Hospital utilization and costs in a cohort of injection drug users

Many injection drug users (IDUs) use emergency departments as a regular source of care because of the lack of accessible primary care services or their inability to attend set appointment times.1^,2^,3^,4 Poor health status among IDUs because of drug use, HIV infection or hepatitis C is common, and some IDUs present later in the course of an illness and thus require hospital admission.4^,5^,6^,7^,8 IDUs at all stages of HIV infection are admitted to hospital significantly more often than non-IDUs with HIV infection.9

Because hospital admissions among HIV-positive IDUs are largely attributable to complications of injection drug use10 that are not directly related to the HIV infection, it is unclear whether community-based HIV-negative IDUs have similar hospital utilization patterns as IDUs with early HIV disease. Other studies of health care use by IDUs have been in the setting of drug treatment or clinics and reflect a highly selected population.11^,12 We therefore conducted this study to determine the predictors of frequent emergency department use among community-based IDUs. We also examined the predictors of frequent hospital admission and reviewed the primary diagnosis for each encounter. To assess whether there was a difference in hospital utilization costs according to HIV status, we estimated the incremental costs incurred by IDUs with early stage disease relative to costs incurred by HIV-negative IDUs. With the rising prevalence of HIV infection among IDUs in many Canadian cities, including Vancouver,13^,14 these data may be crucial for planning and evaluating programs that provide appropriate health care services to reduce drug-related harm among active IDUs with or at risk of HIV infection.

Methods

We obtained informed consent from 598 participants of the Vancouver Injection Drug User Study (VIDUS) to review their medical records at St. Paul's Hospital, the main teaching hospital for IDUs and HIV care in Vancouver. The hospital's Committee on Human Experimentation approved the study. The research design and methods of the VIDUS have been previously described;13 the primary aim was to determine the risk practices associated with HIV seroconversion. In brief, subjects were eligible to participate in the VIDUS if they had injected illicit drugs in the month before enrolment and resided in the Greater Vancouver region. Most participants (82%) were recruited through word of mouth and street outreach programs. Participants provided blood samples for HIV antibody testing and responded to an interviewer-administered questionnaire semi-annually. Responses from the baseline survey were used for our analyses. We reviewed medical records from St. Paul's Hospital for the 598 participants to determine the frequency of emergency department use and hospital admissions as well as the presenting problem for each emergency department visit and the primary diagnosis for each hospital admission between May 1, 1996, and Aug. 31, 1999.

Hospital utilization costs were estimated according to the inpatient resource-utilization profiles of a random sample of 64 participants who were admitted to hospital during the study period. Our cost estimates were for the first 105 admissions. For each admission, information was abstracted from the medical records, including the nursing ward, medications received, investigations, physicians' visits and length of stay. To distinguish between costs and charges,15 we estimated inpatient unit costs using a model of simultaneous allocation16 of all expenditures including overhead, opportunity cost of hospital resources as well as a 5% global depreciation of capital equipment.

Contingency tables were used to examine associations between frequent users and nonfrequent users of the emergency department and hospital by their HIV status and other sociodemographic and behavioural characteristics. We defined a priori frequent emergency department use as 3 or more visits and frequent hospital admission as 2 or more stays during the study period. We classified HIV status as seroprevalent (n = 166), seroconverted (for HIV seroconversions that occurred during the study period) (n = 51) and negative (n = 381). Unstable housing was defined as living in a single-occupancy room hotel, boarding room, hostel, transition house, jail or on the street in the 6 months before enrolment in the study.3^,13 The Wilcoxon rank-sum test was used to compare continuous variables.

To identify independent predictors of frequent emergency department use, variables significant at the 0.05 level in the univariate analysis were entered into logistic regression models in a stepwise, hierarchical fashion. In the final model, all relevant 2-way interactions were considered. This procedure was also performed to identify predictors of frequent hospital admission. We also compared self-reported hospital admissions in the 6 months before entry into the study and at follow-up (1999) by HIV status to assess the extent to which HIV-negative IDUs could have been admitted to hospital elsewhere.

To estimate the incremental difference in hospital utilization costs between IDUs with early HIV infection and HIV-negative IDUs, the average daily cost was multiplied by the median length of stay and then the annualized frequency of hospital admission per person by HIV status.17 We performed 2-way sensitivity analyses using the lower and upper limits of the 95% confidence interval (CI) for the hospital utilization costs and simultaneously varying the length of stay by calculating the difference in length of stay by HIV status using the 25th and 75th percentiles.

Results

Interpretation

We found that the HIV-positive IDUs visited the emergency department and were admitted to hospital substantially more often than the HIV-negative IDUs. This may have been due to continued high-risk injection behaviours, particularly among the IDUs who became HIV positive during the study period. Our annualized frequency of emergency department use among the HIV-negative IDUs was higher than the estimate reported by French and associates4 (1.4 v. 0.78); the same was true for the annualized frequency of hospital admissions (0.52 v. 0.32). These differences may have been due to differences in population, health care systems and ascertainment of health care use (self-report versus actual).

Our finding that female sex was an independent predictor of frequent hospital admission is consistent with previous findings.6^,18^,19 The recent HIV Costs and Services Utilization Study in the United States reported that the suboptimal pattern of care among women and IDUs was largely mitigated by adjusting for insurance coverage and race or ethnic background.20 Comparisons with US findings are difficult given differences in the health care systems and the demographic composition of the study populations. The association between the use of cocaine and unstable housing with frequent emergency department use has been noted elsewhere.21^,22^,23^,24

The predominant reasons for hospital admission (pneumonia and soft-tissue infections) in our study are directly and indirectly related to needle use and highlight the importance of counselling and providing the tools necessary to practise safe injection techniques. In a study by Stein,10 HIV-positive IDUs were admitted to an urban hospital primarily because of injection-related complications. In another study, involving IDUs in drug treatment, those with early HIV infection used more ambulatory and inpatient services than HIV-negative IDUs.12 An increased biological susceptibility to bacterial infections8^,25 or poor hygiene and high-risk drug injection practices13^,26 are possible explanations for the higher frequency of hospital admissions among HIV-positive IDUs.

Our study had several limitations. First, we may have underestimated the emergency department and hospital use because the participants may have received care from other hospitals. Second, we did not capture outpatient clinic visits. Third, we used a relatively small sample of IDUs to estimate the hospital utilization costs. Finally, the study participants may not be representative of all IDUs, because those in the lowest socioeconomic group may have been overrepresented in our study population.

Our cost analysis builds on previous work in that we examined actual hospital costs for IDUs by assigning fully allocated costs of actual resource use. In other studies costs were derived from charge data,4^,10^,27 or aggregated service utilization costs were used to calculate costs per patient-year.28^,29 The incremental hospital utilization costs incurred by the HIV-positive IDUs relative to the costs incurred by the HIV-negative IDUs were $1752 per person; this translates into an additional $197 976 per year for hospital care for the 113 HIV-positive IDUs who were admitted to hospital during our study period.

A recent study reported high hospital use and costs among poor people in Toronto.30 Our results provide some reasons for their findings among the IDU subgroup. Much of the health care use was largely attributable to complications of injection drug use, regardless of HIV status. The use of sterile injecting equipment and safe injection practices might have prevented many of these conditions and, thus, reduced the need for emergency department visits or hospital admissions.31^,32 Programs that integrate harm reduction strategies with primary care and addiction treatment33^,34 should be considered by jurisdictions serving this vulnerable population.