Co-occurrence of lifestyle and mental health issues in adolescents: A cross-sectional study of the YouthCHAT cohort in Aotearoa New Zealand Authors: FanZhen Zhou, James Warren, Felicity Goodyear-Smith Published: 2026 ABSTRACT AbstractPurposeTo identify patterns of co-occurring lifestyle and mental health issues among adolescents aged ≤15 years in Aotearoa New Zealand using data from the YouthCHAT screening tool, and to determine whether distinct clusters of risk profiles are associated with gender or ethnicity.MethodsA cross-sectional analysis was conducted on 11,244 de-identified YouthCHAT electronic psychosocial screens completed between 2019 and 2024 in primary care settings, predominantly serving low socioeconomic school communities. Frequencies of positive screens and help-seeking responses were calculated. Kendall's tau correlations, factor analysis, and k-means clustering were used to examine co-occurrence and derive latent risk clusters. Associations between cluster membership and gender or ethnicity were assessed using chi-squared tests with Bonferroni correction.ResultsThree factors explained 44.6% of variance, representing: (1) psychosocial distress, (2) sexual health and identity concerns, and (3) substance misuse. Four clusters emerged: Minimal Issues (79.8%), Distressed (16.3%), Substance Misuse (3.2%), and High Risk (0.6%). Females were overrepresented in the Distressed cluster (3:1 vs males), and Māori youth were disproportionately represented in the High Risk and Substance Misuse clusters (p < 0.001). Co-occurring issues most frequently involved stress, behaviour, eating/body-image concerns, anxiety, anger, and physical inactivity.ConclusionDistinct clusters of co-occurring lifestyle and mental health issues are evident in early adolescence, with inequitable distribution by gender and ethnicity. YouthCHAT's multi-issue, integrated approach supports early identification of these patterns and aligns with WHO frameworks emphasising comprehensive, holistic adolescent health assessment and targeted support for high-risk groups.Keywords: Adolescent, Mental health, Health risk behaviours, New Zealand, Cluster analysis, Cross-sectional studies, Cohort studies Introduction As they transition from childhood to adulthood, adolescents (young people aged between 10 and 19 years) begin to engage in risky health behaviours, including the use of nicotine, alcohol and other recreational drugs, poor eating habits and physical inactivity, as well as demonstrating increased psychological distress [1]. Sexually active high school adolescents often do not use condoms, hence are at risk of sexually transmitted infections and pregnancies [2]. Adolescents with gender identity and sexual orientation issues are at greater risk of mental health conditions such as suicidal ideation, stress, and anxiety [3], as well as smoking and vaping, alcohol and illicit drug misuse [4]. Sexual abuse of adolescents is a significant health concern worldwide, and early identification means appropriate and timely health services can be provided which may prevent severe and long-term consequences [5]. The overarching goal of the World Health Organization (WHO) is to promote, provide, and protect the health and well-being of all people, everywhere [6]. The WHO's Strategic Objective 6 is “To prevent or reduce risk factors for health conditions associated with use of tobacco, alcohol, drugs and other psychoactive substances, unhealthy diets, physical inactivity and unsafe sex.” Understanding risk factors in combination rather than singly therefore aligns with these WHO objectives. A number of studies have found that these lifestyle and mental health issues occur in clusters. For example, a systematic review demonstrated clustering of low physical activity, high screen time and high soft drink and snack consumption [7]. Studies in Brazil [8], the United States [9] and Malaysia [10] have found a high proportion of adolescents engage in multiple health risk behaviours such as smoking, inadequate fruit intake, alcohol intake and/or physical inactivity, and a study of Vietnamese youth found that unhealthy diet, sedentary behaviour and low fruit/vegetable intake (in females), and alcohol use (in males) were common clusters [11]. Adolescents who engage in multiple health risk behaviours also experience higher levels of depression and anxiety than their peers [12], [13]. Research of young people in Ghana found a high prevalence of adolescents who smoked, drank alcohol and were physically inactive, and that depression was associated with this cluster [14]. YouthCHAT (Youth electronic Case-Finding and Help Assessment Tool) is a self-report web-based application developed in Aotearoa New Zealand (NZ) to assess risky health behaviours and mental distress in young people in primary care settings [15], [16]. Modules address smoking, vaping, alcohol and drug use, gambling, abuse, conduct and behavioural disorders, anger control, eating and body image, physical activity, sexual health and identity, anxiety, depression and external stressors. The number of questions in a module may increase with branching logic depending on the response. For issues with positive responses, adolescents are asked whether this is something with which they would like help, either immediately or later [17]. YouthCHAT is a clinical tool with the screening conduced immediately before a clinical consultation, hence any safeguarding issues for youth with positive screens will be dealt with by the clinician reviewing the YouthCHAT report. In general, socially disadvantaged young people in NZ have a higher prevalence of risky behaviours and mental health issues [16], and YouthCHAT has been researched extensively with indigenous Māori, as well as Pacific and Asian youth [18], [19]. Understanding of the co-occurrence of risky behaviours may provide insight on underlying factors and inequity. Co-occurrence can be defined as an association or correlation of a pair of issues. Clustering is where more than a pair of issues may be correlated. Without intervention, unhealthy behaviours might continue into adulthood. The clustering of risk factors suggests the need for interventions that target multiple health risk behaviours simultaneously rather than single behaviours in isolation [20]. Studies of risk factors often focus on older adolescents, 15 to 18 years.10 13 Risk behaviour increases in teenagers with age, hence understanding early patterns is important. The aim of this study is to describe patterns of co-occurring adolescent lifestyle and mental health issues in NZ in young people 15 years or under in terms of relative frequency of individual and co-occurring factors, and how young people are clustered, based on these factors. Methods Study design and cohort For this cross-sectional study a de-identified extract was taken from the servers of the YouthCHAT system containing all screens conducted from June 2019 to February 2024. The resulting electronic screening (e-screening) cohort consists of young people who received care from primary care nurses and doctors in NZ, predominantly at clinics based in low socio-economic decile schools and public health services, across 70 distinct sites. Year 9 students in low-decile secondary schools in NZ are required to undergo a face-to-face HEEADSSS (Home, Education, Eating, Activities, Drugs and Alcohol, Sexuality, Suicide and Depression, Safety) interview to detect psychosocial issues [21]. A counter-balanced randomised trial found that a YouthCHAT screen performs as well as a HEEADSSS assessment, and an initial YouthCHAT reduces the time needed to conduct a HEEADSSS [22]. Field use of YouthCHAT is therefore highly concentrated on Year 9 adolescents (typically aged 13 or 14 years). We limited the cohort for the present study to respondents aged ≤15 years in keeping with the main usage of the tool. All completed e-screenings in this age range were included for analysis. Ethical approval was granted by the Auckland Health Research Ethics Committee, reference AH26667. The introduction about YouthCHAT includes ‘Your response may be used for research purposes but there is no way that you may be identified’ and youth may then either start the survey or decline. The ethics committee approved a waiver of formal consent for the secondary analysis, given there was minimal risk and the researchers only received de-identified data. Data collection and measurement YouthCHAT e-screening takes approximately 5–15 min and is completed on an electronic device (computer, tablet, phone) before a consultation (for example, at home or in the waiting room) with a general practitioner, nurse or counsellor. The screen is explained with a choice of short videos or animated infographic. Once YouthCHAT is complete, a summary report is immediately available to the health provider to use as a starting point for a discussion with the young person. All responses are stored in a secure encrypted database. Because it is a clinical tool, young people can only access YouthCHAT by invitation from their health povider, as the first step in a consultation. Each respondent is prompted to self-report their age, gender and ethnicity, after which lifestyle and mental health issues are screened sequentially. For each issue, there are one or two ‘gate keeper’ questions. If the answers to these questions do not meet the criteria for a positive screen, questioning proceeds to the next module, providing efficiency for the frequent cases where few, if any, issues are found. For a positive screen, module-dependent follow-up questions are asked, concluding with a question on whether they want help with this issue with options ‘No’, ‘Yes but not today’ or ‘Yes’ [17]. For the depression module, for instance, the PHQ-2 [23] acts as the gate-keeper questions (with a score of ≥2 on either question constituting a positive screen), with the remaining seven questions of the PHQ-9 (Patient Health Questionnaire-9 Modified for Adolescents) and the help question presented when there is a positive screen. This study aims to describe co-occurring lifestyle and mental health issues in adolescents aged ≤15 in terms of individual and co-occurring factors, and how young people are clustered, based on these factors. Table 1 lists the YouthCHAT issues with the positive screen criterion of each (i.e., the responses that trigger asking follow-up, including the help question); the table also gives abbreviations for issue names as used for compactness in Fig. 1, Fig. 2, Fig. 4. The full text of the YouthCHAT questions is provided in Appendix pp. 1–11. Table 1YouthCHAT issues with criteria for asking follow-up (including help question).Issue (abbreviation)Positive screen criterion (condition for follow-up including help question)Anxiety“Yes” to “During the past month have you been worrying a lot about everyday problems?”Smoking nicotine (Smoke)“Yes, currently” to “Have you ever used tobacco (eg cigarettes) or nicotine (e.g., vaping)?” and other than “Never” to “In the past three months, how often have you used tobacco and/or nicotine (e.g., cigarettes, vaping, chewing tobacco, cigars, chop baccie, durries ciggies, durry)?”Drinking alcohol (Alcohol)Other than “Didn't use” to “How often did you use alcoholic drinks (e.g. beer, wine, spirits, premixes, RTDs) in the last month?”Problem gambling (Gamble)“Yes” to “Do you gamble (bet)?”Eating disorder / body-image (Eating)“No” to “Are you happy with your body shape or your weight?” or “Yes” to “Have there been any recent changes in your weight?”Depression“More than half the days” or “Nearly every day” to either of first two questions of PHQ-9Behaviour control (Behaviour)“Yes” to either of “Do you have trouble with any of the following: paying attention at school or work, remembering where you put things or waiting your turn?” or “Do you often break rules or refuse to do as you're told?”Sexual identity concerns (SexID)“Yes” to “Do you have any worries or problems about who you are sexually attracted to or your sexual identity?”Sexually active - risk of sexually transmitted infection (SexSTI)“Yes” to both of sexually active (“Have you been sexually active within the past three months? (This includes sex involving your mouth, vagina, penis or bottom)”) and “Are you at any risk of catching a sexually transmitted infection (STI)? (If you do not use condoms and you or your partner have had more than one partner there is a big risk of this)”Sexually active - risk of pregnancy (SexPregnancy)“Yes” to both of sexually active and “Are you at any risk of getting pregnant or getting your partner pregnant? (If you are not using any protection there is a big risk of getting pregnant or getting your partner pregnant)”Unwanted sexual attention (SexUnwanted)“Yes” to “Has anyone tried to touch you sexually or push sex on you when you have not wanted it?”Exposure to abuse (Abuse)“Yes” to either of “Is there anyone in your life of whom you are afraid or who hurts you in any way?” or “Is there anyone in your life who controls you and prevents you doing what you want?”Anger control (Anger)“Yes” to “Is controlling your anger sometimes a problem for you?”Lack of exercise (Exercise)“No” to “Do you normally do more than 30 min of moderate or vigorous exercise (such as walking or a sport) on 5 days of the week?”StressUnder the prompt “Do any of the following cause you stress or problems?” ticked any of: “Issues at home”, “Issues at school or work”, “Relationships with specific people in your life”, “Being bullied”, “Violence”, “An upsetting event in your life (e.g., accident, illness, loss of a friend/family member, or another hurtful situation)”, “Money”, or “Other” YouthCHAT issues with criteria for asking follow-up (including help question). Positive screening co-occurrence frequencies (with positive screening frequencies of individual issues on the diagonal). Issue score correlations (Kendall's tau). Radar chart of mean issue scores per cluster. The follow-up questions of some YouthCHAT modules have well-known scores and interpretations, such as the PHQ-9 [24]. However, other modules lack comparable scoring, and some are limited to the help question after positive screening. To provide a basis for comparison of screening response intensity across the YouthCHAT modules, we applied a score of 0 for negative screens, and for positive screens of a mental health or behavioural issue, a score based on the help response: 1 for No, 2 for Yes but not today, and 3 for Yes. Data security The production YouthCHAT system (operated by Kekeno) uses National Health Identifier (NHI) to interoperate with clinical systems. Meaningless case numbers are used rather than NHI in the data extract provided by Kekeno to the authors. The authors' extract only contains interview time with age in years, high-level ethnicity and gender (as per the categories reported in the paper) for each participant. This extract is encrypted-at-rest with password never exchanged by email. As a further precaution against re-identification, clinic name was removed from the version that is accessed by the Python code, which is decrypted only into computer memory. Māori data sovereignty In line with Te Mana Raraunga (Māori Data Sovereignty principles), prior to data extraction we contacted our key stakeholders at the Northland Primary Care Organization (where there is a high proportion of Māori youth completing YouthCHAT, and with whom we have conducted considerable YouthCHAT research), asking for them to confirm that we should include their data in our analyses, and they were happy for us to do so. Statistical analysis Frequency of positive screens and associated help responses were computed, as were the frequency of co-occurrence of positive screens, and correlation of issue scores (as Kendall's tau). To validate the approach of having the issue score successively higher based on the help response, the frequency of help responses for positive depression screens was cross-tabulated by PHQ-9, based levels of none-minimal, mild, moderate, moderately severe and severe on cut-offs of 5, 10, 15, and 20, respectively. Chi-squared test of independence of frequency of help response was applied between each adjacent PHQ-9 based level with Bonferroni correction for multiple comparisons (multiplying p-value by 4). To further explore patterns of issue co-occurrence, factor analysis was applied to the issue scores with varimax rotation to aid interpretability of factor loadings, retaining factors with eigenvalues >1 for further analysis. Participants were then clustered using k-means based on the retained factors. Reduction in the sum of squared error (SSE) within clusters guided the selection of the number of clusters for solution to interpret. For each cluster, factor means and mean issue scores were computed. Demographics of each cluster were tallied. Independence of cluster membership for female gender and combined gender diverse / other (compared to male), and for Māori ethnicity (compared to NZ European, ie New Zealanders of European descent) were tested by chi-squared with Bonferroni correction (multiplying p-value by 3). Analysis was conducted using Python with open-source libraries (pandas 1.4.4, NumPy 1.2.4.3, SciPy 1.9.1, and sklearn 1.2.2). Results Of 13,811 YouthCHAT e-screening records, 11,665 were complete, and 11,244 of these were aged ≤15 and retained for analysis. Table 2 shows demographics of this cohort. The majority of incomplete cases ended before completing the initial three demographic questions; 1216 of 2146 had no responses entered. For incomplete cases aged ≤15, where age and gender were available (821) and age and ethnicity available (795) distribution was compared to the retained cohort, finding no difference in gender (p(Χ2) = 0.40), but difference in ethnicity (p(Χ2) < 0.001). Māori were represented in line with the retained cohort (189 observed, 192.8 expected), whereas NZ European had higher than expected completion (231 incomplete, 336.0 expected) and all other ethnicities had lower than expected completion. Help responses of the retained cohort by issue, in total and by ethnicity and gender, are provided in Appendix pp. 12–16. Table 2Demographics of completed YouthCHAT screens aged ≤15 years.n (%, 95% CI)Age 13 and under7495 (66·7%, 65·8%–67·5%) 143363 (29·9%, 29·1%–30·8%) 15386 (3·4%, 3·1%–3·8%)Gender Male5570 (49·5%, 48·6%–50·5%) Female5487 (48·8%, 47·9%–49·7%) Gender Diverse115 (1·0%, 0·9%–1·2%) Gender Other72 (0·6%, 0·5%–0·8%)Ethnicity Māori2731 (24·3%, 23·5%–25·1%) NZ European4857 (43·2%, 42·3%–44·1%) Pacific2018 (17·9%, 17·2%–18·7%) Asian640 (5·7%, 5·3%–6·1%) MELAA182 (1·6%, 1·4%–1·9%) Other Ethnicity816 (7·3%, 6·8%–7·8%)Age 13 and under is the lowest the reporting option offered by YouthCHAT.NZ European is New Zealander of European descent. NZ European as used in NZ largely intersects with ‘White’ as used in the US, but emphasises ancestry over skin colour.MELAA = Middle Eastern, Latin American, and African. Demographics of completed YouthCHAT screens aged ≤15 years. Age 13 and under is the lowest the reporting option offered by YouthCHAT. NZ European is New Zealander of European descent. NZ European as used in NZ largely intersects with ‘White’ as used in the US, but emphasises ancestry over skin colour. MELAA = Middle Eastern, Latin American, and African. Fig. 1 shows positive screening co-occurrence frequencies for pairs of issues, with positive screening frequency of each individual issue on the diagonal. The most frequently observed issues were stress, behaviour and eating/body-image (each around 50% prevalence), followed by anger (36%), anxiety (32%) and exercise/physical-activity (31%). The most frequently co-occurring issue pairs were behaviour with stress (33%), eating issues with stress and behaviour (each 31%), behaviour with anger (27%), and stress with anxiety and anger (each 26%). Fig. 2 shows correlations of pairs of issue scores (ie a score from 0 to 3 incorporating screening result and help response). All the frequently co-occurring issues listed in the previous paragraph are highly correlated, as well as abuse with stress and unwanted sexual attention (0·29 and 0·28, respectively), and anger with depression (0·28). Due to the large sample size, all correlations ≥0·05 have p-values <10−8 (full p-value matrix in Appendix p 17). Table 3 shows frequency of help responses by PHQ-9 level. The independence of help distribution between each successive PHQ-9 level is rejected (adjusted p < 0·05). This can be observed as a consistent progressive shift in help responses with increasing YouthCHAT depression issue score. With each increase in depression level there was a shift away from ‘No’ into positive help responses, and increasing rates of ‘Yes’ (as compared to ‘Yes but not today’) responses. This pattern is consistent with, and hence validates our use of the help response to assign higher issue scores for correlation and factor analysis. Table 3Frequency of help responses by PHQ-9 level (from PHQ-9A) for positive depression screens with chi-squared test for independence of adjacent levels.PHQ-9 levelHelp responseAdjusted p-value compared to level belowNo (score = 1)Yes but not today (score = 2)Yes (score = 3)<5, None-minimal191 (87·5%)8 (7·7%)5 (4·8%)–5–9, Mild207 (68·8%)68 (22·6%)26 (8·6%)0·00310–14, Moderate255 (56·0%)123 (27·0%)77 (16·9%)0·00215–19, Moderately Severe162 (45·1%)120 (33·4%)77 (21·5%)0·03320–27, Severe98 (33·1%)107 (36·2%)91 (30·7%)0·011PHQ-A is a depression screening tool for adolescents, range 0 to 27.1At least one PHQ-2 item must score ≥2 to screen positive and invoke the full PHQ-9A and help question in YouthCHAT; thus, this band is limited to PHQ-9A values 2-4. Frequency of help responses by PHQ-9 level (from PHQ-9A) for positive depression screens with chi-squared test for independence of adjacent levels. PHQ-A is a depression screening tool for adolescents, range 0 to 27. At least one PHQ-2 item must score ≥2 to screen positive and invoke the full PHQ-9A and help question in YouthCHAT; thus, this band is limited to PHQ-9A values 2-4. In factor analysis, three factors had eigenvalues>1 explaining 23·6%, 11·7% and 9·3% of variance in scores, respectively (cumulative 44·6%), with factors loadings as shown in Table 4. The first factor has the highest loadings on stress, anxiety, anger, depression, behaviour and eating-disorder/body-image issues – issues that are individually frequent for positive screens, and that have high levels of co-occurrence as positive screening pairs and score correlation. The second factor has highest loadings on sexual health issues, especially sexually transmitted infection (STI) and pregnancy risk (not especially frequent issues in this young cohort, but with highly-correlated scores), as well as concerns about sexual identity and unwanted sexual attention, along with gambling, alcohol and abuse. The third factor has highest loadings for smoking and alcohol issues, along with unwanted sexual attention and pregnancy risk. Table 4Factor loadings by issue.IssueFactor 1Factor 2Factor 3Anxiety0·7460·0460·034Smoking nicotine0·078−0·0400·797Drinking alcohol0·0770·2450·676Problem gambling0·1720·314−0·018Eating disorder / body-image0·5590·0490·034Depression0·657−0·0470·156Behaviour control0·6200·0930·142Sexual identity concerns0·4490·264−0·074Sexually active - risk of sexually transmitted infection−0·0050·8400·086Sexually active - risk of pregnancy0·0350·8160·216Unwanted sexual attention0·3790·2470·336Exposure to abuse0·5100·2280·179Anger control0·6800·0630·126Lack of exercise0·4050·139−0·185Stress0·7630·0490·068 Factor loadings by issue. Clustering on these factors yielded solutions with score SSE by number of clusters as shown in Appendix p 18. SSE reduction attenuated beyond four clusters, leading to our choice (in keeping with the ‘elbow’ heuristic) to explore the four-cluster solution. This solution had silhouette score 0·57 and cluster means for each of the three factors are shown in Fig. 3. Fig. 4 plots the mean score on each issue for the individuals in each cluster. Characteristics of the four clusters, including the interpretive labels assigned by the authors, were: •‘Minimal Issues’ – 8976 cases (79·8%, 95% CI 79·1%–80·6%). This populous cluster lacked positive means for any of the three factors. ‘Minimal Issues’ – 8976 cases (79·8%, 95% CI 79·1%–80·6%). This populous cluster lacked positive means for any of the three factors. •‘Distressed’ – 1834 cases (16·3%, 95% CI 15·6%–17·0%). Individuals in this cluster had the highest mean score for factor 1 and for stress, anxiety, eating/body-image and depression issues. ‘Distressed’ – 1834 cases (16·3%, 95% CI 15·6%–17·0%). Individuals in this cluster had the highest mean score for factor 1 and for stress, anxiety, eating/body-image and depression issues. •‘High Risk’ – 70 cases (0·6%, 95% CI 0·5%–0·8%). This numerically-small cluster had the highest mean on factor 2, and a substantial positive mean for factor 3, as well as the highest mean on all sexual health issues as well as behaviour, alcohol, anger, exercise and abuse. ‘High Risk’ – 70 cases (0·6%, 95% CI 0·5%–0·8%). This numerically-small cluster had the highest mean on factor 2, and a substantial positive mean for factor 3, as well as the highest mean on all sexual health issues as well as behaviour, alcohol, anger, exercise and abuse. •‘Substance Misuse’ – 364 cases (3·2%, 95% CI 2·9%–3·6%). This cluster had highest mean score for factor 3, a small positive mean on factor 1, highest mean in smoking, and second highest mean in alcohol. ‘Substance Misuse’ – 364 cases (3·2%, 95% CI 2·9%–3·6%). This cluster had highest mean score for factor 3, a small positive mean on factor 1, highest mean in smoking, and second highest mean in alcohol. Cluster means by factor. Table 5 shows gender and ethnicity distribution by cluster membership. Independence of cluster membership by gender – female, and combined gender-diverse/other compared to male – and for Māori compared to NZ European was rejected (each adjusted p < 0·001). Each group showed a reduced frequency in the Minimal Issues cluster versus their comparison group. Females outnumbered males by almost 3:1 in the Distressed cluster; and most gender-diverse/other individuals were outside the Minimal Issues cluster. Māori outnumbered NZ Europeans in the Substance Misuse cluster and constituted half the individuals in the High Risk cluster. Table 5Gender and ethnicity distribution by cluster membership, n (%, 95% CI).Minimal issuesDistressedHigh riskSubstance misuseGender Male4954 (88·9%, 88·1%–89·7%)446 (8·0%, 7·3%–8·7%)22 (0·4%, 0·3%–0·6%)148 (2·7%, 2·3%–3·1%) Female3966 (72·3%, 71·1%–73·4%)1270 (23·1%, 22·0%–24·3%)43 (0· 8%, 0·6%–1·1%)208 (3·8%, 3·3%–4·3%) Gender diverse28 (24·3%, 17·4%–32·9%)78 (67·8%, 58·8%–75·7%)4 (3·5%, 1·4%–8·6%)5 (4·3%, 1·9%–9·8%) Other28 (38·9%, 28·5%–50·4%)40 (55·6%, 44·1%–66·5%)1 (1·4%, 0·2%–7·5%)3 (4·2%, 1·4%–11·5%)Ethnicity Māori2097 (76·8%, 75·2%–78·3%)426 (15·6%, 14·3%–17·0%)36 (1·3%, 1·0%–1·8%)172 (6·3%, 5·4%–7·3%) NZ European4052 (83·4%, 82·4%–84·4%)680 (14·0%, 13·1%–15·0%)20 (0·4%, 0·3%–0·6%)105 (2·2%, 1·8%–2·6%) Pacific1538 (76·2%, 74·3%–78·0%)400 (19·8%, 18·1%–21·6%)11 (0·5%, 0·3%–1·0%)69 (3·4%, 2·7%–4·3%) Asian515 (80·5%, 77·2%–83·4%)121 (18·9%, 16·1%–22·1%)1 (0·2%, 0·0%–0·9%)3 (0·5%, 0·2%–1·4%) MELAA1140 (76·9%, 70·3%–82·4%)40 (22·0%, 16·6%–28·5%)0 (0·0%, 0·0%–2·1%)2 (1·1%, 0·3%–3·9%) Other634 (77·7%, 74·7%–80·4%)167 (20·5%, 17·8%–23·4%)2 (0·2%, 0·1%–0·9%)13 (1·6%, 0·9%–2·7%)1MELAA = Middle Eastern, Latin American, and African. Gender and ethnicity distribution by cluster membership, n (%, 95% CI). MELAA = Middle Eastern, Latin American, and African. Discussion Assessment of risky health behaviours and mental distress in young people (aged 15 and under) using a self-report web-based tool in primary care settings in NZ revealed patterns of co-occurring issues. We found the highest counts, both in frequency of issues individually and frequency of co-occurring pairs, for stress, behaviour, eating/body-image issues, anger, anxiety and physical inactivity. Correlation analysis, with intensity of issue based on help responses, revealed many further significant associations, including abuse with both stress and unwanted sexual attention. Factor analysis of issue scores shows the most frequent issues dominating the first of three factors. The other two factors largely align with unsafe sexual activity (factor 2) and substance misuse (factor 3). We identified four clusters of young people based on the factor analysis. Pleasingly, 80% indicated few risky health behaviours or mental health issues (‘Minimal’ cluster). However 16% were in the ‘Distressed’ cluster, with high scores for stress, anxiety, eating/body-image and depression issues, and 3% in ‘Substance Misuse’ cluster with high scores for smoking and drinking alcohol. A very small proportion (0·6%) were ‘High Risk’ with high scores on the risk of STI or pregnancy from unprotected sexual activity, concerns about sexual identity, exposure to abuse, difficulty controlling anger, problematic alcohol drinking and physical inactivity. Disparities in cluster membership based on gender and ethnicity were observed. We found an increased frequency of clustered lifestyle and mental health issues in females, Māori and gender-diverse young people. Mental health and substance misuse inequities in Māori youth have been found previously, but suggests there appears to be significant under-reporting, assessment, and treatment of emotional conditions for Māori compared with non-Māori [25]. A recent self-reported mental health and wellbeing data in a population sample of NZ secondary school students looked at clustering of anxiety, depression past-year self-harm, suicide ideation and suicide attempt, and similarly found that female, Māori and rainbow students were overrepresented in higher severity clusters [26]. Māori are disproportionately affected by modifiable social, environmental, and financial determinants of health, with colonisation, racism, and structural inequities considered to be root causes of health disparities [27]. Gender-diverse youth may experience mental health inequities from lack of access to affirming health care and discrimination [28]. Strengths of this study The study analysed self-reported data on 15 risky health behaviours and mental health issues from a large sample of over 11,200 young New Zealanders. The unique YouthCHAT ‘help’ question was validated against depression severity responses to the PHQ-9 questions. Scoring of issues by positive screening and help response supported correlation, factor analysis and clustering. Limitations This study used an opportunistic sample based on current users of the clinical tool YouthCHAT aged under 16 years with a single cross-section of self-report data. It therefore cannot be seen as a representative sample of NZ adolescents at large. In the 2023 Census, NZ European made up 58%, Māori 20% and Pacific 12% of the NZ population aged 12 to 16 years [29]. Our sample over-represents Māori and Pacific, more likely to be high-risk, at 24% and 18% respectively. There is a demographic bias of YouthCHAT data, as the tool is predominantly used at clinics in low socio-economic decile schools and public health services. Adolescents from more affluent backgrounds are hence less represented in our data. A further limitation is that people from different cultures and genders vary in how likely they are to ask for help. Analysis of incomplete cases showed NZ European more likely to complete YouthCHAT once initiated; the direction of this bias does not undermine our observation of overrepresentation of Māori in some clusters. Our issue scoring assumes uniformity of participants' willingness to ask for help across demographics. Further, our study assessed immediate outcomes rather than long-term behavioural changes. Intra-site correlation was not modelled, potentially overestimating precision. Alternate factor rotations or numbers of factors or clusters beyond the reported methods were not explored. Implications for research, policy and practice As Teh et al. indicate, a behaviour intervention programme which addresses multiple risky behaviours should be advocated and targeted to this high-risk sub-population, rather than siloed approaches focusing on single issues [10]. Most studies still focus on interventions for single lifestyle issues [30], [31], or for mental health [32], however there is growing evidence that smoking cessation, dietary interventions and physical activity can reduce depression and anxiety, and mind-body therapies such as yoga and mindfulness-based interventions can be effective across multiple domains [33]. The implications of our findings are that early interventions are needed to address multiple health risk behaviours simultaneously rather than single behaviours in isolation, and that vulnerable high-risk populations should be targeted. Further research could look at the distribution of healthy behaviours and mental health in a representative sample of youth to further determine how these issues may cluster. Sequential research could help identify the progression of these factors as youth age. A prospective study that evaluates these factors over time would allow exploration of these aspects in greater depth. Prospective studies might build on our findings, such as tracking the progression of clusters over time, and evaluating intervention efficacy. Fleming et al. propose an equity framework to promote NZ youth mental health, particularly Māori and Pacific [34]. Targeting interventions for high-risk youth at an early age has greater chance of success, before behaviours become entrenched. Because the adolescent brain is still developing, it is more vulnerable to stress, trauma, and poor decision-making. Intervening early can help healthier coping mechanisms and decision-making skills develop before the harmful patterns solidify [32]. We propose an integrated and inter-disciplinary approach that recognises risky health behaviours and mental health have co-directional influences and that family and social situations may require addressing. YouthCHAT explores a broader range of mental health issues and behavioural problems than most other approaches, and the ‘help’ question identifies which issues an adolescent wishes to address, now and later. This can help guide the focus of a personalised intervention [17]. The intervention required should not just address a single issue but provide a holistic line of action to assist youth make positive changes in the context of their lives. YouthCHAT will detect issues in Māori and gender-diverse youth which national school surveys show often go under-identified. 25 26 Customised multi-issue behaviour intervention programmes should target these high-risk populations. Our study indicates that these problems are already emerging for many young adolescents aged under 16. The evidence for effective multiple risk behaviour intervention is currently sparce. Multi-agent interventions involving teachers, families, and community members have been found to reduce disruptive behaviours, depression, and anxiety, and improve social skills and well-being in adolescence [35]. Early initiation of risk behaviours such as smoking, use of alcohol and early sexual activity has been associated with lower educational attainment, and early intervention has been shown to improve both educational and health outcomes [36]. Even a small benefit of such interventions can have high public health benefit [30]. While other studies have looked at the clustering of mental health and risky lifestyle behaviours in youth, our study examines a greater number of risk factors (including unhealthy eating and body-image issues, conduct disorders, exposure to abuse, sexual issues, and physical inactivity) as well as mental health issues. There is substantial evidence from neuroscience, psychology, and epidemiology that risky behaviour increases with age during adolescence [37], peaking in late adolescence (ages 15–19) [38]. Identifying patterns of risky behaviour early allows for targeted interventions, before these behaviours become entrenched or harmful [32]. Because risk behaviour increases with age, most research focuses on older adolescents. However, understanding early patterns is important; hence our focus is on those under 16 years. Although the majority of the young people had minimal issues, 16% were in a ‘distressed’ cluster with depression, anxiety, eating and body image issues and external stressors, and 3% in a cluster of substance misuse. A small number, but importantly more females and Māori, were in a high-risk cluster with numerous factors including sexual issues. This work aligns closely with WHO adolescent health frameworks,including AA-HA! guidance on comprehensive adolescent health assessments, the WHO Mental Health Action Plan, and WHO strategies targeting NCD risk behaviours, by demonstrating how YouthCHAT's multi-issue, integrated screening approach operationalises these principles in primary care to support holistic, coordinated adolescent care. Using a PROGRESS-Plus equity lens, the transferability of this intervention to underserved populations may be influenced by factors such as place of residence, ethnicity, socioeconomic status, and gender diversity, all of which can shape access, relevance, and engagement. Ensuring equitable implementation will require culturally responsive adaptation and attention to structural barriers that differentially affect these groups. Conclusion Globally, youth from low-income families, those with lower educational attainment, who are unemployed, from rural backgrounds, indigenous or from ethnic minorities, are more likely to engage in unhealthy behaviours and suffer mental health issues [32], [39], [40]. Our findings of multiple psychosocial issues in under-served adolescents in NZ are therefore relevant and potentially transferable and generalisablet to underserved adolescents in comparable settings internationally, but generalisability beyond similar contexts should be interpreted with caution. Declaration of generative AI and AI-assisted technologies in the manuscript preparation process Nothing to disclose. CRediT authorship contribution statement FanZhen Zhou: Writing – review & editing, Formal analysis. James Warren: Writing – review & editing, Supervision, Formal analysis, Data curation, Conceptualization. Felicity Goodyear-Smith: Writing – review & editing, Writing – original draft, Methodology, Conceptualization. Funding This research did not receive any specific funding. Declaration of competing interest Felicity Goodyear-Smith reports a relationship with Kekeno that includes: equity or stocks. Felicity Goodyear-Smith and Jim Warren are named Contributors (Inventors) of YouthCHAT for which the company Kekeno has a licence to operate. They are not involved in the operation of YouthCHAT nor collection of the data. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Footnotes Supplementary data to this article can be found online at https://doi.org/10.1016/j.dialog.2026.100310. Supplementary data Supplementary material YouthCHAT questions & further analyses. Data availability The data that support the findings of this study are available on request from co- author, JW. The data are not publicly available due to their containing information that could compromise the privacy of research participants. References Associated Data Supplementary Materials Supplementary material YouthCHAT questions & further analyses. Data Availability Statement The data that support the findings of this study are available on request from co- author, JW. The data are not publicly available due to their containing information that could compromise the privacy of research participants.