Latent Growth Mixture Modeling Research Articles

Discriminating Heterogeneous Trajectories of Resilience and Depression After Major Life Stressors Using Polygenic Scores

Major life stressors, such as loss and trauma, increase the risk of depression. It is known that individuals show heterogeneous trajectories of depressive symptoms following major life stressors, including chronic depression, recovery, and resilience. Although common genetic variation has been associated with depression risk, genomic factors that could help discriminate trajectories of risk vs resilience following adversity have not been identified. To assess the discriminatory accuracy of a deep neural net combining joint information from 21 psychiatric and health-related multiple polygenic scores (PGSs) for discriminating resilience vs other longitudinal symptom trajectories with use of longitudinal, genetically informed data on adults exposed to major life stressors. The Health and Retirement Study is a longitudinal panel cohort study in US citizens older than 50 years, with data being collected once every 2 years between 1992 and 2010. A total of 2071 participants who were of European ancestry with available depressive symptom trajectory information after experiencing an index depressogenic major life stressor were included. Latent growth mixture modeling identified heterogeneous trajectories of depressive symptoms before and after major life stressors, including stable low symptoms (ie, resilience), as well as improving, emergent, and preexisting/chronic symptom patterns. Twenty-one PGSs were examined as factors distinctively associated with these heterogeneous trajectories. Local interpretable model-agnostic explanations were applied to examine PGSs associated with each trajectory. Data were analyzed using the DNN model from June to July 2020. Development of depression and resilience were examined in older adults after a major life stressor, such as bereavement, divorce, and job loss, or major health events, such as myocardial infarction and cancer. Discriminatory accuracy of a deep neural net model trained for the multinomial classification of 4 distinct trajectories of depressive symptoms (Center for Epidemiologic Studies-Depression scale) based on 21 PGSs using supervised machine learning. Of the 2071 participants, 1329 were women (64.2%); mean (SD) age was 55.96 (8.52) years. Of these, 1638 (79.1%) were classified as resilient, 160 (7.75) in recovery (improving), 159 (7.7%) with emerging depression, and 114 (5.5%) with preexisting/chronic depression symptoms. Deep neural nets distinguished these 4 trajectories with high discriminatory accuracy (multiclass micro-average area under the curve, 0.88; 95% CI, 0.87-0.89; multiclass macro-average area under the curve, 0.86; 95% CI, 0.85-0.87). Discriminatory accuracy was highest for preexisting/chronic depression (AUC 0.93), followed by emerging depression (AUC 0.88), recovery (AUC 0.87), resilience (AUC 0.75). The results of the longitudinal cohort study suggest that multivariate PGS profiles provide information to accurately distinguish between heterogeneous stress-related risk and resilience phenotypes.

Resilience Predicts the Trajectories of College Students' Daily Emotions During COVID-19: A Latent Growth Mixture Model.

The objective of this study was to examine the association between resilience and trajectories of college students’ negative and positive affect during the COVID-19 pandemic. A total of 391 college students recruited from China completed a daily online negative and positive affect scale for 1 week, and their resilience was also measured. Profiles of brief trajectories of negative and positive affect over time were identified using the latent growth mixture model, and the effect of resilience on these trajectories was further explored. Two latent profiles of negative affect were found: a constant high negative affect profile and a slowly decreasing low negative affect profile, while three latent profiles of positive affect were identified: a slowly increasing high positive affect profile, a rapidly decreasing medium positive affect profile, and a constant medium positive affect profile. The optimism dimension of resilience predicted the membership in the various profiles significantly, whereas the prediction of tenacity and strength dimensions of resilience was not significant. Activities that promote resilience, especially optimism, should be included to improve the daily emotions of college students during COVID-19.

Trajectories of depression and anxiety in Chinese high school freshmen: Associations with Internet addiction

BackgroundMost previous studies focused on the associations of depression and anxiety with Internet addiction (IA) have used variable-centered approaches. This study aims to explore the distinct developmental trajectories of depression and anxiety, and their association with IA based on person-centered approaches. MethodsA total of 437 Chinese high school freshmen at the baseline were followed across one year. Latent class growth analysis (LCGA) and growth mixture modeling (GMM) were used to identify the heterogeneity of individual trajectories of depression and anxiety. ResultsFor depression, there were three distinct trajectories identified, namely, the escalating group (n=60, 13.7%), recovery group (n=31, 7.1%), and the low-stable group (n=346, 79.2%). For anxiety, the subgroups consisted of the escalating group (n=50, 11.4%), the recovery group (n=34, 7.8%), and the low-stable group (n=353, 80.8%). The probability of IA in the depression low-stable group was significantly smaller compared to those in either the recovery group (χ2=10.794, P=0.001) or the escalating group (χ2=19.314, P<0.001). The probability of IA in the anxiety low-stable group was found to be significantly smaller than that in the recovery group (χ2=17.359, P<0.001) and the escalating group (χ2=7.752, P=0.005). LimitationsThe sample was limited to students from one specific high school and had a one-year follow-up. ConclusionsThe findings of the study suggest the necessity of early prevention and intervention strategies for those students with depression and anxiety, which are at elevated risk of developing IA.

Sex specific trajectories of central adiposity, lipid indices, and glucose level with incident hypertension: 12 years Follow-up in Tehran lipid and glucose study

AimsTo identify sex specific trajectories of waist circumference (WC),triglyceride (TG), total cholesterol (TC), high density lipoprotein cholesterol (HDL-C) and fasting plasma glucose (FPG) during adulthood and examine their associations with incident hypertension.MethodsThe cohort consisted of 5030 participants (2051 males) with at least 2 repeated measurement during a median of 12 years follow up. We identified trajectory groups using latent class growth mixture model, their association with hypertension was examined using multivariate Cox-regression analysis.ResultsWe found 997 cases of hypertension (483 male). For both exposures, three distinct trajectory groups were identified in both genders. For WC, in women: low-increasing, 82.4%; high-stable, 13.4%; high-increasing, 4.2% and in men: stable, 94.6%; low-increasing, 3.6% and for high- increasing, 1.7%. For TG, in women: stable, 91.3%; decreasing, 5.9%; inverse U-shape, 2.8%; in men: stable, 89.7%; inverse U- shape, 6.2% and for decreasing, 4.1%.Regarding WC, high stable and high-increasing trajectories were associated with hypertension in the multivariate model [(hazard ratio (HR) = 1.66 (95% CI 1.26–2.20) and 2.78(1.79–3.60), respectively]. Among men, this association was shown only for the low-increasing trajectory [2.76: 1.49–5.10]. For TG, among women decreasing and inverse U-shape trajectories were significantly associated with hypertension in the multivariate model [1.32:1.01–1.76] and [2.23:1.58–3.23, respectively].We did not find any impact of increasing trajectories of FPG and HDL-C on incident hypertension. Considering TC, all individuals followed a stable trajectory.ConclusionWC dynamic changes in both gender and TG trajectory among women were significantly associated with incident hypertension.

Gestational systolic blood pressure trajectories and risk of adverse maternal and perinatal outcomes in Chinese women

BackgroundAssociations between trajectories of systolic blood pressure (SBP) during pregnancy and pregnant outcomes remain unclear and disparate.MethodsData of 20,353 mothers without chronic hypertension and who delivered live singletons between January, 2014 and November, 2019, was extracted from Taicang register-based cohort. Based on SBP measured during 10 to 40 weeks of gestation, SBP trajectories were explored using latent class growth mixture model, and their associations with maternal and neonatal outcomes were assessed by logistic regression analyses.ResultsSix heterogeneous SBP trajectories were identified: low delayed-increasing (7.47%), low reverse-increasing (21.88%), low-stable (19.13%), medium-stable (21.64%), medium reverse-increasing (16.47%), and high stable (13.41%) trajectories. The high-stable trajectory had SBP around 125 mmHg in the 10th gestational week, and increased slightly onwards. When compared with the low-stable trajectory, the high-stable trajectory had maximally adjusted odds ratio (95% confidence interval) of 5.28 (2.76–10.10), 1.30 (1.13–1.50), 1.53 (1.12–2.08), 1.32 (1.06–1.65) and 1.64 (1.08–2.48) for gestational hypertension (GH), early-term delivery (ETD), preterm delivery (PTD), small for gestational age and low birth weight (LBW), respectively. Besides, the medium reverse-increasing trajectory showed significantly increased risk of GH and ETD, while the medium-stable trajectory had significantly elevated risk of ETD and PTD. Notably, SBP trajectories slightly but significantly improved risk discrimination of GH, ETD and LBW, over traditional risk factors.ConclusionWomen with different SBP trajectories were at varied risk of adverse maternal and fetal outcomes. Meanwhile, our study suggested that BP monitoring during pregnancy is necessary, especially for women with high SBP in early pregnancy or upward trajectory.

Child-to-adult body mass index trajectories and the risk of subclinical renal damage in middle age.

Although it is well established that obesity is a risk factor for chronic kidney disease, the impact of distinct long-term body mass index (BMI) developmental patterns on renal function in later life is poorly understood. This study utilized data derived from the Hanzhong Adolescent Hypertension Cohort, a prospective cohort followed over 30 years. We used latent class growth mixture modeling method to identify the BMI trajectories of participants who had received BMI measurements at least three times from childhood (age: 6-15 years) to adulthood (age: 36-45 years). The modified Poisson regression model was used to identify potential associations between BMI trajectories and subclinical renal damage (SRD) in midlife. Within a total of 2162 individuals, we identified four distinct long-term BMI trajectories: stable normal (54.72%), moderately increasing overweight (32.42%), resolving (10.27%), and progressively increasing obese (2.59%). By the latest follow-up in 2017, a total of 257 (13.1%) individuals were diagnosed with SRD. Compared with the stable normal group, the moderately increasing overweight group and the progressively increasing obese group exhibited significantly a higher urinary albumin-to-creatinine ratio and a higher odd of existing SRD in 2017 (risk ratio [RR], 1.70 [95% confidence interval (CI), 1.33-2.19] and 4.35 [95% CI, 3.00-6.30], respectively). However, individuals who resolved their elevated BMI in early life had a similar risk for SRD as those who had never been obese or overweight (RR, 1.17 [95% CI, 0.77-1.79]). Child-to-adult BMI trajectories that worsen or persist at high levels were associated with an increased risk for SRD in midlife. Maintaining a normal BMI or reversing an elevated BMI in early life may be beneficial to renal function over the long term.

Understanding Variation in Longitudinal Data Using Latent Growth Mixture Modeling.

This article guides researchers through the process of specifying, troubleshooting, evaluating, and interpreting latent growth mixture models. Latent growth mixture models are conducted with small example dataset of N = 117 pediatric patients using Mplus software. The example and data show how to select a solution, here a 3-class solution. We also present information on two methods for incorporating covariates into these models. Many studies in pediatric psychology seek to understand how an outcome changes over time. Mixed models or latent growth models estimate a single average trajectory estimate and an overall estimate of the individual variability, but this may mask other patterns of change shared by some participants. Unexplored variation in longitudinal data means that researchers can miss critical information about the trajectories of subgroups of individuals that could have important clinical implications about how one assess, treats, and manages subsets of individuals. Latent growth mixture modeling is a method for uncovering subgroups (or "classes") of individuals with shared trajectories that differ from the average trajectory.

Trajectories of depression in psychotherapy: How client characteristics predict clinical improvement.

The current study aims to ascertain the trajectories of psychotherapy clients' symptom change and identify client factors that predict treatment outcome. We conducted a latent growth mixture model (LGMM) to identify the change trajectories of 44 clients' depression scores during psychotherapy. Client characteristics were then explored to determine whether any were associated with change trajectories. We examined whether the number of physician visits and/or client self-concealment scores predict 63 clients' improvement after controlling for initial symptom severity. Two trajectories of clients' symptom change were identified: nonimprovers (52.3%) and improvers (47.7%). Nonimprovers had higher levels of self-concealment and baseline depression than improvers. The number of physician visits was associated with higher depression scores at baseline and greater clinical improvement during psychotherapy. Clients showed distinct trajectories of symptom change in psychotherapy. Early identification of clients at risk for treatment failure may increase the probability of therapeutic success.

Housing first, connection second: the impact of professional helping relationships on the trajectories of housing stability for people facing severe and multiple disadvantage

BackgroundDespite the accumulating evidence on the role of professional helping relationships for highly disadvantaged populations, methodological shortcomings have made it difficult to establish a robust relationships-outcomes link. This study sought to establish the impact of professional helping relationships on the trajectories over 24 months of housing stability for 2141 people facing severe and multiple disadvantage using data from the Housing First controlled trial in Canada.MethodThe study used a mixed method design. Latent growth curve and growth mixture models assessed the impact of working alliance across the sample as a whole and within subgroups with different patterns of housing stability. Thematic analysis explored the factors that may affect the quality of working alliances within different subgroups.ResultsThree distinct trajectories of housing stability emerged (i.e., Class 1: “sharp rise, sustained, and decline housing”; Class 2: “hardly any time housed”; Class 3: “high rise, sustained, and decline housing”) with professional helping relationships having different effects in each. The analysis revealed structural and individual circumstances that may explain differences among the classes.ConclusionsThe findings underscore the role of professional helping relationships, as distinct from services, in major interventions for highly disadvantaged populations, and draws new attention to the temporal patterns of responses to both the quality of relationship and targeted interventions.

Profiles of competence development in upper secondary education and their predictors.

This registered report protocol elaborates on the theory, methods, and material of a study to identify latent profiles of competence development in reading and mathematics among German students in upper secondary education. It is expected that generalized (reading and mathematical competence develop similarly) and specialized (one of the domains develops faster) competence profiles will be identified. Moreover, it is hypothesized that students' domain-specific interest and educational history will predict membership of these latent profiles as these factors influence the students' learning environments. For this study, we will use data from the German National Educational Panel Study, including students from ninth grade in secondary schools (expected N = 14,500). These students were tracked across six years and provided competence assessments on three occasions. The latent profiles based on the students' reading and mathematical competences will be identified using latent growth mixture modeling. If different types of profiles can be identified, multinomial regression will be utilized to analyze whether the likelihood of belonging to a certain competence development profile is influenced by students' domain-specific interest or educational history. As this protocol is submitted before any analyses were conducted, it will provide neither results nor conclusions.

Big five personality influences trajectories of information seeking behavior

This article reports results of a study observing how big-five personality traits influence the trajectories of information-seeking adaptive behaviors among university freshmen. Data are collected from 409 freshmen at a Chinese university at 3, 5, 7, and 9 months after university enrollment. A latent growth mixture model is applied to reveal four trajectories of information-seeking behaviors: high or low maintaining, downward or upward. When the information-seeking trajectories are related to personality traits, openness and agreeableness are associated with high maintaining, while agreeableness is associated with downward trajectories. The study provides strong empirical evidence supporting the Minnesota theory of work adjustment and provides important insights to practitioners who want to enhance newcomer adjustment at all organizations.

Patterns of change in cardiovascular risk assessments and ankle brachial index among Puerto Rican adults.

BackgroundPuerto Rican adults have higher odds of peripheral artery disease (PAD) compared with Mexican Americans. Limited studies have examined relationships between clinical risk assessment scores and ABI measures in this population.MethodsUsing 2004–2015 data from the Boston Puerto Rican Health Study (BPRHS) (n = 370–583), cross-sectional, 5-y change, and patterns of change in Framingham Risk Score (FRS) and allostatic load (AL) with ankle brachial index (ABI) at 5-y follow-up were assessed among Puerto Rican adults (45–75 y). FRS and AL were calculated at baseline, 2-y and 5-y follow-up. Multivariable linear regression models were used to examine cross-sectional and 5-y changes in FRS and AL with ABI at 5-y. Latent growth mixture modeling identified trajectories of FRS and AL over 5-y, and multivariable linear regression models were used to test associations between trajectory groups at 5-y.ResultsGreater FRS at 5-y and increases in FRS from baseline were associated with lower ABI at 5-y (β = -0.149, P = 0.010; β = -0.171, P = 0.038, respectively). AL was not associated with ABI in cross-sectional or change analyses. Participants in low-ascending (vs. no change) FRS trajectory, and participants in moderate-ascending (vs. low-ascending) AL trajectory, had lower 5-y ABI (β = -0.025, P = 0.044; β = -0.016, P = 0.023, respectively).ConclusionsFRS was a better overall predictor of ABI, compared with AL. Puerto Rican adults, an understudied population with higher FRS over 5 years, may benefit from intensive risk factor modification to reduce risk of PAD. Additional research examining relationships between FRS and AL and development of PAD is warranted.

Latent profile transition analyses and growth mixture models: A very non-technical guide for researchers in child and adolescent development.

Developmental scientists are often interested in subgroups of people who share commonalities in aspects of development; these subgroups often cannot be captured directly but instead must be inferred from other information. Mixture models can be used in these situations. Two specific types of mixture models, latent profile transition analyses and growth mixture models, are highly relevant to developmental science because they can identify subgroups of people who are similar in their patterns of change. This guide highlights foundational aspects of these two types of models and is intended for readers who have not previously conducted either an LPTA or a GMM, or perhaps no mixture model analyses at all. It includes four primary sections. The first focuses on understanding mixture models conceptually and applying that knowledge to identifying appropriate research questions. The second section addresses data requirements, including planning for data collection or evaluating the suitability of previously collected data, and data preparation. The third section focuses on conducting analyses, with step-by-step instructions and syntax, and the final section discusses presenting the results. I illustrate these concepts and procedures with an example data set and research questions derived from the Five Cs model of positive youth development.

The Role of Peer Support in the Growth Trajectory of Pathological Internet Use Among Youth: A Protective Factor.

Despite the escalating public concern about pathological use of the Internet among youth, little research has investigated the longitudinal trajectory of the problem and the role of peer support in it. This study aimed to determine the growth trajectory of pathological Internet use (PIU), identify latent subgroups presenting heterogeneous growth trajectories of PIU, and examine the initial status of and the rate of change in peer support as predictors of the growth trajectory of PIU among children and adolescents. Using four-wave longitudinal data from 3,079 Singaporean youth at ages 7-19, we performed Latent Growth Curve Modeling, Latent Class Growth Analysis and Latent Growth Mixture Modeling. Results revealed a downward trajectory of PIU among youth. We identified six latent subgroups, three of which were in pathological status for at least a year. The more rapid improvement in peer support, but not the initial level of peer support, predicted the faster reduction of PIU symptoms during the 3-year period. Findings suggest that although the symptoms of PIU among children and adolescents would decline over time, improving peer support could further alleviate youth PIU. We recommended that intervention programs include a peer support component to mitigate PIU among youth.

Latent Class Trajectory and Growth Mixture Models in the Study of Physical Resilience

After a stressor, individuals may experience different trajectories of function and recovery. One potential explanation for this variation is differing trajectories may be indicators of differing classes or levels of resilience to the stressor. Latent Class Trajectory (LCTA) and Growth Mixture models (GMM) are two similar approaches used to discover the number and types of trajectories in a study population. Class membership may determine the shape and level of recovery, which may be predicted by individual characteristics. In this talk, we present some insights to using these models to successfully identify the number of classes of trajectories, membership of trajectory classes, and the functional form of the trajectory. We will identify methods for deciding class enumeration, indices for assessing fit quality, and, importantly, the importance of proper model specification. Real life and simulated examples will be shown to compare and contrast differences between GMM and LCTA results. Part of a symposium sponsored by Epidemiology of Aging Interest Group.

Gender-Specific Determinants of Instrumental Activities and Associations With Cognitive Decline in TILDA

Detecting functional decline is critical in recognizing clinical progression into mild cognitive impairment and dementia as people age. This study aimed to investigate differences in Instrumental Activities of Daily Living (IADL) and associations with cognition by gender. Data are from five waves of TILDA, a nationally representative cohort of 8,175 aged ≥50 years, measured every 2 years between 2009 and 2018. IADL assessed were preparing a hot meal, doing household chores, shopping for groceries, making telephone calls, taking medications, and managing money. Cognition was assessed using the Mini-Mental-State Examination (MMSE). Latent class analysis and growth mixture modelling techniques were used to identify IADL and cognitive latent classes and examine change trajectories. IADL trajectories differed by gender and were associated with cognitive decline. Functional decline discrepancies between men and women may be driven by domestic tasks within IADL which are associated with social gender roles, rather than determining differential cognitive trajectories.

The prevalence rate of sexual violence worldwide: a trend analysis

BackgroundThe purpose of the present study is to showcase the image of Sexual Violence (SV) temporal trends through exploring differences in its prevalence rates during 1990–2017 across 195 countries and territories.MethodsThe SV prevalence rates were derived from the Global Burden of Disease (GBD) database during 1990–2017, worldwide. First, the Latent Growth Model (LGM) was employed for assessing the change in SV prevalence rate over time in Asia, Africa, Europe, North America, South America, Australia & Oceania, for men and women separately. Then, the change in SV prevalence rate over time was determined within countries with high and low Human Development Index (HDI). Finally, the Latent Growth Mixture Models (LGMM) were applied for identifying classes where countries within each class have similar trend of SV prevalence rate over time.ResultsThe SV prevalence was higher among women than men and decreased in both genders over time across the world. The declining trend in SV prevalence against men is visible in both countries with high and low HDI, but SV prevalence against women in countries with low HDI shows an increase. The findings of LGMM identified six classes of SV prevalence trajectories. LGMM allocated Bermuda into the class with the highest decrease in SV prevalence against men, and Equatorial Guinea and Luxembourg into the class with the highest increase. Other countries had very slow declining trends. In terms of SV prevalence against women, LGMM allocated China, North Korea, and Taiwan into the class with the most increase among the countries in the world. Bermuda, Guyana, Mexico, Nigeria, and Saint Lucia were placed into the class which witnessed the largest decline and Angola, Congo, and Equatorial Guinea were ranked next. The trend in other countries was mostly decreasing.ConclusionGiven the high economic and social burden that SV has on victims and societies, the rate of SV in most countries does not seem to have dropped remarkably and requires special attention by relevant policymakers. The SV prevalence rate is highly heterogeneous among world countries which may be due to the definitions and tools used, and more importantly, the culture norms.

Trajectories of depressive symptoms and associated patterns of cognitive decline

The aim was to investigate the pattern and rate of cognitive decline across distinctive trajectories of depressive symptoms in older adults. In this prospective multinational cohort study on 69,066 participants (on average 64 years at baseline, 55% women), assessments of cognitive functions (immediate recall, delayed recall, verbal fluency) and depressive symptoms (EURO-D scale) were conducted at 2-year intervals. The trajectories of depressive symptoms were obtained using latent growth mixture modelling, cognitive decline was assessed using smoothing splines and linear mixed effects models. Four distinct trajectories of depressive symptoms were identified: constantly low (n = 49,660), constantly high (n = 2999), increasing (n = 6828) and decreasing (n = 9579). Individuals with increasing and constantly high depressive symptoms showed linear cognitive decline, while those with constantly low and decreasing depressive symptoms had fluctuating cognition. Participants with increasing depressive symptoms had the fastest decline, while those with decreasing symptoms were spared from decline in cognition. This study suggests that the pattern as well as the rate of cognitive decline co-occurs with specific patterns of changes in depressive symptoms over time. The most pronounced cognitive decline is present in individuals, in whom depressive symptoms increase late in life. Unique mechanisms of cognitive decline may exist for subgroups of the population, and are associated with the trajectory of depressive symptoms.

Pain catastrophizing as a treatment process variable in cognitive behavioural therapy for adults with chronic pain.

Interdisciplinary cognitive behavioural therapy (CBT) for chronic pain is effective at improving function, mood and pain interference among individuals with disabling chronic pain. Traditionally, CBT assumes that cognitive change is an active therapeutic ingredient in the determination of treatment outcome. Pain catastrophizing, a cognitive response style that views the experience of pain as uncontrollable, permanent and destructive, has been identified as an important maladaptive cognition which contributes to difficulties with the management of chronic pain. Consequently, pain catastrophizing is commonly targeted in CBT for chronic pain. To examine change trajectories in pain catastrophizing during treatment and assess the relevance of these trajectories to outcomes at posttreatment. Participants included individuals with chronic pain (N=463) who completed a 3-week program of interdisciplinary CBT. Pain catastrophizing was assessed weekly over the 3weeks of treatment and latent growth curve modelling was used to identify trajectories of change. Findings indicated the presence of two classes of linear change, one with a significant negative slope in pain catastrophizing (i.e. improved class) and the other with a non-significant slope (i.e. unchanged class). Next, latent growth mixture modelling examined treatment outcome in relation to class membership. These results indicated that individuals in the 'improved' PCS class had significantly greater improvement in pain interference and mood, as well as physical and mental quality of life compared to the 'unchanged' class. Implications for our findings, in relation to the CBT model, are discussed.

Trajectories of functioning in a population-based sample of veterans: contributions of moral injury, PTSD, and depression.

Although research has shown that exposure to potentially traumatic and morally injurious events is associated with psychological symptoms among veterans, knowledge regarding functioning impacts remains limited. A population-based sample of post-9/11 veterans completed measures of intimate relationship, health, and work functioning at approximately 9, 15, 21, and 27 months after leaving service. Moral injury, posttraumatic stress, and depression were assessed at ~9 months post-separation. We used Latent Growth Mixture Models to identify discrete classes characterized by unique trajectories of change in functioning over time and to examine predictors of class membership. Veterans were assigned to one of four functioning trajectories: high and stable, high and decreasing, moderate and increasing, and moderate and stable. Whereas posttraumatic stress, depression, and moral injury associated with perpetration and betrayal predicted worse outcomes at baseline across multiple functioning domains, moral injury associated with perpetration and depression most reliably predicted assignment to trajectories characterized by relatively poor or declining functioning. Moral injury contributes to functional problems beyond what is explained by posttraumatic stress and depression, and moral injury due to perpetration and depression most reliably predicted assignment to trajectories characterized by functional impairment over time.