Hide table of contents

TL;DR: A significant fraction of depressions is caused by an underlying “softer” form of bipolar disorder, which is notoriously underdiagnosed. This delays access to correct treatment, causing a significant loss of DALYs. Improving diagnostic accuracy should be possible and cost-effective.

Epistemic status

I am fairly confident about the scale and neglectedness of the problem. As discussed later, bipolar spectrum disorders and medications to treat them are understudied, and the existing literature tends to use discrepant definitions and methodology, which makes it difficult to analyse how effective the interventions I suggest in the Tractability section would be. I have no medical background nor experience in policy change, which might have caused me to overestimate the effectiveness of the proposed solutions, or underestimate the difficulty of implementing the interventions. Overall, my analysis of the tractability of this problem is the area I have the least confidence in, and would most appreciate input on.


Between “regular” depression and what most people imagine when they hear the term “bipolar disorder”[1], there exists a whole spectrum of conditions, bipolar spectrum disorders[2], characterised by less extreme “highs” compared to bipolar disorder type I. It is worth noting that these disorders are not a “milder” version of BP-I in terms of DALYs lost since the main symptom of BSD are depressive episodes, which are on average longer and more frequent.

These disorders affect at least 1.1% of the population worldwide (more in the developed world, e.g. 2.2% in the USA[3]) , causing the sufferers to undergo recurrent depressions which do not get alleviated by usual antidepressants (SSRIs). The fact that the elevated moods of these patients are less extreme than in those with BP-I causes them and their doctors to usually not notice these moods as potential symptoms, which leads to them being misdiagnosed on a massive scale, usually as unipolar depression. On average, sufferers of BSD have to wait 7.5 years for a correct diagnosis, and at least 27% of hospitalised patients initially diagnosed with a major depressive episode will later on be diagnosed with BSD.

In this report, I analyse the scope and neglectedness of this cause, while proposing a potentially cost-effective approach to solving the problem. This intervention would result in significantly shortening the waiting time until correct diagnosis and treatment. I believe that it could be achieved mainly through diagnostic policy change, though I also investigate some other promising approaches.

Even though this cause is not as large in scope as some other suggested causes (e.g. reducing obesity or violence against women), it is much more neglected, while its scope is still not insignificant: I estimate it causes the loss of at least 3 million DALYs per year. My analysis also suggests that it should be tractable in a cost-effective way.[4]


A much more thorough and nuanced analysis of the problem and its potential solutions (targeted mostly to clinicians) can be found in “A Spectrum Approach to Mood Disorders: Not Fully Bipolar but Not Unipolar-Practical Management” by James Phelps, MD. If the topic interests you on a personal level, I recommend “Why Am I Still Depressed? Recognizing and Managing the Ups and Downs of Bipolar II and Soft Bipolar Disorder” by the same author.



There are several types of bipolar disorder listed in the DSM-5.[5] A brief summary of the main types:

  1. Bipolar Disorder type 1 (BP-I) is characterised by at least 1 manic episode. Depressive periods are usually present, but not necessary for diagnosis. A manic episode is a period of abnormally elevated, expansive, or irritable mood and abnormally goal-directed behaviour or energy, which is either:
    1. sufficiently severe to cause marked impairment in social or occupational functioning, or
    2. necessitates hospitalisation to prevent harm to self or others, or
    3. is accompanied by psychosis (e.g. delusions).
  2. Bipolar Disorder type 2 (BP-II) is characterised by at least 1 major depressive episode and at least 1 hypomanic episode, i.e. a period of abnormally elevated, expansive, or irritable mood and abnormally goal-directed behaviour or energy lasting at least 4 days and not meeting the diagnostic criteria for mania (which means it is not severe enough to cause marked impairment in functioning, does not warrant hospitalisation and is not characterised by psychosis).
  3. Cyclothymia is characterised by numerous subsyndromal hypomanic episodes and numerous depressive episodes not severe enough to meet the criteria for a major depressive episode. To warrant diagnosis, this period of mood shifts must have lasted at least 2 years.
  4. Specified bipolar disorder, unspecified bipolar disorder. These categories together encompass all presentations in which symptoms characteristic of a bipolar disorder cause clinically significant distress or impairment, but for some reason do not meet the criteria for any of the above diagnoses (e.g. a hypomanic episode lasting only 3 days, cyclothymia lasting less than 2 years, etc.)


In this analysis, I am focusing on all forms of bipolar disorder except for BP-I (so BP-II, cyclothymia, specified and unspecified bipolar disorder – for more nuance see the Diagnosis of BSD section below). For brevity, in what follows I will refer to these as bipolar spectrum disorders[6] (BSD). The reason for this focus is twofold:

  • The main source of suffering in BSD are the depressive episodes, with the effects of elevated moods being on average negligible in comparison, while in BP-I the main problem are the manic episodes.
  • BP-I requires different treatment than BSD.

One of the symptoms of BSD are occasional elevated moods, characterised by e.g. more energy, happiness, self-esteem, drive and/or irritability compared to the patient’s baseline (non-depressed) state. There is a variety of ways in which these can manifest: from very obvious hypomanias, to subclinical hypomanias which e.g. are shorter than 4 days or only some of the symptoms are present, to so-called mixed states where both depressive and manic symptoms are present at the same time, to mild hyperthymia. I will refer to all of these as elevated moods

Assumptions, methodology and limitations

In this analysis, I will completely disregard any DALYs lost due to elevated moods and focus solely on DALYs lost due to depressive episodes. In general, elevated moods in BSD are less extreme than in BP-I and, as described above, manifest in a very broad range of intensity and effects. Sometimes their effects can of course cause significant harm to the quality of life of the patient, e.g. a hypomanic state causing someone to overspend or damage their personal or professional relationships, but it is much harder to quantify and seems to be largely overshadowed by the DALYs lost due to depression.


I focus this analysis on the state of the problem in the UK,[7] but cursory research suggests that it would apply to the majority of the developed world as well. My approach to solving the problem would most likely start in the UK and scale up to the rest of the world.

In what follows, I tried to find the most accurate data when it comes to the prevalence of symptoms, treatment effectiveness etc.; however, many studies do not clearly specify which forms of bipolar disorder they discuss. Most research is focused on BP-I, with BSD being understudied. The sample sizes are usually quite small and the populations studied tend to be somewhat discrepant between studies I found (e.g. how much does BSD prevalence among patients hospitalised for a major depressive episode tell us about such prevalence among patients diagnosed with it in primary care, etc.). Often, evidence is relatively weak or contradictory between studies; e.g. there is still some disagreement whether BSDs are under- or over-diagnosed, and whether prescribing SSRIs is harmful or helpful. This means that the error margins in my analysis are quite significant and even just estimating them would require further research.[8]

Diagnosis of BSD

There are multiple reasons why the diagnosis of BSD is difficult: there are many different types of it, it seems to be a dimensional rather than categorical condition, there are common comorbidities, and there are several other disorders with a significant overlap in symptoms (ADD, borderline personality disorder, anxiety, schizophrenia). However, the main reason seems to be that the symptom that the patients seek treatment for is depression, and it seems almost impossible to distinguish BSD from unipolar depression using current methods (a 10-minute appointment with a GP typically unfamiliar with BSD).

Currently, the diagnostic journey of a person suffering from BSD in the UK, backed by the current NICE guidelines, looks roughly like follows:

  1. They see their GP for a 10 minute visit, seeking help for what they perceive as depression. The GP will briefly go with them through a depression questionnaire, and if the depressive symptoms seem prominent enough will prescribe them an antidepressant (SSRI). They might or might not ask them whether they had a manic episode, but they are actively discouraged by NICE guidelines from looking for hypomanic symptoms using a questionnaire. Unless the person is actively suicidal,[9] they are unlikely to get a psychiatrist referral at this stage.
  2. If the antidepressant doesn’t work, in the next visit several months later the GP advises them to continue on the current antidepressant a couple more months, or prescribes them a different SSRI. They might or might not issue a psychiatrist referral just yet.
  3. If the second antidepressant does not work, the GP will probably prescribe a third SSRI and also refer the patient to a psychiatrist and possibly to a group CBT therapy. The waiting time for either ranges from half a year to a year.
  4. The psychiatrist might or might not diagnose BSD at first, especially if the patient themself is unaware of the condition and does not spontaneously mention past hypomanias (which to most people just look like periods when they felt happier and had more energy). They are likely to stick to the unipolar depression diagnosis and trial other kinds of antidepressants (SNRI, tricyclics, MAOIs etc.). On a subsequent visit, half a year to a year later, the patient is likely to be seen by a different psychiatrist, not familiar with the case.
  5. On average, it takes 7.5 years to reach the correct diagnosis: perhaps when all other options have been tried, or when the patient finally stumbled upon some resource which helped them self-diagnose, or when the patient’s mood cycling worsens to the extent that their elevated moods finally become obvious (there is some evidence suggesting that SSRI usage in BSD patient leads to heightening their elevated moods even further, pushing some BSD patients into full-blown BP-I manias).

I suspect that even the sufferers who could theoretically afford a visit to a private psychiatrist to get a second opinion on the diagnosis often do not do that, because they have the illusion of already having a correct diagnosis and receiving treatment.

Treatment of BSD

Currently, there exists no “cure” for BSD, but there are treatments which can significantly reduce the symptoms or even eliminate them completely.

NICE guidelines for pharmaceutical treatment of an acute bipolar depression episode recommend as first line treatment either fluoxetine (Prozac) combined with olanzapine, or quetiapine on its own. Second line treatment is lamotrigine on its own. Guidance for maintenance treatment is either continuing the medications which worked in acute treatment, or replacing them with lithium.


There are therapies effective in treatment of BSD, e.g. Interpersonal and Social Rhythm Therapy, which focuses on helping patients maintain a regular daily routine.[10] However, due to the low availability of specialist therapies within the NHS, the patient would most likely get referred to either group or individual CBT.


There are also some dietary supplements, e.g. fish oil, which have some evidence of stabilising mood.


The worldwide lifetime prevalence of BSD is estimated to be at least 2%, with higher prevalence in the developed world (e.g. 3.5% in the USA). 

On average, the patients have to wait 7.5 years for a correct diagnosis, and they spend 30% of that time being depressed, with various degrees of severity and including mixed episodes  – for the sake of this analysis let’s use the common estimate of QALY reduction of 0.5/year due to depression, which can be approximately translated to 0.5 DALY/year. Multiplying these numbers suggests that each person with BSD loses 1 DALY on average prior to receiving a correct diagnosis, which translates to 0.13 DALY lost/year, for a total of 3 million DALYs per year lost worldwide, 230,000 in the US, and 27,000 in the UK.

In addition, BSD commonly causes functional impairment resulting in unemployment and loss of productivity. This creates a significant economic burden.


If a theoretical intervention could lead to patients receiving the correct diagnosis and effective treatment on their first GP visit when seeking help for depression, all the DALYs calculated in the previous section could be saved. However, no currently known treatment is fully effective for all patients, and there is no comprehensive analysis of the average effectiveness of treatments the patients currently get prescribed for BSD. Additionally, there is controversy whether SSRIs are harmful or helpful, which makes it difficult to compare BSD-specific treatments to the default treatment the patient would have received if diagnosed with unipolar depression. I assume that having the correct diagnosis would on average lead to 10% improvement in the patient’s wellbeing – I think this is a conservative assumption, but I fully admit it is probably the weakest point of my analysis.

Suggested diagnostic and treatment journey

I am considering several interventions, which could contribute to the ultimate goal of transforming a BSD patient’s diagnostic journey to look like this:

  1. A patient with depressive symptoms attends a GP visit. At this point, they might have some idea about BSD, because first Google hits for “depression” include at least some mentions of BSD and links to further information (see Intervention 3).
  2. The GP administers both a standard depression questionnaire and a questionnaire meant to flag people with BSD (Intervention 1.a, see also Limitation 1). If the patient scores low on the BSD questionnaire, they are funnelled into the standard unipolar depression pipeline – perhaps that would also need to happen for medium scores, but in that case at least a note should be made in the patient's file.
  3. If the patient scored high, they get a lamotrigine prescription (Intervention 1.b), a psychiatrist referral, and information about appropriate self-help methods.
  4. It should be apparent within a few weeks if lamotrigine was effective; it needs to be titrated slowly, but effects are usually felt within days of reaching the full dose, and quite often earlier.
  5. If lamotrigine wasn’t effective, other drugs might be trialled (this requires further research), and perhaps the diagnosis of BSD might be reevaluated.

In defence of primary care diagnosis

A significant counterargument to the proposed interventions is that there are some specialists who are sceptical of similar existing work when it has been implemented. NICE explicitly updated their guidance away from diagnostic questionnaires for bipolar, due to large rates of false positives. Phelps et al discuss how these false positive rates are not inherent to diagnostic questionnaires, and argue they were due to deficiencies in the particular questionnaire used - essentially, that its specificity was too low, because it did not use non-manic markers, such as sleep patterns; patient history; or family history. 


Existing literature suggests a Bipolarity Index assessment can achieve sensitivity and specificity of ~90% in a population of patients referred to a psychiatrist with potential mood disorders including depression. The predictive value will greatly vary with the prevalence of bipolar spectrum disorders in the population - in the population considered in that research, ~40% of patients had BSD or BP-I (which is consistent with the previous estimates of prevalence of BSD/BP-I patients within patients hospitalised for a major depressive episode). At those specificities and sensitivities, indicative numbers are as follows: 35% of patients were correctly diagnosed as bipolar; 5% of patients incorrectly had their bipolar missed; 5% of patients were incorrectly diagnosed with bipolar, and 55% of patients were correctly diagnosed as non-bipolar.

A potential compromise would be limiting the usage of the questionnaire to patients who already unsuccessfully tried one or more antidepressants, especially since that provides further evidence which could help decide if the patient might be have BSD:

  1. Unsuccessfully trying antidepressants increases the likelihood of BSD.
  2. The patient’s exact reaction to the antidepressant can be analysed:
    1. SSRIs can exacerbate hypomanic states, making them noticeable to the patient.
    2. An experience of the SSRI initially helping, but then its effects fading and the depression returning, is a prominent non-manic marker indicating BSD.
    3. Reaction to discontinuing antidepressants can be indicative: e.g. antidepressant withdrawal inducing hypomanias, or an improvement in mood.


A somewhat consequentialist approach would be: ultimately, it doesn’t matter what you “have”, what matters is whether BSD-specific treatments will help you. All of the treatments we are proposing are quick to try and safe – none of the bipolar interventions I am proposing would cause any direct harm to a person who has unipolar depression (I acknowledge the potential indirect harm caused by the “bipolar” stigma in the Limitations section).


Admittedly, this intervention would slow down the recovery of the subset of patients who do not improve on BSD-specific treatments and would have improved on an antidepressant and we need to estimate the DALY cost of that. We can approximate this number to be equal to the false positives of the Bipolarity Index (it’s very unclear how the number of unipolar patients helped with BSD treatments compares to the number of BSD people helped with antidepressants, so we disregard these groups). As discussed, this fraction can be estimated as 5%, and their diagnostic journey would be slowed down by around 3 months compared to the current one where they automatically get a diagnosis of unipolar depression and are prescribed an antidepressant. Let’s assume that these patients on average suffer an extra 3 months of moderate depression at 50% DALY loss. Quick calculations show that this effect should be very small compared to DALYs gained by BSD patients through this intervention.

The case for lamotrigine

Among the medications commonly used for BSD, lamotrigine stands out in terms of the efficacy versus tolerability ratio. This makes it perfect for our intervention: if the goal is to encourage prescriptions of a BSD-specific drug in primary care, it would need to be something as safe and easy to use as possible.

Some benefits of lamotrigine:

  • Its tolerability is well-studied[11]: it is commonly prescribed for epilepsy as well as for bipolar disorder, with 10 million prescriptions in the US.
  • As opposed to other medications for BSD, it does not have side effects significantly affecting the patient’s quality of life (weight loss, cognitive impairment and/or loss of sex drive are quite common side effects of all the other available drugs). This makes patients less likely to discontinue treatment, which is a common problem with other medications.
  • As opposed to most other medications for BSD, it does not require frequent blood tests.
  • It is a cheap generic medication.
  • It is relatively quick to manifest results.[12]
  • There is some evidence suggesting it might help with other problems often comorbid with BSD, e.g. eating disorders, OCD, PTSD.
  • It might even help with treatment-resistant unipolar depressions, especially  with comorbid anxiety or borderline personality disorder.


There are two notable issues with lamotrigine which need to be addressed:

  • Risk of SJS. Lamotrigine can have a rare but potentially very dangerous side effect: Stevens-Johnson syndrome. It can be avoided with correct titration.
  • Not enough good quality evidence of effectiveness. As a cheap, generic drug, lamotrigine is understudied. Combined with the fact that BSD is understudied, this means that there aren’t many studies investigating lamotrigine use in BSD. Clinical practice suggests that the drug is effective, but research is lagging behind. This is the most comprehensive meta-analysis.

Overall, this suggests that trying lamotrigine in a patient with suspected BSD for a few weeks might help them and won’t hurt them.

For a more detailed exploration of the topic, including dealing with SJS risk, see Phelps 2016, Chapter 7.

Potential interventions

I think that a combination of the following interventions could lead to a significant improvement of the situation of BSD patients in the UK, and could be adapted to other developed (and possibly developing) countries.

  1. Persuade NICE to update its guidelines to:
    1. recommend the use of the Bipolar Index questionnaire in addition to the depression questionnaire by GPs on patients reporting with depression (or patients who have already unsuccessfully tried one antidepressant, see discussion above).
    2. advise GPs to prescribe lamotrigine (with the appropriate titration schedule) instead of antidepressants to patients who scored highly on the Bipolar Index, and issue them a psychiatrist referral.
  2. Conduct an informational campaign to inform GPs of the new guidelines and provide them with essential information on BSD.
  3. Make information about BSD more visible when one searches online for information or self-help materials for depression, giving people a chance to identify it in themselves. That would involve liaising with mental health charities and organisations with big online presence, trying to persuade them to include information about BSD on their webpages about depression.

Here are some other interventions that might be worth considering, but which we will not analyse in this report.

  1. “Offshoring” access to psychiatric care. One can book a video appointment with an English-speaking private psychiatrist based in e.g. Poland for a fraction of the price of a UK-based private psychiatrist, and a prescription obtained that way would most likely be accepted by a UK GP. Creating a website to facilitate that could be a potential idea for a for-profit company, but of course one would need to analyse the cost-benefit ratio of offshoring these services: would that drive the prices up in the other country, making it harder for the local population to access these services?
  2. Reworking the DSM to reflect the spectral nature of bipolar disorders, and possibly rename them to help reduce stigma (but see Limitations below).
  3. A public awareness campaign – I suspect it wouldn’t be cost-effective, and might lead to increased rates of incorrect self-diagnosis.

Plan of action and cost-effectiveness analysis

A tentative plan of action would involve one or two people in the UK:

  1. Doing further desk research.
  2. Contacting experts for advice, and to add credibility to the cause.
  3. Preparing materials to convince various stakeholders.
  4. Contacting other mental health charities for support in liaising with the decision-making stakeholders (anecdotally, only big charities can exert any influence on the government), and, if applicable, to add information about BSD alongside information about depression to their websites.
  5. Approaching NICE and/or the government trying to effect policy change.

Hopefully, it would become apparent within half a year if these efforts have any chance of success. This stage could be done as a volunteer effort, or the person could pay themself a small salary if given funding for that. Let’s assume an expenditure of £10,000.

If NICE is amenable to changing their diagnostic guidance, the next stage would involve:

  1. Creating and distributing training materials for GPs to help them use the questionnaire and better identify patients with BSD: documents, websites, videos, potentially physical materials that would need to be printed and posted. We would need to prioritise making these materials very streamlined and easy to understand since we wouldn’t want to make the GPs dedicate too much of their valuable time to it.
  2. If deemed cost-effective, printing out and sending informative leaflets for patients to be available in waiting rooms.
  3. Implementing, getting approval for and distributing diagnostic and treatment decision making software to be used by GPs for patients with depression.
  4. Raising funding to enable the execution of the listed steps.

A ballpark estimate of the costs involved would be in the range of £200,000-300,000, assuming it costs around £5 to reach one GP. On the completion of this stage, we would consider our intervention complete, so assuming our initial assumptions were correct, including the pessimistic assumption of only 10% improvement in wellbeing on receiving a correct diagnosis, that would have led to saving 2,700 DALYs/year in the UK for a one-off cost of £300,000, which amortised over 15 years[13] would translate to 0.1 DALY saved per dollar, which compares favourably to GiveWell top charities. 

We need to consider a course of action if NICE is resistant to policy change. Depending on a careful and thorough analysis of cost-effectiveness analysis at that point, we would need to choose between the following:

  1. Gathering more evidence in hope of convincing NICE. That would probably involve running clinical trials to prove the overall benefits of using the Bipolar Index, or the effectiveness of lamotrigine. Sadly, this would lead to costs in the range of millions of dollars.
  2. Give up, if it seems that the intervention wouldn’t be cost-effective enough anymore.

Even assuming only 10% probability of the first stage of our intervention being successful, i.e. 90% probability of NICE being too resistant to change its guidelines, or us discovering serious flaws in our research, and us giving up at that stage, we would only have wasted half a year of one person’s time and potentially their small salary.

The interventions would need to be country-specific, aimed to effect policy change within a specific healthcare system, though hopefully with some rolling stone effect. An intuitive interpolation would suggest that scaling up to other developed countries should have a similar or higher cost-effectiveness to the UK implementation.[14] I lack the background to estimate the cost-effectiveness of scaling up to LMIC, but it could be a fascinating direction of further research.


In what senses is this problem neglected?

  1. Spectrum disorders are not a natural fit for the DSM, which complicates the diagnostic process. Rather than being seen as a unified major problem responsible for a significant fraction of depression cases, BSD are relegated to a list of obscure disorders with complex and confusing diagnostic criteria.
  2. BP-I is more attention-grabbing, and also leads to stigmatisation against spectrum disorders. This makes patients less likely to identify the disorder in themselves, and more likely to reject the diagnosis when they receive it. Plausibly, this could have led to the prevalence rates being underestimated, which in turn leads to lack of interest from both the public and the private sector.
  3. Pharmaceutical treatments are generic drugs, with scarce potential profits. Hence, pharmaceutical companies are unwilling to conduct more conclusive clinical trials on e.g. lamotrigine.
  4. In most countries, psychiatric appointments are oversubscribed and much frontline treatment is delegated to primary care physicians.
  5. In general, there is a lack of cost-effective charities within the mental health space. The existing charities dedicated to bipolar disorder seem to focus on self-help, individual counselling and support groups for people who already know they have bipolar rather than clinical research, policy change or raising awareness of BSD.

There are several psychiatrists writing books about bipolar spectrum disorders and trying to raise awareness of it both among patients and practitioners, but I am not aware of any organisations trying to effect policy change.

Limitations and potential obstacles

I identified some limitations of my approach, and I am sure there are some more I haven’t thought of. I list them here, with some solution ideas.

  1. A 10 min GP appointment might not be enough to administer both the depression questionnaire and the Bipolar Index questionnaire. Potential solutions:
    1. Ask the patient to fill out the questionnaire in the waiting room.[15]
    2. Only administer the Bipolar Index questionnaire to patients for whom this is a second or subsequent GP visit for depression and who have unsuccessfully tried at least 1 antidepressant – this might be the better approach to the problem anyway, as discussed in the Tractability section.
  2. The word “bipolar” carries a lot of stigma, since it makes people think primarily about manic symptoms of BP-I. This might lead to patients rejecting the diagnosis, or second-order negative effects in their lives (e.g. an employer who would be understanding of a “depressed” employee might be prejudiced against a “bipolar” one). Potential solution:
    1. Renaming (or adding “also known as”) bipolar spectrum disorders in the DSM to something less scary and suggesting it is mostly a variant of depression, e.g. “cyclic mood-variant depression”. I’m not sure how feasible that would be; bipolar disorder I has already been renamed once[16] precisely in order to destigmatise it, which might suggest the authors of the DSM are willing to do further renaming, or quite the opposite – they might consider the problem solved already.
  3. NICE might be reluctant to change its guidelines, especially since it has previously explicitly discouraged using a questionnaire for diagnosing bipolar in primary care, and since the clinical trial evidence for lamotrigine is lacking. Potential solution:
    1. Fund more clinical trials on lamotrigine and the Bipolar Index questionnaire – this would be a significant expense, which would require further estimation of whether the intervention would still be cost-effective.


In general, I see the suggested approach as an imperfect solution for an imperfect world: optimally, every person with BSD would be able to see a psychiatrist and get effective treatment within days of the onset of symptoms, but I don’t see it happening anytime soon.

Future plans

I would like to do further research on the subject, including reaching out to experts. If after more consideration this intervention still seems promising, I will pursue it as a part-time personal project or, if it seems warranted, a small charity. I would be very happy for it to become a project under the umbrella of an already established cost-effective mental health charity. At this point, I would appreciate:

  • Criticism of my approach – there definitely are things I have overlooked.
  • Practical advice on how to achieve my goals; I would especially appreciate insights from GPs, psychiatrists, NICE committee members, or other people with BSD.
  • Mentoring from people who succeeded in accomplishing a similar kind of intervention.

Thank you for reading!

Karolina Soltys (with contributions from David Owen)


Aiken CB, Weisler RH, Sachs GS. The Bipolarity Index: a clinician-rated measure of diagnostic confidence. J Affect Disord. 2015 May 15;177:59-64. doi: 10.1016/j.jad.2015.02.004. Epub 2015 Feb 12. PMID: 25745836.

Benazzi, F. Bipolar II Disorder. CNS Drugs 21, 727–740 (2007). https://doi.org/10.2165/00023210-200721090-00003

Benedetti F. Antidepressant chronotherapeutics for bipolar depression. Dialogues Clin Neurosci. 2012 Dec;14(4):401-11. doi: 10.31887/DCNS.2012.14.4/fbenedetti. PMID: 23393416; PMCID: PMC3553570.

Berk, M. and Dodd, S. (2005), Bipolar II disorder: a review. Bipolar Disorders, 7: 11-21. https://doi.org/10.1111/j.1399-5618.2004.00152.x

Brown EB, McElroy SL, Keck PE Jr, Deldar A, Adams DH, Tohen M, Williamson DJ. A 7-week, randomized, double-blind trial of olanzapine/fluoxetine combination versus lamotrigine in the treatment of bipolar I depression. J Clin Psychiatry. 2006 Jul;67(7):1025-33. doi: 10.4088/jcp.v67n0703. PMID: 16889444.

Dilsaver SC. An estimate of the minimum economic burden of bipolar I and II disorders in the United States: 2009. J Affect Disord. 2011 Mar;129(1-3):79-83. doi: 10.1016/j.jad.2010.08.030. PMID: 20888048.

Frank E, Swartz HA, Boland E. Interpersonal and social rhythm therapy: an intervention addressing rhythm dysregulation in bipolar disorder. Dialogues Clin Neurosci. 2007;9(3):325-32. doi: 10.31887/DCNS.2007.9.3/efrank. PMID: 17969869; PMCID: PMC3202498.

Gazalle FK, Andreazza AC, Ceresér KM, Hallal PC, Santin A, Kapczinski F. Clinical impact of late diagnose of bipolar disorder. J Affect Disord. 2005 Jun;86(2-3):313-6. doi: 10.1016/j.jad.2005.01.003. PMID: 15935253.

Geddes JR, Calabrese JR, Goodwin GM. Lamotrigine for treatment of bipolar depression: independent meta-analysis and meta-regression of individual patient data from five randomised trials. Br J Psychiatry. 2009 Jan;194(1):4-9. doi: 10.1192/bjp.bp.107.048504. PMID: 19118318.

Ghaemi SN, Sachs GS, Chiou AM, Pandurangi AK, Goodwin K. Is bipolar disorder still underdiagnosed? Are antidepressants overutilized? J Affect Disord. 1999 Jan-Mar;52(1-3):135-44. doi: 10.1016/s0165-0327(98)00076-7. PMID: 10357026.

Goodwin GM; Consensus Group of the British Association for Psychopharmacology. Evidence-based guidelines for treating bipolar disorder: revised second edition--recommendations from the British Association for Psychopharmacology. J Psychopharmacol. 2009 Jun;23(4):346-88. doi: 10.1177/0269881109102919. Epub 2009 Mar 27. PMID: 19329543.

Jia H, Zack MM, Thompson WW, Crosby AE, Gottesman II. Impact of depression on quality-adjusted life expectancy (QALE) directly as well as indirectly through suicide. Soc Psychiatry Psychiatr Epidemiol. 2015 Jun;50(6):939-49. doi: 10.1007/s00127-015-1019-0. Epub 2015 Feb 7. PMID: 25660550; PMCID: PMC4590980.

Judd LL, Akiskal HS, Schettler PJ, Coryell W, Maser J, Rice JA, Solomon DA, Keller MB. The comparative clinical phenotype and long term longitudinal episode course of bipolar I and II: a clinical spectrum or distinct disorders? J Affect Disord. 2003 Jan;73(1-2):19-32. doi: 10.1016/s0165-0327(02)00324-5. PMID: 12507734.

Merikangas, Kathleen R.; Jin, Robert; He, Jian-Ping; Kessler, Ronald C.; Lee, Sing; Sampson, Nancy A.; Viana, Maria Carmen; Andrade, Laura Helena; Hu, Chiyi; Karam, Elie G.; Ladea, Maria (March 2011). "Prevalence and correlates of bipolar spectrum disorder in the world mental health survey initiative". Archives of General Psychiatry. 68 (3): 241–251. doi:10.1001/archgenpsychiatry.2011.12. ISSN 1538-3636. PMC 3486639. PMID 21383262.

National Institute for Health and Care Excellence. Bipolar disorder: assessment and management. Clinical guideline [CG185]. Sep 2014

Ng F, Hallam K, Lucas N, Berk M. The role of lamotrigine in the management of bipolar disorder. Neuropsychiatr Dis Treat. 2007 Aug;3(4):463-74. PMID: 19300575; PMCID: PMC2655087.

Phelps, James. A Spectrum Approach to Mood Disorders: Not Fully Bipolar But Not Unipolar--Practical Management. United States, W. W. Norton, 2016.

Prabhavalkar KS, Poovanpallil NB, Bhatt LK. Management of bipolar depression with lamotrigine: an antiepileptic mood stabilizer. Front Pharmacol. 2015 Oct 23;6:242. doi: 10.3389/fphar.2015.00242. PMID: 26557090; PMCID: PMC4615936.

Solmi, M., Veronese, N., Zaninotto, L., Van der Loos, M., Gao, K., Schaffer, A., . . . Correll, C. (2016). Lamotrigine compared to placebo and other agents with antidepressant activity in patients with unipolar and bipolar depression: A comprehensive meta-analysis of efficacy and safety outcomes in short-term trials. CNS Spectrums, 21(5), 403-418. doi:10.1017/S1092852916000523

Zavodnick AD, Ali R. Lamotrigine in the treatment of unipolar depression with and without comorbidities: a literature review. Psychiatr Q. 2012 Sep;83(3):371-83. doi: 10.1007/s11126-012-9208-4. PMID: 22322995.

Zimmerman M, Galione JN. Screening for bipolar disorder with the Mood Disorders Questionnaire: a review. Harv Rev Psychiatry. 2011 Sep-Oct;19(5):219-28. doi: 10.3109/10673229.2011.614101. PMID: 21916824.

  1. ^

    I.e. bipolar disorder type I (BP-I)  with manic episodes characterised by highly erratic, often dangerous behaviour frequently combined with psychosis/delusions.

  2. ^

    Throughout this report, I refer to them as BSD for short.

  3. ^

    12-month prevalence; lifetime prevalence is 2% worldwide, 3.5% in the USA.

  4. ^

    I estimate the scores of this cause according to the 80,000 Hours scale, neglectedness, solvability framework to be: scale 8, neglectedness 10, solvability 4, overall 22.

  5. ^

    The most current version of the Diagnostic and Statistical Manual of Mental Disorders, which is the official source of psychiatric diagnoses in healthcare systems worldwide.

  6. ^

    It is worth noting that this term is used in some literature to refer to all forms of bipolar disorder, in some other literature – to conditions with bipolar symptoms not prominent or specific enough to qualify for one of the DSM diagnoses, and in yet another literature, confusingly, to conditions whose symptoms have some overlap with bipolar disorder like borderline personality disorder or schizophrenia.

  7. ^

    To familiarise international readers with the UK-specific terminology:

    • The NHS is the public healthcare system.
    • NICE is the organisation issuing diagnostic and treatment recommendations to primary and secondary care practitioners. This guidance is not mandatory.
    • GPs in the UK have autonomy to use their own judgement when it comes to diagnoses and are allowed to prescribe all kinds of medication.
    • The standard GP visit lasts 10 minutes, with a waiting time of up to a month. The waiting time for a NHS-funded psychiatrist can be longer than half a year; one needs a GP referral to access this service.
    • The cost of a private psychiatrist appointment is around £350/$425, which puts it out of reach of many sufferers.
  8. ^

    I decided that it would be better to get some preliminary expert feedback at this stage before I dive even deeper.

  9. ^

    This means answering “Yes” to “Are you planning to kill yourself?”; answering “Yes” to “Do you often wish you were dead?” and “Do you often think about suicide?” usually does not warrant the “actively suicidal” label in the current system.

  10. ^

    The link between sleep patterns and bipolar is quite fascinating, both as a symptom (bipolar depressions tend to be characterised by hypersomnia, as opposed to unipolar depressions, more often characterised by insomnia), and as potential therapy.

  11. ^

    This includes understanding of long-term effects: it has been prescribed for more than 30 years.

  12. ^

    If the patient is to improve on lamotrigine at all, it will happen soon after the target dose is reached (5 weeks according to the safe titration schedule), and sometimes sooner.

  13. ^

    A rough estimate of how often we would need to reach out to GPs with new informational materials.

  14. ^

    It would involve hiring a person speaking the local language and able to liaise with local stakeholders, and potentially sending out informational materials to a number of GPs proportional to the local population.

  15. ^

    This would require using a more patient-readable bipolar spectrum questionnaire, e.g. MoodCheck.

  16. ^

     It was previously known as a “manic-depressive disorder”.

Sorted by Click to highlight new comments since:

Disclaimer: I was personally affected by this problem in the past, and this is how I became aware of it as a potential cause. However, I consider myself cause-impartial: I chose to present this particular cause because I think it is overlooked and potentially promising, and because I happened to have some insight into it. I did my best to find impartial sources, and in cases of conflicting studies I tended to go with those which gave less support to my overall claims. (I wasn’t 100% comfortable disclosing this under my full name on a public forum in the main post, but I’m somehow more at ease with it being a comment ;) ). 


I think you can get involved with the development of NICE guidelines as someone who has used these services: https://www.nice.org.uk/about/what-we-do/our-programmes/nice-guidance/nice-guidelines/how-we-develop-nice-guidelines


You would be part of a panel along with professionals/experts reviewing the evidence. Might be worth considering!

If I understood your post correctly, you have considered the cost-effectiveness of recommending a policy change to NICE, and have not included the cost of treatment. 

I think if the cost of treatment were included, then this intervention might well be cost-effective for the UK NHS and worth them supporting - but would be much less cost-effective than GiveWell top charities.

Thank you for your comment and sorry for the late response!

As far as I understand, there is no pre-existing pathway for laypeople to suggest topics for new panels, there is only the opportunity to contribute to ongoing ones. Taken from their website:

Decisions on which new topics to develop guidelines on, and in what order, are based on factors such as:

the priority given to the topic by commissioners and professional organisations, and organisations for people using services, their families and carers

the health and care burden, and the potential to improve outcomes and quality of life.

A topic selection oversight group at NICE considers topics for guideline development, taking these factors into account. NICE then discusses topics identified in this way with NHS England, the Department of Health and Social Care, and Public Health England, and a prioritised list is agreed by these 3 bodies.

I am currently trying to figure out the best way of tapping into this process.


When it comes to factoring in treatment cost -- the point of my intervention is identifying which patients would benefit from being treated with lamotrigine (which is a generic drug) rather than SSRIs (which is what they are currently prescribed). My guess is that it should be cost-neutral (it's replacing one generic drug with another), but I haven't checked how much the respective medicine costs the NHS. I agree it would be worth doing.

A very interesting read. I didn't know much about BSD.

I was wondering if you had a spreadsheet or something equivalent where you put all the numbers you mention together? Prevalence, loss due to misdiagnoses, false positives, costs, etc. It could help with understanding the predicted cost-effectiveness of the different actions you mention. Apologies if I missed it in the core text.

Here is the spreadsheet. I would like to add error margins and more detailed epistemic status analysis at some point. The main variables affecting the cost effectivenes and somewhat uncertain are:

  • Effectiveness of treatment - I chose to assume that getting a correct diagnosis and treatment with lamotrigine leads to recovery of 10% of patients, compared to being treated with SSRIs. This is just a guesstimate based on the best evidence available, which is still insufficient. Perhaps the first step to the success of this intervention would be gathering more evidence in support of lamotrigine, but the cost involved may render it no longer cost-effective.
  • Percentage of patients with BSD among those seeking primary care treatment for depression (or, in a variant of the intervention, among patients who failed to recover on their first SSRI). The data exists for patients hospitalised for a major depressive episode (27%), and for patients referred to a psychiatrist for mood disorders (36%), but these numbers may be larger than what we need. On the other hand, it is roughly consistent with the 5% 12-month prevalence of depression  and 2% prevalence of BSD.

Thank you!  I have some more detailed notes on the calculations I made, and the resulting numbers are in the post, but creating a proper spreadsheet with it is a good idea! I'll do it on Monday (edit: Tuesday-Wednesday, life's too busy!) and update the post.

thank you, really love this post! i see that some of your recommendations hinge on using the bipolar diagnostic questionnaire as a diagnostic tool. I think that the questionnaire is actually part of the problem. I suspect that many people with treatment-resistant depression don't become obviously hypomanic or manic enough for the doctor to be able to diagnose bipolar type II even if they have it, and so they keep trying wrong meds for years.

I suspect that the solution to this is to update the diagnostic questionnaire to check for differences in how people react to low sleep. The author of Unquiet Mind also wrote a textbook about this, and bipolar is super related to sleep. Most people feel terrible on low sleep, but bipolar people tend to feel great, and can get hypomanic or manic from low sleep. For example, I bet that guy who wrote that debunked post on the LessWrong forum about how people need less sleep has bipolar (https://www.lesswrong.com/posts/HvcZmKS43SLCbJvRb/theses-on-sleep).

For those trying to hack treatment-resistant depression who suspect they have bipolar, the right thing to do might be to ask to try Lamotrigine, if you've already tried a few other drugs that didn't work. I'm less sure about this, but if your physician isn't willing to do that for you, you could ask for something like Effexor which can trigger a hypomanic episode, and then get your bipolar type II diagnosis so that you can get lamotrigine.

Thank you for your comment and sorry for the late response!

The questionnaire that I suggest does factor in sleep patterns; agreed that it is a very meaningful symptom that needs to be factored into the diagnosis.

I would strongly advise against your final suggestion though -- triggering a hypomanic episode on purpose seems like a really dangerous idea! Even if you succeeded in triggering hypomania as opposed to mania, you're unlikely to manage to get seen by a GP while in that state. If you're that desperate, it's probably safer to lie to your doctor and say you have had a hypomanic episode than actually trigger it on purpose in order to be truthful...

ah, i guess it depends on the reliability of your health care... when i was thinking the most about this i was not in the USA. thanks, this is a good and balanced point.

Thanks for this really thorough and insightful proposal. I'm sympathetic with your sense of the scope of the problem. My readings suggest that the total burden of bipolar disorders is around 9 million DALYs per year, so the proposal that BSD misdiagnosed as depression is around 3 million DALYs per year seems plausible. So it's a major problem. I have some questions about implementation, especially outside of developed countries:

• My primary concern is that this is one of a family of problems arising from a lack of psychiatrists, even in developed countries. A lot of your proposal aims at improving the psychiatric expertise/education of GP's, but even your positive proposal includes a referral to a psychiatrist. My thought is that this might be workable in the UK, but it would be unworkable at the global level where the availability of psychiatrists is even lower. You address this in your proposal about "offshoring," but of course that just defers the problem (if the UK offshores to Poland, who does Poland offshore to, etc.).

I wonder whether resources might be better invested in a trans-diagnostic approach, which increases training and licensing for mid-level psychiatric experts (these are psychiatric NP's in the US system). This would potentially address the variety of psychiatric conditions that are not ideally treated in primary care, especially with an eye towards the global level.

• It bears noting that (as I understand it) this is ultimately a problem for pharmeceutical approaches. The standard medications for unipolar depression don't work for BSD, so misdiagnosis leads to mistreatment. But (to my knowledge) depression is responsive to CBT/ACT whether or not it is unipolar or bipolar. So, for non-pharmeceutical approaches to depression, the question of misdiagnosis is less significant. Given the effectiveness of those treatment (and their low-cost and scalability), shouldn't thinking about BSD lead us to allocate a greater percentage of resources to non-pharmaceutical approaches to depression?

Thanks again for this illuminating post, I hope these preliminary reactions are useful.

Thank you for your comment, you make some really good points!

When it comes to the lack of psychiatrists: I agree it's a very difficult issue and I'm not trying to solve it here, but rather sidestep it. I do include a psychiatrist referral in my proposal, but I completely understand that the best one could hope for there is helping the patient see an overworked NHS psychiatrist e.g. in a year rather than two years.  The main element of my proposal is having the GP be able to directly prescribe bipolar-appropriate medication, without necessitating the involvment of a psychiatrist. This is what already happens with unipolar depression and antidepressants, and the fact that it's possible in primary care averts a massive amount of suffering and lessens the load on the secondary care psychiatric system.

I haven't thought about including NPs in my analaysis, thanks for pointing out this possibility. As far as I understand, in the UK they mostly work in mental health institutions, so a BSD patient would be unlikely to encounter them unless they make a suicide attempt. I need to look into how it works in other countries.

It's really difficult to determine how effective therapy is compared to medications; I'd love to see a large clinical trial comparing BSD patients' outcomes  on placebo, lamotrigine, SSRIs, a few other medications, therapies like CBT or IPSRT, following the cohort for a good few years, studying whether people who didn't improve on one type of treatment improve on another etc. I don't think we'll ever get such comprehensive data; we're stumbling in the dark and it's quite frustrating.

I definitely agree that many depressions, unipolar or bipolar, improve on CBT, but I'm quite sure there is a percentage of patients with BSD (or depression, for that matter), where the problem is almost purely physiological (some kind of chemical imbalance in the brain) rather than psychological -- so relying solely on therapy would be like trying to therapise away cancer or diabetes. How big is this percentage, and how does it compare to BSD cases amenable to therapy? We don't have enough data to say. 

I also think that pursuing non-pharmaceutical approaches to depression is somewhat less neglected than advocating for improving access to medication, including in EA circles, hence my focus on the latter.

I get this!

My initial idea for the Cause Exploration Prize was essentially the same thing, targetted at ADHD.

Some thoughts:

  • Calculating outcome uncertainty - Overall, a problem I ran into was cost calculation for every step of the process. Going from pre-existing undiagnosed disorder  to diagnosed and well-managed disorder involves a lot of complex, highly uncertain steps. For example, even if you sit every GP down for a 1-hour deep dive on BSD, there's a lot of reasons why the knowledge wouldn't translate to improved outcomes. Replicate that uncertainty for the entire treatment process, and it gets super messy.

Anyway, about your proposals:

  1. Your first solution makes sense. It sounds logical, so I do wonder why it isn't already a thing. I suppose local mental health legislation is sometimes just arbitrary. Regarding the questionnaire, the mental health professionals I've talked to do seem quite ... passive. Half the session turns into me slowly discussing my mood problems and them telling me it's not that big a deal instead of following my train of thought. While this only covers BSD, I suppose I'd much rather have a questionnaire shoved in my face that could lead somewhere rather than trying the same generic advice that has limited effect for anyone with underlying disorders.
  2. Your second solution - Fundamentally, I'm not convinced it's easy to educate professionals into providing better care. What I mean by that is educators and mental health professionals who should know how to handle neurodivergent people inevitably struggle to empathise with every disorder. Everyone starts off with stereotypes of certain disorders and how they "should" act, and I'm not sure I've seen a consistent way to debunk these stereotypes. In my case, I'd interacted with dozens of mental health professionals and educators whose job description entailed understanding ADHD, yet I don't think any of them would have helped at all without my aggressive prompting. It's almost like teaching people not to be racist, no one is going to think they have bigoted beliefs. Secondly, if you yourself had to prescribe care for people with undiagnosed BSD, you might do really well because you have firsthand experience and plenty of personal research. However, if I asked you to do that for every disorder, I think anyone would struggle. In other words, I'd rely as little on GP behaviour as possible.
  3. I'm most interested in your third solution. In my opinion, interacting with other people who have the same problems I do is far more helpful than trying to communicate with someone who's basing their judgement on a few lectures they had back in medical school. I find that so much mental wellness advice that should work on neurotypicals needs to be completely adapted for those with mental disorders. Having two normal people with the same disorder talk about what does and doesn't work for them is more helpful than anything short of medication, in my opinion. 

    Hence, I really like the focus on personal access to information. A few months ago, I thought about developing scaleable support groups for people with mental disorders. The idea is that people naturally use social media to interact with like-minded people, and the algorithms work in your favour if you want to reach someone with relatable content. This is by no means an easy hypothesis to test, but I posit that a methodical attempt at creating authentic social media communities centred around living with mental disorders (advice, venting, discussing weird quirks etc) would be extremely cost effective in tackling disease burden, by sheer volume of people sharing helpful ideas and seeking help when they otherwise wouldn't have.

    Overall, I think your first suggestion has unknown barriers but is worth attempting, I am very skeptical of your second suggestion and I think your third suggestion is the most interesting. I'd love to discuss this further as I and another person are brainstorming how to create a working intervention in this space. Of course, that's only because we both have a lot of background building anonymous support forums lol.

And if you're curious, I eventually scrapped the ADHD idea for this prize and went with advocating for Social Relationships as a metric for EA to consider, because I thought it was a good compromise between neglected emotional impact, universal, provably impactful and relatively easy to quantify.

Thanks for the insightful comment!

Completely agreed re uncertainty when it comes to intervention outcomes, trying to get any concrete numbers in order to calculate cost-effectiveness almost made me give up, and what I ended up going with is still very nebulous.

When it comes to intervention 2: I absolutely don't expect any realistic degree of GP education to suddenly make them competent in giving BSD patients fully adequate support. Perhaps I should have better phrased it as "efficiently informing GPs that the guidelines have changed and that they should follow them" rather than "educating GPs". In this context, I see educating GPs as a neccessary evil: they are the ones gatekeeping access to medication and who will choose whether to follow NICE guidelines or not. I suspect many of them don't keep up to date with changes in guidelines, so in order for intervention 1 to have any effect it would need to be followed up with somehow translating it into clinical practice. It would be even better if it could be done on autopilot, where the GP can be a Chinese room completely unaware of what BSD even is, just following a flowchart along the lines of "Depressed? Antidepressants didn't work? Fill out this questionnaire... Scored high? Here's lamotrigine, please read the leaflet, next!".  This is why implementing, getting approval for and distributing diagnostic and treatment decision making software to be used by GPs for patients with depression could be very effective - I listed it among the concrete steps to take, but perhaps I should have given it more prominence in the interventions list.

Intervention 3 on its own could definitely contribute to GPs taking the correct action - a plucky patient saying "I have BSD, give me lamotrigine!" would hopefully trigger a flow like "BSD? Let me look that up... Okay, I'm supposed to ask you to fill out a questionnaire.  Scored high? Here's lamotrigine, please read the leaflet, next!".

I have to confess I'm somewhat biased towards a pharmaceutical approach when it comes to BSD. This makes me see the intervention you suggested (connecting patients with people who have a similar problem) specifically in case of BSD more as a means towards an end (the patient getting enough information to convince a GP to prescribe correct medications) rather than a goal in itself (I imagine in many patients, and in my personal experience, no amount of talking with other sufferers, emotional support, group therapy, self-help info etc. would do much to help, while the correct medication would make the problem dissapear completely - though I admit other sufferers might have the opposite experience, which loops back into the whole outcome uncertainty issue). 

Specifically in case of BSD, there would also be the issue that by default the patient would seach for "depression support groups" rather than "BSD support groups". Actually, it would be brilliant if your solution could help people correctly self-diagnose by making them go through a  questionnaire-based decision tree, to later direct them to the appropriate group. Perhaps it wouldn't even need to be categorical or DSM-based, but use some ML solution grouping people according to the weights they've given to various symptoms or other markers.

That being said, I think your idea is very interesting! I'd love to talk more about it and potentially get involved - let's connect :) 

Congrats on the social relationships as a metric proposal, it was a fascinating read!

Curated and popular this week
Relevant opportunities