Skip to content

Epidemiology and course

TL;DR — Cross-national household surveys estimated lifetime prevalence of 0.6% for bipolar I, 0.4% for bipolar II and 1.4% for subthreshold bipolar disorder, but survey algorithms can nearly double an estimate (Merikangas 2011, PMID 21383262; Mitchell 2013, PMID 22906117). Course is recurrent and heterogeneous: adults in a three-year life-chart cohort were euthymic only about half the time, while young people in a two-year cohort had syndromal or subsyndromal symptoms during 60% of follow-up (Joffe 2004, PMID 14996142; Birmaher 2006, PMID 16461861). Depression, subthreshold symptoms and cognitive/functional impairment account for much of the longitudinal burden, and employment rates across reviewed bipolar cohorts ranged from 40% to 75% (Dominiak 2022, PMID 36090375). Conversion from an initial major-depression diagnosis is real but sample-dependent: 5.84% over 13 years in a Swedish population register, 7.4% over 15 years after psychiatric hospitalization, and 12.4% over three years in a Brazilian cohort (Rhee 2023, PMID 37427550; Baryshnikov 2020, PMID 32385906; Oliveira 2021, PMID 33493732). Epidemiological numbers therefore require the case definition, ascertainment setting, observation window and age structure beside the estimate.

Prevalence is definition-dependent

Estimate Population and method Interpretation Source
Lifetime: bipolar I 0.6%, bipolar II 0.4%, subthreshold 1.4%, spectrum 2.4%; 12-month spectrum 1.5% 61,392 adults, 11 countries; household WMH-CIDI interviews Common method across countries, but lay-administered retrospective assessment Merikangas 2011, PMID 21383262
12-month bipolar disorder 0.9% recalibrated versus 1.7% unrecalibrated Australian national survey; same WMH-CIDI data with two algorithms A technical algorithm choice almost doubled prevalence and changed clinical composition Mitchell 2013, PMID 22906117
Lifetime 1.2%; 12-month 0.6% Singapore national survey, first wave 69.4% had another lifetime mental disorder; 52.6% had a chronic physical condition Subramaniam 2013, PMID 23017543
Lifetime spectrum 3.1%: bipolar I 1.5%, bipolar II 0.03%, subthreshold 1.6% Singapore 2016–2018, n=6,126, 69.5% response Large change from the earlier survey illustrates sampling and algorithm sensitivity Teh 2020, PMID 32469825
Lifetime 0.1%–1.83% Systematic review of African community surveys Sparse data; missed-diagnosis estimates reached 36.2% Esan 2016, PMID 26155900
Approximately 1%–1.5% compromise estimate Older synthesis of community versus treated-case studies Community estimates risk false positives; treated samples miss untreated cases Bebbington 1995, PMID 8560330

These estimates should not be averaged. They answer different questions: categorical bipolar I/II prevalence, a spectrum that includes subthreshold syndromes, diagnoses recorded in care, or a structured-interview approximation. Role impairment in WMH data was similar across spectrum subtypes even though symptom severity and suicidality increased from subthreshold presentations to bipolar I (Merikangas 2011, PMID 21383262).

Incidence and age at onset

A Dutch primary-care record cohort of about 800,000 people estimated an overall recorded incidence of 0.70 per 10,000 person-years (95% CI 0.57–0.83), including 0.43 (0.34–0.55) for bipolar I and 0.19 (0.13–0.27) for bipolar II (Kroon 2013, PMID 23531096). Peaks appeared at ages 15–24 and 45–54, but the later peak may reflect recognition and recording rather than biological onset. Incidence was higher in deprived areas and did not differ materially by sex or urbanicity (Kroon 2013, PMID 23531096).

Age/course observation Number Boundary Source
Most first service contact 15–45 years Administrative contact is not symptom onset Almeida 2002, PMID 12475092
First contact at ≥65 years 492/6,182 patients (8%) Late contact did not produce a clearly bimodal distribution Almeida 2002, PMID 12475092
Recorded organic mental disorder, late versus earlier onset 2.8% vs 1.2% Difference was small; most late-onset patients lacked such a record Almeida 2002, PMID 12475092
Youth cohort mean age 13 years Specialty sample with bipolar I, II and NOS, not population incidence Birmaher 2006, PMID 16461861

Late first mania warrants attention to neurological, medication and medical causes, but epidemiology does not justify assuming that every late-onset presentation is secondary. Conversely, early mood symptoms do not guarantee a bipolar trajectory.

Time spent ill

The course is not adequately described by counting hospitalizations. In 138 adults followed with detailed life charts for about three years, bipolar I and II participants were euthymic approximately half the time; much of the remainder comprised minor or subsyndromal depressive and manic symptoms (Joffe 2004, PMID 14996142). In 711 adults followed across 13,191 visits over seven years, about half of visits recorded depressive, manic or hypomanic symptoms; women had more depressive visits, explained statistically by higher rapid-cycling and anxiety rates (Altshuler 2010, PMID 20231325).

Prospective cohort Follow-up Course signal Source
Adults with bipolar I/II, n=138 Mean ~3 years About half the time euthymic; subsyndromal/minor states dominated symptomatic time Joffe 2004, PMID 14996142
Youth with bipolar spectrum disorders, n=263 Mean 2 years 70% recovered from index episode; 50% had a syndromal recurrence; 60% of time symptomatic Birmaher 2006, PMID 16461861
Stanley network, n=711 7 years / 13,191 visits Symptoms present at about half of visits; euthymic visits increased over study participation Altshuler 2010, PMID 20231325
First-episode bipolar I, n=128 Mean 5.7 years; 6.5 episodes/person Most individual cycle-length slopes were random; no general progressive shortening Baldessarini 2012, PMID 21943930
Jorvi cohort, n=191 5 years Baseline predominant polarity: 16% manic, 36% depressive, 48% none Pallaskorpi 2019, PMID 30634112

The classic idea that every recurrence necessarily accelerates future cycles was not supported in the first-episode cohort: early and late euthymic intervals were similar, and only minorities showed acceleration or slowing (Baldessarini 2012, PMID 21943930). Treatment may modify observed course, so this is not proof against sensitization mechanisms; it is evidence against presenting cycle acceleration as inevitable.

Predominant polarity

Predominant polarity describes whether depressive or manic episodes dominate an individual’s observed history. In the five-year Jorvi study, the manic-polarity group spent more time euthymic, less time in major depression and had fewer suicide attempts than depressive/no-predominance groups, but classification changed with the timeframe used (Pallaskorpi 2019, PMID 30634112). It is a group-level course descriptor, not a stable biological subtype or guaranteed treatment selector.

Diagnostic delay and conversion

Bipolar disorder frequently begins with depression, so diagnosis may be revised only after later mania or hypomania. “Conversion” estimates are not interchangeable: a hospitalized psychotic-depression cohort has a different starting risk from community MDD.

Initial population Follow-up Conversion to bipolar disorder Predictors reported Source
Swedish first major-depression registration 13 years 5.84% (95% CI 5.72–5.96) Bipolar family genetic-risk score HR 2.73 (2.43–3.08); inpatient care HR 2.64 (2.44–2.84); psychotic depression HR 2.58 (2.14–3.11) Rhee 2023, PMID 37427550
Finnish first psychiatric hospitalization for unipolar depression, n=43,495 Up to 15 years 7.4% (95% CI 7.0–7.8) Psychotic depression SHR 2.0 (1.5–2.7); risk highest in first year Baryshnikov 2020, PMID 32385906
Brazilian prospective MDD cohort 3 years 12.4% Younger first depression, bipolar family history, illicit-substance use and lower education Oliveira 2021, PMID 33493732
Taiwan newly diagnosed MDD, n=2,820 About 10 years 19.0% Model relied heavily on early treatment and service-use variables; these may reflect clinician suspicion/severity Hu 2020, PMID 32242821
Youth presenting to early-intervention services, n=2,330 Longitudinal 4.3% new full-threshold bipolar disorder Mania-like experiences, poorer function, suicide attempt, childhood depression/anxiety and older age Carpenter 2022, PMID 33121545

Prediction remains probabilistic. Family history, psychosis, early onset and severe/recurrent depression can enrich risk, but none converts a depressive episode into bipolar disorder without a qualifying manic or hypomanic history (Nierenberg 2023, PMID 37815563).

Recurrence and developmental course

Youth in the Course and Outcome of Bipolar Youth cohort showed frequent symptom and polarity shifts: 20% of bipolar II participants converted to bipolar I, and 25% of bipolar-NOS participants converted to bipolar I or II during roughly two years (Birmaher 2006, PMID 16461861). This specialty cohort supports longitudinal reassessment but cannot be used as a population transition probability.

In 2,231 Japanese outpatients, 29.1% experienced mania/hypomania over one year. Lower baseline functioning, rapid cycling, personality disorder, bipolar I, substance abuse, and a manic/mixed baseline state predicted occurrence; baseline antidepressant prescription did not (Tokumitsu 2021, PMID 34972188). The result describes association in routine care, not a causal drug-safety experiment.

Disability and function

Symptomatic remission and functional recovery are not synonyms. A systematic review of 74 studies found bipolar-spectrum disorders adversely affected occupational status, performance, cost and salary; employment estimates for bipolar disorder ranged from 40% to 75% (Dominiak 2022, PMID 36090375). Heterogeneous definitions, welfare systems and sampling prevent a single global employment rate.

Functional driver Quantitative evidence Interpretation
Subthreshold depression Dominated symptomatic time in longitudinal cohorts Maintenance outcomes should include symptoms below episode thresholds (Joffe 2004, PMID 14996142)
Manic morbidity Episode density predicted later FAST impairment: β 6.54 (95% CI 0.43–12.65) More mania/hypomania independently tracked worse psychosocial function (Lomastro 2021, PMID 33792890)
Executive function Phonological-fluency deficit predicted FAST score: β −2.49 (−3.98 to −0.99) Cognition and episode burden made independent contributions (Lomastro 2021, PMID 33792890)
Comorbidity Three-quarters of WMH spectrum cases had another disorder Treatment systems organized around one diagnosis underestimate burden (Merikangas 2011, PMID 21383262)
Treatment contact Fewer than half with lifetime spectrum disorder received mental-health treatment; 25.2% in low-income countries The treatment gap is part of observed natural history (Merikangas 2011, PMID 21383262)

Global burden and measurement caveats

GBD 2019 estimated that DALYs from 12 mental disorders rose from 80.8 million (95% uncertainty interval 59.5–105.9) in 1990 to 125.3 million (93.0–163.2) in 2019, while age-standardized rates remained broadly stable (GBD 2019 Mental Disorders Collaborators 2022, PMID 35026139). The analysis warned that its years-of-life-lost estimates did not capture most premature mortality associated with mental disorders because deaths were assigned to downstream causes.

A bipolar-specific GBD 2021 analysis reported rising absolute incidence counts from 30.2 million in 1990 to 53.9 million in 2021 and regional differences by sociodemographic index (Jiang 2025, PMID 40499832). These model-derived counts should be kept distinct from observed household-survey prevalence and clinical incidence.

Rules for interpreting a course estimate

Reported quantity Required companion information Why it matters
Prevalence Diagnostic algorithm, spectrum boundary, sampling frame and response rate Recalibration changed one national 12-month estimate from 1.7% to 0.9% (Mitchell 2013, PMID 22906117)
Incidence Whether onset means first symptom, first episode, first diagnosis or first service contact Later recorded peaks may represent delayed recognition (Kroon 2013, PMID 23531096)
Time ill Sampling interval and whether subsyndromal weeks count Episode counts omit much of the observed morbidity (Joffe 2004, PMID 14996142)
Recurrence Index polarity, recovery definition and treatment exposure Enriched clinical cohorts cannot describe untreated natural history
Conversion Initial setting and competing diagnostic outcomes Psychotic inpatient depression carries different prior risk from community MDD (Baryshnikov 2020, PMID 32385906)
Function Instrument, welfare/employment context and current symptoms Cross-study employment estimates span 40%–75% (Dominiak 2022, PMID 36090375)
Burden Observed versus modeled data and treatment of associated mortality GBD disability estimates do not directly count most downstream deaths (GBD 2019 Mental Disorders Collaborators 2022, PMID 35026139)

No single number is the “true” course. The most transportable estimates state the population, case definition, observation schedule, treatment context and uncertainty interval.

Age, family history and course are probabilistic

Across 192 population studies (n=708,561), mood disorders as a block had 2.5% onset before age 14, 11.5% before 18 and 34.5% before 25; the block's peak was 20.5 years and its median 31 years. Bipolar disorder fell within a late-overlapping group with a median onset in the 30–35-year range, illustrating how pooled retrospective age-of-onset definitions can differ from clinical high-risk cohorts centered on adolescence and early adulthood (Solmi 2022, PMID 34079068). Both estimates can be true because they summarize different denominators and definitions.

Family aggregation is strong but cannot define a population screening strategy by itself. A Danish study followed 3,048,583 people for 80.4 million person-years and found progressively higher same-disorder risk with closer affected kinship; it also showed, across disorders, that most cases arise without an affected close relative (Pedersen 2025, PMID 40675715). The implication is dual: family history materially changes prior probability, but absence of known family history does not rule bipolar disorder out.

Prospective recurrence markers: signal before utility

In a 12-month cohort of 189 outpatients, 88 (46%) relapsed. Greater baseline activity-rhythm robustness was associated with lower recurrence (MESOR HR per count/min 0.993, 95% CI 0.988–0.997; amplitude HR 0.994, 0.988–0.999), while each hour later onset of the most-active 10-hour period was associated with depressive relapse HR 1.109 (1.001–1.215) (Esaki 2021, PMID 34645802). These small per-unit effects require calibration before clinical use.

A larger South Korean cohort followed 495 people with major depression or bipolar I/II for a mean 279.7 days; 270 episodes occurred in 135 participants. Internally evaluated three-day prediction AUCs were 0.937 for depression, 0.957 for mania and 0.963 for hypomania (Lee 2023, PMID 36146953). Because diagnoses were combined and external validation was not reported in the abstract, these values are proof of cohort-level predictability, not transportable bedside performance. More broadly, cognition and disability should be separated: stable cognitive deficits may constrain functioning even when symptoms remit, but the bipolar evidence base has historically measured real-world capacity less rigorously than schizophrenia research (Harvey 2010, PMID 20636633).

Open questions

  • What proportion of apparent regional prevalence variation remains after identical culturally validated interviews, sampling and recalibrated algorithms? (Mitchell 2013, PMID 22906117; Esan 2016, PMID 26155900)
  • Can first-depression conversion models retain calibration across community, outpatient and inpatient populations without using treatment as a proxy for clinician suspicion? (Hu 2020, PMID 32242821; Rhee 2023, PMID 37427550)
  • Which intervention reduces subsyndromal time and restores function rather than only delaying syndromal recurrence? (Joffe 2004, PMID 14996142; Dominiak 2022, PMID 36090375)
  • Is predominant polarity stable enough over decades to guide maintenance treatment at the individual level? (Pallaskorpi 2019, PMID 30634112)
  • How can burden models attribute premature deaths associated with bipolar disorder without double counting downstream causes? (GBD 2019 Mental Disorders Collaborators 2022, PMID 35026139; Biazus 2023, PMID 37491460)

References

  1. Bebbington P, et al. The epidemiology of bipolar affective disorder. Soc Psychiatry Psychiatr Epidemiol. 1995;30:279–292. PMID 8560330
  2. Almeida OP, et al. Bipolar disorder: similarities and differences between patients with illness onset before and after 65 years. Int Psychogeriatr. 2002;14:311–322. PMID 12475092
  3. Joffe RT, et al. A prospective, longitudinal study of percentage of time spent ill in bipolar I or II disorders. Bipolar Disord. 2004;6:62–66. PMID 14996142
  4. Birmaher B, et al. Clinical course of children and adolescents with bipolar spectrum disorders. Arch Gen Psychiatry. 2006;63:175–183. PMID 16461861
  5. Altshuler LL, et al. Gender and depressive symptoms in 711 patients with bipolar disorder evaluated prospectively. Am J Psychiatry. 2010;167:708–715. PMID 20231325
  6. Merikangas KR, et al. Prevalence and correlates of bipolar spectrum disorder in the World Mental Health Survey Initiative. Arch Gen Psychiatry. 2011;68:241–251. PMID 21383262
  7. Baldessarini RJ, et al. Episode cycles with increasing recurrences in first-episode bipolar-I disorder patients. J Affect Disord. 2012;136:149–154. PMID 21943930
  8. Mitchell PB, et al. Bipolar disorder in a national survey: impact of differing diagnostic algorithms. Acta Psychiatr Scand. 2013;127:381–393. PMID 22906117
  9. Subramaniam M, et al. Prevalence, correlates, comorbidity and severity of bipolar disorder: Singapore Mental Health Study. J Affect Disord. 2013;146:189–196. PMID 23017543
  10. Kroon JS, et al. Incidence rates and risk factors of bipolar disorder in the general population. Bipolar Disord. 2013;15:306–313. PMID 23531096
  11. Esan O, et al. Epidemiology and burden of bipolar disorder in Africa: a systematic review. Soc Psychiatry Psychiatr Epidemiol. 2016;51:93–100. PMID 26155900
  12. Pallaskorpi S, et al. Predominant polarity in bipolar I and II disorders: a five-year follow-up study. J Affect Disord. 2019;246:806–813. PMID 30634112
  13. Hu YH, et al. Predictors for early detection of conversion from major depression to bipolar disorder. JMIR Med Inform. 2020;8:e14278. PMID 32242821
  14. Baryshnikov I, et al. Diagnostic conversion from unipolar depression: a 15-year register study. Bipolar Disord. 2020;22:582–592. PMID 32385906
  15. Teh WL, et al. Prevalence and correlates of bipolar spectrum disorders in Singapore. J Affect Disord. 2020;274:339–346. PMID 32469825
  16. Oliveira JP, et al. Predictors of conversion from major depressive disorder to bipolar disorder. Psychiatry Res. 2021;297:113740. PMID 33493732
  17. Lomastro MJ, et al. Manic morbidity and executive function impairment as determinants of long-term psychosocial dysfunction. Acta Psychiatr Scand. 2021;144:72–81. PMID 33792890
  18. Tokumitsu K, et al. Real-world clinical predictors of manic/hypomanic episodes among outpatients. PLoS One. 2021;16:e0262129. PMID 34972188
  19. Carpenter JS, et al. Predicting emergence of full-threshold bipolar and psychotic disorders in young people. Psychol Med. 2022;52:1990–2000. PMID 33121545
  20. GBD 2019 Mental Disorders Collaborators. Global burden of 12 mental disorders, 1990–2019. Lancet Psychiatry. 2022;9:137–150. PMID 35026139
  21. Dominiak M, et al. The impact of bipolar spectrum disorders on professional functioning: a systematic review. Front Psychiatry. 2022;13:951008. PMID 36090375
  22. Rhee SJ, et al. Predictors of diagnostic conversion from major depression to bipolar disorder. Psychol Med. 2023;53:7805–7816. PMID 37427550
  23. Biazus TB, et al. All-cause and cause-specific mortality among people with bipolar disorder. Mol Psychiatry. 2023. PMID 37491460
  24. Nierenberg AA, et al. Diagnosis and treatment of bipolar disorder: a review. JAMA. 2023;330:1370–1380. PMID 37815563
  25. Jiang J, et al. Global, regional, and national burden of bipolar disorder, 1990–2021. J Affect Disord. 2025;389:119638. PMID 40499832
  26. Solmi M, et al. Age at onset of mental disorders worldwide: large-scale meta-analysis of 192 epidemiological studies. Mol Psychiatry. 2022;27:281–295. PMID 34079068
  27. Lee HJ, et al. Prediction of impending mood episode recurrence using real-time digital phenotypes in major depression and bipolar disorders in South Korea: a prospective nationwide cohort study. Psychol Med. 2023;53:5636–5644. PMID 36146953
  28. Esaki Y, et al. Association between circadian activity rhythms and mood episode relapse in bipolar disorder: a 12-month prospective cohort study. Transl Psychiatry. 2021;11:525. PMID 34645802
  29. Harvey PD, et al. Cognition and disability in bipolar disorder: lessons from schizophrenia research. Bipolar Disord. 2010;12:364–375. PMID 20636633
  30. Pedersen CB, et al. Absolute and relative risks of mental disorders in families: a Danish register-based study. Lancet Psychiatry. 2025;12:590–599. PMID 40675715