This study has several limitations which need to be considered prior to interpreting the findings. The first limitation applies to the generalizability of the findings to all partners of substance users as those included in the present study were all attending treatment with the patient. Furthermore, although consistent with the literature, there was an under-representation of male partners with the majority of partners in the present study being female. Despite this the results from the PCSES do not seem to indicate that there are gender differences operating. In addition to this partners of drug-users were grossly under-represented in this sample given that most patients presented with an alcohol problem. The sample of partners and patients from Study 1 may have under-represented many other groups of partners and patients including :
-Those who had a drug and alcohol-using partner who was not able to attend outpatient treatment for a variety of reasons (e.g. not ready to change their substance use, work commitments, transport problems).
-Those whose partner does not want to be involved in the drug and alcohol-using patient’s treatment for any reason (e.g. stigma, relationship discord, fear of domestic violence).
-Those whose partner may be unable to attend treatment with the patient for a number of logistical reasons (e.g. work or child care commitments, transport problems).
-Those persons with a drug or alcohol problem who are able to change their drug and alcohol use on their own without requiring treatment (“self-changers”).
-Those patients and partners who did not reside in the Western Sydney Area geographic region and therefore lived under different socioeconomic levels.
-Those patients and partners who are unable to read or speak English or those who suffer from comorbid psychiatric disorders.
The selection bias evident in this study may therefore limit the generalizability of the findings. Given that patients were recruited from a public health service treatment facility it is possible that patients from this sample may have been more dependent on drugs and alcohol than community samples. Also, given that it was a requirement of the study that patients consent to their partner attending treatment with them to complete the assessment measures, the sample captured may have only included patients with supportive partners or those in more harmonious relationships with their partner. Furthermore, it is also possible that those partners who are not coping at all well with the patient’s problem may be unable to find the motivation or time to
participate in a study whereas those who are coping better may well have more time and better resources to attend a treatment facility to volunteer in a study.
The final sample size of 83 partners was not very large for conducting factor analysis, despite the data meeting all other prerequisites for factor analysis, such as Bartlett’s Test of Sphericity results, results of the Kaiser-Meyer-Olkin measure of sampling adequacy, lack of multicollinearity and singularity and absence of severely non- normally distributed data. It would have been more ideal to have a larger sample of partners to base the initial exploratory factor analysis on however recruiting
participants was rather difficult in the present study. This was in part due to the large number of partners who themselves displayed a drug and/or alcohol problem, and also due to patients missing scheduled appointments. This difficulty in recruiting drug and alcohol patients with their partners in research studies has been often cited, with Hurcom, Costello & Orford (1999) citing similar recruitment problems resulting in a final sample of only 29 female partners of patients in order to assess partner coping in their study. The sample size for the predictive validity section of the study was also rather low (N=42). Also problematic is the fact that reported drug or alcohol use in the follow-up telephone interview was not verified with the partner or another informant due to practical issues and time constraints. The failure to find evidence of the PCSES’s predictive validity from relapse rates may be due to the fact that relapse is likely to be due to a number of different factors other than partner self-efficacy. A more direct test of predictive validity would have been the prediction of subsequent coping behaviours that were directly measured by the PCSES rather than relapse per se, however, this was beyond the scope of the present study.
A number of measures used in Study 1, such as the Alcohol Abstinence Self-Efficacy Scale, the Drug Abstinence Self-Efficacy Scale and the Support Scale, have not themselves been published nor have their psychometric properties been studied in depth. The internal consistency, as assessed by the Cronbach alpha for each of
Abstinence Self-Efficacy Scale produced a coefficient alpha value of 0.81, the Drug Abstinence Self-Efficacy Scale produced a coefficient alpha value of 0.77 and the Support Scale produced a coefficient alpha value of 0.79. These particular scales were included in the present study due to their clinical utility and due to the often notable absence of alternative validated published scales which appear to measure similar constructs. However, future studies should be conducted to test their
usefulness in other settings and with other populations.
Evidence of the PCSES’s construct validity was based on correlational data therefore one cannot draw solid conclusions about causality. Although these variables appear to be closely related to partner self-efficacy one cannot firmly conclude how and why they are related from the correlational nature of the results. Future longitudinal studies and treatment outcome studies may provide more information about why partner self-efficacy and variables such as depression and anxiety for instance are related. Furthermore, the relationship between anxiety and depression and the PCSES is only one indication of construct validity. Future research needs to focus on gathering stronger evidence in support of the PCSES’s construct validity.
The present study failed to assess the PCSES’s discriminative validity, that is its ability to discriminate between partners of substance users and partners who do not have a substance using partner. Future research may wish to examine whether the PCSES has adequate discriminative validity. The present study did not directly examine the clinical utility of the scale.
Although most of the psychometric properties of the PCSES were good to excellent, internal consistency for Factor 2 “Reinforcing Abstinence” produced the worst results out of all the factors of the PCSES. The internal consistency Cronbach alpha
coefficient was found to be 0.634 for Factor 2 of the PCSES which is just considered acceptable. One possible way of improving a scale factor’s reliability and validity is by increasing the number of items that the scale assesses. Currently Factor 2 consists of 7 items which is not considered minimal, thus it may be that the items which make up this factor are not very applicable to most of the partners in the present sample. As a consequence future research may wish to consider including different items to this factor and ensuring that they are personally relevant to most partners of
substance users.
There is evidence to suggest that the PCSES has good face validity however participants are more likely to fake good their responses on scales which have good face validity. In the present study it is possible that partners may have faked good their answers as the scale was administered in a face-to-face setting. However the fact that very few respondents endorsed extremely high levels of confidence on the PCSES in this study suggests that social desirability bias may not have been a major issue with this particular sample. For research purposes it would have been ideal to administer a social desirability scale to determine if this particular sample of partners tended to respond in a socially desirable way on the PCSES. Unfortunately due to time constraints a social desirability scale was not added to the assessment package as it would have greatly increased the already lengthy testing session time.