Epidemiology and Base Rates

Epidemiology and Base Rates: How Common Is It, and Why That Matters

Epidemiology asks how often a disorder happens, who gets it, and when. The EPPP tests it two ways. First, the vocabulary: incidence, prevalence, and risk. Second, the numbers: which disorders are common, which sex they lean toward, and what age they start. Then it asks you to use those numbers, because how common a disorder is in the room changes what a test score means.

Why This Matters for Psychologists

Two psychologists give the same borderline personality screen. One works in a primary care clinic. The other works on an inpatient unit. A positive score means something very different in each building. The test didn't change. The base rate did. Base rates also tell you which diagnosis to suspect first, what tends to show up alongside it, and which age group to watch.

Incidence vs. Prevalence

Incidence counts new cases. It is the number of people who develop a disorder during a set time, divided by the population at risk: people who didn't have it when the clock started. Incidence tells you how fast new cases appear.

Prevalence counts all existing cases, old and new, as a share of the population. It can be taken at one moment or across a period. Prevalence tells you how big the problem is, which matters for planning services (Saha et al., 2005).

Most DSM-5-TR mood and anxiety chapters report 12-month prevalence. The wider the window, the bigger the number. Lifetime prevalence is always at least as large as 12-month prevalence, which is at least as large as point prevalence. Each wider window catches everyone in the narrower one, plus more.

Lifetime morbid risk runs higher than lifetime prevalence, because a 25-year-old who is well today still has decades of risk ahead. In the NCS-R, lifetime prevalence of any disorder was 46.4%, but projected risk by age 75 was 50.8% (Kessler et al., 2005a). For schizophrenia, review medians were 4.0 per 1,000 for lifetime prevalence and 7.2 per 1,000 for lifetime morbid risk (Saha et al., 2005).

Worked Example: One Town, One Year

Picture a town of 10,000 adults, and follow depression for one calendar year.

• On January 1, 400 adults have major depressive disorder. Point prevalence = 400 / 10,000 = 4%.
• During the year, 96 of the 9,600 people who were well on January 1 develop it. Incidence = 96 / 9,600 = 1% per year. The 400 existing cases stay out of the bottom number (the denominator). They can't become new cases.
• Anyone depressed at any time that year = 400 + 96 = 496. 12-month prevalence = 496 / 10,000, about 5%.
• If 1,600 adults say they have ever had depression, lifetime prevalence = 16%.

The trap: old cases count toward prevalence but never toward incidence.

Prevalence ≈ Incidence × Duration

In a stable population, prevalence, incidence, and duration are linked, so any two give you the third (Freeman & Hutchison, 1980). The classic shorthand is prevalence ≈ incidence × duration (Oman et al., 1999).

Think of a bathtub. Incidence is the faucet pouring in new cases. Recovery and death are the drain. Prevalence is the water level. Prevalence rises as new cases appear and falls as people recover or die.

Quick math: if 10 new cases per 1,000 people appear each year, and each case lasts 2 years on average, about 20 per 1,000 (2%) have the disorder at any moment.

What the formula predicts:

• Long-lasting disorders pile up. Low incidence plus a long course can still mean high prevalence.
• Brief disorders look rare in a snapshot, even when many people get them.
• A treatment that shortens episodes lowers prevalence without changing incidence.
• A treatment that keeps people alive but doesn't cure them raises prevalence.

A real case: older women had higher prevalence of physical disability than older men, but their incidence was not significantly different. The authors concluded that longer duration could explain the gap, through lower recovery and death rates in women (Oman et al., 1999). So to ask whether a group is more likely to develop a disorder, compare incidence, not prevalence. Schizophrenia shows the same split: men have higher incidence (about 1.4 to 1), yet reviews find no sex difference in prevalence (McGrath et al., 2008; Saha et al., 2005). DSM-5-TR notes a related pattern for PTSD: women in the general population have it for a longer duration than men do (American Psychiatric Association [APA], 2022).

Caveat: the shortcut assumes a steady population whose rates aren't shifting much (Freeman & Hutchison, 1980).

Measuring Risk: Relative Risk, Odds Ratio, Attributable Risk

Risk just means probability: the share of a group who develop the outcome (Richardson et al., 2023). Three numbers compare people exposed to a possible risk factor with people who weren't exposed.

Relative risk (RR), also called the risk ratio, divides the exposed group's risk by the unexposed group's risk. RR = 1 means no difference. RR = 2 means double the risk. RR says nothing about absolute risk; doubling a tiny risk still leaves a small one (Tenny & Hoffman, 2023b).

Odds ratio (OR) divides two odds instead of two risks. Odds compare the chance something happens with the chance it doesn't (Richardson et al., 2023). The standard exam pairing: RR comes from a cohort study, which follows exposed and unexposed people forward and counts new cases, like the made-up study below. OR comes from a case-control study, which starts with people who already have the disorder, finds similar people who don't, and looks back for exposures (Tenny et al., 2023; Tenny & Hoffman, 2026). The researcher picks the mix of cases and controls (say, two controls per case), so the study can't measure risk directly. It compares the odds of past exposure instead (Tenny et al., 2023; Persoskie & Ferrer, 2017).

Attributable risk (AR) subtracts the unexposed group's risk from the exposed group's risk. It is also called the risk difference. If the factor truly causes the outcome, AR tells you how many extra cases the factor produces (Persoskie & Ferrer, 2017). Because it is an absolute number, it speaks to real-world impact (Richardson et al., 2023).

Made-up study: researchers follow 1,000 teens who were bullied and 1,000 who weren't, and count who has depression by age 25.

• RR = .20 / .10 = 2.0. Bullied teens had twice the risk.
• OR = .25 / .111 = 2.25.
• AR = .20 - .10 = .10. If bullying were the cause, it would add 100 cases per 1,000 bullied teens.

Notice the OR came out bigger than the RR. When an outcome is rare, odds and risk are nearly the same, so the OR roughly equals the RR. When the outcome is common, the OR grows faster, and reading it as an RR overstates the effect (Persoskie & Ferrer, 2017; Richardson et al., 2023). Caveat: in a case-control study, whether the OR can stand in for the RR depends on how the controls were picked (Persoskie & Ferrer, 2017).

A real example: in the Epidemiologic Catchment Area study, people with a mental disorder had an OR of 2.7 for also having an addictive disorder (Regier et al., 1990).

Remember that a link is not a cause. Case-control studies can show an association between an exposure and a disorder, but they can't prove the exposure caused it (Tenny et al., 2023).

Comorbidity

Comorbidity means having two or more disorders at once. In the community, it is common, and it piles up in a minority of people.

• In the original National Comorbidity Survey, the 14% of people with three or more lifetime disorders accounted for more than half of all lifetime disorders. That small group also held most of the severe cases (Kessler et al., 1994).
• In the NCS-R, 55% of people with a 12-month disorder had only one diagnosis, 22% had two, and 23% had three or more. Serious cases clustered in the highly comorbid few (Kessler et al., 2005b).
• 72.1% of people with lifetime major depressive disorder also had another disorder (Kessler et al., 2003).
• In the ECA, 47.2% of people with lifetime major depression also met criteria for an anxiety disorder. Among them, anxiety tended to start years earlier: average onset 16.4 for any anxiety disorder vs. 23.2 for depression (Regier et al., 1998).

It's like traffic. Most roads flow fine, but a few intersections hold most of the gridlock.

Comorbidity is even higher in clinics. In the ECA, people treated in specialty mental health and addiction settings had significantly higher odds of comorbid disorders (Regier et al., 1990). When a substance problem and another disorder overlap, it can be hard to tell which is causing which symptoms. Observing the person after use and withdrawal have stopped helps sort it out (Spielman et al., 2020). Each disorder's common comorbidities are covered in its own diagnosis lesson.

The Big Surveys

Epidemiologic Catchment Area (ECA) Study

• A National Institute of Mental Health program; its landmark reports appeared in 1984.
• About 20,000 people across five U.S. sites, including people living in institutions such as prisons.
• Lay interviewers used the Diagnostic Interview Schedule (DIS) to make DSM-III diagnoses (Robins et al., 1984).
• People were re-interviewed one year later, so the ECA could estimate incidence, not just prevalence (Regier et al., 1984).

Key findings: 15.4% of adults met criteria for a disorder in the past month (Regier et al., 1988). Lifetime prevalence was 22.5% for any non-substance mental disorder, 13.5% for alcohol abuse or dependence, and 6.1% for other drugs (Regier et al., 1990). In the first three sites, the most common lifetime diagnoses were alcohol abuse and dependence, phobia, major depressive episode, and drug abuse and dependence. Antisocial personality and alcohol problems leaned male; depression and phobias leaned female. Young adults (25 to 44) had the highest rates of most disorders (Robins et al., 1984).

National Comorbidity Survey (NCS) and Its Replication (NCS-R)

The NCS was the first U.S. survey to use a structured psychiatric interview with a nationally representative (probability) sample. It covered ages 15 to 54 and used DSM-III-R. Nearly 50% reported at least one lifetime disorder and close to 30% a 12-month disorder. The most common disorders were major depressive episode, alcohol dependence, social phobia, and simple phobia. Fewer than 40% of people who had ever had a disorder had ever gotten professional treatment (Kessler et al., 1994).

The NCS-R interviewed 9,282 English-speaking adults from 2001 to 2003. It used a fully structured interview, the World Mental Health version of the Composite International Diagnostic Interview (CIDI), and DSM-IV criteria (Kessler et al., 2005a).

Sources: Kessler et al. (2005a, 2005b).

The standard exam answer: anxiety disorders are the most common class, lifetime and 12-month. For 12-month, the order is anxiety, then mood, then impulse-control, then substance use. For lifetime, impulse-control edges out mood. Caveat: NCS-R impulse-control disorders (ODD, conduct disorder, ADHD, intermittent explosive disorder) were mostly assessed only in people aged 18 to 44 (Kessler et al., 2005b).

More NCS-R anchors:

• The most common single 12-month disorders were specific phobia (8.7%), social phobia (6.8%), and major depressive disorder (6.7%) (Kessler et al., 2005b).
• Lifetime major depressive disorder was 16.2%, and only 21.7% of 12-month cases got adequate treatment (Kessler et al., 2003).

Caveat: numbers move with the interview and criteria. A later survey using DSM-5 criteria found 12-month and lifetime major depressive disorder at 10.4% and 20.6% (Hasin et al., 2018).

Age of Onset

The NCS-R found most disorders start young. Median age of onset was 11 for anxiety and impulse-control disorders, 20 for substance use, and 30 for mood disorders. Half of all lifetime cases start by age 14, and three-quarters by 24 (Kessler et al., 2005a).

DSM-5-TR fills in the disorder-level picture (APA, 2022):

• Early childhood: autism spectrum disorder is usually recognized at 12 to 24 months. ADHD is most often identified in elementary school. Specific phobia has a median onset of 7 to 11, mostly before age 10.
• Early adolescence: social anxiety disorder has a median onset of 13.
• Late teens to 20s: OCD (mean 19.5), panic disorder (U.S. median 20 to 24), bipolar I (U.S. mean 22), schizophrenia (late teens to mid-30s), anorexia and bulimia (adolescence or young adulthood), alcohol use disorder (peaks late teens to mid-20s). Major depressive disorder becomes much more likely after puberty, and incidence peaks in the 20s.
• Later: generalized anxiety disorder has a mean onset of 35 in North America, later than the other anxiety disorders.
• By definition: antisocial personality disorder can't be diagnosed before 18.

Sex shifts the clock for some disorders. Men develop OCD earlier, nearly 25% before age 10. Schizophrenia peaks in the early to mid-20s for men and the late 20s for women, and women show a second peak in midlife.

Think of a dance set. Phobias drop in the intro. Social anxiety comes in with middle school. The big mood and psychosis tracks hit in the 20s. GAD walks in once the set is already rolling.

High-Yield Base Rates for the Most-Tested Disorders

All figures below come from DSM-5-TR prevalence sections (APA, 2022) unless marked. U.S. figures unless noted. Criteria live in each diagnosis lesson.

How to read the sex column:

• Lean female: depression and most anxiety disorders (about 2:1), PTSD, and eating disorders.
• Lean male: ADHD, autism, alcohol use disorder, and antisocial PD.
• Near even: bipolar I, schizophrenia prevalence, and borderline PD in community samples.

Age matters too. U.S. 12-month depression is three times higher at ages 18 to 29 than at 60 and older. Twelve-month alcohol use disorder (DSM-IV criteria) was 16.2% at ages 18 to 29 but 1.5% at 65 and older (APA, 2022).

Social Class and Place

Many disorders are more common at lower socioeconomic status (SES) (Kessler et al., 1994). Two classic explanations compete. Social causation: the stress and hardship of poverty raise risk. Social selection, also called downward drift: the disorder pushes people down the economic ladder. In a classic Israeli birth-cohort study, selection fit schizophrenia better, while causation fit depression in women and antisocial PD and substance use disorders in men (Dohrenwend et al., 1992). Place matters too. Migrants have higher schizophrenia incidence and prevalence, with a median prevalence ratio of 1.8 vs. native-born people, and DSM-5-TR links urban living to higher rates (Saha et al., 2005; McGrath et al., 2008; APA, 2022).

Base Rates Change What a Positive Score Means

In assessment, the base rate is how common a condition is in the group you are testing. The formulas for sensitivity, specificity, positive predictive value (PPV), and negative predictive value (NPV) are covered in the Criterion-Related Validity lesson. Here is the epidemiology side.

Imagine one borderline personality screen, with made-up sensitivity of .80 and specificity of .90, used on 1,000 patients in each of three settings. The base rates come from DSM-5-TR: about 6% in primary care, 10% in outpatient mental health, and 20% in psychiatric inpatients (APA, 2022).

Same test. Same sensitivity and specificity. In primary care, about two of every three positive screens are false alarms. As the base rate rises, PPV climbs and NPV slips. The rule: when a condition is rare in your setting, trust negatives and doubt positives. The Criterion-Related Validity lesson also notes that sensitivity and specificity can shift between settings.

Base Rate Neglect

Base rate neglect (the base rate fallacy) is ignoring how common something is and judging from the vivid details of one case. The Social Cognition lesson on errors, biases, and heuristics covers it in depth. The clinical version: treating a positive screen as a diagnosis in a setting where the disorder is uncommon.

The fix is to think in counts. People, experts included, struggle to combine probabilities, but they do much better when the same facts come as natural frequencies (Hoffrage et al., 2000). "34% PPV" is hard to feel. "Out of 142 people who screened positive, 48 actually have it" is easy to picture.

Base Rates Depend on the Setting

Your caseload is not the population. DSM-5-TR shows how far base rates move (APA, 2022):

• Borderline PD: median 2.7% in the community, about 6% in primary care, 10% in outpatient clinics, and 20% on inpatient units.
• Antisocial PD: 0.6% to 3.6% in the community, but over 70% among men with the most serious alcohol use disorders, and in substance clinics, prisons, and other forensic settings.
• PTSD: one-third to more than half of people who survived rape, combat, captivity, or internment and genocide that targeted people for their ethnic group or politics. Rates are also higher in jobs with trauma exposure, such as police, firefighters, and emergency medical staff.
• ADHD: higher in foster children and correctional settings.

Sex ratios shift with setting too. Borderline PD looks female-dominated in clinics but shows no sex difference in the community, which may reflect more help-seeking by women. Social anxiety disorder is more common in women in the community, yet men are equal or slightly ahead in clinics. In GAD, about 55% to 60% of clinic patients are women, compared with about two-thirds in community studies (APA, 2022).

Judging how common a disorder is from your caseload is like judging how often people break bones by standing in an emergency room. You only see the people who came in.

EPPP Traps and Common Misconceptions

Misconception 1: "Incidence and prevalence are the same thing."

• Reality: Incidence counts only new cases among people at risk. Prevalence counts everyone who has the disorder, old and new.

Misconception 2: "A disorder with high prevalence must have high incidence."

• Reality: Prevalence ≈ incidence × duration. A long course can make a low-incidence disorder common.

Misconception 3: "An odds ratio of 3 means three times the risk."

• Reality: That's an RR. ORs roughly match RRs only when the outcome is rare; with common outcomes, ORs overstate risk.

Misconception 4: "Mood disorders are the most common class."

• Reality: Anxiety disorders lead, lifetime and 12-month. Specific phobia is the most common single 12-month disorder in the NCS-R.

Misconception 5: "A test with good sensitivity and specificity gives trustworthy positives anywhere."

• Reality: PPV falls as the base rate falls. In low base rate settings, most positives can be false.

Misconception 6: "The sex ratio in my clinic is the sex ratio in the population."

• Reality: Help-seeking changes who shows up. Borderline PD is female-heavy in clinics but even in the community.

Misconception 7: "Schizophrenia affects 1% of people."

• Reality: DSM-5-TR gives lifetime prevalence of 0.3% to 0.7%, and reviews put median lifetime morbid risk near 0.7% (McGrath et al., 2008). The family-risk table in the Schizophrenia Spectrum lesson uses a 1% general-population figure; for a DSM-5-TR prevalence item, pick 0.3% to 0.7%.

Memory Aids

• "IN-cidence = NEW people coming IN." Prevalence is everyone already in the room.
• The bathtub: faucet (incidence) × time in the tub (duration) = water level (prevalence).
• "Half and a quarter; 14 and 24": about half of adults have a lifetime disorder, about a quarter in a year; half of cases start by 14, three-quarters by 24.
• NCS-R class order: 12-month is "A MIS" (Anxiety, Mood, Impulse-control, Substance). Lifetime swaps the middle two: "AIMS." Anxiety always comes first.
• Sex patterns, "the four A's": ADHD, Autism, Alcohol use disorder, and Antisocial PD lean male. Depression, anxiety, PTSD, and eating disorders lean female. Bipolar I and schizophrenia prevalence sit near even.
• "Rare means positives lie." Low base rate, low PPV.

Key Takeaways

• Incidence = new cases per population at risk over time. Prevalence = all existing cases at one moment (point prevalence) or over a span (period, 12-month, lifetime). Lifetime morbid risk projects forward and runs higher than lifetime prevalence.
• Prevalence ≈ incidence × duration in a steady population. Shorter episodes lower prevalence; keeping people alive without curing them raises it.
• RR compares risks (the cohort statistic), OR compares odds (the case-control statistic), and AR is the risk difference, the absolute number of extra cases if the factor is causal. ORs overstate RRs when outcomes are common.
• Comorbidity is common and concentrated: serious cases cluster in a small, highly comorbid group, and clinics see more of it than the community.
• ECA: about 20,000 people, five sites, DIS, DSM-III, lay interviewers, one-year follow-up for incidence. NCS-R: 9,282 adults, CIDI, DSM-IV, lifetime any disorder 46.4%, 12-month 26.2%, anxiety the most common class.
• Onset: median 11 for anxiety and impulse-control, 20 for substance, 30 for mood; half by 14, three-quarters by 24. GAD starts later than other anxiety disorders.
• Sex patterns: depression, anxiety, PTSD, and eating disorders lean female; ADHD, autism, alcohol use disorder, and antisocial PD lean male; bipolar I and schizophrenia prevalence are near even.
• SES: social causation (hardship raises risk) vs. social selection or downward drift (the disorder lowers SES). Drift fits schizophrenia best.
• Base rates drive PPV: the same screen gave 34% PPV in primary care and 67% on an inpatient unit. Watch for base rate neglect.
• Setting changes base rates and sex ratios: borderline PD rises from about 6% in primary care to 20% inpatient; antisocial PD tops 70% in some forensic and addiction samples.

Now cover the base-rate table and rebuild the three-setting screen example from memory. If you can explain why PPV moved, you own this topic.