Predicting Access to Medical Care Among Out of Treatment Drug Users

1
Predicting Access to Medical Care Among Out of
Treatment Drug Users

• Catherine Woodstock Striley, MSW, LCSW, Ph.D.,
  NIMH Postdoctoral Trainee
• Linda B. Cottler, Ph.D., Professor
• Washington University
• School of Medicine
• Department of Psychiatry

2

• Abstract
• To test an Andersen-based model of access to
  health care in a drug using street-recruited
  sample including (among other variables tested)
  attitudes and beliefs in the predisposing block
  of variables legal and illegal income in the
  enabling block and drug, mental health, and
  sexually transmitted diseases in the need block
  of variables.
• Demographics, health attitudes/beliefs,
  insurance, income, STDs and HIV, injection drug
  use, smoking and other behaviors were assessed
  using the Risk Behavior Assessment. The
  Diagnostic Interview Schedule was used to elicit
  mental health symptoms and diagnoses, and the
  Composite International Diagnostic Interview -
  Substance Abuse Module was used to elicit drug
  abuse and dependence. A 3 month follow-up
  ascertained the number of health care services
  received since baseline. An Andersen-based
  model of access was tested in multiple regression
  and in negative binomial regression to account
  for the dependent count variable.
• A sample of 1224 street-recruited drug users in
  St. Louis City were enrolled in an NIDA-funded
  two arm HIV intervention (EachOneTeachOne), which
  randomized persons into standard of care versus
  an enhanced, peer-delivered educational program.
  The sample was predominately African American
  (90) and male (61). Age ranged from 18 to 70,
  with a median age of 39.
• Using both regression with backward stepwise
  elimination and negative binomial regression, the
  model explained 9 of the variance in access.
  Being a woman, older age, believing no one could
  help, having insurance, exchanging drugs for sex,
  symptoms of depression, and meeting criteria for
  opiate tolerance, increased access. If the
  person endorsed reducing important activities,
  another criterion for opiate dependence, this
  decreased access.

3

• Many of the variables found to predict access to
  medical care have been found in other studies of
  non-substance abusers, such as female gender,
  depression, and having insurance. However, that
  sex-traders were more likely to have access to
  medical care was surprising, and needs further
  clarification. Substance use has also been found
  to affect access in other studies. Particular
  diagnostic criteria for opiate dependence are
  important to access, and others are not. Greater
  tolerance probably makes it more likely that
  substance users can still negotiate the system
  requirements for access. Reducing activities due
  to drug use may include reducing the type of
  activities specifically needed to negotiate
  access.
• Individual level variables alone are not
  particularly predictive of access in this
  population. Special paths to access for drug
  users may need to be established. Impairment and
  symptoms may interact with system-level barriers
  to care, such as needing to schedule appointments
  in advance, acting together as a barrier to
  medical care. It is important to go beyond
  thinking about drug use alone as an impediment to
  care, and to develop programs based on specific
  problems caused by drug use.

4
Acknowledgements

• Funded by
• NIDA 501-DA08324, L.B. Cottler, PI
• NIMH training grant 5T32-MH17104, L.B.Cottler,
  Director

5
Problem

• Intravenous drug users have high need for medical
  care
• but, they have reduced access to that care
• and more inappropriate access through emergency
  rooms
• Few have studied access in street-recruited drug
  using populations
• Unclear what effect individual mental health and
  substance use habits have on medical care access
  for out of treatment users

6
Aim

• To test an Andersen-like model of predisposing,
  enabling, and need variables to predict realized
  access or use of health care services in a
  population of out of treatment drug users
• that includes mental health problems, substance
  dependence, and enrollment in an HIV enhanced
  prevention project on realized access to health
  services

7
Theoretical Access Model

• Need Variables
• Depression symptoms
• Antisocial personality disorder symptoms
• Smokes
• Injection drug user
• Has an STD
• Has AIDS/ARC
• Has exchanged drugs for sex
• Has exchanged sex for drugs
• Meets criteria for DSM-IV inhalant dependence,
  opiate dependence, cocaine dependence 1-7

Access to care
8
Methods

• EachOneTeachOne
• NIDA funded study
• Investigated the effectiveness of providing HIV
  education to chronic drug and intravenous drug
  users
• The baseline and 3 month follow-up interviews
  used the DIS, the CIDI-SAM, the NIDA RFA, and the
  Health Services Supplement

9
Sample

• 1220 drug users recruited from areas of St. Louis
  City with high drug abuse, crime and prostitution
  
• Store-front satellite health centers
  (HealthStreet) located in two of these
  neighborhoods
• Community Health Outreach Workers went out onto
  the streets in pairs to recruit the study sample.
  
• Recruited individuals received standard HIV pre-
  and post-test counseling and
• were randomized to that or an enhanced HIV
  prevention

10
Descriptive Statistics
11
Descriptive Statistics
12
Independent Variables

• Predisposing
• Gender of participant -- self-reported
• Race of participant -- African-American,
  Caucasian, or other race
• Age
• Education -- highest educational attainment
• Health beliefs and attitudes -- covered 13 of
  most salient possible barriers to care

13

• Enabling
• Health insurance -- included private, VA,
  Medicare, Medicaid, group, prepaid
• Income for last month (range)
• Enhanced HIV prevention program (intent to treat
  analysis)
• Received a checkup in 2 years before baseline

14

• Need
• Symptoms of depression and antisocial personality
  disorder
• Smoked cigarettes
• Has a reported STD
• Has exchanged drugs for sex
• Has exchanged sex for drugs
• Meets DSM IV specific dependence criteria
• for opiates
• for cocaine
• for inhalants (all criteria combined)

15
Dependent Variable

• Number health care visits of any kind in 3 months
  since base line
• A measure of realized access or service
  utilization

16
Analyses

• First, backward elimination in multiple
  regression using SAS PROC REG where predisposing,
  enabling and need variables were entered in
  blocks
• Then, confirmatory testing with binomial negative
  regression using SAS PROC GENMOD (Negative
  Binomial Regression) due to truncated and skewed
  nature of services utilization variable

17
Why Use Negative Binomial Regression?

• Data is discrete
• Data is truncated, with many scores at zero
• Unlike Poisson, estimates third parameter, the
  dispersion parameter
• Expresses contagion
• A "spell" of depression so services will clump
• Services are regular, not really random
• Larger the variance relative to mean, higher
  level of dispersion

18
Health Care Visits Comparing Mean (with CI) for
6 Months Baseline and Follow-up
19
Results Final Model

• Both multiple regression and negative binomial
  regression found the same variables predictive
• of the variance in service utilization or
  realized access was explained by
• gender (female) b.12
• age (increasing) b.11
• regular checkup last 2 years b.09
• having insurance b.16
• meeting depression criteria b.06
• has given drugs for sex b.10
• meets opiate dependence criterion for tolerance
  b.13
• meets opiate dependence criterion for reduced
  activities b -.13
• Standardized Betas are from multiple regression

20
NBR Results Increasing Visits by Significant
Variables
21
Conclusions

• Very little (9) of the service variance in this
  out-of-treatment drug using population can be
  explained with a traditional individual model of
  predisposing, enabling and need variables
• Individual health attitudes and beliefs did not
  contribute to the final model
• Like many access models, being a women, getting
  older, having insurance and a usual source of
  care, increase access
• Mental and behavior health problems contributed

22
Implications

• Models explaining access in this high need
  population need to include provider and
  environmental variables
• Many requirements for planned access to care may
  hinder drug-abusing populations from receiving
  care
• Research needs to include such variables as
  medical care agency policies provider attitudes
  and behavior and barriers between the call to
  access services and actual service