Hospitalisation for an alcohol-related cause among injecting drug users in Scotland: Increased risk following diagnosis with hepatitis C infection

International Journal of Drug Policy 22 (2011) 63–69

Contents lists available at ScienceDirect

International Journal of Drug Policy journal homepage: www.elsevier.com/locate/drugpo

Research paper

Hospitalisation for an alcohol-related cause among injecting drug users in Scotland: Increased risk following diagnosis with hepatitis C infection Scott A. McDonald a,∗ , Sharon J. Hutchinson a,b , Sheila M. Bird b,c , Chris Robertson a,b , Peter R. Mills d , John F. Dillon e , David J. Goldberg a a

Health Protection Scotland, Clifton House, Clifton Place, Glasgow G3 7LN, Scotland, UK Department of Mathematics and Statistics, University of Strathclyde, 26 Richmond Street, Glasgow G1 1XH, Scotland, UK MRC Biostatistics Unit, Institute of Public Health, Cambridge CB2 0SR, Scotland, UK d Gartnavel General Hospital, Glasgow G12 0YN, Scotland, UK e Ninewells Hospital & Medical School, Dundee DD 9SY, Scotland, UK b c

a r t i c l e

i n f o

Article history: Received 3 November 2009 Received in revised form 3 April 2010 Accepted 20 April 2010

Keywords: Alcohol Hospital admissions Injecting drug users Hepatitis C virus

a b s t r a c t Background: The rate of hepatitis C (HCV) related liver disease progression is known to be strongly associated with alcohol consumption, yet there are very few data on alcohol use in injecting drug users (IDUs), who represent 90% of Scotland’s HCV-diagnosed population. To investigate the extent of alcohol use in IDUs, we used hospitalisation with an alcohol-related diagnosis as an indicator for problematic consumption levels, and compared admission rates pre- and post-HCV diagnosis. Methods: Data for 41,062 current/former IDUs attending drug treatment/support services in Scotland from April 1995 to March 2006 were linked to the national hospital discharge database to retrieve alcoholrelated episodes, and to the national HCV Diagnosis database to determine HCV-diagnosed status. Relative risks were estimated using Cox proportional hazards regression for recurrent events. Results: The proportion of IDUs with ≥1 alcohol-related admission following first attendance at drug services was greater among those diagnosed with HCV by the end of follow-up (16%) compared with those who were not (6%). For the 9145 IDUs who had been diagnosed with HCV by 31 March 2006, there was a 1.5-fold increased relative risk of an alcohol-related admission >30 days post-HCV diagnosis (95% CI: 1.2–1.7) compared with >30 days pre-HCV diagnosis, adjusted for sex, age, and deprivation. Conclusions: IDUs diagnosed with HCV infection have an increased risk of subsequent hospital admission for an alcohol-related cause. Because of the synergistic effect of HCV infection and excessive alcohol intake on the development of cirrhosis, it is imperative that alcohol intake is addressed in the management of chronic HCV infection in this population. © 2010 Elsevier B.V. All rights reserved.

Introduction Injecting drug use is the most important risk behaviour for infection with the hepatitis C virus (HCV) in resource-rich countries. The sharing of injecting equipment among drug users is a known pathway for transmission, and high prevalences of HCV among injecting drug user (IDU) populations have been observed (Hagan & Des Jarlais, 2000; Roy et al., 2002). In Scotland, 44% of IDUs are estimated to be infected with HCV (Roy et al., 2007), and of the 50,000 persons infected with HCV, approximately 90% are current or former IDUs (Hutchinson et al., 2006). The rate of HCV-related liver disease progression is known to be strongly associated with alcohol consumption (Freeman, Law,

∗ Corresponding author. Fax: +44 141 300 1170. E-mail address: [email protected] (S.A. McDonald). 0955-3959/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.drugpo.2010.04.003

Kaldor, & Dore, 2003; Hutchinson, Bird, & Goldberg, 2005a; Monto et al., 2004; Ostapowicz, Watson, Locarnini, & Desmond, 1998), yet there are very few data on levels of alcohol use in the HCV-infected IDU population (Campbell et al., 2006). To assess the potential impact of alcohol on the future course of the HCV epidemic, it is important to gather information about consumption in conjunction with other epidemiological risk factors in HCV-infected populations; these data are required by projection models to estimate the future burden of HCV-related liver disease (Hutchinson, Bird, & Goldberg, 2005b). Our study therefore extends previous research such as NTORS, (Gossop, Marsden, Stewart, & Kidd, 2003; Gossop, Marsden, Stewart, & Rolfe, 2000) which examined drinking behaviour (among other outcomes, but not including HCV diagnosed status) in 753 drug misusers using a prospective design. Previous analysis of Scottish Health Survey data on 23,183 individuals demonstrated a strong association between self-reported alcohol consumption and hospitalisation for an alcohol-related

64

S.A. McDonald et al. / International Journal of Drug Policy 22 (2011) 63–69

cause (McDonald, Hutchinson et al., 2009); the relative risk of a first-time alcohol-related admission following interview ranged from 2.6 (95% CI: 1.3–5.3) to 19.0 (95% CI: 9.9–36.5) for male respondents who reported drinking 1–7 units and 50+ units/week, respectively. Employing record-linkage methodology and using data on inpatient hospital admissions with an alcohol-related diagnosis, we examined the extent of and factors associated with problem alcohol use (i.e., leading to hospitalisation) in Scotland’s current/former IDU population. Because of the potential for alcohol use (even in moderation) both to accelerate the progression of HCV-related liver disease (Hutchinson et al., 2005a; Ostapowicz et al., 1998), and to reduce the effectiveness of antiviral therapy, chronically infected individuals should be advised to reduce their alcohol intake (Scottish Intercollegiate Guidelines Network (SIGN), 2006). It is however not known whether this guideline has had any impact in terms of minimising consumption in the diagnosed HCV infected population, or whether the HCV diagnosis (with or without advice on alcohol use) itself has an impact on alcohol use. Therefore, our goal was to compare the risk of hospitalisation for an alcohol-related cause pre- and post-diagnosis with HCV infection. Although the SIGN guideline was published in December 2006, after our data collection period, similar guidelines pre-dated the SIGN guideline as early as 1999 (Consensus Panel, 1999; Department of Health, 2001, 2002). Hence, we also compared the risk of alcohol-related hospitalisation pre- and post-HCV diagnosis, but restricted analysis to the post-1999 portion of our study period. Methods

Health Protection Scotland maintains a database of all persons who have been diagnosed HCV positive in Scotland since testing commenced in 1991 (Shaw et al., 2003); detection of HCV antibody is a requirement for inclusion. This database contains the following non-named information: surname Soundex code, forename initial, date of birth, sex, and the postcode district of residence, as well as data concerning risk activities and the date of the earliest positive specimen. The database contained records for 20,588 persons as of 31 March 2006. Linkage procedure Approval by NHS National Services Scotland Privacy Advisory Committee was obtained to link data from the above four national sources. Linkage of records between the SDMD, the HCV diagnosis, and the SMR01 and GROS data sources was carried out by ISD using probabilistic record-linkage techniques (Kendrick & Clarke, 1993) to match individuals on the SDMD with those on the HCV diagnosis database and the previously created linked dataset of the SMR01 and the national death registry. A preliminary step using exact (deterministic) matches identified attendances within the SDMD associated with the same individual. Then, ISD’s probabilistic method involved calculating a score for each SDMD attendance record as a potential match to each HCV diagnosis record and each SMR01/GROS record; an individual on the SDMD was successfully linked if the score for their top ranked matching record exceeded a predetermined threshold value. ISD reported that 78.4% of the records (episodes) in the SDMD successfully linked to the SMR01/GROS. The linked dataset was anonymised before transfer to Health Protection Scotland for analysis.

Design Data analysis The design was a retrospective cohort study, using recordlinkage between four national databases to determine alcoholrelated hospitalisation rates and HCV-diagnosed status in a large cohort of current/former IDUs. Study population and data sources The study population consisted of current/former IDUs in contact with drug treatment and support services, including general practitioners, hospitals, specialist drug clinics, and non-statutory agencies, and reported to the Scottish Drug Misuse Database (SDMD) held by Information Services Division (Information Services Division (ISD), 2007). These agencies report information on new contacts (defined as first presentation or repeat presentation if it has been at least 6 months since last attendance) to the SDMD. IDU status was defined according to self-report: if at any attendance at drug services the client reported having either injected drugs in the past month or having ever injected, they were classified as a current/former IDU. The SDMD contains limited identifying information: sex, date of birth, forename initial, first and fourth letter of surname, and postcode sector of residence, as well as behavioural data such as age first injected drugs. Data were available for 41,062 unique IDUs who attended drug services between 1 April 1995 and 31 March 2006. The Scottish Morbidity Records (SMR01), also held by ISD, are an episode-based patient record of all acute inpatient and day case hospital discharges from non-obstetric, non-psychiatric specialities; the period of data used was 1 January 1981 to 31 March 2006. Discharge diagnosis is coded according to the World Health Organisation’s International Classification of Diseases (ICD) Ninth Revision for discharges before 1996, and Tenth Revision for discharges between 1996 and 2006. ISD routinely combine the SMR01 data with death registrations held by the General Register Office for Scotland (GROS) to form a linked dataset.

Outcome measures For each IDU, the occurrence of one or more alcohol-related hospital episodes was coded using an indicator variable. This involved searching the linked hospitalisation records for alcohol-related discharge diagnosis (ICD) codes in either the main or a supplementary diagnosis field. The relevant codes comprised liver-related diagnoses: alcoholic liver disease (ICD-10 K70; ICD-9 571.0-571.2), and non-liver-related codes: alcohol use (ICD-10 Z72.1), mental and behavioural disorders due to use of alcohol (ICD-10 F10; ICD-9: 291, 303, 305), degeneration of nervous system due to alcohol (ICD-10 G31.2, G62.1, G72.1, I42.6, K29.2; ICD-9 357.5, 425.5, 535.3), toxic effects of alcohol (ICD-10 T51.0, T51.9; ICD-9 980.0), alcohol-induced chronic pancreatitis (ICD-10 K86.0), evidence of alcohol involvement (ICD-10 Y90-1), finding of alcohol in blood (ICD-10 R78.0; ICD-9 790.3), alcohol rehabilitation (ICD-10 Z50.2), and accidental or intentional self-poisoning by and exposure to alcohol (ICD-10 X45, X65; ICD-9 E860.0, E860.9). Some analyses were conducted for liverrelated and non-liver-related outcomes separately. Epidemiological risk factors Additional risk factors coded for each IDU were sex, age, HCVdiagnosed status, and social deprivation score (Carstairs 2001 quintile) (Carstairs & Morris, 1990), derived from the postcode sector of residence provided at the most recent attendance at drug services. Age when first injected was self-reported. However, this variable was not necessarily consistently reported across episodes; consequently both the across-episode mean and the earliest reported age first injected were derived. The mean reported age first injected is used in the analyses below; results did not change appreciably if earliest reported age first injected was substituted. Date of injecting debut was estimated from the mean

S.A. McDonald et al. / International Journal of Drug Policy 22 (2011) 63–69

reported age first injected. Time-at-risk analyses included a threelevel time-dependent covariate for period around HCV diagnosis (>30 days preceding HCV diagnosis [including those never and ever diagnosed], ±30 days around diagnosis [if ever diagnosed], >30 days after [if ever diagnosed]).

Table 1 Characteristics of all persons in the SDMD (April 1995 to March 2006) who have ever reported injecting drugs, and the number and proportion of those who had been admitted to hospital with an alcohol-related diagnosis subsequent to first attendance at drug services (since April 1995). ≥1 Alcohol-related admission N (column %)

Statistical analysis Cox proportional-hazards regression was used to estimate the relative risk of a first alcohol-related hospital admission associated with sex, current age, a time-dependent covariate for HCV diagnosis, and deprivation quintile among all IDUs. For this analysis, time at risk commenced at the date of the IDU’s first attendance at drug services (i.e., their first episode recorded on the SDMD), and ended at the earliest of the first subsequent alcohol-related admission, date of death, or the right-censoring date (31 March 2006). The relative risk of multiple hospital admissions with mention of alcohol for those IDUs ever diagnosed with HCV infection was also estimated using Cox proportional-hazards regression for recurrent data (Andersen & Gill, 1982); robust standard errors were computed in order to adjust for the non-independence of multiple episodes for the same person. (Twisk, Smidt, & de Vente, 2005) For this analysis, each risk period was defined to begin either at the date of the IDU’s first attendance at drug services or the date of discharge of the previous hospitalisation, and to end at either the admission date of the subsequent hospitalisation, date of death, or the rightcensoring date (31 March 2006). Thus, time at risk excluded stays in hospital. Covariates included sex, current age, HCV diagnosis (timedependent), and deprivation quintile. Missing data were few and were treated by removing the case from the regression analysis. Alcohol-related bed-days were computed as the sum over all alcohol-related stays in hospital. Multiple admission and bed-day rates were estimated under that assumption that admissions followed a negative-binomial distribution (Glynn & Buring, 1996), to overcome known deficiencies in the events per person-years measure (Windeler & Lange, 1995). All statistical analyses were carried out using R version 2.4.0 (R Development Core Team, 2006). Results Characteristics of the study population

65

N (row %)

Total

41,062 (100.0)

3332 (8.1)

Female Male

11,877 (28.9) 29,185 (71.1)

821 (6.9) 2511 (8.6)

Age at 31 March 2006 <20 20–29 30–39 40–49 50+ Dead

469 (1.1) 14,298 (34.8) 18,401 (44.8) 5255 (12.8) 625 (1.5) 2014 (4.9)

18 (3.8) 886 (6.2) 1382 (7.5) 1511 (9.7) 78 (12.5) 457 (22.7)

Sex

Years since injection debut if not known dead, relative to 31 March 2006 <2 1375 (3.5) 49 (4.0) 2–3 3257 (8.4) 137 (4.6) 4–5 4710 (12.5) 300 (6.7) 6–7 6026 (16.1) 357 (6.2) 8–9 5926 (15.8) 434 (7.6) 10+ 16,848 (43.7) 1398 (8.7) No information 2920 Hepatitis C diagnosed status on 31 March 2006 Yes 9145 (22.3) No 31,917 (77.7) Carstairs 2001 deprivation quintile 1 2 3 4 5 No information

1493 (5.6) 2325 (8.8) 4060 (15.3) 5913 (22.3) 12,741 (48.0) 14,530

1499 (16.4) 1833 (5.7) 152 (10.2) 269 (11.6) 483 (11.9) 745 (12.6) 1594 (12.5)

however, the mean age at first admission was slightly lower for females (28.9 years, SD = 7.6) compared with males (30.6, SD = 7.5). Expected alcohol-related admission rates were lower in the periods preceding the 2-month window around the date of HCV diagnosis (e.g., 2–24 months before: 4.1 admissions/100 personyears, 95% CI: 3.6–4.8) compared with after diagnosis (e.g., 2–24 months after: 8.9, 95% CI: 7.3–10.7), with the highest rate (29, 95%

The majority of the IDU study population was male (71%), with a mean age of 26.8 years (SD = 6.9 years) at first attendance at drug services (Table 1). On average, IDUs made their injecting debut at 21.9 years (SD = 5.5); 44% of IDUs had an injecting career spanning 10 years or longer by 31 March 2006. The records for 9145 persons (22% of all IDUs) were linked to the HCV Diagnosis database (i.e., diagnosed HCV antibody-positive). 3332 individuals (8.1% of all IDUs) were hospitalised at least once for an alcohol-related condition between their first attendance at drug services and 31 March 2006 (Table 1). A greater proportion (21.9%, 8998/41,062) of all IDUs had at least one alcohol-related admission, considering the entire hospital discharge database (including the period prior to first attendance at drug services). The proportion of IDUs who had at least once alcohol-related admission subsequent to first attendance at drug services was far greater among those diagnosed with HCV by the end of follow-up (31 March 2006) compared with those who were not (16.4% and 5.7%, respectively). First and multiple alcohol-related hospital admissions The cumulative probability of a first-time alcohol-related admission subsequent to first attendance at drug services increased at roughly the same rate for males and females with age (Fig. 1);

Fig. 1. Cumulative probability of a first-time hospital admission with mention of alcohol subsequent to first attendance at drug services as a function of current age and sex. The 5- and 10-year cumulative probabilities of a first-time alcohol-related admission subsequent to first attendance at drug services for all IDUs were 15% and 24%, respectively.

66

S.A. McDonald et al. / International Journal of Drug Policy 22 (2011) 63–69

Table 2 Alcohol-related admissions subsequent to first attendance at drug services among HCV-diagnosed IDUs, expected admission rates (per 100 person-years, computed under the assumption that admissions were negative-binomially distributed), and expected alcohol-related bed-days in hospital per 100 person-years in relation to HCV diagnosis date, among 9140 HCV-diagnosed IDUs in the SDMD in relation to the date of HCV diagnosis, where time-at-risk begins at the date of first attendance at drug services. Five HCV-diagnosed IDUs were excluded because of obvious record-linkage errors (death date preceded date of first attendance at drug services). HCV diagnosis

>4 years before 3–4 years before 2–24 months before ±1 month 2–24 months after 3–4 years after >4 years after

Person-years

5046 6505 10,281 1049 12,157 11,104 18,010

N

173 222 414 268 917 736 1276

Rate (SE)

3.6 (1.1) 3.5 (1.1) 4.1 (1.1) 29.4 (1.1) 8.9 (1.1) 7.7 (1.1) 8.3 (1.1)

Alcohol-related bed-day rates (SE) Liver

Non-liver

All

0 (–) 0.5 (1.9) 1.2 (1.6) 95.6 (1.4) 47.0 (1.3) 27.2 (1.3) 33.8 (1.3)

11.4 (1.2) 11.7 (1.3) 13.0 (1.1) 262.6 (1.3) 31.2 (1.1) 30.4 (1.2) 223.8 (2.4)

11.4 (1.2) 12.3 (1.3) 14.6 (1.1) 331.4 (1.3) 57.8 (1.3) 40.4 (1.3) 223.8 (2.4)

Note: N = total admissions. Rates are estimated admissions/bed-days per 100 person-years of follow-up, assuming counts are negative-binomially distributed (Glynn & Buring, 1996). SE = standard error of rate. Liver refers to admissions with mention of alcoholic liver disease (ICD-10 K70; ICD-9 571.0-3); Non-liver to admissions with an alcohol-related code other than ICD-10 K70 or ICD-9 571.0-3.

CI: 23–38) observed in the period spanning 1 month before and 1 month after the date of HCV diagnosis (Table 2). The mean length of hospital stay for an alcohol-related condition was 4.2 days (SD = 6.3). The rate of bed-days for non-liver-related, as well as liver-related, alcohol admissions was higher in the period after HCV diagnosis, compared with the period before. Relative risk of alcohol-related hospitalisation Cox regression analysis indicated an increased relative risk of a first-time alcohol-related admission for males (HR = 1.3, 95% CI: 1.2–1.9), for IDUs aged < 20 years (HR = 1.3, 95% CI: 1.1–1.5) and 40+ years (HR = 1.7, 95% CI: 1.5–1.9), compared with 20–29 years,

and in the 60 days proximal to HCV diagnosis (±30 days) (HR = 5.8, 95% CI: 4.9–6.9) compared with >30 days pre-HCV diagnosis date, adjusting also for deprivation (Table 3a). Restricting to the first admission with a non-liver alcohol-related diagnosis resulted in similar hazard ratios (Table 3b). In the subpopulation of IDUs that were diagnosed with HCV infection by 31 March 2006, Cox proportional hazards regression analysis for recurrent events indicated an elevated relative risk for males (Table 4a), with an adjusted hazard ratio of 1.6 (95% CI: 1.3–1.9). The relative risk of alcohol-related admission increased with current age of 30 years and over, considering 20–29 as the reference category (30–39 years: HR = 1.7, 95% CI: 1.4–1.9; 40+ years: HR = 2.6, 95% CI: 2.1–3.3). Deprivation score was associated with a

Table 3 Relative risk of a first-time admission with an alcohol-related diagnosis among all IDUs (N = 41,062), where time-at-risk begins at the date of first attendance at drug services (as recorded on the SDMD) and ends at the earliest of date of first subsequent alcohol-related admission, (ii) date of death, or (iii) 31 March 2006. Factor

Level

(a) First-time admission for any alcohol-related condition Sex Female Male

N

Rate

Hazard ratio

95% CI

P

1.23–1.44

<0.00001

821 2511

2.4 3.2

Ref. 1.33

<20 20–29 30–39 40+

182 1505 1216 429

3.9 2.7 2.8 4.7

1.32 Ref. 1.05 1.73

1.12–1.54

0.0007

0.97–1.13 1.54–1.92

0.24 <0.00001

Pre ±30 days Post

2348 131 853

2.9 16.5 2.8

Ref. 5.81 0.99

4.87–6.94 0.92–1.08

<0.00001 0.89

1 152 2.8 2 269 3.2 3 483 3.2 4 745 3.3 5 1594 2.7 (b) First-time admission with mention of a non-liver-related alcohol-related condition Sex Female 908 2.4 Male 2457 3.2

Ref. 1.14 1.16 1.16 0.95

0.94–1.39 0.96–1.39 0.97–1.38 0.81–1.12

Ref. 1.32

1.22–1.44

<0.00001

Current age

HCV diagnosis period

Deprivation quintile

Current age

HCV diagnosis period

Deprivation quintile

0.19 0.12 0.10 0.56

<20 20–29 30–39 40+

182 1500 1187 397

3.9 2.7 2.8 4.3

1.32 Ref. 1.03 1.61

1.12–1.54

0.0008

0.95–1.12 1.44–1.81

0.44 <0.00001

Pre ±30 days Post

2314 136 816

2.9 9.3 2.7

Ref. 4.72 0.98

3.89–5.73 0.90–1.06

<0.00001 0.56

1 2 3 4 5

148 266 476 729 1558

2.8 3.1 2.2 3.2 2.6

Ref. 1.16 1.17 1.26 0.96

0.95–1.42 0.97–1.41 0.98–1.39 0.81–1.13

0.15 0.094 0.094 0.62

Note: N = number of first-time alcohol-related admissions; Rate = number of first-time alcohol-related admissions per 100 person-years of follow-up; CI = confidence interval. HCV diagnosis period is a time-dependent variable, coded as ‘Pre’ for the period before HCV diagnosis date minus 30 days, ‘±30 days’ for the 30 days before and after HCV diagnosis, and ‘Post’ for the period subsequent to HCV diagnosis date plus 30 days.

S.A. McDonald et al. / International Journal of Drug Policy 22 (2011) 63–69

67

Table 4 Rate of hospitalisation with (a) any alcohol-related diagnosis or (b) a non-liver-related alcohol-related diagnosis among all HCV-diagnosed IDUs (N = 9140), where time-at-risk begins at the date of first attendance at drug services, with results of Cox proportional-hazards regression analysis for recurrent events (P values and 95% CIs are based on robust standard errors). Factor

Level

N

(a) Hospitalisation with mention of any alcohol-related condition Sex Female 870 Male 3136 Current age

HCV diagnosis period

Deprivation quintile

95% CI

P

4.5 8.3

Ref. 1.56

1.31–1.85

<0.00001

67 1028 2002 909

3.6 4.0 7.6 14.2

1.10 Ref. 1.66 2.62

1.42–1.94 2.08–3.29

<0.00001 <0.00001

Pre ±30 days Post

809 268 2929

3.9 29.9 9.3

Ref. 6.28 1.45

5.18–7.61 1.24–1.69

<0.00001 <0.00001

9.4 13.2 12.0 12.6 7.8 22.6

Ref. 1.16 1.12 1.46 1.21 1.38

0.80–1.68 0.79–1.58 1.05–2.03 0.89–1.64 0.93–2.07

3.9 6.8

Ref. 1.49

1.27–1.76

1 132 2 235 3 417 4 873 5 2250 Not known 99 (b) Hospitalisation with mention of a non-liver-related alcohol-related condition Sex Female 766 Male 2603

HCV diagnosis period

Hazard ratio

<20 20–29 30–39 40+

Deprivation quintile

Current age

Rate

0.80–1.52

0.78–1.48

0.55

0.43 0.52 0.023 0.23 0.11

<0.00001

<20 20–29 30–39 40+

67 982 1672 648

3.6 3.9 6.4 9.8

1.07 Ref. 1.51 2.07

0.66

1.30–1.76 1.66–2.58

<0.00001 <0.00001

Pre ±30 days Post

788 230 2351

3.8 24.9 6.7

Ref. 5.75 1.30

4.70–7.03 1.12–1.52

<0.00001 0.0007

1 2 3 4 5 Not known

110 219 359 728 1855 98

7.3 11.5 10.6 10.1 6.2 22.5

Ref. 1.29 1.15 1.45 1.17 1.63

0.90–1.85 0.83–1.59 1.06–1.99 0.88–1.57 1.02–2.41

0.17 0.38 0.020 0.28 0.014

Note: N = number of alcohol-related admissions; Rate = expected admissions per 100 person-years of follow-up, assuming counts are negative-binomially distributed (Glynn & Buring, 1996). CI = confidence interval. HCV diagnosis period is a time-dependent variable, coded as ‘Pre’ for the period before HCVdiagnosis date −30 days, ‘±30 days’ for the 30 days before and after HCV diagnosis, and ‘Post’ for the period subsequent to HCV diagnosis date plus 30 days.

significantly elevated relative risk for the 4th quintile only, compared with the 1st quintile (HR = 1.5, 95% CI: 1.1–2.0). Compared with the period >30 days preceding HCV diagnosis, there was a 6.3fold increased relative risk of admission in the 30-day period before and after diagnosis with HCV infection (95% CI: 5.2–7.6), and a 1.5fold increased relative risk >30 days after HCV diagnosis date (95% CI: 1.2–1.7). A supplementary analysis conducted on the entire IDU study population (i.e., not restricting to those ever diagnosed with HCV infection), resulted in larger hazard ratios for HCV diagnosis period; for the 60-day period around HCV diagnosis date: HR = 6.4 (95% CI: 5.1–7.9); >30 days after HCV diagnosis: HR = 1.8 (95% CI: 1.5–2.0). For non-liver-related outcomes (i.e., the full set of alcoholrelated ICD codes excluding alcoholic liver disease), the adjusted hazard ratios for sex, age, and deprivation quintile were slightly smaller than those obtained in the Cox model applied to the full set of alcohol-related outcomes (Table 4b). The relative risk of a nonliver-related alcohol admission for the period >30 days after date of HCV diagnosis was also slightly smaller to that obtained for all alcohol-related conditions (HR = 1.3, 95% CI: 1.1–1.5). A recurrent Cox model was also fitted separately to the nonliver-related alcohol-related admissions data for HCV diagnoses made from 2000 onwards. The hazard ratios for the 60-day period around the HCV diagnosis date (6.6, 95% CI: 5.2–8.4) and for the subsequent period (1.5, 95% CI: 1.2–2.0) were larger than the hazard ratios obtained using the entire dataset (see Table 4b).

Discussion Our results show a clear distinction in the proportion of IDUs with at least one alcohol-related hospitalisation subsequent to first attendance at drug services between those who have ever been diagnosed HCV antibody-positive (16%) and those who have not (6%). We attempted to address possible ascertainment bias by including a time-dependent variable for period in relation to date of HCV diagnosis, and found a sixfold increased relative risk of a first-time alcohol-related admission within the 30-day period around HCV diagnosis (consistent with an increased likelihood of testing/diagnosis associated with hospital admission), but no evidence for an increased risk of a first-time admission >30 days after HCV diagnosis (HR = 1.0), compared with >30 days before. However, recurrent event regression analysis applied to the subpopulation of HCV-diagnosed IDUs, with adjustment for sex, age and deprivation, indicated a 1.5-fold increased relative risk of alcohol-related hospitalisation >30 days subsequent to being diagnosed HCV antibody-positive compared with >30 days before. Removing alcoholic liver disease from the outcome event definition – thus examining the risk of hospitalisation for a non-liver alcohol-related condition (i.e., not likely to be a consequence of HCV infection) – reduced the hazard ratio only slightly, to 1.3. This suggests differences in problem alcohol use before and after HCV diagnosis. Evidence for excessive alcohol use at a young age is provided by the observation that 16% of IDUs younger than 20 years old at

68

S.A. McDonald et al. / International Journal of Drug Policy 22 (2011) 63–69

the end of the follow-up period had one or more hospital admissions with mention of alcohol. This finding is consistent with the results of a computer-administered interview study (Campbell et al., 2006) showing a high prevalence (72%) of moderate to risky drinking in younger (18–35 years) HCV-positive IDUs in the USA. Alcohol-related hospitalisations in the <20-year-old Scottish IDUs may be largely attributed to single-episode behavioural problems or alcohol poisoning, whereas liver disease as a consequence of high levels of cumulative alcohol intake might be increasingly responsible for hospitalisation in the older groups. In support, of the first-time admissions in IDUs aged < 20 years, 62% listed mental and behavioural problems (ICD-10 F10), 31% listed toxic effects of alcohol (ICD-10 T51.0, T51.9), and 0% listed alcoholic liver disease (ICD-10 K70). In comparison, the distribution over these three categories for first-time admissions in IDUs aged 40+ years was 69%, 10% and 13%. We lacked data on alcohol consumption among our IDU population. However, as a proxy variable for alcohol consumption levels, the rate of hospitalisation with an alcohol-related diagnosis appears to have some validity, as demonstrated in a study indicating a positive dose-response relationship between self-reported ‘usual’ consumption and the risk of first-time alcohol-related admission in a large Scottish cohort (McDonald, Hutchinson et al., 2009). Furthermore, data from a repeated cross-sectional survey of 808 Glasgow IDUs in 2005 and 2007, showed that 26% of all IDUs and 29% of HCV-infected IDUs reported drinking over the UK weekly guidelines of 14 and 21 units for females and males, respectively (M. O’Leary, personal communication). There were an estimated 90,000 current/former IDUs living in Scotland during 2005 (Hutchinson et al., 2005b). The SDMD, representing roughly 45% of the current/former IDU population living in Scotland, has allowed estimates of hospital usage associated with alcohol within this population to be generated, and problematic alcohol use in IDUs to be inferred. Previous studies of alcohol use by IDUs, restricted to samples of size 94–753, have reported the prevalence of excessive consumption in this group to range from 15% to 53% (Campbell et al., 2006; Gossop et al., 2000; Watson et al., 2007). Based on linked alcohol-related hospital discharge records, 8% (3332/41,062) of our IDU population may have had a problematic drinking episode(s) or a history of heavy intake which led to hospitalisation (subsequent to first attendance at drug services). If so, this figure reflects a lower bound on prevalence of problem drinking, since not all cases will be admitted to hospital as an inpatient. This indicator of problematic alcohol intake was higher for the HCV-diagnosed IDU subset – 16% compared with 6% – suggesting that excessive alcohol consumption is more prevalent in persons (ever) diagnosed with HCV infection (but we recognise that the directionality of this interpretation cannot be assumed). Although Scotland in general has a larger alcohol problem (reflected in cirrhosis mortality rates) compared with other European countries (Leon & McCambridge, 2006), the observations reported here on IDUs and HCV-infected IDUs are likely to reflect IDU populations elsewhere, as the extent of alcohol consumption among IDUs in Scotland (M. O’Leary, personal communication) is comparable to those in other countries (Stein et al., 2000). Our findings are disturbing given that heavy alcohol use is known to accelerate the rate of progression to cirrhosis and end-stage liver disease in HCV-infected individuals (Freeman et al., 2003; Hutchinson et al., 2005a; Monto et al., 2004). A further concern, in particular with regards to the acquisition of HCV infection through unsafe practices such as needle sharing, is the reported association between excessive alcohol use and risky injection behaviours (adjusted odds ratio of 2.3 [95% CI: 1.2–4.4] for alcohol abuse and reported needle sharing in the last 6 months) (Stein et al., 2000). Addressing problematic alcohol use

among all IDUs, including those who are HCV-seronegative, may also have a beneficial effect on the transmission of HCV within this population. One possible interpretation of the observed increased relative risk of an alcohol-related admission subsequent to HCV diagnosis within those IDUs ever HCV-diagnosed is that receipt of a positive diagnosis, despite any advice to minimise consumption, may have had an adverse effect on drinking habits. However, it is not possible to distinguish the diagnosis event itself from other factors which may underlie both HCV infection and subsequent positive diagnosis and an increase in the rate of alcohol-related hospital admissions. Furthermore, if an HCV test is more likely to be offered to a patient presenting with problem alcohol use, then presentation bias could also partly explain the elevated relative risk. Our study has certain limitations. First, the proportion of diagnosed HCV infection in our IDU cohort (22%) is much lower than the estimated national seroprevalence of 40–50% derived from anonymised unlinked studies (Roy et al., 2007). This is due – in the main – to an estimated two-thirds of HCV-infected individuals remaining undiagnosed, and – to a lesser extent – to the limitations of record-linkage using impoverished identifiers. In the latter case, our estimate of the relative risk of alcohol-related hospital admission associated with an HCV diagnosis may be lower than reality, because those IDUs who have actually been diagnosed but not identified as such may exhibit an increased admission rate but contribute person-time to the undiagnosed group; this issue is eliminated in the analysis restricting only to HCV-diagnosed IDUs. We were also not able to distinguish individuals who were HCV Ab+ (criterion for inclusion in the HCV Diagnosis database) and RNA+ from those that had cleared the virus (HCV Ab+/RNA−); it is the former group of chronically infected individuals representing the majority (around 75% (Micallef, Kaldor, & Dore, 2006)) of those HCV Ab+ to whom advice on minimising alcohol intake following an HCV diagnosis, should be directed. Second, relative risks may also be slightly underestimated due to the competing risk issue (Satagopan et al., 2004): because of the higher mortality rate and younger age at death in IDUs compared with the Scottish population (due mainly to drug-related causes of death (McDonald, Donaghy et al., 2009)), death is an informative censoring event in that it alters the probability of observing an alcohol-related outcome. Finally, the study population was restricted to those IDUs attending drug treatment services between 1995 and 2006. Some proportion of the IDU population will not have come into contact with services, meaning that presentation bias must be considered when interpreting our results (individuals with more chaotic lifestyles may have a reduced likelihood of attending services but also may have a higher chance of being hospitalised for reasons associated with their lifestyle, such as problem alcohol use). A study of 441 current IDUs recruited from Glasgow street sites in 2004 indicated that a high proportion (90%, 395/441) had ever been in treatment (Taylor et al., 2005). However, ‘help-seeking’ IDUs may still be over-represented in both the SDMD and Taylor et al’s study. Behavioural factors also supply an alternative explanation for the increased rate of alcohol-related hospitalisation among those with an HCV diagnosis, if risk-taking behaviour underlies both the acquisition of HCV infection (leading to a positive diagnosis) and excessive alcohol use (leading to hospitalisation). In conclusion, hospital admission rates for an alcohol-related cause are very high in Scotland’s IDU population. IDUs who are diagnosed with HCV infection appear to have a subsequent increased risk of hospital admission for an alcohol-related cause. Elevated alcohol intake following HCV diagnosis is one of a number of factors that could underlie the increased risk of hospital admission for an alcohol-related cause subsequent to a positive diagnosis. Alcohol may also be considered a risk factor for HCV transmission –

S.A. McDonald et al. / International Journal of Drug Policy 22 (2011) 63–69

if those IDUs who consume large quantities of alcohol also take more risks when injecting – for example, by inadvertent or careless needle-sharing. Because of the known synergistic effect of HCV infection and heavy alcohol consumption on the development of liver disease, it is imperative that alcohol intake is addressed in the management of chronic HCV infection in this population, and the in guidelines enforced by medical practitioners. Clearly, it is not possible to manage and treat HCV-infected persons without considering any drug or alcohol problem needs. Thus, an integrated approach to the management of HCV-infected persons – integrating HCV specialist treatment services with those for social care, mental health, and addiction – has been proposed, and is currently being developed, in Scotland (Scottish Government Health Department, 2008). Acknowledgements Funding was provided by a grant from the Chief Scientist Office of the Scottish Executive. SMB was funded by Medical Research Council, WBS number U.1052.00.002.00001.01. The authors thank ISD for performing the probabilistic record-linkage work. Conflict of interest statement SMB owns stock in GlaxoSmithKline. References Andersen, P. K., & Gill, R. D. (1982). Cox’s regression model for counting processes: A large sample study. Annals of Statistics, 10, 1100–1120. Campbell, J., Hagan, H., Latka, M., Garfein, R., Golub, E., Coady, M., et al. (2006). High prevalence of alcohol use among hepatitis C virus antibody positive injection drug users in three US cities. Drug Alcohol Depend, 81(3), 259–265. Carstairs, V., & Morris, R. (1990). Deprivation and health in Scotland. Health Bulletin, 48(4), 162–175. Freeman, A., Law, M., Kaldor, J., & Dore, G. (2003). Predicting progression to cirrhosis in chronic hepatitis C virus infection. Journal of Viral Hepatitis, 10(4), 285–293. Glynn, R. J., & Buring, J. E. (1996). Ways of measuring rates of recurrent events. British Medical Journal, 312(7027), 364–367. Gossop, M., Marsden, J., Stewart, D., & Kidd, T. (2003). The National Treatment Outcome Research Study (NTORS): 4–5 year follow-up results. Addiction, 98(3), 291–303. Gossop, M., Marsden, J., Stewart, D., & Rolfe, A. (2000). Patterns of drinking and drinking outcomes among drug misusers. 1-Year follow-up results. Journal of Substance Abuse Treatment, 19(1), 45–50. Hagan, H., & Des Jarlais, D. (2000). HIV and HCV infection among injecting drug users. Mount Sinai Journal Of Medicine, 67(5–6), 423–428. Hutchinson, S., Bird, S., & Goldberg, D. (2005a). Influence of alcohol on the progression of hepatitis C virus infection: A meta-analysis. Clinical Gastroenterology and Hepatology, 3(11), 1150–1159. Hutchinson, S., Bird, S., & Goldberg, D. (2005b). Modeling the current and future disease burden of hepatitis C among injection drug users in Scotland. Hepatology, 42(3), 711–723.

69

Hutchinson, S., Roy, K., Wadd, S., Bird, S., Taylor, A., Anderson, E., et al. (2006). Hepatitis C virus infection in Scotland: Epidemiological review and public health challenges. Scottish Medical Journal, 51(2), 8–15. Information Services Division (ISD). (2007). Drug misuse statistics Scotland 2007. Edinburgh: Common Services Agency. Kendrick, S., & Clarke, J. (1993). The Scottish record linkage system. Health Bulletin (Edinburgh), 51, 72–79. Leon, D. A., & McCambridge, J. (2006). Liver cirrhosis mortality rates in Britain from 1950 to 2002: An analysis of routine data. Lancet, 367(9504), 52–56. McDonald, S. A., Donaghy, M., Goldberg, D. J., Hutchinson, S., Robertson, C., Bird, S., et al. (2009). A population-based record linkage study of mortality in hepatitis C-diagnosed persons with or without HIV coinfection in Scotland. Statistical Methods in Medical Research, 18(3), 271–283. McDonald, S. A., Hutchinson, S., Bird, S., Graham, L., Robertson, C., Mills, P., et al. (2009). Association of self-reported alcohol use and hospitalisation for an alcohol-related cause in Scotland: A record-linkage study of 23,183 individuals. Addiction, 104(4), 593–602. Micallef, J. M., Kaldor, J. M., & Dore, G. J. (2006). Spontaneous viral clearance following acute hepatitis C infection: A systematic review of longitudinal studies. Journal of Viral Hepatitis, 13(1), 34–41. Monto, A., Patel, K., Bostrom, A., Pianko, S., Pockros, P., McHutchison, J., et al. (2004). Risks of a range of alcohol intake on hepatitis C-related fibrosis. Hepatology, 39(3), 826–834. Ostapowicz, G., Watson, K., Locarnini, S., & Desmond, P. (1998). Role of alcohol in the progression of liver disease caused by hepatitis C virus infection. Hepatology, 27(6), 1730–1735. R Development Core Team. (2006). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Roy, K., Hay, G., Andragetti, R., Taylor, A., Goldberg, D., & Wiessing, L. (2002). Monitoring hepatitis C virus infection among injecting drug users in the European Union: A review of the literature. Epidemiology and Infection, 129(3), 577–585. Roy, K., Hutchinson, S., Wadd, S., Taylor, A., Cameron, S., Burns, S., et al. (2007). Hepatitis C virus infection among injecting drug users in Scotland: A review of prevalence and incidence data and the methods used to generate them. Epidemiology and Infection, 135(3), 433–442. Satagopan, J., Ben-Porat, L., Berwick, M., Robson, M., Kutler, D., & Auerbach, A. (2004). A note on competing risks in survival data analysis. British Journal of Cancer, 91(7), 1229–1235. Scottish Government Health Department. (2008). Hepatitis C action plan for Scotland: Phase II: May 2008–March 2011. Edinburgh: Scottish Government. Scottish Intercollegiate Guidelines Network (SIGN). (2006, December). Management of hepatitis C. A national clinical guideline (SIGN publication; no. 92, 47pp.). Edinburgh: Scottish Intercollegiate Guidelines Network (SIGN). Shaw, L., Taylor, A., Roy, K., Cameron, S., Burns, S., Molyneaux, P., et al. (2003). Establishment of a database of diagnosed HCV-infected persons in Scotland. Commun Dis Public Health, 6(4), 305–310. Stein, M., Hanna, L., Natarajan, R., Clarke, J., Marisi, M., Sobota, M., et al. (2000). Alcohol use patterns predict high-risk HIV behaviors among active injection drug users. Journal of Substance Abuse Treatment, 18(4), 359–363. Taylor, A., Allen, E., Hutchinson, S., Roy, K., Goldberg, D., Ahmed, S., et al. (2005). Evaluation of the lord advocate’s guidance on the distribution of sterile needles and syringes to injecting drug users. Edinburgh: Scottish Executive (Effective Interventions Unit). Twisk, J., Smidt, N., & de Vente, W. (2005). Applied analysis of recurrent events: A practical overview. Journal of Epidemiology and Community Health, 59(8), 706–710. Watson, B., Conigrave, K., Wallace, C., Whitfield, J., Wurst, F., & Haber, P. (2007). Hazardous alcohol consumption and other barriers to antiviral treatment among hepatitis C positive people receiving opioid maintenance treatment. Drug and Alcohol Review, 26(3), 231–239. Windeler, J., & Lange, S. (1995). Events per person year—A dubious concept. British Medical Journal, 310(6977), 454–456.