When a given OTP displayed missing values for these variables, we imputed values by calculating predicted values using multiple regression analysis based on dummy variables for NDATSS survey wave and the observed values of these variables within the same OTP program in other waves, We compared results obtained with imputed and nonimputed data and found that imputation had no substantive impact on our point estimates but increased our sample size in pooled regression analysis from 339 to 370 programs.
Comparisons with client-level surveys indicate high validity of NDATSS data; therefore, this limitation should not bias the results.
More specifically, each wave of the NDATSS involved several steps.
We have conducted several studies of NDATSS data reliability and validity.
Further, to evaluate the validity of NDATSS methadone dose data, we compared them with data obtained from two other national studies, DSRS (described above), and a study by Kleber et al.
In one respect, we find a different pattern in the 2005 data from those observed in previous NDATSS survey waves.
For this outcome, opposition to syringe exchange and support for abstinence-based approaches were more important predictors in 2004/2005 than in previous NDATSS waves.
Further, the NDATSS survey data show that between 2000 and 2005 the proportion of patients who received below 40 mg/day actually increased, though not significantly, from 13.
The nature of the data from the NDATSS suggests caution in generalizing study results.
The 1990 NDATSS data were compared with the Drug Services Research Study (DSRS), an independent national study of treatment units and clients conducted in 1990 (Batten et al.
Some units of the 2000 NDATSS were not asked about women's services because they had been fielded a subset of study questions for another study.