Transitions in the AUD status, though not associated with changes in physical functioning, were strongly associated with changes in mental/psychological functioning—driven primarily by changes in the SF-12 domain representing mental health. Although many of our initial expectations were supported by the results of this study, a number of findings were unexpected. First, there were no significant decreases in functioning associated with the onset of alcohol abuse only, suggesting that abuse might more appropriately be considered a maladaptive behavioral pattern than a disorder characterized by physical and psychological disability. In addition, whereas NR from abuse was associated with a small and marginally significant degree of improvement in mental functioning, the opposite was true for AR from abuse. The worsened pain, vitality, social functioning and limitation in activities due to emotional problems that were associated (albeit marginally) with AR from abuse signal the need for a better understanding of the factors that might lead to this AUD transition.
The net decline in mental/psychological functioning associated with the development of dependence was substantially reduced by controlling for the excess baseline levels of mental and physical problems experienced by individuals who developed dependence. This suggests that much of the reduction in QOL associated with becoming dependent reflects the worsening of pre-existing mental conditions. Although these findings preclude causal attribution, they do suggest that treatment and prevention of alcohol dependence should address the issues surrounding comorbid or precursor mood and anxiety disorders.
The improvements in QOL associated with remission from dependence in this study were smaller than those reported in some previous studies of treatment samples (e.g. Garg et al
; Feeney et al
; Morgan et al
). This discrepancy is not surprising. Treatment samples are likely to be selective not only of individuals with high baseline levels of severity and impairment (thus with a great deal of room for improvement in QOL) but also of those who through their entry into treatment have indicated a commitment or willingness to change their drinking patterns. Moreover, the follow-up interval for these treatment studies was in the range of 3 months to a year, shorter than the period examined in this study, and they did not adjust for the full range of covariates examined in this study. Additional analyses of the longitudinal NESARC sample could examine changes among individuals who had received treatment in the follow-up interval or include treatment/12-step participation as a covariate in the models used to assess the magnitude and significance of changes in QOL.
The one transition that was strongly associated with a change in physical functioning in this study was drinking cessation among individuals who did not meet the criteria for an AUD at baseline. This change in drinking behavior, although not a transition in the AUD status per se
, was strongly associated with decreases in numerous aspects of physical and mental/psychological functioning, and thus is consistent with the ‘sick quitter’ hypothesis, i.e. the argument that individuals who stop drinking are selective of those with adverse physical and mental conditions and, accordingly, that inclusion of former drinkers with lifetime abstainers may bias tests of the health effects of moderate drinking (Shaper et al
; Fillmore et al
). However, the exclusion of individuals who had already stopped drinking at baseline means that only a few of the sick quitters would be observed in this analysis, so it should not be considered a formal test of that hypothesis. In addition, because changes in the AUD status and QOL were measured in the same time period, the analysis cannot rule out the possibility that the decreases were the result rather than the cause of the drinking cessation.
One of the unique contributions of this study was its inclusion of changes in QOL associated with developing alcohol dependence, a subject that to our knowledge has not received attention in the previous literature. In association with the onset of dependence, we found a net change in the NBMCS score for mental/psychological functioning of −2.48 for individuals with no baseline AUD and of −3.48 for those with abuse only at baseline. To put these changes into perspective, they exceeded the net changes associated both with starting to smoke (−1.54 and −1.05, respectively) and with starting to use illicit drugs (−1.70 and −1.84, respectively), as estimated in the models from which the QOL changes were derived (full models not shown).
In a study of primary care patients that used measures of alcohol dependence and QOL comparable to those used in this study (Volk et al
), the cross-sectional difference in the mean NBMCS score between individuals with no AUD and those with dependence was more than three times as great as the longitudinal decline in NBMCS associated with developing alcohol dependence in the current study (−8.5 versus −2.48). Several factors may account for this discrepancy. First, the scores reported by Volk et al
. adjusted for age, sex, race/ethnicity and cigarette use, but they did not control for other mental and medical conditions, nor for marital status and education. Prior to these adjustments, the decrease in QOL found in this study was greater, although still only about half as large as the difference reported by Volk et al
. More importantly, individuals who developed alcohol dependence during the 3-year follow-up interval between Waves 1 and 2 of the NESARC represent individuals in the early stages of dependence, among whom consumption levels were still relatively low (Dawson et al
). With a longer follow-up interval, dependence likely would have increased in severity and thus have been associated with a greater decline in mental and psychological functioning. Indeed, when the unadjusted baseline NBMCS scores for individuals with dependence are compared with those for individuals with no AUD (averaging across the appropriate transition categories in Table ), it is evident that these cross-sectional differences at baseline are greater than the longitudinal changes in QOL associated with the onset of dependence.
This study has a number of methodological strengths that increase our confidence in the validity of these findings. Because of the longitudinal design of the NESARC, both transitions in the AUD status and changes in QOL were measured directly by comparison of past-year data at two points in time. Accordingly, they were not subject to the level of recall error that might bias retrospective reports of AUD and QOL, nor were they subject to potential bias arising from asking respondents to directly report subjective changes in QOL. In addition, the past-year AUD and QOL measures, as well as the measures used as model covariates, have demonstrated high levels of reliability in test–retest and validity studies.
Despite these methodological strengths, there are limitations to this analysis. First, because this study conditioned its outcome measure of change in QOL upon baseline QOL, it shares limitations common to lagged dependent variable (LDV) models, including possible bias resulting from a measurement error in the outcome variable and from insufficient control for all possible time-invariant confounders (Norström, 2008
). One method for assessing the impact of these limitations is comparison of the LDV model parameters with those from within-subjects change in score (CS) models (Johnson, 2005
). Because CS models have their own limitations, including statistical inefficiency and failure to address genuine causal relationships between the outcome measures at times 1 and 2, including possible ceiling effects, they should not be considered the gold standard for validating this study's results; however, comparison of the two models may reveal differences that can enrich the interpretation of this study's findings.
When the changes in QOL examined in this analysis were reanalyzed using CS models, many of the marginally significant changes in Tables 2–4 became nonsignificant. Although this is to be expected given the relative inefficiency of the CS models, it underscores the need for caution in interpreting these marginally significant findings. The great majority of significant changes in QOL remained significant or marginally significant, and their magnitudes were within the sampling error of those presented in Tables 2–4, varying in both directions (further details of the model comparison available upon request).
Another possible limitation of this study is that it did not consider the potentially confounding effects of changes over the follow-up period in all model covariates. Changes in mental disorders and physical conditions were excluded by design, because they were considered intervening variables, but changes in the marital or educational status are more problematic. They might represent intervening variables too, but they could also be seen as potential confounders that could be controlled. It would be interesting to see what effect it would have on the results of this study to control for marital and educational transitions.
Finally, the design of this analysis precluded causal inferences with respect to the associations between changes in QOL and AUD status, as these occurred within the same 3-year time frame and we have no information as to their temporal ordering within that period. This is important to bear in mind when considering possible explanations for the decreased psychological functioning that was observed in association with AR from alcohol abuse, and it extends to any conclusions that onset and offset of AUDs lead to (as opposed to result from) changes in psychological functioning. Moreover, our inability to study changes in QOL among individuals who died or became incapacitated over the follow-up interval limits our ability to draw inferences as to the magnitude of the changes or their associations with the course of AUD in the full Wave 1 sample. However, because the analysis examined changes conditional upon baseline mental and physical functioning, any selectivity in terms of sample attrition with respect to baseline QOL should have little impact on the findings.
Despite these limitations, this study is an important complement to earlier studies of clinical and patient samples. It demonstrates that even among the less severely dependent members of the general population, many of whom may not have realized that they had recently met the criteria for an AUD, the development of alcohol dependence is accompanied by significant reductions in mental and psychological functioning. Its findings suggest that AR and NR are associated with comparable levels of improvement in the general population, and that even partial remission is associated with some improvement. By documenting the wide range of psychological and general health harms that were at least marginally associated with developing dependence, it illustrates some of the potentially important implications of preventing and treating of alcohol dependence—implications that need to be verified using a design permitting causal attribution. A greater understanding of this process might be gained through future analyses examining changes in QOL as a function of the first incidence of specific dependence criteria or as a function of the number of new symptoms experienced.