Homophilic selection is the tendency for an individual to select friends based on shared characteristics. Within the context of mental health diagnosis—and self-diagnosis—the existing literature is limited. The objective of this review is to identify existing literature on the subject and to argue in support of further research in this area 1.

To substantiate this, two systematic reviews were conducted: the first investigated the question, ‘Do individuals with a diagnosed psychiatric condition exhibit homophilic friendship selection?’, and the second, ‘Do individuals with a self-diagnosed psychiatric condition exhibit homophilic friendship selection?’.

The paper describes both reviews in turn (providing a background, method, results, and conclusion) and ends by describing the potential of future research.

Systematic review 1: Do individuals with a diagnosed psychiatric condition exhibit homophilic friendship selection?

The phenomenon that people are attracted to those similar to them has been well acknowledged since the mid twentieth century, a tendency ascribed to the term homophily [2]. Homophily is further categorised into 2 types: socialisation (people tending to influence each other so that they become more similar over time in each other’s presence), and selection (people tending to select friends based on a first impression that they are similar to one another) [3]. A sub-category of selection is deselection, where people instead deselect and avoid making friends with those they immediately recognise as different to them. Research of this tendency within psychiatry is relatively recent. In the introduction to his own paper in 2011, Matteo Giletta summarised all the existing data involving homophilic selection and depressive symptoms up until that point: ‘Only three studies tested selection effects; two reported that both male and female adolescents were more likely to select friendship groups with similar levels of depressive symptoms [4, 5], but one found no evidence of selecting a best friend based on similarity in depressive symptoms [6]. Including those papers, there are now 17 papers investigating homophilic selection and preference. The first aim of the review is to consider all the existing literature in order to address the titular question, and the second aim is to identify areas where the research is insufficient, according to predetermined parameters.

Method

Studies were identified through an initial database search of PsycINFO, Medline, and Embase on December 7th 2023 and a subsequent search on August 7th 2024 via Ovid, using MeSH (Medical Subject Heading) terms, keywords and phrases, with the purpose of ensuring the search was kept up to date. The most technically accurate search term for this review would be ‘homophilic selection’, although this yielded only 6 results from PsycINFO, none of which also referred to a mental health condition, and the phrase is not a MeSH term. For this reason, the search strategy was kept broad [8 major psychiatric diagnoses] + [variations of homophily, friendship, and social network analysis].

Due to its nature the research question could not be reduced to non-repeating concepts, however this was overcome by screening a large set of papers using a relatively strict eligibility criteria. Papers were searched from three databases—PsycINFO, Medline, and Embase—selected based on relevance and availability on Ovid. The full search strategy is listed in the “Appendix”.

All 3 sets of papers (PsycINFO = 19,399, Medline = 5029, Embase = 7988, Total = 32,416) were uploaded to SR-Accelerator for deduplication and screening. 9536 duplicates were removed, leaving 24,546 papers screened by evaluating the titles and abstracts against an eligibility criteria.

Exclusion criteria

The paper cannot refer to social behaviour or relationships in a clinical or therapeutic setting. The aim is to understand how people behave in their ‘everyday lives’ rather than their behaviour in response to external support. Furthermore, such relationships are not amongst peers but amongst a patient and a practitioner, which is not the type of relationship being investigated.

The paper cannot intend to measure to emotional contagion or depressive contagion etc. since the aim is to investigate how social factors affect mental health issues over time but rather how existing mental health issues affect initial social behaviour.

The paper cannot refer to peer support programs; if ‘support’ is mentioned it must be in the context of mutual relationships.

The paper cannot assess a demographic where everyone has the same medical condition (e.g. social network analysis of psychiatric patients with HIV) since such a demographic and their behaviour isn’t representative of people with mental health diagnoses in general.

Inclusion criteria

The paper must record the mental health status of one individual AND of the individuals around them (via any method), OR record the mental health status of many individuals considered together as a group (via any method).

The paper must collect comparative data considering the social or friendship behaviour of those with mental illness and the social or friendship behaviour of those without mental illness.

The paper must refer to voluntary close relationships formed by groups of people where they have been given the opportunity to organise themselves.

The paper must refer to a specific mental health diagnosis or symptom or trait. Equivalent research into shyness, aggression, drug-use, deviant behaviour, etc., will not be included.

3 authors independently conducted an abstract screening using SR-Accelerator of the 24,546 papers, producing a consensus of 29 papers which met the eligibility criteria. 3 were unavailable for full-text retrieval. The full text of the remaining 26 papers were independently screened by the authors against the eligibility criteria to produce a final set of 14 papers which were accepted into the data extraction stage of the review. (The “Appendix” provides with explanation a list of the publications removed after screening). Inter-reviewer discrepancies were small and resolved by structured discussion.

3 additional papers found from a citation search of the 14 were also screened and accepted. 2 of these 3 papers were on PsycINFO but not screened by the original search. This may highlight a weakness of the search, although these are the 3 oldest papers to be accepted in the review (all pre-1995), so it may also be the case that these papers were tagged with more dated keywords that fell through the search instead. Notably, an assessment of bias was not conducted.

ResultsSummary of papers

Reference

Source

Name of first author

Date published

Type of research design

Country

Qualitative data

Quantitative data

Diagnoses investigated

Number of participants

Participant age summary

[7]

Screening search

Gerbert J. T. Haselager

November 12 1982

Social network analysis

Netherlands

No

Yes

Depressive symptoms

192

Mean = 11.08

[8]

Screening search

Maarten Van Zalk

July 2010

Social network analysis

Sweden

No

Yes

Depressive symptoms

847

Mean = 14.29

[9]

Screening search

Jacob E. Cheadle

6 June 2012

Social network analysis

USA

No

Yes

Depressive symptoms

798

Cohort range = 12–18

[10]

Screening search

Noona Kiuru

30 June 2011

Social network analysis

Finland

No

Yes

Depressive symptoms

949

Lowest = 16

[6]

Screening search

Matteo Giletta

30 May 2011

Social network analysis

Netherlands

No

Yes

Depressive symptoms

974 (after all restrictions to the sample)

mean = 13.77

[11]

Screening search

Natalie P. Goodwin

19 October 2011

Social network analysis

USA

No

Yes

Depressive symptoms

367

Cohort range = 12–17

[12]

Screening search

Sterett H. Mercer

9 November 2010

Social Network Analysis

USA

No

Yes

Depression, social anxiety

1016

Cohort range = 8–9

[4]

Citation search

Aaron Hogue

1 November 1995

Social network analysis

USA

No

Yes

Internalised distress (depression, anxiety, tension, headaches and stomachaches)

6357

Cohort range = 14–17

[13]

Screening search

Nejra Van Zalk

3 May 2011

Social network analysis

Sweden

No

Yes

Social anxiety and depression

834

Mean = 14.29

[14]

Citation search

Abram Rosenblatt

April 1991

Inter-participant interview

USA

Yes

Yes

Depressive symptoms

24

Lowest = 18

[15]

Screening search

Rebecca A. Schwartz-Mette

5 December 2018

Quantitative interview

USA

No

Yes

Depressive symptoms

228

Mean = 19.54

[16]

Citation search

Kathleen Ries Merikangas

12 November 1982

Quantitative interview

USA

No

Yes

Primary affective disorder (depression)

56 (plus partners)

Mean = 43.39

[17]

Screening search

Catherine J. Crompton

7 March 2020

Qualitative interview

Scotland

Yes

No

Autism (specifically no social anxiety diagnoses)

12

Mean = 33.58

[18]

Screening search

Emma Pritchard-Rowe

28 August 2023

Qualitative interview

England

Yes

No

Autism

22

Mean = 39

[19]

Screening search

Collette Sosnowy

11 March 2019

Qualitative interview

USA

Yes

No

Autism

20

Mean = 23.5

[20]

Screening search

Cynthia Maya Beristain

4 September 2020

Qualitative interview

Canada

Yes

No

ADHD

9

Cohort range = 16–18

[21]

Screening search

Imola Marton

11 September 2012

Social network analysis

Canada

No

Yes

ADHD

92

Cohort range = 8–12

Assessment of diversity inclusion of papers

Reference

Participant gender (%)

Are gender differences investigated?

Ethnic background

Are ethnic differences investigated?

How long have participants had their diagnosis

[7]

50% female, 50% male

Yes

Not recorded, but the context population's was 89.5% Dutch/Caucasian, and 10.5% was from Suriname, the Netherlands Antilles, Indonesia, Turkey, and Morocco

No

Not stated

[8]

42% female, 58% male

Yes

10% ethnic minorities

No

Not stated

[9]

Not stated

Yes

'Mostly white'

No

At least 1 year (present at 2 points in time)

[10]

56% female, 44% male

Yes

99% Finnish speaking

No

At least 1 year (present at 2 points in time)

[6]

49% female, 51% male

Yes

Born in the Netherlands (95.5%) and had at least one parent who was born in the Netherlands or in another European country (93.4%). Approximately 11% of the participants were ethnic minorities from Turkey (2%), Surinam and Dutch Antilles (1.7%), Morocco (1%), or elsewhere (6.3%)

No

Not stated

[11]

51% female, 49% male

Yes

51% females, 64% White, 20% African American, 12% Latinos and 1% of youth from other ethnic origins

No

At least 2 years (present at 3 points in time)

[12]

51% female, 49% male

Yes

68% White, 20% African American,

5% Hispanic, 5% Asian, and 2% Multiracial

No

Not stated

[4]

56% female, 44% male

Yes

10.3% Black, 59.9% White, 14.2% Asian, 15.6% Hispanic

No

At least 1 year (present at 2 points in time)

[13]

41% female, 59% male

Yes

8% first generation immigrants

No

For 70% of candidates; at least 2 years (present at 3 points in time)

[14]

Not stated

No

Not stated

No

Not stated

[15]

54% female, 46% male

Yes

82.2% European American, 10.5% African American, 4.6% Asian, 2.7% American Indian/Alaskan Native, and 8.8% Hispanic/Latino(

No

Not stated

[16]

64% female, 36% male

Yes

Not stated

No

Mean 4 months of hospitalisation

[17]

83% female, 17% male

No

Not stated

No

Mean 4.91 years

[18]

45% female, 45% male, 9% non-binary

No

91% White, 4.5% Mixed/multiple ethnic groups, 4.5% other

No

Not stated

[19]

35% female, 55% male, 10% non-binary

No

90% non-Hispanic white, 5% Asian, 5% African American

No

Not stated

[20]

56% female, 44% male

No

Not stated

No

At least 9 years (DSM-IV states symptom onset must occur before age 7, all participants were at least 16)

[21]

28% female, 72% male

Yes

Not stated

No

At least 1 year (DSM-IV states symptom onset must occur before age 7, all participants were at least 8)

Results extracted from papers

Reference

Did the participants indicate a homophilic preference?

Homophilic selection?

Homophilic socialisation?

Do the participants have any other relationships? (friends without diagnoses, family, marriage, clinical staff…)

[7]

Yes but not to a statistically significant degree

Not investigated

Not investigated

Yes

[8]

Yes

Yes (as well as heterophilic deselection)

Yes

Yes, these relationships were used as controls

[9]

Yes

Yes (more so in boys) (and no evidence for deselection)

Yes

Yes

[10]

Yes

Yes

Yes

Not investigated

[6]

Yes

No (statistically insignificant compared to controls)

Yes (more so in females)

Yes but these relationships were not investigated

[11]

Yes

Yes

Yes

Not investigated

[12]

Yes

No

Yes

Not investigated

[4]

Yes

Yes

Yes for boys, no for girls (potential to increase distress in boys but not reduce it)

Not investigated

[13]

Yes

'In sum, then, social anxiety could be a selection criterion for choosing friends'

Yes

Yes but these relationships were not investigated

[14]

Yes

Yes (as well as heterophilic deselection)

No; the interaction was short and amongst strangers

Not investigated

[15]

Yes

Yes

No; the interaction was short and amongst strangers

Not investigated

[16]

Yes

Yes

Yes

Yes (patients tended to have a family history of affective disorder)

[17]

Overall yes

Yes (befriending new people in novel autistic spaces)

Yes

Yes (peers without autism), but for the majority of participants these friendships were less fulfilling. For a minority they were more fulfilling

[18]

Overall yes

Ambiguous

Yes

Yes (peers without autism), but for the majority of participants these friendships were less fulfilling. For a minority they were more fulfilling

[19]

Yes

Yes (befriending new people in autistic spaces)

Yes

Yes (peers without autism), but these friendships were less fulfilling

[20]

Most did, 3 + indicated the opposite (to avoid engaging in deviant behaviour)

More so heterophilic deselection

Not investigated

Yes (peers without ADHD); these were fulfilling as long as they were understanding and accepting

[21]

Overall yes (to friends with a 'learning or behavior problem')

Yes (in the summer camp)

Yes (those with ADHD have shorter lasting friendships than controls)

Yes (peers without ADHD), and these are less common than ADHD-ADHD friendships

Results by diagnosis and consideration of limitations

12 papers found participants exhibiting homophilic selection;