Validation of self-report utilization measures
We conducted validity analyses of the self-report utilization measures using a 7-year follow-up of this cohort. Using question wording similar to that used to ask about non-KP utilization, we asked about patients’ past-year use of KP ER visits, hospitalizations, and outpatient doctor/nurse practitioner visits and compared them to KP utilization databases. For ER visits and outpatient medical visits there was 90%, and 89% agreement, respectively, between the self-report and KP database measure of having any visit in the prior year, and the Kappa coefficients were .62 and .79, respectively. For having had a hospital visit there was 95% agreement and the Kappa was .62. All Kappa coefficients were considered indicative of substantial agreement (Landis and Koch, 1977
). For the questions about the number of ER visits, hospital days, and doctor/NP visits, the Pearson correlations between the self-report and database measures were .77, .71, and .60, respectively (all were p<0.0001).
describes characteristics of the study sample, including demographic factors, percentage of participants with SAMC and non-SAMC medical conditions, and substance use measures. Forty percent of the sample were alcohol dependent only, 31% were drug dependent but not alcohol dependent, and 18% were both alcohol and drug dependent. The large majority of the sample (58%) had multiple substances which they either used regularly for 5 or more years or for which they met dependence criteria (not shown). In addition, most of those who were dependent on alcohol used other drugs as well, and the same pattern of mixed use was observed for those who were dependent on other substances. Prevalent substances of dependence were alcohol (58%), stimulants (24%), marijuana (17%), narcotic analgesics (9%), and cocaine (9%) (not shown).
Demographics, Intervention Factors, and Life Context among Participants responding to 5-year Follow-up
The mean and median numbers of SAMCs per participant were 1.7 and 2.0, respectively; the minimum was zero and the maximum was eight (not shown).
Based on ASI legal problem severity composite items, 12% of participants were on probation or parole at baseline, 10% were awaiting charges, trial, or sentencing, 7% had been in jail in the month prior to baseline, and 7% participated in illegal activities for profit in the month prior to baseline (not shown). To address the possibility that our results were confounded by participants’ legal problems, we re-ran the models controlling for baseline ASI legal problem severity, but found no differences in the pattern of results.
A total of 338 of the 598 (57.5%) in the sample met criteria for remission in the year prior to 5-year follow-up. presents bivariate analyses of the demographic, life context, and intervention variables for the total sample and by remission status at 5 years. Those who remitted were more likely to be aged 50 and older and less likely to be aged 18–34 (p=.002) than aged 35 – 49, and had higher levels of medical problem severity (p=.023). They were marginally more likely to have been randomized to integrated care in the index CD treatment episode (p=.067), and had slightly lower levels of baseline ASI drug problem severity (p=.080), and slightly fewer friends who were heavy substance users (p=.088), although these relationships only approached statistical significance.
presents results of the final logistic regression models. In preliminary models, ethnicity, education, employment, income, the two social support variables, and coping did not approach significance (i.e., alpha <.20); thus, they were dropped from the final models. SAMC conditions did not predict remission, nor did other conditions. However, medical problem severity was associated with higher odds of remission (OR=2.0; 95% CI=1.14, 3.54). Assignment to integrated care in the original randomized trial was predictive of a higher odds of remission at 5 years (OR=1.48; 95% CI=1.04, 2.13).
Logistic Regression Model Predicting Remisson in Study Subjects responding to 5-year Follow-up
In post-hoc analyses, we also stratified the sample by age group and examined the model within each age category (not shown). SAMC predicted higher odds of remission among the 18–34 age group (N=228) (OR=2.19 p=.01), but there was no significant relationship in the 35–50 age group (N=313) or the 51 and older age group (N=57). We also examined cross-tabulations of remission by SAMC status within each age group. In younger individuals, 52% of those with SAMCs remitted compared to 35% of those without SAMCs (p=0.02). In those aged 34–50, 56% of those with SAMCs and 62% of those without SAMCs remitted. Due to the small cell size of those aged 51 and older, we had insufficient power (.203) to test the significance of this relationship, and cross-tabs showed that 70% of those with a SAMC were remitted compared to 80% of those without.
In post-hoc analyses we examined the effect of SAMC conditions by gender, because prior studies found that medical conditions predicted better substance use outcomes for men but not women (Green et al., 2004
; Romelsjo et al., 1991
). There was no relationship of SAMCs to remission for either women or men.
also presents model results in those with SAMCs (to test our second hypothesis). Among those with SAMCs, those having 2–10 visits in the 5 years after intake had almost three times the odds of remission at 5 years. As hypothesized, integrated care predicted higher odds of remission among those with SAMCs.
A post-hoc analysis used the same model and also controlled for number of ER visits and psychiatric visits, as well as whether or not the participant had CD readmissions. This model showed the same pattern of findings – those with 2–10 primary care visits had 3.1 times the odds of remission compared to those with 1 or 0 visits (95% CI = 1.28, 7.64) (not shown).
Finally, we conducted post-hoc analysis to assess whether psychiatric versus medical SAMCs influenced outcome (not shown). We examined the relationship of each to remission; neither measure was significantly related to remission in the full sample. However, among those who did not have psychiatric SAMCs, having medical SAMCs was significantly related to remission; in this subgroup (n=271) 64% of those with medical SAMCs were remitted at 5 years versus 51% of those without medical SAMCs (p=.047) (not shown).