The Majority of DSM Diagnoses Are Never Used...

 

The landscape of medical and psychiatric diagnoses that are actually used by clinicians has always interested me.  Diagnostic classifications like the DSM and the ICD are generally used for the purpose of billing and generating statistics.  There is also an implicit research function that is probably why the number of diagnoses are so expansive.  I wrote a brief comment about this on my other blog almost exactly 3 years ago.  In a study of 1,260,097 psychiatric diagnoses reported from hospital care between 2001-2007 only 16 or 4.2% of the available diagnoses accounted for 50% of the reported activity (1).  Forty-nine diagnoses accounted for accounted for 75% of the activity and 108 diagnoses accounted for 95% of the activity.  Of the total diagnoses available most were used infrequently if at all.  In a separate abstract, 32 diagnoses were not used at all and 121 diagnoses were used in less than 0.1% of cases (2).  This is an important issue that I intend to use in further posts about diagnostic reasoning in psychiatry.  This is intended as an update and moving the concept over to my main blog.

The first question when it comes to either DSM or ICD diagnoses in different clinical settings is – how many are there?  In the case of the DSM – I personally counted the diagnoses and came up with 281 diagnoses using the methods outlined in that post.  Since then, I have encountered a reference that lists the total diagnoses as 245 (3).  In an earlier DSM-III study of 11,292 general psychiatric admission 296 of 329 available diagnoses were used and the 9 most frequent accounted for 35.8% of all diagnoses (4).

Surveys of Psychiatric Diagnoses Used In Practice
N	Classification	Used/Available (%)	Skew
11,292 adults	DSM-III	296/329 (90%)	73% of diagnoses were from 6 diagnostic categories with major depression the predominate category at 23%
214,206 adults	ICD 9/10-CM	----	mood disorders (22%), anxiety disorders (21%), and substance use disorders (16%) together accounted for the majority of documented psychiatric diagnoses
13,684,154 children and adolescents	ICD 9/10 – grouped as 13 diagnostic groups and 1 other	-----	Diagnostic groups were trauma/stressor-related disorders (27%), anxiety disorders (19%), and depressive disorders (17%)
7,076 adults	DSM-III-R	------	41.2% of the adult population under 65 experienced at least one DSM-III-R disorder in their lifetime, 23.3% within the preceding year. Depression, anxiety, and alcohol abuse and dependence were most prevalent
1:  Mezzich JE, Fabrega H Jr, Coffman GA, Haley R. DSM-III disorders in a large sample of psychiatric patients: frequency and specificity of diagnoses. Am J Psychiatry. 1989 Feb;146(2):212-9. doi: 10.1176/ajp.146.2.212 2:  Barr PB, Bigdeli TB, Meyers JL. Prevalence, Comorbidity, and Sociodemographic Correlates of Psychiatric Diagnoses Reported in the All of Us Research Program. JAMA Psychiatry. 2022;79(6):622–628. doi:10.1001/jamapsychiatry.2022.0685 3:  Mojtabai R, Olfson M. Trends in Mental Disorders in Children and Adolescents Receiving Treatment in the State Mental Health System. J Am Acad Child Adolesc Psychiatry. 2025 Aug;64(8):906-920. doi: 10.1016/j.jaac.2024.08.008. Epub 2024 Aug 28. PMID: 39214290. 4:  Bijl, R., Ravelli, A. & van Zessen, G. Prevalence of psychiatric disorder in the general population: results of the Netherlands Mental Health Survey and Incidence Study (NEMESIS). Soc Psychiatry Psychiatr Epidemiol 33, 587–595 (1998). https://doi.org/10.1007/s001270050098

 I have listed several additional surveys of diagnoses in various samples.  Comparison across studies is complicated by the classification system used and whether specific diagnoses are counted or diagnostic groups.  If only groups are counted it is more difficult to illustrate the skew by weighting.  Large healthcare systems have these statistics but I am not aware of any of that data being published.  Having worked for one of those systems the data is often considered proprietary.  The data would also be affected by the clinical populations being treated.  I would expect safety net hospitals to have a much higher percentage of disability associated diagnoses than private hospitals.  I would expect the same skew between acute care settings (inpatient units and acute psychiatric services) to have a different distribution of diagnoses than outpatient clinics.  Of the 3 studies that looked at this issue above using DSM criteria – most DSM diagnoses are used infrequently if at all.

What about the criticism of the proliferation of diagnoses?  I expect to see the usual discussion of this issue and the DSM-6 is hyped as a controversial topic over the coming years.  We already know the answer to the question but everyone will need to pretend that we don’t. By my count the DSM diagnoses peaked with the DSM-IV.  A lot of the controversy about diagnostic proliferation will start by saying the DSM-I had 106 diagnoses in 1954 and that number has more than doubled.  Nobody will say that most clinicians are using a set of diagnoses so limited that they have the numerical codes memorized so they do not have to keep looking them up.  A comparable look at the ICD shows that it started out in in 1893 as the International List of Causes of Death (or the Bertillon Classification of Causes of Death).  There were 44, 99, or 161 codes that could be used depending upon the reporting capabilities of the country.  The 161-code version became the ICD and in 1898 the American Public Health Association (APHA) recommended that Canada, Mexico, and the US adopt it and revise it every 10 years based on advancements in medical knowledge.  The current version ICD-11 has 55,000 codes up from the previous version (ICD-10) 14,000 codes.  

Any comparison of numbers of diagnoses is problematic for several reasons.  First, the authority proposing the classification system is averse to reporting them. That is true whether it is the DSM or the ICD. When I counted them, I provided the methodology and you can replicate it yourself.  With the DSM there are occasional isolated counts close to mine – but no explanations.  With the ICD – things are more complex and estimated range from 10,000 – 15,000 diagnoses that would be recognized as unique.  In the ICD-11 those diagnoses are included with other biomedical terms in the underlying Foundation of the ICD.  The Foundation is technically a semantic database of terms including symptoms and other findings. 

Before getting into how these codings work relative to diagnoses – a brief introduction to ICD coding terminology since it is impossible to separate out what physicians typically consider diagnoses.  In the example below, I have produced a hierarchical tree diagram that is considered the basis for the ICD.  In the example I am following how an episode of recurrent depressive disorder-severe without psychotic features is coded.  The top category is the grouping of all medical disorders into 28 categories.  The next group is all mental, behavioral, and neurodevelopmental disorders grouped into 24 categories.  From there a mood disorder group, depressive disorder group and recurrent depressive disorder group follow.  The final grouping is the variant of 15 recurrent depressive disorder possibilities that we are looking for.  In ICD jargon, that final groups is called a leaf code because it is the ultimate result of the hierarchy and it cannot be split any farther.  The branching about that level is called stem codes.   

 

A more interesting comparison is how the diagnostic codes in the rest of medicine have increased.  

Version	Approximate Leaf Codes	Notes	References
ICD-10 (WHO)	~10,607	Base international version	[1]
ICD-10-CM (US)	~71,932	US clinical modification with extensive granularity	[1]
ICD-11-MMS	~14,622	Moderate increase over ICD-10; post coordination expands expressivity	[1]
ICD-11 Foundation	Much larger	Includes 5,500+ rare diseases; serves as semantic knowledge base	[2-3]
1:  Fung KW, Xu J, Bodenreider O. The new International Classification of Diseases 11th edition: a comparative analysis with ICD-10 and ICD-10-CM. J Am Med Inform Assoc. 2020 May 1;27(5):738-746. doi: 10.1093/jamia/ocaa030. PMID: 32364236; PMCID: PMC7309235. 2:  Feinstein JA, Gill PJ, Anderson BR. Preparing for the International Classification of Diseases, 11th Revision (ICD-11) in the US Health Care System. JAMA Health Forum. 2023;4(7):e232253. doi:10.1001/jamahealthforum.2023.2253 3: Chute CG. The rendering of human phenotype and rare diseases in ICD-11. J Inherit Metab Dis. 2018 May;41(3):563-569. doi: 10.1007/s10545-018-0172-5. Epub 2018 Mar 29. PMID: 29600497; PMCID: PMC5959961.

   

The table shows a direct comparison between the ICD-10 and ICD-11.  The conclusion is that there has been a moderate increase in codes.  Leaf codes can undercount and overcount the diagnoses and are not necessarily strict representations of diagnoses.  For example, a code of type 2 diabetes mellitus can generate many additional codes depending on the complications.  The only equivalent in the DSM are the modifier codes.  Medicine can also code symptoms rather than a specific diagnosis – so those codes like “neck pain, cough, constipation, etc) also generate codes that have no DSM equivalent.  There is residual or not-otherwise-specified (NOS) codes in the ICD that meet no diagnostic criteria.  The DSM-5-TR has replaced NOS codes with other specified disorder or unspecified disorder that are probably not much better.  The ICD-11 added complexity codes for severity, histopathology and other features to increase specificity.

The structure of the ICD is relevant to counting diagnoses.  The basic a hierarchical tree structure that can be viewed at the following link.  In this case the diagram illustrates the hierarchy Mental, Behavioral, or Neurodevelopmental disorders (category) -> Mood Disorders (3) -> Bipolar Disorder -> Bipolar Type 1 Disorder (16) -> Bipolar type I disorder, current episode manic, without psychotic symptoms (32) -> Bipolar type I disorder, current episode manic, without psychotic symptoms, with prominent anxiety symptoms (32).  The numbers in parentheses indicate the total branching of the hierarchical tree diagram.  The branching is graphically represented in the center panel.

A comparison of leaf codes in the DSM is possible by estimating leaf codes as 3-5-digit total billable codes.  That would include about 350 leaf code like endpoints and 150 environmental codes for total of about 500.  That is only about 3% of the total ICD-11 codes in the about table – a number made more significant by the fact that the DSM includes diagnoses that the ICD codes in other categories – most notably neurocognitive disorders.   

In conclusion – all of the controversy about the proliferation of diagnoses (or codes) in the DSM as excessive does not match the reality of how diagnoses in general have increased in the rest of medicine. If anything, it seems to be lagging.  It also misses the point why this happens in the first place as it was well put by the American Public Health Association in 1898 – to revise the ICD every 10 years “based on advancements in medical knowledge.”   

 

George Dawson, MD, DFAPA

 

Supplementary 1:  Leaf code approximation:  I counted all of the diagnoses listed in the chapter "Numerical Listing of DSM-5 Diagnoses and Codes (ICD-10-CM)” Total codes listed in that appendix are 760 but it is a mapping of DSM diagnoses onto the ICD-10 and that is not an exact match.  As a result, there are 148 duplicate codes bringing the total down to 612.  The list also contains parent or sub-stem codes such as F79 (Unspecified intellectual disability) that requires an additional digit to become a leaf code.  There are 23 sub-stem codes bringing the total number of leaf codes to 589.  

 

References:

1:  Munk-Jørgensen P, Najarraq Lund M, Bertelsen A. Use of ICD-10 diagnoses in Danish psychiatric hospital-based services in 2001-2007. World Psychiatry. 2010 Oct;9(3):183-4. doi: 10.1002/j.2051-5545.2010.tb00307.x. PMID: 20975866; PMCID: PMC2948730. 

 

2:  Müssigbrodt H, Michels R, Malchow CP, Dilling H, Munk-Jørgensen P, Bertelsen A. Use of the ICD-10 classification in psychiatry: an international survey. Psychopathology. 2000 Mar-Apr;33(2):94-9. doi: 10.1159/000029127. PMID: 10705253.

3:  Leucht S, van Os J, Jäger M, Davis JM. Prioritization of Psychopathological Symptoms and Clinical Characterization in Psychiatric Diagnoses: A Narrative Review. JAMA Psychiatry. 2024;81(11):1149–1158. doi:10.1001/jamapsychiatry.2024.2652

4:  Mezzich JE, Fabrega H Jr, Coffman GA, Haley R. DSM-III disorders in a large sample of psychiatric patients: frequency and specificity of diagnoses. Am J Psychiatry. 1989 Feb;146(2):212-9. doi: 10.1176/ajp.146.2.212. PMID: 2783540.

5:  Chute CG, Çelik C. Overview of ICD-11 architecture and structure. BMC Med Inform Decis Mak. 2022 May 16;21(Suppl 6):378. doi: 10.1186/s12911-021-01539-1. PMID: 35578335; PMCID: PMC9109286.

6:  Harrison JE, Weber S, Jakob R, Chute CG. ICD-11: an international classification of diseases for the twenty-first century. BMC Med Inform Decis Mak. 2021 Nov 9;21(Suppl 6):206. doi: 10.1186/s12911-021-01534-6. PMID: 34753471; PMCID: PMC8577172.

7:  Quan H, Steinum O, Southern DA, Ghali WA. Coding mechanisms for main condition in ICD-11. BMC Med Inform Decis Mak. 2025 Jul 10;21(Suppl 6):387. doi: 10.1186/s12911-025-03069-6. PMID: 40640794; PMCID: PMC12243148.