Associations between affect and alcohol consumption in adults: an ecological momentary assessment study

ABSTRACT

Background: Consuming alcohol for coping with negative affect (NA) or enhancing positive affect (PA) may lead to risky drinking patterns. Previous research has yielded mixed findings regarding these affective drinking associations.

Objectives: To examine support for the self-medication and expectancy models of alcohol use in an adult community sample, by examining reciprocal associations between alcohol consumption and NA and PA within and between persons.

Methods: During seven consecutive days, 162 adults from the community (109 female) reported their affective experiences and alcohol consumption, following a signal contingent ecological momentary assessment protocol on their smartphones.

Results: Within-person daily NA preceding the first drinking event was associated with increased likelihood of same-day alcohol consumption. Within-person momentary NA was associated with a decrease in the amount of next-moment alcohol consumption. Within-person momentary PA was positively associated with likelihood of next-moment alcohol consumption. Between persons, levels of daily and momentary NA and PA were not associated with any index of alcohol consumption. The intercepts and slopes of NA were not significantly different before and after alcohol consumption. The intercept of PA was higher after alcohol consumption, whereas the slope of PA decreased after alcohol consumption.

Conclusion: In the current sample affective drinking was a within-person process (i.e. persons were sensitive to their varying levels of affect). Some support was found for the self-medication and expectancy models. People may drink for coping with NA, but may also be at risk for developing affective drinking patterns in response to PA.

KEYWORDS:

• Alcohol consumption
• affect
• smartphone
• ecological momentary assessment

Introduction

In 2016, approximately 9% of the Dutch adults were classified as excessive drinker, drinking more than 14 (women) and 21 (men) weekly alcoholic consumptions (1). Excessive drinking may lead to severe negative health outcomes such as several types of cancer, cardiovascular diseases, type 2 diabetes, liver diseases, mental disorders, and injuries (2–5). Therefore, identifying persons who are at risk for developing hazardous drinking patterns (i.e. between-person factors), and situations when a person is most at risk for hazardous drinking behavior (i.e. within-person factors) is important for public health.

Drinking to cope with negative affect (NA), or to enhance positive affect (PA), may lead to risky alcohol use patterns (6–10). The self-medication hypothesis (11) and expectancy theory (12) propose how people may resort to alcohol consumption to alleviate NA, or to augment PA. These models imply a temporal relationship, with affect preceding alcohol consumption, and propose how learned associations may develop from repeatedly experiencing alcohol’s sedative and stimulating properties (13–15). Accordingly, research has established a link between these emotional drinking motives and alcohol consumption. However the findings are mixed. For example, Strahan et al. (16) found that social anxiety was positively associated with alcohol consumption in men but not in women, and that positive alcohol outcome expectancies (i.e. relief from adversity) predicted levels of drinking in both genders. Swendsen et al. (17) found a positive relationship between nervous mood states and alcohol consumption. In another study (18) stress was associated with next-day alcohol consumption in men but not in women. In the opposite direction, alcohol consumption was associated with a decrease in next-day stress in both genders. This association was stronger in women than in men (18). Pedrelli et al. (19) found that low distress tolerance in female students was associated with heavy drinking to cope with NA. In male students, only the intensity of acute stress was associated with heavy drinking (19). Papachristou et al. (20) found an inverse direct effect of social anxiety on alcohol consumption. However, there was a positive indirect effect of social anxiety on alcohol consumption, which was mediated by positive alcohol outcome expectancies. Anthenien et al. (21) found an association between enhancement motives and alcohol consumption, which was mediated by positive alcohol outcome expectancies. In the same study (21), the association between coping motives and alcohol consumption was not mediated by alcohol outcome expectancies. In another study, positive alcohol outcome expectancies were associated with alcohol consumption in men, but not in women (22). Birch et al. (23) demonstrated that a positive mood induction activated reward-alcohol cognitions in enhancement-motivated drinkers. Conversely, a negative mood induction did not activate relief-alcohol cognitions in persons characterized as coping-motivated drinkers. The self-medication hypothesis (11) and expectancy theory (12) propose under which circumstances persons may consume alcohol (i.e. within-person processes). The aforementioned studies have largely examined differences between persons. Although these studies are valuable and informative, they may not always reflect the within-person dynamics as proposed by the motivational models of alcohol consumption (24,25).

Ecological momentary assessment (EMA) (26) is well suited for studying experiences and behavior in real time, in natural settings (27). With recurrent assessments throughout the day, EMA allows for capturing within-person processes (28) and for examining between-person effects. Nevertheless, EMA studies on affect-alcohol associations have also yielded mixed findings. Dvorak et al. (29) found that within persons, NA was not associated with alcohol consumption. According to Peacock et al. (30) within persons NA was associated with alcohol consumption depending on the time frames which were analyzed. NA preceding the first daily drinking event was not associated with alcohol consumption. However, there was a negative significant association between whole-day NA and same-day alcohol consumption (30), suggesting attenuation of NA after alcohol was consumed. Simons et al. (31) found that within persons, NA was associated with alcohol consumption, while between persons, NA was not associated with alcohol consumption. Three studies demonstrated that within persons, PA was positively associated with alcohol consumption (29,31,32). However, in two of these studies, between persons, PA was respectively not associated (31) and negatively associated (32) with alcohol consumption. These mixed findings with regard to types of affect (i.e. negative or positive) and person-level of analysis (i.e. within or between), make it difficult to draw conclusions.

Research on affect-related alcohol consumption has for a large part been conducted among college students or alcohol-dependent samples. Students often have relatively stable patterns of alcohol use (33), which may not reflect all emotionally triggered drinking behavior. Equally, drinking patterns of alcohol-dependent persons may not reflect affective drinking behavior. It is conceivable that alcohol-dependent persons may have relatively fixed drinking patterns, regardless of their emotional status. We wish to corroborate to the existing knowledge by examining associations between NA and PA, and two indices of alcohol consumption (likelihood and quantity) in the daily lives of an adult community sample, parsing within- and between-person effects.

In previous EMA research, associations between affect and alcohol consumption have mainly been studied with daily aggregates of affect, composed of repeated daily EMAs (29–32). In the current study, we wish to discriminate between daily aggregates of affect and momentary affective states. Daily aggregates encompass multiple affective states, perhaps resembling what is often conceptualized as mood (34). However, research has demonstrated that affect is a variable phenomenon (35). Insight into how momentary affective states are associated with alcohol consumption may provide further valuable information for developing ecological momentary interventions (EMIs). Therefore, in the current study, affect-alcohol associations will be examined at daily and momentary level.

We conducted a series of analyses to examine support for the self-medication and expectancy models in a community sample. We hypothesized that within and between persons, daily NA would be positively associated with likelihood and quantity of alcohol consumption. Additionally, we hypothesized that within and between persons, daily PA would be positively associated with likelihood and quantity of alcohol consumption. We examined this further in momentary data, hypothesizing that NA in the previous assessment interval (approximately 90 minutes) would be positively associated with next-interval likelihood and quantity of alcohol use. Additionally, we hypothesized that PA in the previous assessment interval (approximately 90 minutes) would be positively associated with next-interval likelihood and quantity of alcohol use. The self-medication and expectancy models propose responses based on learned associations. To examine further support for these models, analyses were conducted to test alcohol’s effect on NA and PA. We hypothesized that within days, NA parameters (i.e. intercepts and slopes) would significantly decrease after alcohol consumption, and PA parameters would significantly increase after alcohol consumption.

Material and methods

Sample

One hundred sixty-two participants, (109 female), 20 to 50 years old, were drawn from a comprehensive study in the Netherlands investigating snacking behavior in a community sample (36). Further details of the sample are provided in the supplementary materials.

Procedure

Respondents were recruited via social media (Facebook), institutional newsletters, and within the social networks of master thesis students from the Open University of the Netherlands. Participants filled out an online questionnaire and followed an EMA protocol, with ten daily audio prompts and a self-initiated evening assessment, during seven consecutive days. A detailed description of the procedure is provided in the supplementary materials.

Instruments

Online questionnaire

The online questionnaire consisted of assessments on psychosocial variables (not used in this study), and demographic and general questions (36).

Ecological momentary assessment measures

Affective states were assessed with items derived from the PANAS (37) and previous EMA studies (38–41). Statements about negative and positive adjectives were rated on 7-point Likert scales, ranging from 1 (not at all) to 7 (very). The NA measure was composed by averaging the items ‘I feel insecure’, ‘I feel anxious’, ‘I feel gloomy’, and ‘I feel nervous’ for every momentary assessment. PA was constructed by averaging ‘I feel cheerful’, ‘I feel relaxed’, ‘I feel content’, and ‘I feel happy’ for every momentary assessment. Daily NA and PA were computed by averaging momentary NA and PA per participant per day. Alcohol consumption was sampled at every momentary assessment and with the self- initiated evening assessment. The smartphone app provided a pre-structured food database derived from the Dutch Food Composition Database (42) with a search facility, and an option for open ended reporting. Momentary likelihood of alcohol consumption was a binomial coded variable (i.e. 0 = no alcohol, 1 = alcohol). These momentary registries were summed per participant per day, and dichotomized to compute daily likelihood of drinking. Alcoholic consumptions were coded into standard alcohol glasses. A Dutch standard alcohol glass contains approximately 10 grams of alcohol. Standard alcohol glasses were summed per participant per momentary assessment and per day, to obtain measures for respectively momentary and daily amount of alcohol consumption. Further details of the EMA measures are provided in the supplementary materials.

Statistical analyses

Multilevel models were analyzed in STATA 14 (StataCorp, 2015) (43). These models take the hierarchical structure of the data into account. The XTMELOGIT and XTMIXED commands were used for examining daily and momentary associations between NA and PA and respectively likelihood and quantity of alcohol consumption. Daily affect encompassed aggregates of all momentary assessments, prior to the first drinking event. Associations between momentary affects and likelihood of alcohol consumption were examined across all EMAs. The analyses with the momentary amount of alcohol consumption as the outcome measure were limited to the periods after alcohol consumption was initialized.

A piecewise multilevel analysis (44) with the first daily alcoholic consumption as the transition point was conducted in MLwiN 3.03 (45) to compare the intercepts and slopes of NA and PA before and after the first daily drinking event. For comparability, these analyses were repeated for non-drinking days, using the mean time of the first drink on drinking days as a transition point. All analyses were controlled for gender, age, level of education, BMI, and day of the week. The significance level was denoted at p < .05. Further details of the analyses are provided in the supplementary materials.

Results

Descriptives and correlations

Descriptives of the study variables are presented in Table 1. Correlations of the affective variables are displayed in Table 2.

Table 1. Descriptives for study variables

Variable		M	SD	Min.	Max.
Gender Female n(%) Male n(%)	109(67.28) 53(32.72)
Level of education* Low education level n(%) Middle education level n(%) High education level n(%)	5(3.09) 57(35.18) 100(61.73)
Age		36.07	9.23	20	50
BMI		24.45	3.91	17.18	39.56
Daily negative affect^a		1.43	0.60	1	4.70
Daily positive affect^a		5.35	0.92	2.54	7
Momentary negative affect^b		1.41	0.72	1	6
Momentary positive affect^b		5.40	1.13	1	7

Notes: *Low education level: none, elementary school, and lower general education. Middle education level: intermediate general and intermediate vocational education, higher general secondary or pre-university education. High education level: higher vocational or academic education. ^a Descriptives are based on aggregates per subject per day of momentary raw scores. ^b Descriptives are based on raw scores of the data.

Table 2. Bivariate correlations among affect variables

		Daily pre-consumption variables
	Variable	1	2	3	4
1	Within-person negative affect^a	–
2	Within-person positive affect^a	− 0.44***	–
3	Between-person negative affect^b	0.00	0.00	–
4	Between-person positive affect^b	0.00	0.00	− 0.62***	–
		Momentary variables
		1	2	3	4
1	Within-person negative affect^a	–
2	Within-person positive affect^a	− 0.31***	–
3	Between-person negative affect^b	0.00	0.00	–
4	Between-person positive affect^b	0.00	0.00	− 0.61***	–

Notes: ^a For within-person variables values were grand-mean centered and subsequently person-mean centered. ^b Between-person variables were person-means of the grand-mean centered data.

***p < .001.

Preliminary analyses

Compliance was calculated by dividing the number of retained assessments (9025) by the number of possible observations (162 participants x 11 daily assessments (10 daily signals plus 1 self initiated evening questionnaire) x 7 days = 12474), resulting in a 72% compliance rate. Compliance did not differ significantly on the second and third day compared to the first day of participation (b = −.09, SE = .19, p = .62; b = −.12, SE = .19, p = .53, respectively). There were significant reductions in compliance on the forth, fifth, sixth, and seventh day compared to the first day (b = −.51, SE = .19, p = .006; b = −.72, SE = .19, p < .001; b = −.77, SE = .19, p < .001; b = −.60, SE = .19, p = .002, respectively). Alcohol consumption was reported on 580 momentary assessments (6.43%), with an average of 2.15 standard alcohol glasses (SD = 1.52) per momentary assessment. A Dutch standard alcohol glass contains approximately 10 grams of alcohol (42,46–48). Alcohol was consumed on 434 (39%) of the 1119 assessment days. Across drinking days, an average of 2.88 (SD = 2.30) standard alcohol glasses was consumed. Across the entire assessment period (drinking and non-drinking days), with Tuesday as reference day (32), persons were more likely to consume alcohol on Friday, Saturday, and Sunday (OR = 3.48, CI[2.08–5.28], p < .001; OR = 5.54, CI[3.28–9.35], p < .001; OR = 3.28, CI[1.96–5.50], p < .001, respectively). Additionally, persons consumed more alcohol on Friday, Saturday, and Sunday, compared to Tuesday (b = .82, SE = .20, p < .001; b = 1.40, SE = .20, p < .001; b = .87, SE = .20, p < .001, respectively).

Associations between negative affect and alcohol consumption

Within persons, daily NA was positively associated with likelihood, but not quantity of alcohol consumption. Between persons, daily NA was not associated with alcohol consumption. Within persons, momentary NA was negatively associated with next-moment quantity, but not likelihood of alcohol consumption. Between persons, momentary NA was not associated with alcohol consumption. Results are displayed in Table 3.

Table 3. Associations between affect variables and alcohol consumption

	Negative affect
	Likelihood of alcohol use			Quantity of alcohol use
Within persons^a	OR	p	95% CI	B	p	SE
Daily negative affect	1.98	0.02	1.11, 3.53	.09	0.56	.15
Momentary negative affect	1.11	0.28	.92, 1.35	‒ .16	0.04	.08
Between persons^b
Daily negative affect	.89	0.62	.56, 1.42	.09	0.69	.22
Momentary negative affect	.99	0.92	.75, 1.30	.23	0.14	.15
	Positive affect
	Likelihood of alcohol use			Quantity of alcohol use
Within persons^a	OR	p	CI	B	p	SE
Daily positive affect	1.33	0.15	.90, 1.98	.21	0.08	.12
Momentary positive affect	1.21	0.01	1.05, 1.40	.02	0.82	.07
Between persons^b
Daily positive affect	1.00	0.99	.74, 1.35	.08	0.43	.10
Momentary positive affect	.96	0.62	.80, 1.14	.07	0.52	.10

Notes: ^a Within-person daily and momentary affect were person-mean centered to reflect respectively daily- and momentary-level deviations from a person’s typical level of daily and momentary affect. ^b Between-person daily and momentary affect are based on individuals’ means across all days and momentary assessments respectively. Between-person daily and momentary affect were grand-mean centered.

Associations between positive affect and alcohol consumption

Within and between persons, daily PA was not associated with alcohol consumption. Within persons, momentary PA was positively associated with next-moment likelihood, but not quantity of alcohol consumption. Between persons, momentary PA was not associated with alcohol consumption (Table 3).

Intercepts and slopes of affect before and after alcohol consumption

There were no significant differences between the intercepts and slopes of NA before and after alcohol consumption. The intercept of PA was significantly higher for post-drinking periods than for pre-drinking periods. There was a significant decrease in the slope of PA after alcohol was consumed (see Figure 1(a,b)).

Figure 1. (a) The intercepts (b = −.04, p = .21) and slopes (b = .35, p = .21) of negative affect were not significantly different before and after alcohol consumption. For comparability, the intercepts and slopes of negative affect on non-drinking days are displayed in the figure. The mean time of the first drink on drinking days was used as a transition point to split non-drinking days into “pre-drinking” and “post-drinking” periods. The dotted line indicates the time of the first alcoholic drink/transition point. Negative affect before the transition point was not significantly different from that after the transition point (intercepts_b = .09, p = .15; slopes_b = −2.18, p = .07) on non-drinking days. (b) The intercepts (b = .17, p = .001) and slopes (b = −1.33, p = <.001) of positive affect were significantly different before and after alcohol consumption. For comparability, the intercepts and slopes of positive affect on non-drinking days are displayed in the figure. The mean time of the first drink on drinking days was used as a transition point to split non-drinking days into “pre-drinking” and “post-drinking” periods. The dotted line indicates the time of the first alcoholic drink/transition point. Positive affect before the transition point was not significantly different from that after the transition point (intercepts_b = −.08, p = .42; slopes_b = .83, p = .64) on non-drinking days

Discussion

As hypothesized, within persons, daily NA was positively associated with likelihood of same-day alcohol consumption. Against expectations, within persons, daily NA was not associated with the quantity of same-day alcohol consumption, and between persons, overall levels of daily NA were not associated with alcohol consumption. Against expectations, within persons, momentary NA was not associated with next-moment likelihood of alcohol consumption, and negatively associated with next-moment quantity of alcohol consumption. Additionally, between persons, overall levels of momentary NA were not associated with alcohol consumption.

In support of the self-medication hypothesis (11), daily NA was associated with increased likelihood of subsequent same-day alcohol consumption. Daily affect measures were averaged over all momentary assessments up to the first alcoholic consumption. Increased daily NA may therefore reflect the incapacity of a person to cope effectively with multiple negative momentary states. Perhaps multiple negative states deplete self-control, leading to self-medication. However, a decrease in next-moment quantity of alcohol consumption in response to a negative momentary state does not support this finding. Sheppes et al. (49) found that low-intense negative emotions may elicit adaptive coping responses such as cognitive reappraisal. In the current study the reported levels of negative affect were low and may have evoked effective coping strategies rather than resorting to alcohol consumption. Persons in the current study may also have delayed their responses to momentary NA to more appropriate drinking times later on the day.

Contrary to what was hypothesized, within and between persons, there were no significant associations between daily PA and alcohol consumption. As hypothesized, within persons, momentary PA was positively associated with likelihood of next-moment alcohol consumption. Against expectations, within persons, momentary PA was not associated with next-moment quantity of alcohol consumption. Contrary to expectations, between persons, overall levels of momentary PA were not associated with momentary alcohol consumption.

The absence of associations between within-person daily PA and alcohol consumption do not support expectancy theory. However, increased likelihood of alcohol consumption in response to momentary PA is in line with expectancy theory (12). It is conceivable that for persons who are not alcohol dependent, drinking is an end-of-the day or leisure time activity. The daily affect measures were averaged over all momentary assessments up to the first alcoholic consumption. It is therefore unknown if these daily uplifts occurred at times or in settings when the current sample would normally consume alcohol. Future research may benefit from taking the context into account. Perhaps elevated day-levels of PA were less salient than momentary uplifts. More insight may be obtained by examining the effects of consecutive higher than average momentary PA states on day levels of alcohol consumption, after alcohol consumption is initialized. The finding of a marginally non-significant (p = .08) association between within-persons daily PA and quantity of alcohol may be the result of the relative small sample size. Research with bigger samples is therefore warranted. Perhaps the momentary measures involved time spans which were too short to find effects on the quantity of alcohol consumed.

Discordant with some previous studies (31,32) there were no between-person effects of affect on alcohol consumption. Thus, in the current sample, individual differences in overall levels of affect were not associated with alcohol consumption. The self-medication hypothesis (11) and expectancy theory (12) describe within-person processes (24) which can not automatically be translated to the between-person level (50). Moreover, the current sample comprised adults from the community from 20 to 50 years. Learned associations between alcohol consumption and affect may not hold for all individuals in a community sample which was perhaps more heterongeneous than many samples in previous studies. Additionally, features of the current study may explain this finding. Previous research has demonstrated that a variety of person-level factors such as impulsivity and coping strategies (29,51,52) may influence alcohol consumption. In the current study, alcohol consumption was modelled across all individuals, regardless of such personal characteristics. Future research may benefit from including such personal characteristics to provide more insight into person-level differences in drinking behavior.

It was hypothesized that after initializing alcohol consumption, we would observe a decrease and increase of respectively NA and PA. However, there were no significant differences between the intercepts and slopes of NA before and after consuming alcohol on a given day. These findings do not support the propositions of the self-medication model and suggest that participants in the current study did not experience relief from adversity by consuming alcohol. Levels of NA were low in the current sample. This may have resulted in floor effects, where information about true differences at the lowest value is lost (53). These null findings may also be the result of the relative small post-drinking time window. In the current study, alcohol was mainly consumed at the end of the day. Perhaps the post-alcohol analysis was underpowered to detect small effects, due to the low number of assessments after alcohol consumption was initialized.

As expected, the intercept of PA was higher after alcohol consumption. Contrary to what was hypothesized, the slope of PA decreased after alcohol consumption. Because time was the independent variable and the first daily alcoholic drink was used to split days in pre-drinking and post-drinking periods, the increase in intercept suggests a jump in PA at the time of the first alcoholic drink. However, this effect did not persist. Perhaps participants experienced anticipatory pleasure when alcohol consumption was initialized and gradually returned to their normal levels of PA. Another explanation for the observed effects may be that PA followed its natural diurnal variation (54). The choice for the transition point in the piecewise multilevel analyses was based on the theoretical models of alcohol consumption, and not determined by a curve fitting procedure. Hence, the jump we observed in PA after alcohol consumption was initialized could also be an artefact of the transition point we imposed.

The results of the current study are in line with Peacock et al. (30) who demonstrated that daily PA prior to the first alcoholic drink was not associated with alcohol consumption. However, contrary to Peacock et al. (30) in the current study, within persons, daily NA prior to the first drinking event was associated with higher likelihood of subsequent alcohol consumption. The findings in this study are in opposition of Dvorak et al. (29), who demonstrated that within persons, pre-drinking NA was not associated with subsequent alcohol consumption, while pre-drinking PA was associated with both likelihood and intensity of drinking. The results also deviate from Simons et al. (31), who found that within persons, NA preceding alcohol consumption was associated with increased drinking levels, and pre-drinking PA was associated with drinking frequency and intensity. Individuals in our sample drank less frequently and smaller amounts of alcohol than a community and two student samples in some EMA studies we cited (29,31,32), but more frequent and greater amounts of alcohol on drinking days than the community sample in the study of Peacock et al. (30). These disparities need to be taken into account when making comparisons.

Limitations

Previous alcohol research has largely been conducted among student or clinical samples. It is conceivable that community samples have different drinking patterns than student or alcohol-dependent samples. Support for self-medication through alcohol consumption in non-clinical samples is limited (52). We used a nonstandard method of assessing alcohol use, without any biological verification, and need to take into account that self-reports of alcohol consumption may be biased (55). In the current study, affective states were sampled ten times a day. Mohr et al. (32) collected EMAs on affect thrice daily, while Peacock et al. (30) sampled data event-contingent. Our 72% compliance rate was lower than in the studies of Simons et al. (31), Dvorak et al. (29), Mohr et al. (32), and Peacock et al. (30) (compliance rates 79%-93%). The disparity in compliance may reflect differences in the samples and limit the comparability of the studies. The composite measures of affect differed across previous research and the current study. Specific types of NA (i.e. anger, sadness) have been associated differently with alcohol consumption (56,57,58). Moreover, across previous research and the current study, the items used for composing affect measures differ in the extent to which they evoke arousal and negative or positive feelings as proposed by the circumplex model of affect (59). According to this model, three of our four NA items would evoke arousal. In this sense our NA measure is different from Dvorak et al. (29), Peacock et al. (30), and Mohr et al. (32), whose constituent items are more in balance with regard to arousal versus passivity. Arousal evoking adversity may be associated in a different way with alcohol consumption than negative feelings which evoke passivity. Equally, PA measures differ with regard to their constituents across previous research and the current study. Our composite measure encompasses two items which are presumed to evoke arousal, and two items which are presumed to be associated with calmness (59). The PA measure in the study of Dvorak et al. (29) may evoke more arousal than ours. The composite PA measures in the studies of Peacock et al. (30) and Mohr et al. (32) are more in line with our measure with regard to the balance between arousal and passivity. Finally, the adjectives used at item level for the affective EMAs differ substantially across all studies. Although many are synonyms, small differences in perception of what is asked may eventually influence the reports. This may be true in particular when numerous assessments are conducted, as is often the case in EMA research. Future research could benefit from examining whether item-level analyses yield different results from analyses conducted with composite measures. Additionally, if specific items are associated with alcohol consumption, this could lower participant burden, as shorter questionnaires would then suffice.

The sample in the current study was predominantly female and higher educated than the majority of the Dutch adult population. Education level has been associated with different drinking patterns (60) and gender may influence affect-related alcohol consumption (29,56). Participants in the current study were aware of its nutritional focus, and may have been more preoccupied with health aspects than the majority of the Dutch adults. Additionally, EMA protocols come with considerable respondent burden (28). It is conceivable that individuals who completed this study are a selective sample. Therefore the results cannot be generalized to the general population. The ethnicity of the participants in the study was unknown. Given the multicultural Dutch population, this is a limitation. When the data for the current study were gathered, 74%-86% of the Dutch adults aged 25–45 were in possession of a mobile phone or smartphone in respectively 2012 and 2013, and 50%-63% of the Dutch adults aged 45–65 were in possession of a mobile phone or smartphone in respectively 2012 and 2013 (61). Therefore, the older participants in the current study may be a more selective sample than the younger participants. At the start of the data-collection (October 2012) 61% of the Dutch smartphones used the Android platform (62).

Conclusion

Some support for self-medication and mood enhancement through alcohol consumption was found. The findings suggest that persons were sensitive to their varying levels of affect. Days which are characterized by higher than average levels of NA may pose a risk for drinking to cope. Moments which are experienced as more positive than average, may also increase the likelihood of alcohol consumption. The data provided insights into how persons respond in real-time. These insights could be incorporated in EMIs (63) to intervene in those days and moments which may pose a risk for affective drinking. EMIs could deliver just-in-time tailored messages to create self-awareness and to provide alternative strategies for coping with NA or for enhancing PA.

Conflict of Interest

The authors report no relevant financial conflicts

Supplementary Material

Supplementary data for the article can be accessed here.

Additional information

Funding

This work was supported by the European Research Advisory Board under Grant EA1634