We collected data from individuals with ATS disorders who presented as outpatients at drug counseling centers or as inpatients in hospital based drug treatment programs. Of the 133, 11 contacted drug-counseling centers (participation rate: 8.3%), 1 of the 13 contacted cessation therapy clinics specialized in crystal meth addiction (participation rate: 7.7%), 2 contacted addiction departments, and 1 forensic department of Regional Psychiatric Hospitals took part in the study. The Ethics Council of the University of Bamberg granted permission to conduct the study (July 28, 2014/June 06, 2016).

Only patients who were primarily addicted to illicit ATS were asked whether they were interested in participating in the study. The patients were briefed on the goals and contents and were informed about their right to withdraw from the study at any time. Participation was voluntary for all patients and anonymity was guaranteed. The participants were asked to complete questionnaires on their ATS use motives (AMQ), patterns of ATS use and demographic data. All outpatients received a 5€ remuneration for taking part in the study. The data were collected between January 2015 and July 2016.

The original sample consisted of 248 individuals with ATS disorders. Participants who did not use ATS as their primary drug of choice (N = 8, 3.2%) were excluded as were those who failed to answer more than half of the questions on ATS use motives (N = 6, 2.4%) and one participant who took part in the study on two occasions (0.4%). The remaining data included 233 participants. There were no missing values on gender or age. The full information maximum likelihood (FIML) option of the statistical software AMOS (41) was used to account for the remaining missing values (up to 2.1% per item). The analyzed data consisted of 173 men (74.2%) and 60 women (25.8%). The mean age was 31.1 years (SD = 7.9; age range 18–54). Drug counseling centers provided 113 data sets (48.5%), 49 data sets (21.0%) were provided by addiction departments of Regional Psychiatric Hospitals, 43 (18.5%) by a cessation therapy clinic, and 28 (12.0%) by a forensic department of a Regional Psychiatric Hospital.

The 12 motive items in the AMQ were adapted from the DMQ-R SF (36) with 3 items per dimension (enhancement, social, coping, and conformity). As the use motives seem to differ between the beginning of ATS use and, e.g., after the development of an ATS addiction (15, 38) the participants were asked to specify how often they used ATS for two periods of time: in their first month of use and their latest month of use. Possible answers were “never” (coded as 1), “seldom” (coded as 2), “sometimes” (coded as 3), “most of the time” (coded as 4), and “always” (coded as 5). The phrasing of all items is given in Table 2.

The patient questionnaire is based on the “Deutscher Kerndatensatz Klienten (KDS-K) zur Dokumentation im Bereich der Suchtkrankenhilfe” (German core data set of clients for documentation in the field of addiction treatment) (42) which has been applied in German drug-counseling centers and in inpatient drug treatment centers since 2007. It provides, amongst others, information about the age, gender (1 = female, 2 = male) and frequency of use in the first and the latest month of use, coded as monthly frequency, i.e., several times a day = 60, once a day = 30, three to six times a week = 19, one to two times a week = 6, two to three times a month = 3, once a month = 1, and less than once a month = 0.

We used confirmatory factor analysis (CFA) estimated in statistical software SPSS (AMOS Version 23), in order to confirm the assumed four-factor structure of the AMQ (a) for ATS use in the first month and ATS use in the latest month, separately. Since the formulation of two item pairs was more similar than the third item used to constitute a given latent factor (e.g., the two items “to get high” and “because it’s fun” as enhancement motives) the errors of the two item pairs were allowed to correlate as to compensate answer tendencies (43). We used the Comparative Fit Index (CFI), the Trucker Lewis Index (TLI), and the Root Mean Square Error of Approximation (RMSEA) to assess the model fit. The CFI and TLI compare the tested model with a null or independent model; recommended thresholds for a good model fit are a CFI and a TLI of ≥0.95 (44). The RMSEA represents the mean deviation of the data from the model per degree of freedom. A RMSEA of <0.1 (44) is desirable. To evaluate the internal consistencies of the AMQ, we used Cronbach’s Alpha. Values of 0.7 are considered as satisfactory, of 0.8 as good and of 0.9 as excellent (40).

To test the equivalence of the factor structure in different subgroups (b), we estimated nested models of confirmatory factor analyses with increasing constraints for gender and age (median split: younger users up to 30 years vs. older users over 30 years). First, we tested the configural invariance for the grouping variables gender and age which is supported if the unconstrained model has an acceptable fit to the data. Second, we tested the metric invariance for the grouping variables which requires equivalent factor loadings between the groups (λ-constrained model). Furthermore, we analyzed whether the fit indices remained equivalent when we constrained the variances in addition to the factor loadings (third model) and, finally, when we constrained the factor loadings, the variances and the correlations between the groups (fourth model) (43, 45). We compared the CFI, the TLI and the RMSEA of the fixed models with the fit indices of the freely estimated models.

Furthermore, independent sample t-tests conducted in SPSS 23 were used to test mean differences in the motive dimensions (c) between the gender and age subgroups. Therefore, we computed mean scores for the four motive dimensions.

To confirm the concurrent validity (d), we estimated a structural equation model with the latent four motive factors as independent variables and the ATS use frequencies of the first and the latest months of use as dependent variables. All analyses were conducted with SPSS AMOS Version 23 (41) by using the FIML option of AMOS.

For the first month of ATS use, the CFA provided highly significant factor loadings reaching from 0.48 to 0.95 (all p < 0.001; Table 2). The social factor had the lowest and the highest item loadings. The conformity factor had consistently high item loadings from λ = 0.71 to λ = 0.85 with an internal consistency of 0.85. All other internal consistencies were 0.72 or higher. Except between enhancement and social, all interfactor correlations were smaller than 0.20. There was a good fit to the data with a CFI of 0.945, a TLI of 0.907, and a RMSEA of 0.071.

The CFA of the latest month of use also provided highly significant factor loadings (all p < 0.001; Table 2) reaching from 0.54 to 0.88. The enhancement factor had the lowest item loading (λ = 0.54) with an internal consistency of α = 0.72. The highest item loading was found on the social factor (0.88), its internal consistency was α = 0.84. As in the first month of use, we observed the highest correlation between the enhancement and the social factor; all others were smaller. There was a good fit to the data with a CFI of 0.940, a TLI of 0.898 and a RMSEA of 0.077.

In order to test the four-factor structure of the four-factor model according to gender and age groups, we performed nested models of confirmatory factor analyses with increasing degrees of freedom for the subgroups. For the first month of ATS use configural invariance was given for all subgroups (unconstrained models) with CFI and TLI values above 0.9 and RMSEA below 0.1 (Table 3). Metric invariance was also given as the model fit only differs slightly between the unconstrained and λ-constrained models. Furthermore, the fit indices in the two subgroups did not change considerably when we constrained the factor loadings and variances (third model), or the factor loadings, variances and correlations (fourth model). All CFI and TLI values were above 0.9 and all RMSEA values below 0.1.

For the latest month of ATS use, configural invariance was given in the unconstrained model for the subgroup gender with CFI and TLI values above 0.9 and RMSEA below 0.1 and almost given for the subgroup age with a CFI value of 0.935, a TLI value of 0.890, and RMSEA below 0.1. All other tested models (λ-constrained model, plus variance constrained model, and plus correlations constrained model) displayed similar CFI values above 0.9, TLI values very close to 0.9, and RMSEA values below 0.1 in both subgroups gender and age (Table 3).

The results basically remained the same when the sample was split at age 25 and 35, respectively. The resulting model fit indices can be obtained from the authors on request.

In the first month of ATS use, men scored higher on positive reinforcement (enhancement and social motives). However, this difference was only significant for enhancement motives (Table 4). Women scored somewhat higher on negative reinforcement (coping and conformity motives) but the differences were not statistically significant. The subgroup age revealed no significant differences in the four motive dimensions. Younger ATS users scored slightly higher on enhancement, social, and conformity motives and lower in coping motives than older ATS users but none of these differences were significant.

In the latest month of use, men scored higher on positive reinforcement (enhancement and social motives) and on conformity motives with a significant difference in enhancement motives (Table 4). Women and men scored identically on coping motives. Also for the latest month of ATS use, the subgroup age revealed no significant differences in the four motive dimensions.

The estimation of the structural equation model revealed that higher coping motives are connected with higher frequency of ATS use for the first and the latest month of ATS use. A further positive association was found between social motives and frequency of ATS use for the latest month of use (Table 5). The explained variance was 12.6 and 15.6% for the first and the latest month of ATS use, respectively. In a final step, the link between the four motive factors and frequency of ATS use was adjusted for gender and age effects. However, the results (not shown but available from the authors on request) remained basically unchanged due to non-significant gender and age effects.

Since the study relies on the retrospective recall of use motives, memory bias constitutes an important limitation (48) which may partly explain the sufficient but not optimal model fit. The sample consisted of a heterogeneous group of patients which differed in the length of their abstinence: some presented themselves at drug counseling centers and were not abstinent at that time, some were in cessation therapy clinics and were abstinent from 0 to 6 months and others again were in the forensic department of a Regional Psychiatric Hospital and were abstinent from a few months to a few years. Depending on the recency of their last ATS use (49–51) and history of treatments (52), patients had different opportunities to deal with their relapses, to develop abstinence skills and to reflect on their motives of ATS use. Particularly the latter aspect may have had an effect on the answers given in the questionnaire. While mean levels of motive endorsement may be subject to recall bias, the four-factor classification of motives according to the valence and the source is not affected as studies in the field of alcohol use have shown (37, 45). We, thus, consider the recall bias as reasonable trade-off between feasibility of data collection and accuracy. In addition, some important aspects in the field of ATS use had not been considered in the 12-item AMQ: individuals with ATS disorders reported using ATS to improve sex (15–17), to enhance their cognitive performance (10, 17–21), to lose weight (17), to be able to stay awake, and as a self-medication (22) for symptoms of ADHD. Although the aforementioned reasons seem to be less common compared to items in the AMQ such as getting high or having fun (17), further studies should examine the importance of other amphetamine-specific motives of ATS use. Considering the intricate circumstances in this clinical setting—e.g., approaching ATS users and convincing therapists of the advantages of participating in the study—the sample size of 233 people can be regarded as sufficient. Larger sizes of clinical samples are difficult to achieve. This becomes apparent in the generally low response rates reported in the study design section which may have had an effect on the results. Moreover, it should be examined whether the AMQ, which was developed and validated in this study using a clinical sample in Southern Germany, can also be applied to non-clinical samples and in other countries. It would also be worthwhile to apply the AMQ to primary substances like illicit prescription stimulants or MDMA and related drugs. Furthermore, longitudinal studies are needed to be able to prospectively evaluate the development of ATS motives over time. Finally, we recommend including additional outcome variables in future research to extend the concurrent validity of the AMQ.

The application of the Motivational Model of Alcohol Use (23, 29) to ATS use is an important step toward a better understanding of why people use ATS particularly since most use the substance in spite of knowing the high potential for dependence and the multiple somatic and mental consequences (2–6). Moreover, as a theory-based instrument, the AMQ closes a gap in ATS research and may assist in generating targeted preventive and therapeutic treatments (26–28) as it can reveal recalled former and current motives of ATS use. For the assessment of use motives of patients, the AMQ can help understand why people begin and continue to use ATS and in a second step to personalize the treatment which has to manage the resulting changes related to ATS abstinence (17). In this study, the AMQ was shown to be succinct and viable in a clinical setting and has demonstrated its construct validity and reliability. It also constitutes a basis for future research, in particular to study changes in the motive structure between the first and latest month of ATS use.