When Is Drinking Alcohol Morally Worth It?

This report builds a single, reproducible utility model for alcohol consumption that combines positive effects (benefits) and negative effects (harms) into a common welfare unit. The model is then evaluated via Monte Carlo simulation across scenarios, producing distributions of lifetime outcomes rather than single point estimates.

A central result is scenario-dependent: the median net utility (p50) is negative in the baseline scenario but becomes positive when drinking and driving is removed. See Figure 1 (baseline histograms) and Figure 2 (never drink-and-drive histograms).

A secondary, smaller decision-relevant criterion is social bonding and affiliation effects: these can increase the positive side in some contexts, but their expected magnitude in this model is substantially smaller than the effect of eliminating drink-and-drive risk.

A utilon is defined as:

1 utilon = a 1-point change on a 0–10 life satisfaction scale sustained for 1 person for 1 year.

This is a WELLBY-style unit, enabling consistent aggregation across short-lived experiences (hours/days) and long-lived outcomes (years), including premature mortality and chronic disability when expressed as life-satisfaction-point-years.

For any simulated life-course realization:

• Positive utilons ($U_{+}$) capture enjoyment, mood lift, and social benefits (modeled as life-satisfaction gains over time).
• Negative utilons ($U_{-}$) capture acute risks (injury/violence/poisoning), next-day impairment, chronic health burdens, and alcohol use disorder (AUD) state burdens (modeled as life-satisfaction losses over time).

The model reports:

$U_{\text{net}} = U_{+} - U_{-}$

Utilities are discounted over time using a continuous discount rate (r):

$U = \int_0^{T} e^{-rt},\Delta LS(t),dt$

In implementation with daily or annual time steps, this becomes a discounted sum over time slices. Core sensitivity includes ($r \in {0,,0.015,,0.03,,0.05}$).

This section maps directly to sim.cpp implementation choices. The code uses a global script config with:

• 60-year horizon (years = 60),
• 365 days/year,
• baseline intake drinks_per_day = 1.5,
• day-count model default poisson, and
• simulation mode default expected with an alternative daily mode.

The report outputs shown elsewhere were mostly generated in expected mode for speed at large Monte Carlo sizes (20k runs), while the model selected for mechanistic explanation here is daily mode because it preserves event timing and state transitions day-by-day.

Alcohol exposure is parameterized in grams of ethanol, with “standard drinks” treated as a conversion convenience. Two conventions are supported:

• 10 g ethanol per drink
• 14 g ethanol per drink

Each simulation generates an exposure stream, commonly at a daily level:

• drinks_today: integer count of drinks for that day
• g_today = drinks_today * grams_per_drink

Daily drinks can be generated using one of:

• Constant: drinks_today = drinks_per_day
• Poisson: drinks_today ~ Poisson(drinks_per_day)
• Two-point mixture: drinks_today = 0 with probability ($p_0$), otherwise a high value chosen to preserve the mean

This matters because many harms are highly nonlinear in dose and concentrate in heavy episodic days.

In sim.cpp, these are implemented in sample_drinks_today(...) and drinks_pmf(...):

• sample_drinks_today is used when the simulator needs realized day-level trajectories (the daily mode path).
• drinks_pmf is used when the simulator needs expectations over dose without sampling every day-level benefit event (the expected mode path).

So both modes share the same exposure assumptions, but they differ in whether the model propagates full day-level randomness through each channel.

Two binary indicators are derived:

• binge_today = 1[drinks_today >= binge_threshold] with threshold in {4, 5}
• high_intensity_today = 1[drinks_today >= high_intensity_multiplier * binge_threshold] with multiplier in {2, 3}

These markers route exposure into acute-risk channels and the AUD risk process.

This section summarizes alcohol-linked benefits as short-lived or social increments in life satisfaction (LS), then converts these increments into utilons (LS-point-years).

Core channels included in this report are:

• Momentary hedonic uplift while drinking (minutes to hours), informed by experience-sampling evidence that people report higher immediate happiness at drinking moments.[5]
• Social bonding and affiliation effects during shared drinking episodes, including experimentally observed increases in perceived bonding in alcohol-administration group settings.[3]
• Sociability and reduced social anxiety in some contexts, where effects may depend on setting and subgroup (e.g., socially anxious populations).[6][7]
• Expectancy-related confidence/affect effects, where believing one consumed alcohol can itself shift reported experience even without ethanol exposure.[11]

The model treats these channels as potentially real but generally time-limited, so most positive utilons are generated through frequent small gains rather than large persistent LS shifts.[5][9]

A central finding from the positive-evidence review is that not all observed benefit in “drinking occasions” is pharmacological ethanol benefit. The model therefore separates attribution into:

• Context effect: benefits from the social situation itself (gathering, ritual, group interaction), which may occur even with non-alcoholic substitution.[7][10]
• Expectancy effect: benefits caused by believing alcohol was consumed.[11]
• Pharmacological ethanol effect: incremental effect of ethanol beyond expectancy (alcohol vs placebo contrast).[3]

For conservative moral accounting, the strictest ethanol-attributable estimate emphasizes the third component. Broader occasion-attributable estimates include context and expectancy components but apply attenuation weights for substitution risk.

The conversion framework uses a common accounting identity:

$U_{k} = N \times \Delta LS_{k} \times D_{k} \times A_{k}$

where:

• ($N$): number of affected persons,
• ($\Delta LS_{k}$): LS increment on the 0–10 scale,
• ($D_{k}$): duration in years,
• ($A_{k}$): attribution weight in ($[0,1]$).

Mapping rules used in practice:

• If LS is directly observed on a 0–10 scale, conversion is direct (preferred case).[1]
• If momentary affect is on another scale (e.g., 0–100), rescale and apply a cross-walk (or sensitivity band) to translate to LS-equivalent changes before integrating over time.[5]
• If outcomes are intermediate psychosocial constructs (bonding, anxiety), use cross-walk coefficients to LS or treat as scenario-specific proxy effects with explicit uncertainty.[3][6]

The evidence review adds four implementation guardrails that are carried into this report:

• Exposure should be parameterized in grams of ethanol, with “standard drink” used only as a jurisdiction-specific convenience unit.[4][58][59]
• Discounting should be explicit and sensitivity-tested because some benefits are immediate while a subset of social/economic channels may persist.[1]
• Double counting should be avoided when multiple proxies measure overlapping constructs (e.g., mood + social confidence during the same episode).[5][9]
• Health-benefit claims at low doses are high-uncertainty and should not be treated as robust positive utilons in the base case given recent reassessments of protective-effect claims.[2][38]

Given this evidence structure, the baseline benefits estimate is intentionally modest at the annualized LS level: it allows frequent short-run gains but constrains long-run persistent uplift unless supported by stronger causal evidence. This keeps positive utilons empirically grounded while preserving uncertainty bands for plausible social and expectancy effects.[3][5][11]

The harms side consumes the same daily exposure stream as the benefits side, then maps alcohol exposure into discounted LS losses via four non-overlapping buckets: acute events, next-day impairment, chronic health burden, and AUD state burden. The implementation choice is designed to keep channels interpretable while minimizing double counting between health-state losses and direct wellbeing losses.[1][39]

Acute risk is modeled at the day level using ethanol dose in grams (g_today = drinks_today * grams_per_drink) and steeply increasing relative-risk curves. A practical form used in the model is log-linear in dose:

$RR(g) = RR_{10}^{g/10}$

with daily event probability bounded as:

$p(d) = 1 - (1-p_0)^{RR(g(d))}$

This structure is aligned to ED-based dose-response evidence for alcohol and injury risk from case-crossover analyses.[56]

To calibrate absolute daily probabilities, the model anchors baseline rates to administrative ED incidence and then scales by dose-response: injury ED visits from NHAMCS, assault-injury ED visits from NCHS NHAMCS briefs, and ethanol-poisoning ED visits (ICD-10 T51.0x) from national HCUP/NEDS analyses.[54][55][57]

A key empirical detail is that these endpoints are not mutually exclusive in raw data (e.g., assault and ethanol-poisoning visits can be subsets of injury visits), so the simulation either treats them as overlapping burdens with careful accounting or derives approximately non-overlapping categories where needed.[54][55][57]

Hangover is modeled as a direct LS decrement on subsequent day(s), triggered primarily by binge/high-intensity drinking days. This is intentionally separated from labor-income channels so productivity impacts are not counted twice. Empirical motivation comes from evidence that hangovers are associated with measurable absenteeism/presenteeism and functioning decrements, but the model translates these into direct LS losses in this channel.[44]

Core sampled inputs are: P(hangover|binge), LS loss per hangover day, and duration (typically 1–2 days), with higher trigger probability under high-intensity episodes.[58][63][64]

Chronic harm depends on cumulative/lagged intake rather than only same-day dose. This section uses smoothed exposure (EMA with half-life menu) and applies excess-risk mappings for major chronic outcomes (e.g., cancer, cirrhosis, AF/BP-mediated cardiovascular burden), then converts health burden into utilons using a QALY-to-WELLBY factor.[33][34][35][36]

$EMA_y = EMA_{y-1}e^{-\ln(2)/H} + \bar{g}_y\left(1-e^{-\ln(2)/H}\right)$

$U_{\text{health}} = \text{DALYs/QALYs lost} \times k_{\text{QALY}\rightarrow\text{WELLBY}}$

The base conversion menu includes k in {5,6,7,8}, centered on 7, consistent with Green Book supplementary wellbeing guidance for bridging health utility and LS-point-years.[1]

AUD is modeled as a Markov process (NoAUD -> AUD -> Remission, with relapse). Transition intensities are linked to risk-drinking frequency (5+/4+ day frequency), following prospective NESARC evidence showing sharply increasing dependence incidence with more frequent risk drinking, and substantial recurrence risk among remitters who resume risk drinking.[39][40]

AUD harm is represented through disability weights / LS decrements while in-state, with uncertainty over severity calibrated to GBD-style disability-weight evidence and newer severity-state syntheses.[41][42]

In distributional terms, this channel contributes modestly at median outcomes but materially to upper-tail negative realizations, because sustained AUD episodes compound over time and interact with other harm channels.[39][41]

At a high level, each run does:

1. Sample one positive-utility person profile (PosPerson) from discrete menus.
2. Sample one negative-parameter draw (NegParams) from discrete menus.
3. Branch on --mode:
   • daily: call simulate_life_rollout(...).
   • expected: call the hybrid annualized path inside simulate_one_person().
4. Return a SimOut tuple containing total positive, total negative, net utility, and component breakdowns.
5. Aggregate over --runs and print percentile summaries plus decile contribution tables.

This makes the Monte Carlo outer loop identical across modes; only the within-person utility accounting engine changes.

Instead of single best-guess parameters, the model uses discrete parameter menus and samples uniformly across:

• discount rate
• grams per standard drink
• binge thresholds and high-intensity multipliers
• acute risk curves and baseline event rates
• hangover probability and severity
• chronic latency half-lives and risk slopes
• QALY→WELLBY conversion factors
• causal-weight knobs for observational channels
• AUD onset/remission/relapse and disability weights

This yields distributions over ($U_{+}$), ($U_{-}$), and ($U_{\text{net}}$), enabling percentile reporting and sensitivity comparisons across scenario variants.

The scenario set modifies the simulated world in targeted ways:

• Baseline: full model as specified
• Never drink and drive: removes (or sharply reduces) the driving-related acute harm channel while leaving other components intact
• No binge: removes binge/high-intensity structure (or forces it absent), reducing acute tails and AUD risk triggers
• Abstinence: eliminates both benefits and harms from alcohol exposure

For this write-up, the chosen structural model is --mode daily, because it is the most behaviorally explicit representation of utility over time:

• It simulates each day in sequence over the life horizon.
• It discounts each day at continuous-time factor exp(-r t) using the day midpoint.
• It realizes whether that day is social, whether acute events occur, whether hangover carries to the next day, and whether AUD state changes at month boundaries.
• It updates lagged chronic exposure states (EMA) every day before computing chronic disutility.

In utility terms, daily mode is closest to the conceptual object

$U_{\text{net}} = \sum_{d=1}^{D} e^{-r t_d}\left(\Delta LS_{+,d} - \Delta LS_{-,d}\right),$

with path-dependent state variables (AUD state, hangover days remaining, alive/dead status, exposure EMAs).

By contrast, expected mode is a computational shortcut that:

• uses expected daily positive utility from the drink-count PMF,
• computes annual expected acute/hangover/chronic terms from PMF-based probabilities and yearly realized binge/high-intensity frequencies,
• simulates AUD at annual transitions rather than monthly/day-linked state updates.

So expected remains useful for large sweeps and sensitivity grids, while daily is preferred when we want the utility narrative to explicitly reflect within-life timing and dependence across events.

Inside simulate_life_rollout(...), daily accounting proceeds as:

1. Drinking draw (possibly AUD-modified intensity): sample drinks_today.
2. Positive utility: compute day-level life-satisfaction uplift via a saturating function of drinks, plus social-context modifier; add discounted \Delta LS / 365.
3. Acute harms: draw Bernoulli events for traffic injury, non-traffic injury, violence (binge-linked), and poisoning (high-intensity-linked); convert event DALY burden to utilons.
4. Hangover utility loss: on binge-triggered days, stochastically start/extend hangover spell; apply next-day LS loss while spell lasts.
5. Chronic harms: update chronic EMA exposure states and compute daily equivalent chronic utility loss.
6. AUD burden and transitions: apply daily AUD disability equivalent when active; update AUD state monthly using risk-day-dependent onset/remission/relapse hazards.
7. Mortality stop rule: if a fatal acute event occurs, terminate future utility flow.

The resulting net utility is therefore not just “expected dose effect,” but a discounted path integral over realized stochastic life histories.

Figure 1 (baseline histograms) and Figure 2 (never drink-and-drive histograms) visualize the joint outcome distributions across key components. The main comparison is the net utility shift:

• Baseline: median net utility is negative
• Never drink and drive: median net utility becomes positive

The shift is driven primarily by a large reduction in the acute harm distribution (notably the high-variance tail), while the positive distribution remains similar.

• Positive utilons (p50) are stable across baseline vs never drink-and-drive (≈10.4 utilons).

• Negative utilons (p50) drop sharply when drinking and driving is removed (≈11.9 → 7.6).

• Net utilons (p50) flip sign:

  • baseline −1.41
  • never drink-and-drive +2.29

All values are utilons; net is ($U_{+} - U_{-}$). Percentiles are taken from Monte Carlo outputs.

Interpretation notes:

• The lower tail (e.g., net p05) is strongly negative in non-abstinence scenarios, reflecting rare-but-severe harm realizations.
• The median improves materially when driving-related harms are removed, indicating that a concentrated risk channel dominates the sign at p50 in the baseline.

Component medians (p50) for negative utilons:

The acute channel explains most of the median reduction under never drink-and-drive, while hangover and chronic proxies are largely unchanged at p50. AUD is typically zero at the median but contributes to the upper tail (p75+), consistent with a heavy-tailed state-burden process.

Across multiple baseline seeds, the net p50 remains close to the baseline median (roughly −1.7 to −1.1), suggesting the sign flip under never drink-and-drive is not a random-seed artifact but a structural effect of removing a high-impact harm pathway.

Under this framework, alcohol use is “morally worth it” (net-positive expected welfare in the median sense) when the exposure pattern avoids high-impact harm pathways and concentrates any benefits without triggering heavy episodic risk.

The scenario results identify two dominant levers:

1. Eliminating drinking and driving is the largest single improvement observed, flipping median net utility from negative to positive (Figures 1–2; Table 7.2).
2. Eliminating binge structure improves net outcomes and reduces tail risk, but does not match the median shift produced by removing driving-related harms (Table 7.2).

A secondary, lower-magnitude criterion is whether social bonding and affiliation effects are plausibly present in a specific context (e.g., supportive group settings). In this model, those effects can increase positive utility at the margin, but they are less decision-relevant than avoiding severe harm channels such as drink-and-drive risk.

Abstinence is a neutral baseline in this utilon accounting (0 net by construction) and provides a reference point for interpreting whether any modeled drinking pattern clears a “net-positive” bar under chosen ethical aggregation rules.

These figures support the distributional shift: net p50 changes from negative to positive when drinking and driving is removed.

The table below inventories the configurable parameters used by sim.cpp, with defaults and/or sampled ranges plus an interpretation column. The final citation mapping is left as requested.

1. Frijters, P., Krekel, C., Sanchis, R., & Santini, Z. I. (2024). The WELLBY: a new measure of social value and progress. Humanities and Social Sciences Communications, 11, Article 736. (Nature: https://www.nature.com/articles/s41599-024-03229-5)
2. Zhao, J., Stockwell, T., et al. (2023). Association Between Daily Alcohol Intake and Risk of All-Cause Mortality: A Systematic Review and Meta-analyses. JAMA Network Open, 6(3), e236185. https://doi.org/10.1001/jamanetworkopen.2023.6185 (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC10066463/)
3. Sayette, M. A., Creswell, K. G., Dimoff, J. D., Fairbairn, C. E., Cohn, J. F., Heckman, B. W., Kirchner, T. R., Levine, J. M., & Moreland, R. L. (2012). Alcohol and Group Formation: A Multimodal Investigation of the Effects of Alcohol on Emotion and Social Bonding. Psychological Science, 23(8), 869–878. https://doi.org/10.1177/0956797611435134 (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC5462438/)
4. Kerr, W. C., & Stockwell, T. (2012). Understanding standard drinks and drinking guidelines. Drug and Alcohol Review, 31(2), 200–205. https://doi.org/10.1111/j.1465-3362.2011.00374.x (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC3276704/)
5. Baumberg Geiger, B., & MacKerron, G. (2016). Can alcohol make you happy? A subjective wellbeing approach. Social Science & Medicine, 156, 184–191. https://doi.org/10.1016/j.socscimed.2016.03.034 (PubMed: https://pubmed.ncbi.nlm.nih.gov/27043371/)
6. Himle, J. A., Abelson, J. L., Haghightgou, H., Hill, E. M., Nesse, R. M., & Curtis, G. C. (1999). Effect of alcohol on social phobic anxiety. American Journal of Psychiatry, 156(8), 1237–1243. https://doi.org/10.1176/ajp.156.8.1237 (PubMed: https://pubmed.ncbi.nlm.nih.gov/10450266/?utm_source=chatgpt.com)
7. Dunbar, R. I. M., & Launay, J. (2016). Functional Benefits of (Modest) Alcohol Consumption. Adaptive Human Behavior and Physiology, 3(2), 118–133. https://doi.org/10.1007/s40750-016-0058-4 (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC7010365/)
8. Childs, E., O’Connor, S., et al. (2011). Bidirectional interactions between acute psychosocial stress and acute intravenous alcohol in healthy men. Alcoholism: Clinical and Experimental Research, 35(10), 1794–1803. https://doi.org/10.1111/j.1530-0277.2011.01522.x (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC3183385/)
9. Dora, J., Piccirillo, M., et al. (2023). The daily association between affect and alcohol use: A meta-analysis of individual participant data. Psychological Bulletin, 149(1–2), 1–24. https://doi.org/10.1037/bul0000387 (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC10409490/)
10. Castro, F. G., Barrera, M., Jr., et al. (2014). Culture and Alcohol Use: Historical and Sociocultural Themes From 75 Years of Alcohol Research. Journal of Studies on Alcohol and Drugs (Supplement), 75(Suppl 17), 36–49. https://doi.org/10.15288/jsads.2014.s17.36 (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC9838738/)
11. Testa, M., et al. (2006). Understanding Alcohol Expectancy Effects: Revisiting the Placebo Condition. Alcoholism: Clinical and Experimental Research, 30(2), 339–348. https://doi.org/10.1111/j.1530-0277.2006.00039.x (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC1403295/)
12. Caumiant, E. P., Kang, D., Girard, J. M., & Fairbairn, C. E. (2025). Alcohol and Emotion: Analyzing Convergence Between Facially-Expressed and Self-Reported Indices of Emotion Under Alcohol Intoxication. Psychology of Addictive Behaviors, 39(8), 677–688. https://doi.org/10.1037/adb0001053 (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC12238295/?utm_source=chatgpt.com)
13. (Worksite wages paper). (1995). Is moderate alcohol use related to wages? Evidence from four worksites. (ScienceDirect / Elsevier journal page). (ScienceDirect: https://www.sciencedirect.com/science/article/pii/016762969590921R)
14. Peters, B. L., & Stringham, E. P. (2006). No booze? You may lose: Why drinkers earn more money than nondrinkers. Journal of Labor Research. (Springer: https://link.springer.com/article/10.1007/s12122-006-1031-y)
15. (Wharton working paper / PDF). Impact of Alcohol on Negotiator Behavior. Wharton / University of Pennsylvania (PDF).
16. World Health Organization, Regional Office for Europe. (2022/2023). No level of alcohol consumption is safe for our health. (WHO Europe news-room post; published Dec 28, 2022 on the page.) (World Health Organization: https://www.who.int/europe/news-room/04-01-2023-no-level-of-alcohol-consumption-is-safe-for-our-health?utm_source=chatgpt.com)
17. National Institute on Alcohol Abuse and Alcoholism (NIAAA). (n.d.). Alcohol drinking patterns. (Office for National Statistics: https://www.ons.gov.uk/peoplepopulationandcommunity/wellbeing/methodologies/personalwellbeingsurveyuserguide?utm_source=chatgpt.com)
18. Substance Abuse and Mental Health Services Administration (SAMHSA). (n.d.). National Survey on Drug Use and Health (NSDUH): Data we collect.
19. Centers for Disease Control and Prevention (CDC). (n.d.). Behavioral Risk Factor Surveillance System (BRFSS).
20. National Institute for Health and Care Excellence (NICE). (n.d.). Guide to the methods of technology appraisal (Economic evaluation / reference case discounting).
21. UK HM Treasury (Supplementary Green Book guidance). Wellbeing guidance for appraisal: supplementary Green Book guidance (PDF). https://assets.publishing.service.gov.uk/media/60fa9169d3bf7f0448719daf/Wellbeing_guidance_for_appraisal_-_supplementary_Green_Book_guidance.pdf
22. Frijters, P. et al. WELLBYs, cost-benefit analyses and the Easterlin Discount (LSE eprints PDF). https://eprints.lse.ac.uk/114605/1/Frijters_PR3.pdf
23. NIAAA. Understanding the Dangers of Alcohol Overdose (standard drink, binge & high-intensity definitions). https://www.niaaa.nih.gov/publications/brochures-and-fact-sheets/understanding-dangers-of-alcohol-overdose
24. WHO. Alcohol — Fact sheet. https://www.who.int/news-room/fact-sheets/detail/alcohol
25. CDC. Alcohol and Public Health: Alcohol-Related Disease Impact (ARDI) / facts & stats. https://www.cdc.gov/alcohol/facts-stats/index.html
26. CDC (MMWR). Deaths from Excessive Alcohol Use — United States (MMWR report). https://www.cdc.gov/mmwr/volumes/73/wr/mm7308a1.htm
27. NIAAA. Alcohol Use Disorder (AUD) in the United States: age groups and demographic characteristics. https://www.niaaa.nih.gov/alcohols-effects-health/alcohol-topics/alcohol-facts-and-statistics/alcohol-use-disorder-aud-united-states-age-groups-and-demographic-characteristics
28. Taylor, B. et al. (2010). A systematic review and meta-analysis of how acute alcohol consumption and injury risk relate to alcohol dose and drinking pattern. Drug and Alcohol Dependence (ScienceDirect abstract page). https://www.sciencedirect.com/science/article/abs/pii/S0376871610000712
29. CDC. Impaired Driving: Facts (shares of deaths among drivers vs others, etc.). https://www.cdc.gov/impaired-driving/facts/index.html
30. (HHS) Office of the Assistant Secretary for Health. Alcohol and Cancer Risk (PDF). https://www.hhs.gov/sites/default/files/oash-alcohol-cancer-risk.pdf
31. The Lancet (2018). Alcohol use and burden / no-safe-level style synthesis (full text). https://www.thelancet.com/article/S0140-6736%2818%2931310-2/fulltext
32. The Lancet Oncology. Global alcohol-attributable cancer burden (full text). https://www.thelancet.com/journals/lanonc/article/PIIS1470-2045%2821%2900279-5/fulltext
33. Systematic review/meta-analysis (2023). Alcohol intake and breast cancer risk (prospective cohorts) (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC11629438/
34. Dose-response meta-analysis (2023). Alcohol consumption and liver cirrhosis morbidity/mortality (PubMed record). https://pubmed.ncbi.nlm.nih.gov/37684424/
35. Meta-analysis (AF). Alcohol consumption and atrial fibrillation risk (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC9561500/
36. Dose-response meta-analysis (BP). Alcohol intake and blood pressure (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC10510850/
37. Systematic review/meta-analysis (2014). Alcohol and ischemic heart disease; patterning / reference group issues (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC4203905/
38. “Burden of proof” style synthesis (2024). Alcohol and ischemic heart disease protective-claim uncertainty (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC11094064/
39. NESARC prospective analysis (risk drinking frequency → dependence incidence) (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC2366117/
40. Remission/recurrence follow-up (NESARC-related) (PubMed record). https://pubmed.ncbi.nlm.nih.gov/18034696/
41. GBD disability weights paper (GBD 2013 DWs; includes AUD DWs) (The Lancet Global Health full text). https://www.thelancet.com/journals/langlo/article/PIIS2214-109X%2815%2900069-8/fulltext
42. Meta-analysis (2025). Disability weights for alcohol use disorder severity states (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC12595804/
43. Mendelian randomization study on depression ↔ alcohol problems (PubMed record). https://pubmed.ncbi.nlm.nih.gov/41587741/
44. Study on hangover-related absenteeism/presenteeism and productivity loss (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC10855845/
45. Cohort study on spousal drinking concordance/discordance and divorce (Ovid PDF). https://www.ovid.com/journals/alcer/pdf/10.1111/acer.12029~discordant-and-concordant-alcohol-use-in-spouses-as
46. Review on life events and wellbeing adaptation (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC3289759/
47. Classic adaptation / wellbeing dynamics paper (PubMed record). https://pubmed.ncbi.nlm.nih.gov/16313658/
48. BMJ Open. Alcohol use/problems and employment outcomes (article page). https://bmjopen.bmj.com/content/9/7/e029184
49. Related employment/labor-market paper (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC6451700/
50. Unemployment → life satisfaction effect size (Springer link). https://link.springer.com/article/10.1007/s10902-025-00941-0
51. Meta-analysis / review on alcohol and sexual risk behaviors (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC10270666/
52. Related systematic review (PMC full text). https://pmc.ncbi.nlm.nih.gov/articles/PMC4553138/
53. International Journal of Public Health article (Frontiers / SSPH journal site). https://www.ssph-journal.org/journals/international-journal-of-public-health/articles/10.3389/ijph.2023.1605669/full
54. Cairns, C., & Kang, K. (2022). National Hospital Ambulatory Medical Care Survey: 2019 Emergency Department Summary Tables (NHAMCS). National Center for Health Statistics (CDC). (CDC Stacks: https://stacks.cdc.gov/view/cdc/115748?utm_source=chatgpt.com)
55. Davis, D., & Santo, L. (2023). Emergency department visit rates for assault: United States, 2019–2021 (NCHS Data Brief No. 481). National Center for Health Statistics (CDC). (CDC: https://www.cdc.gov/nchs/products/databriefs/db481.htm?utm_source=chatgpt.com)
56. Cherpitel, C. J., Ye, Y., Bond, J., Borges, G., & Monteiro, M. (2015). Relative risk of injury from acute alcohol consumption: Modeling the dose-response relationship in emergency department data from 18 countries. Addiction, 110(2), 279–288. (PubMed: https://pubmed.ncbi.nlm.nih.gov/25355374/?utm_source=chatgpt.com)
57. Conrad, S. W., Greene, C. R., Dal Pan, G., Callahan, C. L., Meyer, T. E., & Radin, R. G. (2026). United States healthcare encounters for poisoning involving cannabis relative to other substances. The American Journal of Drug and Alcohol Abuse. (Author manuscript in PMC). (PMC: https://pmc.ncbi.nlm.nih.gov/articles/PMC12889077/?utm_source=chatgpt.com)
58. National Institute on Alcohol Abuse and Alcoholism (NIAAA). (n.d.). What is a standard drink? (NIAAA: https://www.niaaa.nih.gov/alcohols-effects-health/what-standard-drink?utm_source=chatgpt.com)
59. Centers for Disease Control and Prevention (CDC). (2024). About standard drink sizes. (CDC: https://www.cdc.gov/alcohol/standard-drink-sizes/index.html?utm_source=chatgpt.com)
60. Australian Government Department of Health, Disability and Ageing. (2024). Standard drinks guide (10g definition). (Department of Health and Ageing: https://www.health.gov.au/topics/alcohol/about-alcohol/standard-drinks-guide?utm_source=chatgpt.com)
61. National Health and Medical Research Council (NHMRC, Australia). (n.d.). Alcohol — What is a standard drink? (10g definition). (NHMRC: https://www.nhmrc.gov.au/health-advice/alcohol?utm_source=chatgpt.com)
62. Cleveland Clinic. (2022). Blood alcohol content (BAC): What it is & levels. (Cleveland Clinic: https://my.clevelandclinic.org/health/diagnostics/22689-blood-alcohol-content-bac?utm_source=chatgpt.com)
63. Kanny, D., Naimi, T. S., Liu, Y., & Brewer, R. D. (2020). Trends in total binge drinks per adult who reported binge drinking — United States, 2011–2017. MMWR, 69(2), 30–34. (CDC: https://www.cdc.gov/mmwr/volumes/69/wr/mm6902a2.htm?utm_source=chatgpt.com)
64. Bohm, M. K., Liu, Y., Esser, M. B., Mesnick, J. B., Lu, H., Pan, Y., & Greenlund, K. J. (2021). Binge drinking among adults, by select characteristics and state — United States, 2018. MMWR, 70(41), 1441–1446. (CDC: https://www.cdc.gov/mmwr/volumes/70/wr/mm7041a2.htm?utm_source=chatgpt.com)