Patterns of Drug Use and Drug-related Hospital Admissions in HIV-positive and -negative Gay and Bisexual Men

Abstract

We aimed to compare rates of illicit drug-related hospitalisations in HIV-negative (HIV−ve) [n=1325] and HIV-positive (HIV+ve) [n=557] gay and bisexual men (GBM) with rates seen in the general male population and to examine the association between self-reported illicit drug use and drug-related hospitalisation. Participants were asked how often they used a range of illicit drugs in the previous six months at annual interviews. Drug-related hospital admissions were defined as hospital admissions for mental or behavioural disorders due to illicit drug use (ICD 10: F11–16, F18, F19); drug poisoning (T40–T45, T50); or toxic effect of gases (T53, T59, T65). Drug-related hospitalisations were 4.8 times higher in the HIV−ve cohort [SIR:4.75 (95%CI:3.30–6.91)] and 3.5 times higher in the HIV+ve cohort [SIR:3.51 (1.92–5.88)] compared with the general population. Periods of weekly drug use [IRR:1.86 (1.01–3.46)], poly-drug use [IRR:2.17 (1.07–4.38)] and cannabis use [low use-IRR:1.95 (1.01–3.77), high use-IRR:2.58 (1.29–5.16)] were associated with drug-related hospitalisation in both cohorts, as was being a consistently high meth/amphetamine user throughout follow-up [IRR:3.24 (1.07–9.83)] and an inconsistent and consistent injecting drug user throughout follow-up [IRR:3.94 (1.61–9.66), IRR:4.43(1.04–18.76), respectively]. Other risk factors for drug-related hospitalisation indicated the likelihood of comorbid drug and mental health issues in GBM hospitalised for drug use.

Introduction

There is substantial evidence that illicit drug use is more prevalent among gay and bisexual men (GBM) globally compared with their heterosexual counterparts (1–4). Research on drug use in Australian GBM comes from the National Drug Strategy Household Survey (NDSHS)(5) and the Gay Community Periodic Surveys (GCPS)(6) conducted in major capital cities. Both the NDSHS and GCPS have consistently shown significantly higher levels of recent (in the last 6–12 months) meth/amphetamine use (8.6–13.9%), ecstasy use (12.8–28.8%), cocaine use (5.9–19.0%) and cannabis use (30.4–38.6%) in GBM compared to heterosexual men (5, 7–11). These rates are higher than other population groups in Australia and are consistent with data from other high income countries (1, 12).

Injecting drug use (IDU) remains generally low in GBM (<5% in most non-purposive samples), though rates are higher than those found in a comparable heterosexual population (13) and tend to be significantly higher in HIV positive (HIV+ve) compared with HIV negative (HIV−ve) GBM (15). Methamphetamines and steroids are reported to be the most commonly injected drugs in GBM in Australia (16). Polydrug use (defined as the consumption of more than one illicit drug during a specific time period) has also been identified as being common in GBM (17–19). Polydrug use has been shown to be both associated with increased risk for toxicity and overdose (20, 21) as well as poorer outcomes in terms of mental and physical health when compared with single drug use.

The implications of illicit drug use in GBM are not well understood. While the physical and mental harms of drug abuse are well documented (22–27) and likely to be present in GBM abusing drugs, the difficulty in distinguishing between illicit ‘use’ and problematic ‘abuse’ of drugs is inherently difficult. Drug abuse is generally distinguished from drug use by measurable physical, social, or psychological harm (28). To-date, much of the substance use research in HIV−ve and HIV+ve GBM has focused on HIV acquisition and transmission respectively as the measureable harm.

This study benefits from the linkage of two cohorts of GBM, with high levels of illicit drug use (29, 30), to a national hospital data register enabling rates of drug-related hospital admissions to be examined over a long period of follow-up. Furthermore, detailed drug-use questionnaires were completed by cohort participants at annual interviews and thus hospitalisations can be analysed with respect to reported drug use. We aimed to compare rates of illicit drug-related hospitalisations in HIV−ve and HIV+ve GBM with rates seen in the general male Australian population and to examine the association between self-reported illicit drug use and drug-related hospital admissions.

Methods

Study population

Our study cohorts included participants recruited to the Health in Men (HIM) (HIV−ve) and Positive Health (pH) (HIV+ve) studies who provided informed consent for their study data to be used for data linkage. Both studies have been described in detail elsewhere (31, 32). Briefly, men were recruited from Sydney, Australia using similar community-based methods. The majority of participants in both studies were recruited through gay community events and venues. Other sources of recruitment included direct recruitment from participants in other relevant studies, personal networking, ‘snowballing’ through friends and acquaintances, direct referrals from medical practitioners, gay press and HIV-positive publications (33, 34). Participants were interviewed face-to-face annually. Enrolment in HIM occurred from 2001 to 2004 and active follow-up ceased in 2007. Enrolment in pH occurred from 1998 to 2006 and follow up ceased in 2007. The HIV-serostatus of participants in both cohorts was confirmed by serological testing at intake and in HIV−ve participants through annual testing thereafter. All participants in both studies either had sexual contact with at least one man during the previous 5 years or self-identified as gay, homosexual, queer or bisexual. In both studies the overwhelming majority of participants identified as gay, homosexual or queer with the remaining largely identifying as bisexual (33, 35).

Drug Use Measures

Participants were asked how often they used a range of drugs in the previous six months for recreational purposes at annual face-to-face interviews. Drugs assessed included cannabis; amyl nitrate; Viagra or other erection medications; cocaine; amphetamines or methamphetamines; MDMA or other forms of MDA; psychedelics or hallucinogens (lysergic acid diethylamide(LSD), mescaline, or phencyclidine(PCP)); downers (barbiturates, tranquilisers or sedatives), rohypnol (flunitrazepam) or ketamine; heroin or other opiates (including methadone); and gamma hydroxybutyrate (GHB) [only in the HIV+ve cohort]. Slang terms were in some cases used in addition to drug names (e.g. ‘Angel dust’ for PCP). Participants were also asked how often they injected drugs (other than steroids) for other than health reasons in the previous 6 months. The possible frequency of use options for illicit and injecting drug use were never, once or twice, about once a month, about once a week, more than once a week, or every day. Frequency of use was re-categorised into low use (once or twice to about once a month) and high use (about once a week to everyday). Polydrug use for the purposes of this analysis was defined as using more than one drug about once a week or more frequently (i.e. answered ‘more than once a week’ or ‘every day’). GHB was excluded from the polydrug use measure as it was not measured in both cohorts. Use of Viagra and other erection medications were also excluded from this measure as they were used ubiquitously to enhance penile function during sexual intercourse and examination of illicit drug use was our primary aim.

Registries and Data linkage

Data linkage included all consenting participants. Probabilistic linkage methods (36) were used to link individuals to the data sources described below.

1. The New South Wales (NSW) Admitted Patient Data Collection (APDC) includes all inpatient admissions from all public (including psychiatric), private and repatriation hospitals, private day procedure centres and public nursing homes in NSW, Australia. Diagnosis and procedure fields are coded according to the 10th revision of the International Classification of Disease-Australian Modification (ICD-10-AM). Patient name has only been recorded since 1 July 2000, so we restricted analysis to admissions from 1 July 2000.

2. The Registry of Births, Deaths and Marriages (RBDM) which reports fact of death was used to censor person-years of observation.

3. The HIV administrative database is a register of HIV, notified to the NSW department of Health by laboratories, hospitals, and medical practitioners. In addition to annual serological testing in the HIM cohort, seroconversions were identified through linkage of participants to the HIV registry.

First name, surname, address, postcode, date of birth and date of last contact were used to probabilistically link participants from the study cohorts to the APDC and RBDM registries using ChoiceMaker software (ChoiceMaker Technologies Inc., New York, US). Deterministic linkage was used to link participants to the HIV/AIDS notifications using two-character surname and given name codes, date of birth, sex and postcode. Linkage was conducted by the NSW Centre for Health Record Linkage, independent of the study investigators. Full details of the linkage process are outlined at (http://www.cherel.org.au/how-record-linkage-works).

Individual consent for data linkage was optional and was collected in addition to consent to participate in the study. Only data from participants who consented to data linkage were included in this analysis (93% of HIM and 74% of pH participants). We found no significant differences between those that consented versus those that declined linkage in examined cohort characteristics. Examined cohort characteristics included ethnicity, education, employment, income, sexual identity, previous exposure to an STI or hepatitis C, self-reported health, Kessler 6 score of psychological distress, relationship status and HIV-serostatus, sexual risk taking behaviour, frequency of exercise, experiences of discrimination and alcohol use (see Appendix A). Ethics approval was granted by the University of NSW and the NSW Population and Health Services Research Ethics Committee.

Drug-related hospital admissions

A hospital admission was defined as an episode of care ending with hospital discharge, death or transfer to another type of care. Duplicate and nested hospital admissions (admissions within the date range of another admission) were excluded (n=55) to ensure only one principal diagnostic code for each admission. Drug-related hospital admissions were defined as hospital admissions with a principal or secondary ICD-10 diagnosis code: mental or behavioural disorders due to drug use (F11–16, F18, F19); drug poisoning (T40–T43, T50); or toxic effect of gases (T53, T59, T65).

Statistical Methods

Time at risk commenced at entry into the study cohort or the opening of database for hospital admissions (1 July 2000), whichever was latest. Incidence rates of events were determined using person-years (PYs) methods with data right censored at death or the end of the follow-up period (31 December 2007). Data from HIV−ve participants who seroconverted (n=51) were excluded from analysis as exact date of seroconversion was unattainable.

The number of hospitalisations with drug use as a primary diagnosis in the cohorts was compared with the expected number using rates for the Australian male population in 2001 (standard population) (37) and summarized as standardised incidence ratios (SIRs). Numbers were calculated in 10 year age groups (18–24, 25–29, etc. to 85+ years), thus adjusting for age. To adjust for correlation between hospitalisations in the same individual, 95% confidence intervals (CIs) for the SIRs were calculated using the method by Stukel et al. (38).

Random-effects Poisson methods were used to assess the relationship between self-reported drug use, other risk factors for drug-related hospital attendance and drug-related hospital attendance. Each risk factor was assessed for its association with drug-related hospitalisation independently in univariate analyses. Drug use was assessed as a patient-level and observation-level, time-updated variable. For the patient-level analysis, participants were categorized as nonusers, inconsistent users (moving between use and non-use), consistently low users (once or twice to about once per month) and consistently high users (about once per week or more frequent) according to their longitudinal pattern of drug use for each drug type. Participants needed to have at least two rounds of interviews to be included in the patient-level analysis otherwise they were included in the analysis as a missing category. For the observation-level, time-updated analysis, participants were categorised on the basis of their self-reported drug use for each drug type in a particular interview year, which was then tested for its association with drug-related hospitalisations in the same year. Other risk factors assessed as fixed effects at entry into the cohort included: ethnicity, country of birth, highest level of education, employment, income, presence of antibodies to hepatitis C, smoking, sexual identity, self-reported general health, prior use of mental health counselling services, and Kessler 6 score of psychological distress. Other risk factors assessed as time-updated covariates included: age, frequency of exercise, alcohol consumption, frequency of gay social engagement, number of gay friends, serostatus of regular relationship, number of sexual partners and frequency of unprotected anal intercourse. The interactions between risk factors for drug-related hospitalisation and HIV status (cohort) were tested. If a significant association was found, indicating that the effect of the risk factor differed by cohort then stratified analyses were done and separate results present for HIV+ve and HIV−ve GBM cohorts. Missing data was included in the analysis as a separate category. Analyses were performed using STATA (version 13; StataCorp LP, College Station, Texas, USA) and SAS (version 9.3; SAS Institute INC., North Carolina, USA).

Results

Study Participants

A total of 1,882 participants were included in the analysis; 1,325 HIV−ve GBM and 557 HIV+ve GBM. At baseline, HIV−ve participants were aged from 18 to 75 years with a median age of 35 years (Table 1). HIV+ve participants were aged from 21 to 69 with a median age of 41 years. Seventy-four percent (73.6%) of HIV−ve men and 60.0% of HIV+ve men had attained some form of tertiary education after secondary school. The majority of men in both cohorts identified as Anglo-Australian or Anglo-Celtic (74.3% of HIV−ve, 76.1% of HIV+ve) and had been born in Australia (68.9% of HIV−ve, 70.4% of HIV+ve). The majority of men self-identified as gay, queer or homosexual (95.2% of HIV−ve, 92.1% of HIV+ve). Of the HIV+ve participants, 44.7% reported a recent CD4 above 500 cells/mm3, the majority were receiving antiretroviral therapy (74.2%) and most participants (76.7%) were diagnosed with HIV prior to 1996 (pre-HAART era).

Table 1.

Baseline characteristics of 557 HIV+ve and 1,325 HIV−ve gay and bisexual men recruited in Sydney, Australia.

	HIV−ve cohort (n=1325)	HIV+ve cohort (n=557)
Age (years), median (IQR)	35.3 (29.6–42.0)	40.9 (36.0–46.7)
Ethnicity
Anglo-Australian/ Anglo-Celtic	985 (74.3)	424 (76.1)
Country of birth
Australia	913 (68.9)	392 (70.4)
Education
Tertiary	975 (73.6)	334 (60.0)
Employment
Unemployed or Receiving Pension/Disability Pension	83 (6.3)	216 (38.8)
Income
<500 per week/ <26,000 per year	266 (20.1)	267 (47.9)
Sexual Identity
Gay, queer or homosexual	1088 (95.2)	443 (92.1)
Bisexual	42 (3.7)	17 (3.5)
Other	9 (0.8)	21 (4.4)
Exposure to Hepatitis C
Yes	52 (3.9)	69 (12.4)
Daily Smoker
yes	420 (31.7)	292 (52.42)
Average number of drinks when drinking
5 to 8	242 (18.3)	68 (12.2)
>9	62 (4.7)	31 (5.6)
Year diagnosed HIV Positive
1980–1986		151 (27.1)
1987–1991		153 (27.5)
1991–1996		123 (22.1)
1996–2006		118 (21.2)
Last CD4 cell count
less than 100		26 (4.7)
100 to 200		28 (5.0)
201 to 350		89 (16.0)
351–500		105 (18.9)
Over 500		249 (44.7)
Antiretroviral Therapy History
never taken		87 (15.6)
currently taking		413 (74.2)
past but now stopped		47 (8.4)

All numbers are N (%) unless otherwise stated.

Abbreviations: IQR=Interquartile Range

There were a total of 5,455 interviews completed in the HIV−ve group and 1,887 interviews completed in the HIV+ve group. Of the 1,325 HIV−ve participants included in the analysis, all men completed at least 1 interview following entry into the study; 89% two interviews, 79% three interviews, 68% four interviews, 48% five, 28% the maximum possible of six interviews. Of the 557 HIV+ve participants included in the analysis, all men completed at least one interview; 78% complete two interviews, 56% completed three interviews, 43% four interviews, 32% five, 18% six, 11% the maximum number of seven interviews. HIV−ve GBM reported no illicit drug use in 24% of interviews, using an illicit drug once or twice was reported in 21% of interviews, using an illicit drug monthly was reported in 24% of interviews, and using an illicit drug at least weekly was reported in 32% of interviews (Table 2). HIV+ve GBM reported no illicit drug use in 17% of interviews, using an illicit drug once or twice was reported in 19% of interviews as was using an illicit drug about monthly, using an illicit drug at least weekly was reported in 46% of interviews. The most consistently used drugs in both cohorts of men were cannabis, amyl nitrate, MDMA or other MDA and meth/amphetamines. One percent of HIV−ve and 6% of HIV+ve men reported consistent injecting drug use with 5% of HIV−ve men and 13% of HIV+ve men reporting intermittent injecting drug use. The most commonly used drugs in polydrug use (at least weekly use of 2 or more drugs) were cannabis and amyl nitrate in both cohorts followed by meth/amphetamines, MDMA or other MDA and downers, Rohypnol or ketamine (Figure 1).

Table 2.

Comparison of prevalence of drug use (%) reported in 5,455 interviews in 1,325 HIV−ve and 1,887 interviews in 557 HIV+ve gay and bisexual men recruited in Sydney, Australia.

	HIV−ve cohort (n=1,325; 5,455 interviews)	HIV+ve cohort (n=557; 1,887 interviews)	P-value
Any recreational drug use†
% of Interviews: None	23.7	16.6
% of Interviews: Once or twice	20.6	18.6
% of Interviews: About 1/month	23.9	19.0
% of Interviews: About 1/week or more	31.8	45.8	<0.001
Weekly drug use
% of Interviews: 0 drugs	68.2	54.2
% of Interviews: 1 drugs	22.9	35.3
% of Interviews: 2 drugs or more	8.9	10.5	<0.001
Any drug use
Nonuser	10.3	4.3
Intermittent user of 1 drug	7.1	7.2
Intermittent user of >1 drug	13.8	14.0
Consistent user of 1 drug	2.6	3.8
Consistent user of 1 drug, intermittent user of other drugs	16.3	18.5
Consistent user of >1 drug	2.3	3.4
Consistent user of >1 drug, intermittent user of other drugs	36.2	21.9	<0.001
Use of cannabis
Nonuser	28.8	19.0
Intermittent user	33.0	25.0
Consistent user	26.9	34.1	<0.001
Use of amyl nitrate
Nonuser	24.9	23.0
Intermittent user	31.9	31.6
Consistent user	31.9	23.5	0.076
Use of Viagra/erection pills
Nonuser	51.3	47.2
Intermittent user	27.6	19.8
Consistent user	9.7	11.0	0.038
Use of cocaine
Nonuser	54.5	55.1
Intermittent user	27.7	21.2
Consistent user	6.5	1.8	0.219
Use of meth/amphetamines
Nonuser	38.6	36.8
Intermittent user	32.5	25.0
Consistent user	17.7	16.3	0.219
Use of MDMA or other MDA
Nonuser	29.0	32.3
Intermittent user	24.2	25.7
Consistent user	35.6	20.1	<0.001
Use of psychedelics or hallucinogens
Nonuser	73.6	61.9
Intermittent user	14.0	14.4
Consistent user	1.1	1.8	0.111
Use of downers, Rohypnol or ketamine
Nonuser	50.6	76.5
Intermittent user	29.1	16.0
Consistent user	16.4	2.2	<0.001
Use of heroin or other opiates
Nonuser	86.2	74.5
Intermittent user	2.3	3.1
Consistent user	0.2	0.5	0.157
Use of GHB
Nonuser	-	65.2
Intermittent user	-	10.6
Consistent user	-	2.3
Injecting drug use
Nonuser	83.2	57.6
Intermittent user	4.5	12.9
Consistent user	1.1	5.8	<0.001

^†Excludes use of Viagra/other erection medications.

Abbreviations: GHB=gamma hydroxybutyrate;

All numbers are % of participants (n) unless otherwise stated.

Figure 1.

Polydrug use (at least weekly use of 2 or more drugs) by type of drug in 5,455 interviews in 1,325 HIV−ve and 1,887 interviews in 557 HIV+ve gay and bisexual men recruited in Sydney, Australia.

Drug-related hospitalisation rates

The maximum follow-up time in the HIV−ve and HIV+ve cohort was 6.6 and 7.5 years with a mean follow-up time of 5.4 and 5.8 years, respectively (Table 3). There were 3 deaths in the HIV−ve cohort and 25 in the HIV+ve cohort during follow-up. We observed 76 drug-related hospitalisations in 1,325 HIV−ve men [rate p. 100PYs 1.07 (95%CI 0.85–1.34)] and 83 drug-related hospitalisations in 557 HIV+ve men [rate p. 100PYs 2.56 (95%CI 2.06–3.17)].

Table 3.

Comparison of drug-related hospitalisation admission in 1325 HIV−ve and 557 HIV+ve gay and bisexual men recruited in Sydney, Australia

	HIV−ve (n=1,325)	HIV+ve (n=557)
Total Follow-up Time, years	7101	3243
Mean Follow-up Time, years	5.4	5.8
Number of Drug-related Discharges, N (%)	76 (3.2)	83 (4.8)
Primary diagnosis drug-related	48	14
Secondary diagnosis drug-related	28	69
Median Length of Stay, days (IQR)	2 (1–6)	4 (2–9)
Number of Participants with Drug-related Discharge, N(%)	37 (2.8)	48 (8.6)
Rate p. 100PYs (95%CI)	1.07 (0.85–1.34)	2.56 (2.06 – 3.17)
Compared with General Australian Male Population:
SIR (95%CI)a	4.75 (3.30–6.91)	3.51 (1.92–5.88)
HIV+ve compared with the HIV−ve Cohort:
IRR (95%CI)b	Reference	3.02 (1.67–5.46)
By diagnosis, N (%)
F11: Mental and behavioural disorders due to use of opioids	1 (1.3)	18 (21.7)
F12:Mental and behavioural disorders due to use of cannabinoids	10 (13.2)	27 (32.5)
F13:Mental and behavioural disorders due to use of sedatives or hypnotics	0	4 (4.8)
F14: Mental and behavioural disorders due to use of cocaine	0	2 (2.4)
F15:Mental and behavioural disorders due to use of other stimulants, including caffeine	11 (14.5)	16 (19.3)
F16:Mental and behavioural disorders due to use of hallucinogens	0	4 (4.8)
F18:Mental and behavioural disorders due to use of volatile solvents	0	0
F19:Mental and behavioural disorders due to multiple drug use and use of other psychoactive substances	7 (9.2)	5 (6.0)
T40:Poisoning by narcotics and psychodysleptics	2 (2.6)	0
T41:Poisoning by anaesthetics and therapeutic gases	5 (6.6)	1 (1.2)
T42:Poisoning by antiepileptic, sedative-hypnotic and antiparkinsonism drugs (includes barbiturates, benzodiazepines)	22 (28.9)	3 (3.6)
T43:Poisoning by psychotropic drugs, not elsewhere classified	10 (13.2)	3 (3.6)
T50:Poisoning by diuretics and other and unspecified drugs, medicaments and biological substances	2 (2.6)	0
T53:Toxic effect of halogen derivatives of aliphatic and aromatic hydrocarbons	5 (6.6)	0
T59:Toxic effect of other gases, fumes and vapours	1 (1.3)	0
T65:Toxic effect of other and unspecified substances	0	0

^aAdjusted for different age distributions in the HIV−ve cohort, HIV+ve cohort and the general male Australian population. SIRs for primary diagnosis only, hospital rates for secondary drug diagnoses were unavailable in the general population.

^bIRRs are age-adjusted.

Abbreviations: n=number of individuals; SIR=Standardized Incidence Ratio; 95%CI=95% Confidence Interval; N=Number of hospitalisations; PYs= Person years; IQR=Interquartile Range; PYs=Person-years; IRR=Incidence Rate Ratio

There was a significantly higher rate of drug-related hospital admissions in the HIV+ve compared with HIV−ve cohort [IRR 3.02 (95% CI 1.67–5.46); p-value<0.001]. Hospitalisation rates with a drug-related primary diagnosis were 4.8 times higher in the HIV−ve cohort [SIR 4.75 (95% CI 3.30–6.91)] and 3.5 times higher in the HIV+ve cohort [SIR 3.51 (95% CI 1.92–5.88)] compared with the Australian male population (Table 3). Hospitalisation for mental and behavioural disorders due to stimulant use was prevalent in both cohorts. Hospitalisation for opioids and cannabinoids was more prevalent in the HIV+ve cohort compared with the HIV−ve cohort. Hospitalisations due to ICD-10 code “poisoning by antiepileptic, sedative-hypnotic and antiparkinsonism drugs”, which were primarily for benzodiazepine use (N=18 (82%)) was more prevalent in the HIV−ve cohort.

Self-reported drug use as a risk factor for hospitalisations

Periods of any weekly single drug use (p-value=0.048) and poly-drug use (2 or more drugs: p-value=0.031) were found to be significantly associated with drug-related hospitalisation in both cohorts, as were periods of injecting drug use (p-value=0.017) (Table 4). Periods of low cannabis use (p-value=0.046) and high cannabis use (p-value=0.008) were also found to be associated with drug-related admissions as were, to a lesser extent, periods of high use of downers, rohypnol or ketamine (p-value=0.053). Both low (p-value=0.001) and high (p-value=0.027) meth/amphetamine use was associated with drug-related hospital admission in the HIV−ve cohort but not the HIV+ve cohort. There was also a strong indication that missing drug-use data for a particular interview year was associated with drug-related hospitalisation (see Table 4).

Table 4.

Univariate analysis of risk factors for drug-related hospitalisations in 1325 HIV−ve and 557 HIV+ve gay and bisexual men recruited in Sydney Australia.

Risk Factor	Category	N	PYs	IRR	95% CI	P- value	Over all P- value
Observation Level, Time-updated Drug Useb
Number of drugs used weekly or more	0	31	4084	1
	1	26	1554	1.86	1.01 – 3.46	0.048
	2+	17	526	2.17	1.07 – 4.38	0.031
	Missing	85	4180	1.93	1.17 – 3.18	0.010	0.0555
Injecting drug use	No	55	5320	1
	Yes	13	310	2.43	1.17 – 5.04	0.017
	Missing	91	4713	1.44	0.95 – 2.20	0.073	0.0375
Use of cannabis	No	24	3102	1
	Low use	24	1921	1.95	1.01 – 3.77	0.046
	High use	26	1139	2.58	1.29 – 5.16	0.008
	Missing	85	4181	2.19	1.26 – 3.76	0.005	0.026
Use of meth/amphetamines in HIV−vea	No	14	3004	1
	Low use	23	1570	3.77	1.69 – 8.39	0.001
	High use	7	203	3.56	1.15 – 11.00	0.027
	Missing	32	2324	3.28	1.52 – 7.06	0.002	0.0071
Use of meth/amphetamines in HIV+vea	No	20	907	1
	Low use	8	424	0.63	0.25 – 1.62	0.342
	High use	2	54	1.10	0.21 – 5.82	0.904
	Missing	53	1857	0.86	0.46 – 1.59	0.628	0.7985
Use of downers, Rohypnol or ketamine	No	51	4458	1
	Low use	15	1543	0.83	0.43 – 1.62	0.589
	High use	8	163	2.41	0.99 – 5.89	0.053
	Missing	85	4181	1.35	0.89 – 2.06	0.158	0.0885
Patient Level Drug Usec
Use of cannabis	Nonusers	32	2659	1
	Inconsistent users	25	3243	0.72	0.33 – 1.56	0.405
	Consistently low users	8	842	0.71	0.23 – 2.22	0.560
	Consistently high users	53	2199	1.98	0.92 – 4.24	0.080
	Missing	41	1401	2.67	1.14 – 6.24	0.024	0.0085
Use of cocaine	Nonusers	64	5703	1
	Inconsistent users	54	2701	1.79	0.96 – 3.33	0.067
	Consistently low users	0	466	-
	Consistently high users	0	73	-
	Missing	41	1401	2.82	1.32 – 6.02	0.007	0.0810
Use of meth/amphetamines	Nonusers	34	3985	1
	Inconsistent users	50	3235	1.61	0.81 – 3.19	0.172
	Consistently low users	16	1132	1.43	0.56 – 3.66	0.452
	Consistently high users	18	591	3.24	1.07 – 9.83	0.037
	Missing	41	1401	3.45	1.51 – 7.89	0.003	0.0302
Use of heroin or other opiates	Nonusers	93	8627	1
	Inconsistent users	18	281	5.21	1.29 – 21.00	0.020
	Consistent users	1	11	14.69	0.38 – 573.00	0.151
	Missing	41	1401	2.85	1.40 – 5.79	0.004	0.0066
Injecting drug use	Nonusers	74	7891	1
	Inconsistent users	32	786	3.94	1.61 – 9.66	0.003
	Consistent users	12	246	4.43	1.04 – 18.76	0.044
	Missing	41	1420	3.14	1.55 – 6.38	0.001	0.0003
Other Risk Factors
Age	18–35	43	2949	1
	34–45	67	4007	1.42	0.85 – 2.37	0.177
	44–55	48	2472	1.11	0.60 – 2.05	0.747
	+55	1	916	0.07	0.01 – 0.62	0.016	0.0314
Education in HIV−vea	10 years of high-school or less	14	759	1
	Completed high school	17	1108	0.86	0.19 – 3.97	0.848
	Tertiary education	5	1521	0.17	0.03 – 0.87	0.033
	University or Postgraduate	40	3707	0.56	0.15 – 2.06	0.385	0.2079
Education in HIV+vea	10 years of high-school or less	35	801	1
	Completed high school	10	540	0.47	0.16 – 1.43	0.184
	Tertiary education	29	753	1.07	0.43 – 2.67	0.888
	University or Postgraduate	9	1149	0.17	0.06 – 0.47	0.001	0.0015
Employment	Employed	86	8348	1
	Unemployed	73	1812	4.21	2.19 – 8.06	<0.001	0.0001
Income per week ($AUD)	<500	91	3055	1
	500–999	37	3559	0.33	0.17 – 0.64	0.001
	1000–1499	15	1946	0.26	0.11 – 0.59	0.001
	>1500	15	1585	0.27	0.12 – 0.64	0.003
	Missing	1	199	0.16	0.01 – 1.94	0.148	0.0009
Had antibodies to Hepatitis C	No	110	8741	1
	Yes	36	720	3.81	1.47 – 9.86	0.006
	Missing	13	884	1.05	0.40 – 2.80	0.915	0.0224
Self-reported Health	Poor	23	707	1
	Fair	46	2939	0.47	0.16 – 1.36	0.164
	Good	72	4539	0.47	0.17 – 1.32	0.154
	Excellent	17	2126	0.26	0.08 – 0.84	0.024	0.2680
Reported as having sought counselling for mental health	No	84	7438	1
	Yes	75	2898	2.51	1.40 – 4.52	0.002	0.0087
Kessler 6 Scale (Psychological Distress)	Low	46	5329	1
	Moderate	44	1841	2.86	1.41 – 5.83	0.004
	High	39	1833	2.79	1.34 – 5.82	0.006
	Missing	30	1341	3.36	1.37 – 8.26	0.008	0.0022
Daily Smoker	No	65	6219	1
	Yes	92	3986	2.01	1.15 – 3.51	0.015
	Missing	2	139	1.45	0.21 – 10.25	0.708	0.0505
Number of Alcoholic drinks consumed in one sitting (on average)	Non-drinker	2	105	1
	1 or 2	6	1812	0.43	0.08 – 2.38	0.332
	3 or 4	16	2046	0.87	0.18 – 4.28	0.868
	5 to 8	8	643	1.28	0.24 – 6.86	0.774
	>9	11	132	6.37	1.13 – 35.88	0.036
	Missing	116	5606	2.09	0.77 – 9.23	0.329	<0.0001
Serostatus of regular relationship	No regular partner	31	2155	1.71	0.85 – 3.47	0.131
	Concordant	15	2297	1
	Discordant	15	699	3.51	1.54 – 7.98	0.003
	Unknown	1	116	1.02	0.13 – 8.28	0.986
	Missing	97	5077	2.29	1.22 – 4.31	0.010	0.0296
Number of sexual partners, 6 months prior	0–1	10	1421	1
	2–5 men	8	1147	0.90	0.33 – 2.50	0.847
	6–10 men	21	1007	2.46	1.02 – 5.92	0.044
	11–50 men	14	1396	1.45	0.57 – 3.69	0.433
	>50 men	13	734	2.44	0.89 – 6.68	0.082	0.1598

^asignificant p-value for interaction term;

^bTime-updated use of amyl nitrate, viagra/erection pills, cocaine, psychedelics or hallucinogens, heroin/other opiates and GHB had p-values>0.1;

^cPatient-level use of amyl nitrate, viagra/erection pills, psychedelics or hallucinogens, downers/rohypnol/ketamine, and GHB all had p-values>0.1;;

Abbreviations: N= Number of drug-related hospital admissions, PYs=Person-years, 95% CI= 95% Confidence Intervals

There was evidence that participants who engaged in consistent high users of meth/amphetamine were more likely to be hospitalised for drug-related reason than nonusers (p-value=0.037). Participants who were inconsistent users of heroin or other opiates (p-value=0.020) were also found to be significantly more likely to be hospitalised for drug-related reasons. Consistent use of heroin/other opiate was extremely infrequent and did not reach statistical significance. While not reaching statistical significance there was some indication that inconsistent users of cocaine (p-value=0.067), and consistent high users of cannabis (p-value=0.080) were more likely to be hospitalised for drug-related reasons. Being an inconsistent or consistent injecting drug user was also associated with drug-related hospitalisation (p-value=0.003, p-value=0.044, respectively). Again there was also a strong indication that missing longitudinal drug data (i.e. only attending one interview post enrolment) was associated with drug-related hospitalisation (see Table 4).

Other risk factors for drug-related hospitalisation included being unemployed at entry in to the cohort (p-value<0.001), having been previously exposed to hepatitis C (p-value=0.006), having previously sought counselling for mental health (p-value=0.002), having a high (p-value=0.006) or moderate (p-value=0.004) score on the Kessler 6 Scale of psychological distress compared to low score, being a daily smoker (p-value=0.015) and drinking more than 9 drinks in one sitting compared to not drinking (p-value=0.036). Being in a HIV-sero-discordant relationship compared to being in a concordant relationship was also found to be associated with drug-related hospitalisation (p-value=0.003) as was having 6–10 sexual partners compared to 0–1 partners in the previous 6 months (p-value=0.044). Being older than 55years (p-value=0.016) and earning more than 500$AUD per week (500–999: p-value=0.001, 1000–1499: p-value=0.001, ≥1500: p-value=0.003), and self-reporting excellent health (p-value=0.024) was found to be protective for drug-related hospitalisation. Having tertiary education in the HIV−ve cohort (p-value=0.033) and having university or postgraduate level of education in the HIV+ve cohort (p-value 0.001) compared to 10 years or less of high-school was also found to be protective for drug-related hospitalisation.

Discussion

We determined the incidence of drug-related hospitalisation over approximately 6–7 years of follow-up among HIV+ve and HIV−ve GBM recruited in Sydney, Australia and found an increased number of drug-related hospitalisations in both cohorts compared to the general Australian male population. A significant proportion of these were for stimulant use and cannabinoids use in both cohorts. We found a significantly greater incidence of drug-related hospitalisation in the HIV+ve compared to the HIV−ve group. There was also significantly more hospitalisation for opioid use and a greater degree of injecting drug use in the HIV+ve cohort.

We further examined patterns of drug use as a risk factor for drug-related hospitalisations in HIV−ve and HIV+ve GBM. There were high levels of self-reported cannabis use, amyl nitrate use, MDMA and other MDA use and meth/amphetamine use in both cohorts. This accords with illicit drug use data published in other cohorts of men who have sex with men globally (2, 39–41). There was also significant polydrug use in both cohorts, with participants reporting using 2 or more drugs weekly in around 10% of interviews. There was evidence in both cohorts that periods of cannabis use, both low use and high use, periods of polydrug use and periods of injecting drug use were associated with a greater degree of drug-related morbidity. There was also evidence that consistent meth/amphetamine users and both inconsistent and consistent injecting drugs users were more likely to be hospitalised than nonusers. While heroin and other opiate use and downers, rohypnol and ketamine use was low in both cohorts, there was some indication that both were associated with drug-related hospitalisation.

Other significant risk factors for drug-related hospitalisation indicate the likelihood of comorbid drug and mental health issues in drug using GBM hospitalised for drug use. Both a higher Kessler distress score and having previously sought mental health counselling were found to be significantly associated with drug-related hospitalisation. Rates of drug-related hospitalisation significantly reduced in the over 55 year age group suggesting that treatment services for drug-using GBM should be targeting towards this under 55 year age bracket. Unsurprisingly, exposure to hepatitis C was positively correlated with drug-related hospitalisation. There is evidence that both injecting drug use and non-injecting illicit drug use are associated with hepatitis C seroconversion (42). Smoking was also shown to be positively correlated with drug-related hospitalisation which is supported by previous literature which has linked smoking and illicit drug use (5). There was also some indication that being more sexually adventurous was associated with drug-related hospitalisation. This perhaps is unsurprising considering the link between drug use and sexual activity previously described in cohorts of men who have sex with men (43, 44). However, it does suggest support services should be aimed at GBM who have a high number of sexual partners (>6 in 6 months) and who also engage in other risky drug use behaviour to reduce drug-related harm. Interestingly being in a HIV-serodiscordant relationship compared to a concordant relationship was found to be positively associated with drug-related hospitalisation. Possibly the heightened stressors associated with being in a HIV-serodiscordant relationship contribute to engagement in risky drug use, further research is required to more critically examine this association.

There are some limitations of this research that must be taken into consideration. Hospitalisation and drug use were assessed from 2000 to 2007 and may not reflect all of current drug use practices in GBM. It is a cohort of GBM recruited in Sydney, Australia and thus the findings may not be applicable in other settings. The patterns of drug use found however very much mimic those shown in other cohorts of MSM globally (13, 42). While the method of recruitment in both cohorts was similar and is a strength of the study, the representativeness of the cohort to the wider HIV+ve and HIV−ve homosexual population in Australia is not known. The HIV+ve cohort was largely comprised of GBM who had lived long-term with HIV and who may not represent more recently diagnosed HIV+ve gay men. Representative samples of gay and other homosexually active men are difficult to attain (33). Despite these limitations, investigation of baseline characteristics in both cohorts showed similarity to those described in other cohorts of HIV+ve and HIV−ve GBM in Australia (6). Also when we looked at baseline drug use in our HIV−ve and HIV+ve cohort it was consistent with the prevalence of drug use reported in the same time period in Sydney GCPS (15, 45).

A further limitation is that we were unable to exclude gay and bisexual men from the general population estimates. However, the proportion of men identifying as gay or bisexual is low in the general Australian male population (1.6 and 0.9% respectively) (46) and would only have biased our relative estimates towards the null. Rates of admissions with a secondary drug-related diagnosis code were also unavailable in the general Australian male population which precluded this comparison. Furthermore participant drug use was measured at annual interviews, which could have varied across the calendar year. Self-reported drug use in the cohorts could also have been affected by social desirability bias. Researchers in the United States, however, found self-reports about recent marijuana use and cocaine use in a cohort of MSM were valid (47). There was a strong indication that participants lost to study follow-up who thus had missing self-reported drug use data were those at greatest risk for drug-related hospitalisation. Fortunately record linkage ensured that the outcomes of participants with missing data could be measured.

A major advantage of this study was the availability of both detailed self-report drug use data and cause-specific hospital admission data in GBM. These enabled the enumeration of drug-related harms in GBM and the investigation of what types of drug use patterns predicted these harms. Drug practices in GBM have long been a public health concern yet to-date limited research has estimated the impact of heightened drug use in GBM on health outcomes unrelated to HIV acquisition or transmission. This study suggests that GBM do suffer greater drug-related morbidity as a result of their drug use. However, due to high prevalence of illicit drug use in GBM, support and interventions in this area need to be tailored to those at greatest risk of harm. Our findings suggest that the greatest at-risk group for drug-related harm are GBM who self-report risky drug use, particularly cannabis and meth/amphetamine use, injecting drug use and weekly single substance or polydrug use, and who are under 55 years old, have comorbid mental health issues, have poorer socioeconomic indicators, consume alcohol at high levels and/or are engaging in high risk sexual activity.

Appendix A

Comparison of patient characteristics by willingness to consent to record linkage

		HIV−ve Cohort			HIV+ve Cohort
Risk Factor	Category	Consented to Linkage	Did not consent to linkage	P- value	Consented to Linkage	Did not consent to linkage	P- value
Ethnicity	Anglo-Australian/Anglo-Celtic	324 (95.3)	16 (4.7)	0.199	430 (79.8)	109 (20.2)	0.344
	Other	1023 (96.8)	34 (3.2)		112 (76.2)	35 (23.8)
Highest level of education	10 years of high-school or less	149 (99.3)	1 (0.7)	0.195	131 (78.4)	36 (21.6)	0.826
	Completed high school	220(96.9)	7 (3.1)		96 (76.2)	30 (23.81)
	Tertiary education	295 (95.8)	13 (4.2)		137 (80.6)	33 (19.4)
	University or Postgraduate	710 (96.0)	30 (4.1)		203 (77.8)	58 (22.2)
Employment	Unemployed	1286 (96.6)	46 (3.5)	0.334	298 (80.8)	71 (19.2)	0.178
	Employed	88 (94.6)	5 (5.4)		224 (76.5)	69 (23.6)
Weekly Income	<500	278 (95.5)	13 (4.5)	0.123	274 (75.9)	87 (24.1)	0.133
	500–999	520 (97.7)	12(2.3)		143 (79.4)	37 (20.6)
	1000–1499	302 (96.8)	10 (3.2)		80 (87.0)	12 (13.0)
	>1500	252 (96.7)	14 (5.3)		57 (78.7)	14 (19.7)
Sexual identity	Gay, queer or homosexual	1308 (96.3)	50 (3.7)	0.630	525 (78.5)	144 (21.5)	0.644
	Bisexual	52 (98.1)	1(1.9)		18 (85.7)	3 (14.3)
	Other	12 (100)	0 (0)		24 (75.0)	8 (25.0)
Self-reported as having had any STI (other than HIV)	No	459 (97.0)	14 (3.0)	0.379	134 (77.0)	40 (23.0)	0.222
	Yes	917 (96.1)	37 (3.9)		87 (70.7)	36 (29.3)
Had antibodies to Hepatitis C	Never	1191 (96.6)	42 (3.4)	0.521	468 (77.4)	137 (22.6)	0.197
	Yes	54 (98.2)	1 (1.8)		71 (83.5)	14 (16.5)
Self-Reported General Health	poor/fair	119 (96.0)	5 (4.0)	0.087	25 (73.5)	9 (26.5)	0.685
	good	439 (97.6)	11 (2.4)		121 (78.6)	33 (21.4)
	very good	552 (95.0)	29 (5.0)		290 (79.7)	74 (20.3)
	excellent	265 (97.8)	6 (2.2)		125 (78.2)	40 (24.2)
Kessler 6 Scale(Psychological Distress)	low	1029 (96.6)	36 (3.4)	0.760	215 (79.3)	56 (20.7)	0.864
	moderate	324 (95.9)	14 (4.1)		297 (78.8)	80 (21.2)
	high	19 (95.0)	1 (5.0)		512 (79.0)	136 (21.0)
Daily Smoker	No	936 (96.7)	32 (3.3)	0.428	257 (77.9)	73 (22.1)	0.716
	Yes	440 (95.9)	19 (4.1)		301 (79.0)	80 (21.0)
In a relationship / have a regular male sex partner	Yes	765 (96.1)	31 (3.9)	0.467	265 (77.7)	46 (22.3)	0.776
	No	610 (96.8)	20 (3.2)		115 (75.7)	37 (24.3)
Sero-status of regular relationship (concordant, discordant, unknown etc.)	no regular partner	646 (96.4)	24 (3.6)	0.451	265 (77.7)	76 (22.3)	0.776
	concordant	496 (95.9)	21 (4.1)		115 (75.7)	37 (24.3)
	discordant	136 (96.5)	5(3.6)		143 (80.3)	35 (19.7)
	unknown	62 (100)	0 (0)		6 (75.0)	2 (25.0)
Number of male partners (last 6 months)	0	22 (100)	0 (0)	0.190	11 (84.6)	2 (15.4)	0.621
	1	244(97.6)	6 (2.4)		35 (87.5)	5 (12.5)
	2–5	356 (98.1)	7 (1.9)		54 (78.3)	15 (21.7)
	6–10	240 (94.1)	15 (5.9)		46 (88.5)	6 (11.5)
	11–50	398 (95.7)	18 (4.3)		63 (86.3)	10 (13.7)
	51–200	103 (96.3)	4 (3.7)		22 (91.7)	2 (8.3)
	>200	12 (92.3)	1 (7.7)		3 (75.0)	1 (25.0)
Frequency of exercise in a week	low	73 (97.3)	2 (2.7)	0.634	-	-
	moderate	447 (97.0)	14 (3.0)		-	-
	active	78+9 (96.2)	31 (3.8)		-	-
	very active	64 (94.1)	4 (5.9)		-	-
Ever reported as experiencing discrimination or harassment	No	607 (97.3)	17 (2.7)	0.128	-	-
	Yes	769 (95.8)	34 (4.2)		-	-
Number of alcoholic drinks consumed in one sitting (on average)	Non-drinker	57 (93.4)	4(6.6)	0.253	65 (78.3)	18 (21.7)	0.633
	1 or 2	415 (95.2)	21 (4.8)		206 (78.9)	55 (21.1)
	3 or 4	568 (96.9)	18 (3.1)		188 (21.3)	51 (21.3)
	5 to 8	253 (97.3)	7 (2.7)		68 (80.0)	17 (20.0)
	>9	67 (98.5)	1 (1.5)		31 (14.1)	14 (31.1)
Number of gay/homosexual male friends	none	5 (100)	0 (0)	0.359	-	-
	a few	119 (98.4)	2 (1.7)		-	-
	some	373 (96.1)	15 (3.9)		-	-
	most	816 (96.0)	34 (4.0)		-	-
	all	63 (100)	0 (0)		-	-