

a

0.4a (0.0, 0.8)


A

B

22 20 18 16 14 12 10 8 6 4 2 0

Rate per 10000

Rate per 10000

(o) Victoria

Females

00 q1 02 q1 04 q1 06 q1 08 q1 10 q1 01 q1 03 q1 05 q1 07 q1 09 q1 Quarterly period (2000q3–2010q2)

(+) (•)

25-44 years 65+ years

22 20 18 16 14 12 10 8 6 4 2 0 00 q1 02 q1 04 q1 06 q1 08 q1 10 q1 01 q1 03 q1 05 q1 07 q1 09 q1 Quarterly period (2000q3 –2010q2)

Rate per 10000

15-24 years (Δ) 45-64 years

(o)

Regional

(+)

Metropolitan

22 20 18 16 14 12 10 8 6 4 2 0 00 q1 02 q1 04 q1 06 q1 08 q1 10 q1 01 q1 03 q1 05 q1 07 q1 09 q1 Quarterly period (2000q3 –2010q2)

Fig. 1 Quarterly rates ( per 10 000 residents) and fitted joinpoint regression lines of trends for alcohol-attributable hospital admissions by (A) all of Victoria, (B) sex, (C) age group and (D) Victorian region, 2000/01 –2009/10.

Main findings of this study

AAMorb has consistently and significantly increased for Victoria between 2000/01 and 2009/10. Although this increase was manifested in the increasing trends seen in subpopulations of all age, sex and region categories, females, age groups 25 –44 and 45– 64 and people living in metropolitan Victoria were identified as broad high-risk groups in terms of their markedly increasing AAMorb. Specifically, the increasing metropolitan trends have been driven mainly by significant increases for metropolitan females and age group 25 –64 in comparison with their regional counterparts. Metropolitan males as well as those aged 65þ in metropolitan Victoria also experienced significantly larger increases in morbidity than their regional counterparts. Our analysis also identified the period from the early 2005 to late 2006 as one where morbidity rates affecting metropolitan females temporarily declined only to increase subsequently. The evidence that significant differences existed in trends between subgroups of the

population by region highlights the relative roles of underlying determinants that include divergent alcohol consumption patterns, access to healthcare facilities, the actual usage of treatment services, socioeconomic factors, state of overall health and perhaps the existence of co-morbidities. Of further interest are the comparable trends between those aged 15– 24 and 65þ—both of these groups experiencing declining morbidity in recent times. In contrast, AAMort trends between 2000 and 07 have remained stable for Victoria and for most subgroups. However, significant and steady declines in mortality were experienced by those aged 15– 24 and 25 – 44, primarily seen amongst those living in metropolitan Victoria, and by regional males aged 45þ. Multivariate analysis also showed varying trends for metropolitan males, with a significant increase in mortality for metropolitan males aged 45þ and a consistent decline after 2001q4 for metropolitan males aged 15– 44. Metropolitan males 45þ are to be recognized as another


D (o)

Rate per 10000

Males

(+)

22 20 18 16 14 12 10 8 6 4 2 0

00 q1 02 q1 04 q1 06 q1 08 q1 10 q1 01 q1 03 q1 05 q1 07 q1 09 q1 Quarterly period (2000q3–2010q2)

C

(o)

403

404


Table 2 Joinpoint comparison of metropolitan and regional QPCs and estimated AQPCs for alcohol-attributable hospital admissions (2000/01 –2009/10) and deaths (2000 –07) by age group and sex for region Metropolitan Segment

Regional QPC (95% CI)

AQPC (95% CI)

Segment

QPC (95% CI)

AQPC (95% CI)

Alcohol-attributable hospital admissions Sex Male

2000q3– 2002q1

2.7a (0.9, 4.5)

2002q1– 2010q2

0.7a (0.6, 0.9)

1.0a (0.8, 1.3)

2000q3 – 2010q2

0.5a (0.3, 0.6)

0.5a (0.3, 0.6)

1.7a (0.5, 2.9)

2000q3 – 2010q2

0.7a (0.6, 0.8)

0.7a (0.6, 0.8)

Female 2000q3– 2005q4 2005q4– 2006q3 2006q3– 2010q2

2.9a (2.5, 3.4) 25.9 (219.5, 10) 1.5a (0.9, 2.2)

Age (years) 2000q3– 2008q4

0.9a (0.7, 1.1)

2008q4– 2010q2

22.2 (24.9, 0.5)

0.4 (20.1, 0.9)

2000q3 – 2001q1

13.9 (27.5, 40.1)

2001q1 – 2002q3

21.6 (25.8, 2.8)

2002q3 – 2010q2

0.7a (0.5, 1.0)

1.0 (20.2, 2.2)

25– 44 2000q3– 2010q2

1.2a (1.0, 1.3)

1.2a (1.0, 1.3)

2000q3 – 2010q2

0.5a (0.3, 0.6)

0.5a (0.3, 0.6)

2.4a (1.9, 2.8)

1.4a (0.3, 2.5)

2000q3 – 2010q2

0.7a (0.5, 0.8)

0.7a (0.5, 0.8)

0.8a (0.6, 1.1)

2000q3 – 2010q2

0.1a (0.0, 0.2)

0.1a (0.0, 0.2)

1.2a (0.5, 1.9)

2000q3 – 2010q2

0.5a (0.5, 0.6)

0.5a (0.5, 0.6)

45– 64 2000q3– 2005q4 2005q4– 2006q3 2006q3– 2010q2

26.0 (218.7, 8.7) 1.6a (1.0, 2.2)

65þ 2000q3– 2007q4

1.2a (1.1, 1.4)

2007q4– 2010q2

20.4 (21.1, 0.4)

Region 2000q3– 2005q4 2005q4– 2006q3 2006q3– 2010q2

1.8a (1.5, 2.1) 22.4 (210.7, 6.6) 1.1a (0.8, 1.5)

Alcohol-attributable deaths Sex Male 2000q1– 2007q3

0.1 (20.3, 0.4)

0.1 (20.3, 0.4)

2000q1 – 2007q3

20.4a (20.8, 20.1)

20.4a (20.8, 20.1)

2000q1– 2007q3

0.1 (20.2, 0.4)

0.1 (20.2, 0.4)

2000q1 – 2007q3

0.3 (20.4, 1.0)

0.3 (20.4, 1.0)

2000q1– 2007q3

22.0a (22.9, 21.0)

22.0a (22.9, 21.0)

2000q1 – 2007q3

20.5 (21.8, 0.9)

20.5 (21.8, 0.9)

2000q1– 2007q3

21.0a (21.7, 20.3)

21.0a (21.7, 20.3)

2000q1 – 2007q3

20.5 (21.4, 0.4)

20.5 (21.4, 0.4)

2000q1– 2007q3

0.3 (20.2, 0.9)

0.3 (20.2, 0.9)

2000q1 – 2007q3

20.7 (21.4, 0.1)

20.7 (21.4, 0.1)

2000q1– 2007q3

0.3 (20.2, 0.7)

0.3 (20.2, 0.7)

2000q1 – 2007q3

20.2 (20.7, 0.2)

20.2 (20.7, 0.2)

2000q1– 2007q3

0.1 (20.2, 0.3)

0.1 (20.2, 0.3)

2000q1 – 2007q3

20.2 (20.5, 0.1)

20.2 (20.5, 0.1)

Female Age (years) 15– 24 25– 44 45– 64 65þ Region

CI, confidence interval. Based on alcohol-attributable hospital admissions per 10 000 residents in Victoria per quarter of a year and alcohol-attributable deaths per 100 000 residents in Victoria per quarter of a year. a

Indicates that the QPC or AQPC is significantly different from zero at a ¼ 0.05.


15– 24


A

B

Rate per 100000

Rate per 100000

(o) Victoria

16 14 12 10 8 6 4 2 0

(o)

Males

(+)

Females

16 14 12 10 8 6 4 2 0

00 q1

02 q1 01 q1

04 q1 03 q1

06 q1 05 q1

00 q1 07 q1

02 q1 01 q1

Quarterly period (2000q1–2007q3)

(o)

(+)

25-44 years

(•)

65+ years

16 14 12 10 8 6 4 2 0 00 q1

02 q1 01 q1

04 q1 03 q1

06 q1 05 q1

06 q1 05 q1

07 q1


D

Rate per 100000

15-24 years (Δ) 45-64 years

04 q1 03 q1

(o)

Regional

(+)

Metropolitan

16 14 12 10 8 6 4 2 0 00 q1

07 q1


02 q1 01 q1

04 q1 03 q1

06 q1 05 q1

07 q1


Fig. 2 Quarterly rates ( per 100 000 residents) and fitted joinpoint regression lines of trends for alcohol-attributable deaths by (A) all of Victoria, (B) sex, (C) age group and (D) Victorian region, 2000– 07.

special high-risk population for alcohol-attributable harm considering their significantly rising trends for both AAMorb and AAMort. What is already known on this topic

Our findings of increasing morbidity and stable mortality in the backdrop of a supposedly stable consumption pattern add a new paradigm to theories underlining interactions between consumption and harm. According to Norstro¨m and Ramstedt,15 when alcohol consumption increased by 30% between 1996 and 2004 in Sweden the mortality attributed to alcohol remained fairly stable (during 1995 –2003) and morbidity increased. Romelsjo16 believes that this could not be explained solely in terms of an increase in moderate alcohol consumption and improvements in health care, but one of rising high consumers and a corresponding rise in alcoholrelated problems. In contrast, alcohol-related mortality in Finland increased during the economic boom of the late 1980s with a concomitant rise in the total alcohol

consumption and decreased during the subsequent recession during the early 1990s with a corresponding decrease in consumption.17 Males and females reached AAMort peaks at different times during the period between 1987 and 2003, and among both males and females, mortality rose markedly between 1987 and 2003 for older age groups while it decreased for younger age groups.17 This previously published data compared trends using a simplistic comparison of percentage changes in rates over time with directional changes and subgroup differences detected by eyeballing. In the present study, any discussion comparing trends in morbidity and mortality with per capita consumption is handicapped by the absence of available per capita data for Victoria. Policies and programmes for the prevention of alcoholrelated harm have identified measures such as regulating physical availability, taxation and pricing, drink-driving countermeasures, and treatment and early intervention as key strategies that have the most potential to be effective.7 In this context, it is pertinent to assume that greater availability and


C

Rate per 100000

405

406


Limitations of this study

Constraints associated with measuring trends of alcohol-attributable harms and alcohol consumption are well documented20 and these findings are to be interpreted in the context of these limitations. First, there are various limitations attributed to secondary data of this nature, especially relating to accuracy, completeness and the differences in coding and classification practices that may have taken place overtime. Secondly, when aetiological fractions are used to derive AAMorb and AAMort, only certain proportions of admissions and deaths which may have otherwise been wholly attributed to alcohol are extracted for analysis. We conducted sensitivity analyses comparing primary diagnosis to aetiological fractions (not presented), which revealed no discernible effect on the trends. When primary diagnoses (without aetiological fractions) were used for overall Victorian morbidity, the pairwise comparison (against aetiological fractions) failed to reject parallelism against trends (P ¼ 0.76) indicating the existence of parallel trends. Similar results were obtained for overall Victorian mortality data (P ¼ 0.89), and for geographical subgroups (not reported). It is possible though for cause-specific trends of AAMorb and AAMort to differ from what is described here. Thirdly, mortality is known to be a less precise measure of alcohol-related harm than morbidity as only a proportion of those who develop liver cirrhosis and other alcohol-related disease actually die from them, and also since mortality data are influenced by changes in treatment.16 However, by using two direct measures of alcohol-attributable harm and by avoiding more complex

surrogate measures that are employed to explore alcohol involvement in harms such as assaults or serious road injuries, we have strived to minimize problems related to the validity of measures. Finally, joinpoint regression can potentially give rise to arbitrary slopes for populations with large random variation but these are denoted by wide confidence intervals.21 We have interpreted our results with caution and have avoided any discussion on such slopes an example being the rapid rise in morbidity for those aged 15–24 in regional Victoria by 13.9% per quarter (95% CI, 27.5 to 40.1) between 2000q3 and 2001q1 in Table 2. Although not a limitation in the present analysis, the non-inclusion of community-based data may have altered the pattern and the level of AAMorb especially in settings where easy access to healthcare may be limited. What this study adds

Our study has identified previously overlooked subgroups of the population as being at greater (and increasing) risk for alcohol-attributable chronic harm in terms of their increasing trends. They include metropolitan populations overall, and females, those aged 25 – 64 and metropolitan males aged 45þ in particular. Our findings inform policies and public health strategies which aim to reduce such harm and suggest such policies and strategies need to target these high-risk populations as well as the more traditional risk groups in future.

Supplementary data Supplementary data are available at the Journal of Public Health online.

Funding M.L. is supported by an Australian National Health and Medical Research Council Early Career Fellowship (#1053029). The Centre for Alcohol Policy Research is funded by the Foundation for Alcohol Research and Education, an independent, charitable organisation working to prevent the harmful use of alcohol in Australia http:// www.fare.org.au. No funding statements to declare for H.J., J.F., B.L., S.M.

References 1 Ju¨rgen R, Colin M, Svetlana P et al. Global burden of disease and injury and economic cost attributable to alcohol use and alcohol-use disorders. Lancet 2009;373:2223 – 33. 2 English D, Holman DA. The quantification of drug caused morbidity and mortality in Australia. Canberra, Australia: Australian Government Publishing Service, 1995.


perhaps increased affordability of alcohol amongst the Victorian population in the preceding 10-year period may have caused an explosion of related harms which are more pronounced in hitherto unrecognized subgroups for such risks whilst hiding an obvious increment in average consumption. Furthermore, with the emergence of high-risk populations previously thought to be at lower risk of harm from alcohol, we may have provided the clearest evidence so far of the societal change speculated by Livingston et al. 4 in terms of the relationship between alcohol consumption and its harms. On the other hand, Chikritzhs et al.,18 in their evaluation of a major alcohol policy in Australia using mortality, have shown the importance of distinguishing between acute and chronic alcohol harms, and that chronic alcohol harm respondents to policy changes on the long term. Similarly, Stockwell et al.19 have shown that chronic alcohol-attributable hospital admissions in Canada had decreased after 2 years of increasing price of alcohol. Therefore, the evidence is there for the potential to reduce AAMorb and AAMort in developed countries using the right strategies and selecting the correct target groups.


407

3 Collins D, Lapsley H. The costs of tobacco, alcohol and illicit drug abuse to Australian Society in 2004/05. Canberra, Australia: National Drug Strategy, 2008.

12 Kim HJ, Fay MP, Feuer EJ et al. Permutation tests for joinpoint regression with applications to cancer rates. Stat Med 2000;19: 335 – 51.

4 Livingston M, Matthews S, Barratt MJ et al. Diverging trends in alcohol consumption and alcohol-related harm in Victoria. Aust NZ J Public Health 2010;34:368 – 73.

13 Hyune-Ju K, Fay MP, Binbing Y et al. Comparability of segmented line regression models. Biometrics 2004;60:1005– 14.

5 Pascal R, Chikritzhs T, Jones P. Trends in estimated alcohol attributable deaths and hospitalisations in Australia, 1996 – 2005. National Alcohol Indicators, Bulletin No. 12. Perth, Australia: National Drug Research Institute, Curtin University of Technology, 2009. 6 Australian Institute of Health and Welfare. 2004 national drug strategy household survey: first results (AIHW CAT. No. PHE 57). Canberra, Australia: AIHW, 2005. 7 Ministerial Taskforce on Alcohol and Public Safety. Victoria’s alcohol action plan—2008– 2013. Melbourne, Australia: State Government of Victoria, 2008.

9 Ridolfo B, Stevenson C. The quantification of drug-caused mortality and morbidity in Australia, 1998. Canberra, Australia: Australian Institute of Health and Welfare, 2001.

15 Norstro¨m T, Ramstedt M. Sweden—is alcohol becoming an ordinary commodity? Addiction 2006;101:1543– 5. 16 Romelsjo A. Can an improvement and increase in treatment explain unexpected trends in alcohol-related harm in Sweden? Addiction 2007;102:669– 71. 17 Herttua K, Ma¨kela¨ P, Martikainen P. Differential trends in alcoholrelated mortality: a register-based follow-up study in Finland in 1987 – 2003. Alcohol Alcohol 2007;42:456– 64. 18 Chikritzhs T, Stockwell T, Pascal R. The impact of the northern territory’s living with alcohol program, 1992 – 2002: revisiting the evaluation. Addiction 2005;100:1625– 36. 19 Stockwell T, Zhao J, Martin G et al. Minimum alcohol prices and outlet densities in British Columbia, Canada: estimated impacts on alcohol-attributable hospital admissions. Am J Public Health 2013. doi:10.2105/AJPH.2013.301289.

10 Dobson AJ, Kuulasmaa K, Eberle E et al. Confidence-intervals for weighted sums of Poisson parameters. Stat Med 1991;10:457– 62.

20 Department of Mental Health and Substance Dependence. International guide for monitoring alcohol consumption and related harm. Geneva, Switzerland: World Health Organization, 2000.

11 Statistical Research and Applications Branch. Joinpoint Regression Program (Ver. 3.5.2). Bethesda, Maryland, USA: National Cancer Institute, 2011.

21 Baade PD, Coory MD, Aitken JF. International trends in prostatecancer mortality: the decrease is continuing and spreading. Cancer Causes Control 2004;15:237– 41.


8 Australian National Centre for Classification in Health. The International Statistical Classification of Diseases and Related Health Problems, tenth revision, Australian Modification (ICD-10-AM). Lidcombe, New South Wales, 2010.

14 StataCorp LP. Stata Statistical Software: Release 11. College Station, TX: StataCorp LP, 2009.