Women made up 37% of the full sample (n = 783), and the mean age was 38.4 years. The sample was 57% white, 30% African American, and 5% Hispanic. Twenty-seven percent had at least some college education, and 41% had an annual income of $25,000 or higher.
Individual characteristics and service use over time were compared between those who entered public treatment programs and those entering private treatment programs at baseline. Relative to the private programs, public patients were younger, fewer were employed, fewer were married, a larger proportion were African American and fewer were white; they had incomes less than $25,000 a year and had fewer years of education (no college) (all < .05, ). There were no significant gender differences. ASI drug severity measures at baseline were higher in the public sample (median 0.106 vs. 0.073 for the public and private samples, respectively, p < 0.001); however, alcohol, psychiatric, and medical severity were similar.
Demographic Characteristics by Initial Treatment Type
A higher proportion of individuals in the private treatment sample had general medical visits over time (56% of the private sample vs. 45% of the public sample, p = 0.011; 50% vs. 41%, p = 0.020; 55% vs. 43%, p = 0.003; 68% vs. 58%, p = 0.023 at 1, 3, 5, and 7 years, respectively). Those in private treatment programs reported higher proportions of any psychiatric services in the prior 12 months throughout 7 years; however they were not significant. Rates of AOD treatment significantly differed between systems at 5 years, with 20% of the public sample reporting an AOD treatment readmission the year prior to 5 year follow-up; 11% of the private sample (p-value = 0.001). Those in public treatment programs also reported higher proportions of AA attendance in the prior 12 months than those in private programs (77% vs. 53%, p < 0.001; 55% vs. 39%, p < 0.001; 47% vs. 32%, p < 0.001; and 51% vs. 32%, p < 0.001 at 1 year, 3 years, 5 years, and 7 years, respectively). ()
Service Use by Initial Treatment Type
In the first step, in the private treatment sample, age and income ($25,000+, <$25,000) were significantly associated with remission over time; education level (some college, no college) was significant in the public treatment sample. Gender, ethnicity (white, other), employment status (employed, unemployed), and marriage status (married, single) were not significantly related to remission in either sample. Therefore, the final baseline models for both public and private initial treatment samples were limited to age, income and education level ( and ).
Odds Ratios from Four Nested Models of 30-day Remission in Private Sample (N=552)
Odds Ratios from Four Nested Models of 30-day Remission in Public Sample (N=231)
In the second step, all four severity measures (alcohol, drug, medical and psychiatric ASI) were added to the model. In the private sample having alcohol ASI scores of 0 at prior time points and having psychiatric ASI scores of 0 at prior time points significantly predicted remission over time. In the public sample both alcohol and psychiatric ASI scores of 0 at prior time points also significantly predicted remission over time.( and ).
Our third model included three components: regular general health visits, psychiatric services as needed determined by the psychiatric ASI score, and AOD treatment as needed determined by the drug and alcohol ASI scores. Based on this definition of chronic or continuing care, 13%, 15%, 13% and 15% in the private sample, and 14%, 15%, 14% and 14% in the public sample, received such care at 1, 3, 5, and 7 years, respectively; there were no significant differences in proportions receiving care between public and private samples over time ().
Multivariate mixed-effects logistic regression models examining the relationships between health care and remission over time (step 3) for the private sample indicated that the odds of remitting for those who received care were 1.50 times the odds for those who did not (adjusted OR = 1.50, 95% C.I. = 1.01, 2.21, p = 0.046) (). However, receiving health care was not associated with remission over time for those in the public sample ().
In the final step we added AA meeting attendance to the model. Individuals who attended at least one AA meeting in the year prior were more likely to be remitted in both public and private samples than those who did not (Private OR = 1.89, CI = 1.34, 2.67, p < 0.001; Public OR = 1.76, CI = (1.15, 2.69), p = 0.01). The health care variable was not significant in either sample after adding AA attendance ( and ).