5.1 Quantitative and Experimental Methods in Psychiatry: Epidemiology

Background & Scope of Psychiatric Epidemiology

Epidemiology = study of disease occurrence, change, and explanatory factors in populations (not individuals)
- Informs prevention, treatment planning, cost estimation
Psychiatric epidemiology applies general epidemiologic models to largely chronic mental disorders rather than acute infectious models
Stuart L. Morris’ 7 classical “uses of epidemiology” (1957) remain central
- Historical study
- Compare disorder patterns across time
- Example: rising projected rates of Alzheimer’s disease as U.S. “baby boomers” age
- Community diagnosis
- Describe morbidity/mortality; identify vulnerable sub-populations; estimate burden (productivity loss, costs, premature death)
- WHO Global Burden of Disease (GBD) project quantifies this (discussed below)
- Health-services utilization
- Clarifies treatment needs vs. actual service demand
- De facto U.S. mental-health system spans specialty, primary care, schools, social services, alternative medicine
- Primary care = most utilized mental-health setting worldwide
- Psychiatric “treatment need” is ambiguous (continuum of distress, spontaneous remission, functional impairment)
- Completing the clinical picture
- Community data illuminate natural history, prodrome, untreated course, risk/protective factors
- Berkson bias: hospitalized samples over-represent comorbidity ➔ potentially misleading risk conclusions
- Identification of syndromes
- DSM-III (1980) introduced operational categories; yet symptom-based, categorical approach shows limits
- DSM-5 added cross-cutting dimensional measures; NIMH RDoC pursues domain-based neuroscience framework
- Epidemiology tests community prevalence, disability, co-aggregation to refine/merge syndromes (e.g., depression–anxiety–somatization cluster)
- Assessing individual risk
- Risk estimates guide clinical/actuarial decisions (e.g., cholesterol, smoking ➔ cardiovascular risk)
- Translation of psychiatric risk knowledge to personalized prevention is nascent
- Identifying causes
- Goal = causal pathways → targeted intervention
- Psychiatric etiologies multifactorial (gene × environment)
- Famous example: 5-HTTLPR polymorphism × stress ➔ adult depression (yet not yet clinically actionable)

Measures of Disease Frequency

Three prerequisite elements: case definition, target population, specified time period
Prevalence = proportion with disorder at/within time frame
- Current/Point or 1-Month
- Point ideal but hard (daily waxing/waning); 1-month often used
- 1-Year
- Aligns with annual service planning; can extrapolate: \text{Expected Cases}=\text{1-yr prevalence}\times\text{population} (if demographics stable)
- Lifetime
- Captures risk over life course; critical for risk-factor studies
- Not useful for short-term planning; vulnerable to recall bias (esp. elderly)
Incidence
- Counts new cases over period; denominator excludes existing cases
- First incidence restricts to first-ever episodes
- Mental disorders are rare events → require large cohorts, long follow-up, person-time denominators to handle dynamic populations
Relationship
- P \approx I \times d where d = mean duration
- Effective cure ↓ duration ↓ prevalence; life-prolonging treatment of chronic disease ↑ prevalence (e.g., AIDS after HAART)
Comorbidity
- Random (expected) vs. non-random (prevalence of A higher in those with B)
- Implications: shared risk, overlapping nosology, or diagnostic artefact (e.g., substance use counted within antisocial PD in DSM-III)

Measures of Effect & Association

Risk factor (↑ likelihood) vs. Protective factor (↓ likelihood)
Relative Risk (RR)
- 2×2 table cells: a (exposed + disease), b (exposed – disease), c (unexposed + disease), d (unexposed – disease)
- RR=\frac{a/(a+b)}{c/(c+d)}
- RR>1 risk; RR<1 protection; RR=1 no effect
Odds Ratio (OR)
- Case–control design approximation: OR=\frac{a\times d}{b\times c}
Confounding: third variable associated with both exposure and outcome, outside causal path (e.g., alcohol confounds smoking–suicide link)
- Control via stratification, regression → “adjusted” estimates
Mediation: intermediate variable explains pathway (e.g., CBT → ↓anxiety → ↓depression)
Effect Modification / Moderation: strength/direction of E→O varies by levels of A (e.g., medication efficacy differs by adherence)
Attributable Fraction (AF)
- Among exposed: AF_E=\frac{RR-1}{RR}
- Population: AFP = \frac{Pe(RR-1)}{1 + Pe(RR-1)} with Pe = prevalence of exposure

Epidemiologic Study Designs

Experimental

Clinical trials (efficacy vs. effectiveness)
- Randomized, controlled, often double-blind; control = placebo or active comparator (ethical shift away from long placebo in severe illness)
Prevention trials
- Enroll at-risk but unaffected; costly (large N, long follow-up); sometimes non-randomized universal interventions

Observational

Analytic
- Cohort: select by exposure → follow for outcome
- Pros: temporal clarity, measure incidence, collect confounders before outcome
- Cons: large/expensive, exposure misclassification risk
- Case-control: select by outcome → ascertain past exposure; OR approximates RR
- Pros: efficient for rare diseases, multiple exposures; Cons: recall bias, control selection
Descriptive
- Cross-sectional, repeated, longitudinal; provide prevalence, service use, sociodemographics; foundation for further studies

Sampling Strategies

Sampling frame = list of population
Probability sampling (generalizable; calculable error)
- Simple random
- Stratified random (ensures subgroup representation, ↑ precision)
Non-probability sampling (no frame; limited generalizability)
- Convenience, consecutive, quota, snowball (hidden populations)

Measurement: Reliability & Validity

Reliability

Test–retest (stability over time)
Inter-rater (consistency across raters)
Statistic: Cohen’s \kappa; 0 \le \kappa \le 1 (>0.4 often ‘adequate’)

Validity

Content: items fully cover construct (expert judgment)
Criterion
- Concurrent vs. Predictive; metrics:
- Sensitivity =\frac{a}{a+c}, Specificity =\frac{d}{b+d}
- Positive Predictive Value =\frac{a}{a+b}, Negative Predictive Value =\frac{d}{c+d}
Construct: theoretical coherence (Robins & Guze validators: clinical description, lab studies, delimitation, follow-up, family studies)

Case Identification & Diagnostic Systems

Dependence on symptom self-report + informants; parallels with other non-lab disorders (e.g., migraine)

DSM Evolution

DSM-I (1952) & DSM-II (1968): glossary, poor reliability
U.S.–U.K. study highlighted diagnostic variability
DSM-III (1980): symptom criteria, atheoretical, spurred DIS & ECA
DSM-III-R ➔ DSM-IV (1994) ➔ DSM-5 (2013) & DSM-5-TR (2022)
- DSM-5 removed ‘clinical significance’ criterion in many disorders but definition notes usual association with distress/impairment

ICD

WHO’s global morbidity standard; ICD-11 (2019) harmonized more closely with DSM-5 (≈30 % identical disorders, ≈10 % minor differences)

Disability & ICF

WHO International Classification of Functioning, Disability & Health (ICF) – components: Body Functions/Structures, Activities & Participation, Environmental, Personal (uncoded)

Assessment Instruments

Adults

DIS (based on DSM-III; fully structured; computer algorithms)
CIDI (WHO; DSM & ICD compatible; computerised)
- UM-CIDI (National Comorbidity Survey version)
- WMH-CIDI (World Mental Health surveys; added new dx, disability modules)
- CIDI 5.0 aligns with DSM-5 / ICD-11
AUDADIS-5 (focus on substance + comorbid; used in NESARC)
SCID-5 (semi-structured; clinician-administered; multiple versions)

Children & Adolescents

DISC-5 (lay-administered; parent/child versions)
CAPA / PAPA / YAPA (interviewer-based; Duke)
K-SADS-PL DSM-5 (semi-structured; lifetime)

Disability Scales

WHODAS 2.0 (36-item or 12-item; ICF aligned; good reliability)
CGAS / BIS for youth global functioning

Major Epidemiologic Surveys

Epidemiologic Catchment Area (ECA) (1980s)

5 U.S. sites; DIS; N ≈ 18.6k household + 2.3k institutional
Found high community prevalence, comorbidity; many onsets in childhood; under-treatment; highlighted Berkson bias; influenced parity laws

National Comorbidity Survey series

NCS (1990-92): UM-CIDI; ages 15–54; N ≈ 8k
NCS-R (2001-03): WMH-CIDI; N ≈ 9.3k
- 1-yr prevalence any disorder \approx 26\%; service use uneven (Table 5.1-8)
NCS-2: longitudinal follow-up of original NCS
NSAL & NLAAS: focused on African American, Afro-Caribbean, Latino, Asian populations ➔ disparity data

NESARC (Alcohol & Related Conditions)

Wave 1 (2001-02) N = 43k; AUDADIS-IV
Wave 2 (2004-05) re-interview N = 34.7k → incidence (Table 5.1-11)
NESARC-III (2012-13) N = 36.3k; AUDADIS-5; includes genetic data (~23k)

NSDUH (annual since 1971)

Face-to-face household, age ≥12; 2019 N ≈ 50.7k adults
- 2019 past-year substance use: marijuana 35 % (18–25); alcohol 72 % etc.

World Mental Health (WMH) Initiative

29 countries, >160k respondents; WMH-CIDI; large cross-national comparisons (Table 5.1-10)

Child / Youth Surveys

MECA (Methods for the Epidemiology of Child & Adolescent Mental Disorders) – methods study, N ≈ 1.3k
NCS-A (2001-04): adolescents 13–17, N ≈ 10k; any 1-yr disorder prevalence 40 %
YRBSS: biennial high-school risk behaviours → proxy mental-health risk
NSCH: parent-reported mental-health history across 0–17 yrs (Fig. 5.1-1)

Global Burden of Disease (GBD)

DALY = YLL + YLD; 1 DALY = 1 lost healthy year
GBD 2019: mental disorders ≈ 5 % total DALYs (after cardiovascular, neoplasms etc.)
- Top mental DALY contributors: depressive > anxiety > alcohol/drug > schizophrenia (Table 5.1-13)
- Trends ↑ since 1990 (except alcohol use disorder) (Fig. 5.1-2)
GBD 2010 added mental disorders as risk factors (e.g., depression → ischemic heart disease; substance injection → HIV/HCV)

Emerging Issue: COVID-19 Pandemic

Declared March 11 2020
Meta-analyses show ↑ suicide ideation, depressive & anxiety disorders, PTSD, burnout, eating-disorder symptoms, violence
Risk factors
- Frontline clinicians: low support, exhaustion, poor health
- General population: sleep disturbance, loneliness, excessive news, screen time
Protective factors
- Structured routines, nature exposure, limited passive media use, adequate sleep
Early UK longitudinal study (N > 8k families) ➔ ↑ parent-reported behavioural & attentional problems during lockdown; stressor dose–response effects
Long-term psychiatric impact still under evaluation

Key Equations & Metrics (Quick Reference)

Prevalence–Incidence relationship: P \approx I \times d
Relative Risk: RR=\frac{a/(a+b)}{c/(c+d)}
Odds Ratio: OR=\frac{a\times d}{b\times c}
Sensitivity: Se=\frac{a}{a+c}; Specificity: Sp=\frac{d}{b+d}
Positive Predictive Value: PPV=\frac{a}{a+b}; Negative Predictive Value: NPV=\frac{d}{c+d}
Attributable Fraction (exposed): AF_E=\frac{RR-1}{RR}
Population-Attributable Fraction: AFP=\frac{Pe(RR-1)}{1+P_e(RR-1)}

Ethical, Philosophical & Practical Considerations

Balancing rigorous diagnosis vs. cultural relevancy and dimensional nuance
Equity: documenting disparities (race, ethnicity, SES) to guide fair resource allocation
Informed consent & translation standards critical in multinational surveys
Ethical limitations on placebo use in severe mental illness trials
Pandemic highlights need for rapid, scalable mental-health surveillance and interventions

Study-Guide Tips

Master disease-frequency measures and formulas; practice converting between prevalence/incidence
Sketch 2×2 tables quickly; compute RR, OR, Se/Sp, PPV/NPV on sample numbers
Compare strengths/limitations of cohort vs. case-control designs; memorize classic biases (Berkson, recall)
Recognize key instruments and which surveys used them (DIS-ECA, CIDI-NCS/WMH, AUDADIS-NESARC, DISC-child)
Link DSM/ICD evolution to epidemiologic comparability challenges
Understand DALY concept; recall that mental disorders drive YLDs more than YLLs
Track COVID-19 emerging data; anticipate exam questions on mental-health impact and protective behaviours