Latent Growth Mixture Modeling Research Articles

Retirement as a predictor of physical functioning trajectories among older businessmen

BackgroundAssociations between retirement characteristics and consequent physical functioning (PF) are poorly understood, particularly in higher socioeconomic groups, where postponing retirement has had both positive and negative implications for PF.MethodsMultiple assessments of PF, the first of which at the mean age of 73.3 years, were performed on 1709 men who were retired business executives and managers, using the RAND-36/SF-36 instrument, between 2000 and 2010. Questionnaire data on retirement age and type of pension was gathered in 2000. Five distinct PF trajectories were created using latent growth mixture modelling. Mortality- and covariate-adjusted multinomial regression models were used to estimate multinomial Odds Ratios (mOR) on the association between retirement characteristics and PF trajectories.ResultsA one-year increase in retirement age was associated with decreased likelihood of being classified in the ‘consistently low’ (fully adjusted mOR = 0.82; 95%CI = 0.70, 0.97; P = 0.007), ‘intermediate and declining’ (mOR = 0.89; 95%CI = 0.83, 0.96; P = 0.002), and ‘high and declining’ (mOR = 0.92; 95%CI = 0.87, 0.98; P = 0.006) trajectories, relative to the ‘intact’ PF trajectory. Compared to old age pensioners, disability pensioners were more likely to be classified in the ‘consistently low’ (mOR = 23.77; 95% CI 2.13, 265.04; P = 0.010), ‘intermediate and declining’ (mOR = 8.24; 95%CI = 2.58, 26.35; P < 0.001), and ‘high and declining’ (mOR = 2.71; 95%CI = 1.17, 6.28; P = 0.020) PF trajectories, relative to the ‘intact’ PF trajectory.ConclusionsAmong executives and managers, older age at retirement was associated with better trajectories of PF in old age. Compared to old age pensioners, those transitioning into disability and early old age pensions were at risk of having consistently lower PF in old age.

Self-critical perfectionism, dependency and entropy during cognitive behavioural therapy for depression.

This study examined whether 'personality vulnerability' (i.e., self-critical perfectionism or dependency) predicts the trajectory of change, as well as variability and instability (i.e., entropy) of symptoms, during cognitive behaviour therapy (CBT) for depression. Study participants were outpatients (N=312) experiencing a primary mood disorder. Participants received CBT for depression group sessions over 15weeks. Self-report measures of self-critical perfectionism, dependency, and depression were collected longitudinally. A latent growth mixture modelling (LGMM) statistical approach was used to evaluate the presence of latent classes of individuals based on their longitudinal pattern of symptom change during CBT and to evaluate whether baseline self-critical perfectionism or dependency predicts class membership. A Latent Acceleration Score (LAS) model evaluated whether perfectionism or dependency led to variability in depression symptom change (e.g., velocity) by considering changes in velocity (e.g., acceleration and/or deceleration). LGMM indicated the presence of two latent classes that represent symptom improvement (N=239) or minimal symptom improvement over time (N=73). Elevated baseline self-critical perfectionism, but not dependency, predicted a greater likelihood of membership in the class of participants who demonstrated minimal symptom improvement over time. The second analysis examined whether baseline self-critical perfectionism also predicts depression symptom variability and instability. The LAS perfectionism model demonstrated that perfectionism accelerates depression symptom change during the first seven sessions of treatment, then has a decelerating effect on depression symptom change. Results indicated that higher baseline self-critical perfectionism predicted higher variability and instability in depression symptoms and variability in acceleration and deceleration, over the course of treatment.

Patients Follow 3 Different Rate-of-Recovery Patterns After Anterior Cruciate Ligament Reconstruction Based on International Knee Documentation Committee Score

To determine whether subgroups of patients exist based on the rate-of-recovery pattern of International Knee Documentation Committee (IKDC) scores after anterior cruciate ligament reconstruction (ACLR) and to determine clinical predictors for these subgroups. Patients who underwent primary or revision ACLR at a single institution from January 2014 to January 2019 were identified. Latent class growth analyses and growth mixture models (GMMs) with 1 to 6 classes were used to identify subgroups of patients based on functional rate-of-recovery patterns by use of preoperative, 1-year postoperative, and 2-year postoperative IKDC scores. A total of 245 patients who underwent ACLR were included in the analysis. A 3-class GMM was chosen as the final model after 6 different models were run. Class 1, showing improvement from preoperatively to 1-year follow-up, with sustained improvement from 1 to 2 years postoperatively, constituted 77.1% of the study population (n= 189), whereas class 2, showing functional improvement between 1- and 2-year follow-up, was the smallest class, constituting 10.2% of the study population (n= 25), and class 3, showing slight improvement at 1-year follow-up, with a subsequent decline in IKDC scores between 1- and 2-year follow-up, constituted 12.7% of the study population (n= 31). Revision surgery (P= .005), a psychiatric history (P= .025), preoperative chronic knee pain (P= .024), and a subsequent knee injury within the follow-up period (P= .011) were the predictors of class 2 and class 3 rate-of-recovery patterns. Patient demographic characteristics, graft type, and concomitant ligament, meniscus, or cartilage injury at the time of surgery were not associated with the different recovery patterns described in this study. Patients may follow different rate-of-recovery patterns after ACLR. By use of the GMMs, 3 different rate-of-recovery patterns based on IKDC scores were identified. Although most patients follow a more ideal rate-of-recovery pattern, fewer patients may follow less favorable patterns. Revision surgery, a history of psychiatric illness, preoperative chronic knee pain, and a subsequent knee injury within the follow-up period were predictive of less favorable rate-of-recovery patterns. Level III, retrospective cohort study.

Accessibility to Digital Mental Health Services among the General Public throughout COVID-19: Trajectories, Influencing Factors and Association with Long-Term Mental Health Symptoms.

Digital mental health services (DMHSs) have great potential for mitigating the mental health burden related to COVID-19, but public accessibility (ease of acquiring services when needed) to DMHSs during the pandemic is largely unknown. Accessibility to DMHSs was tracked longitudinally among a nationwide sample of 18,804 adults in China from before to one year after COVID-19 outbreak. Unconditional and conditional latent growth curve models and latent growth mixture models were fitted to explore the overall growth trend, influencing factors, and latent trajectory classes of accessibility to DMHSs throughout COVID-19. Generalized estimating equation models and generalized linear mixed models were employed to explore the association between accessibility to DMHSs and long-term mental health symptoms. We found that people generally reported increased difficulty in accessing DMHSs from before to one year after COVID-19 outbreak. Males, youngsters, individuals with low socioeconomic status, and individuals greatly affected by COVID-19 reported greater difficulty in accessing DMHSs. Four DMHS accessibility trajectory classes were identified: “lowest–great increase” (6.3%), “moderate low–slight increase” (44.4%), “moderate high–slight decrease” (18.1%) and “highest–great decrease” (31.2%). Trajectory classes reporting greater difficulty in accessing DMHSs were at higher risk for long-term mental symptoms. In conclusion, an overall increase in difficulty in accessing DMHSs is observed throughout COVID-19, and heterogeneity exists in DMHS accessibility trajectories. Our results suggest that easy access to DMHSs should be consistently facilitated. Moreover, access gaps should be reduced across demographic groups, and target populations for service allocation should alter as the pandemic evolves.

Trajectories of Response to PROTECT, a community-based treatment for elder abuse victims with depression

<h3>Introduction</h3> Elder abuse afflicts up to 10% of adults 60 years and older, with approximately one third of elder abuse (EA) victims presenting clinically significant depressive symptoms. When offered EA resolution services, victims underutilize them. In collaboration with community partners, we developed PROTECT, <i>Providing Options to Elderly Clients Together</i>, a brief behavioral psychotherapy for late-life depression among elder abuse victims. PROTECT aims to reduce depression and stress and increase sense of agency and capacity to utilize EA services (Sirey et al. 2021). In a recent expand of PROTECT delivery, we implemented the intervention in collaboration with five New York City elder abuse agencies over three years. We aimed to: a) examine the trajectories of response to PROTECT in a large community sample; b) identify predictors of treatment response in this sample. <h3>Methods</h3> A total of 179 elder abuse victims received PROTECT between 2018 and 2021. PHQ9 scores were collected at every visit during the weekly 9- week treatment. Latent Growth Mixture Modelling (LGMM) was used to derive separate latent trajectories of PHQ9 scores over time. LGMMs were constructed for two trajectories for both linear and quadratic time. The final model was chosen based on a combination of smallest Bayesian information criterion (BIC) and clinical plausibility. Participant characteristics were compared between classes through Chi-squared/Fisher's Exact tests or independent two-sample t-tests/Wilcoxon rank-sum tests. Linear mixed effects models were then used to estimate changes of PHQ9 over 10 weeks for each class. We stratified the data based on group class to obtain a separate coefficient for every trajectory. <h3>Results</h3> The sample comprised of 179 patients, with the majority of patients undergoing current abuse (70%). The mean PHQ9 of the total sample at baseline was 11.7 (std=5.1). PROTECT resulted in an average reduction in PHQ9 of 4.9 (std=4.7) and 53% of patients achieved a clinically relevant reduction in depression severity (PHQ9 reduction of 5 points or more). There was no association between demographic characteristics or type and number of abuse and treatment response. A two-group linear trajectory model was chosen based on minimal BIC and clinical plausibility to examine the relative response based on initial depression severity. Group-based trajectories differed by baseline severity of PHQ9 scores (mean PHQ9 of 9.6 and 17 for each group, respectively). The trajectories showed similar improvement from treatment for each group, suggesting that treatment is effective for all patients regardless of their baseline PHQ scores (mean improvement in PHQ9 was 9.6 > 5.1 and 17 > 11.7 for high baseline and low baseline severity groups, respectively). <h3>Conclusions</h3> PROTECT is effective in reducing depression severity among elder abuse victims. Importantly, PROTECT showed to be effective regardless of demographic differences, type and number of abuse or depression severity at baseline. These results are encouraging, as PROTECT is designed as a community-based intervention for elder abuse victims of diverse backgrounds and abuse histories. <h3>This research was funded by</h3> This project is funded by a contract from the NYC Department for the Aging (#20191406829, PI: Sirey). Dr. Solomonov is supported by a grant from the National Institute of Mental Health K23 MH123864

Long-Term Impact of Grandchild Caregiving Trajectories on Depression in Middle-Aged and Older Chinese People: A Longitudinal Study.

There has been little research investigating the effects of caregiving for grandchildren on grandparents' mental health from a dynamic perspective. The aim of this study was to evaluate the effects on caregivers' depression of changes in grandparenting intensity. The study population included 8,157 respondents obtained from the China Health and Retirement Longitudinal Study (CHARLS). Latent growth mixture modeling was used to group respondents into five classes of trajectory of caregiving intensity as follows: "sharply decreasing", "never or rarely", "slowly decreasing", "increasing", and "continuously high". A generalized additive mixed model (GAMM) and a marginal structural model (MSM) both associated the "continuously high" and "sharply decreasing" intensities with depression. "Continuously high" intensity significantly increased the risk of depression in the male group only. Further research should be conducted to analyze the deep-seated mechanisms of association between grandparenting and mental health, in different cultural contexts and among subgroups with different characteristics.

Abstract P120: Latent Group Trajectory Methods To Study Type 2 Diabetes Epidemiology: A Review Of The Literature

Introduction: The progression of type 2 diabetes (T2D) is unique to each patient and can be depicted through individual temporal trajectories. Latent group trajectory methods (latent growth mixture models [LGMM] or latent class growth analysis [LCGA]) can be used to classify similar individual trajectories in a priori non-observed groups (latent groups), sharing common characteristics. Although increasingly used in the field of T2D epidemiology, many questions remain regarding the utilization of these methods. Objective: To review the literature of longitudinal studies using latent group trajectory methods among individuals with T2D. Methods: MEDLINE (Ovid), EMBASE, CINAHL and Web of Science were searched through August 25 th , 2021. Data were collected on the type of method used (LGMM or LGCA), characteristics of studies and quality of reporting using the GRoLTS-Checklist. Results: From the 4,694 citations screened, a total of 38 studies were included. The studies were published between 2010 and 2021. Total follow-ups ranged from 8 weeks to 11 years, with 95% (36 out of 38) studies with a follow-up of 1 year of more. The characteristics of studies are presented in Table 1. Regarding the quality of reporting, trajectory groups were adequately presented, however many studies failed to report important decisions made for the trajectory group identification. Many studies considered trajectory groups as exposures to a subsequent outcome. Yet, issues in relation with selection bias, immortal bias and residual confounding were suspected and poorly addressed. Conclusions: Although LCGA were preferred, the context of utilization, data sources and research questions were diverse and unrelated to the type of method used. We recommend authors to clearly report the decisions made in trajectory groups identification.

Longitudinal trajectories of treatment burden: A prospective survey study of adults living with multiple chronic conditions in the midwestern United States.

ObjectivesDetermine whether there are different longitudinal patterns of treatment burden in people living with multiple chronic conditions (MCC) and, if so, explore predictors that might reveal potential routes of intervention.MethodsWe analyzed data from a prospective mailed survey study of 396 adults living with MCC in southeastern Minnesota, USA. Participants completed a measure of treatment burden, the Patient Experience with Treatment and Self-management (PETS), and valid measures of health-related and psycho-social concepts at baseline, 6, 12, and 24 months. Latent class growth mixture modeling (LCGM) determined trajectories of treatment burden in two summary index scores of the PETS: Workload and Impact. Multivariable logistic regressions were used to identify independent predictors of the trajectories.ResultsLCGM supported a 2-class model for PETS Workload, including a group of consistently high workload (N = 69) and a group of consistently low workload (N = 311) over time. A 3-class model was supported for PETS Impact, including groups of consistently high impact (N = 62), consistently low impact (N = 278), and increasing impact (N = 51) over time. Logistic regression analyses showed that the following factors were associated with patterns of consistently high or increasing treatment burden over time: lower health literacy, lower self-efficacy, more interpersonal challenges with others, and worse subjective reports of physical and mental health (all p < .05).ConclusionsDifferent longitudinal patterns of treatment burden exist among people with MCC. Raising health literacy, enhancing self-efficacy, and lessening the effects of negative social interactions might help reduce treatment burden.

Triglyceride-glucose index trajectory and arterial stiffness: results from Hanzhong Adolescent Hypertension Cohort Study

BackgroundThe triglyceride-glucose index (TyG index) has emerged as a reliable surrogate marker of insulin resistance associated with arterial stiffness. However, most studies were based on a cross-sectional design, and few studies have evaluated the longitudinal impact of the TyG index on arterial stiffness. This study aimed to investigate the associations of single time point measurement and the long-term trajectory of the TyG index with arterial stiffness in a Chinese cohort.MethodsData are derived from the Hanzhong Adolescent Hypertension Cohort study. A total of 2480 individuals who participated in the 2017 survey was included in the cross-sectional analysis. A sample of 180 individuals from the sub-cohort with follow-up data in 2005, 2013, and 2017 was enrolled in the longitudinal analysis. The TyG index was calculated as ln (fasting triglyceride [mg/dL] × fasting glucose [mg/dL]/2), and arterial stiffness was determined using brachial-ankle pulse wave velocity (baPWV). The latent class growth mixture modeling method was used to identify the TyG index trajectories from 2005 to 2017.ResultsIn the cross-sectional analysis, the median age of the study population was 42.8 (39.8, 44.9) years, and 1351 (54.5%) were males. Each one-unit increment in TyG index was associated with a 37.1 cm/s increase (95% confidence interval [CI] 23.7–50.6 cm/s; P < 0.001) in baPWV, and similar results were observed when the TyG index was in the form of quartiles. In the longitudinal analysis, we identified three distinct TyG index trajectories and found that the highest TyG index trajectory carried the greatest odds of increased arterial stiffness, with a fully adjusted odds ratio (OR) of 2.76 (95% CI 1.40, 7.54).ConclusionsElevated levels of baseline TyG index and higher long-term trajectory of TyG index were independently associated with increased arterial stiffness. Monitoring immediate levels and longitudinal trends of the TyG index may help with the prevention of arterial stiffness in the long run.

Risk and resilience in trajectories of post-traumatic stress symptoms among first responders after the 2011 Great East Japan Earthquake: 7-year prospective cohort study.

First responders to disasters are at risk of developing post-traumatic stress disorder (PTSD). The trajectories of post-traumatic stress symptom severity differ among individuals, even if they are exposed to similar events. These trajectories have not yet been reported in non-Western first responders. We aimed to explore post-traumatic stress symptom severity trajectories and their risk factors in first responders to the 2011 Great East Japan Earthquake (GEJE) - a historically large earthquake that resulted in a tsunami and a nuclear disaster. A total of 55 632 Japan Ground Self-Defense Force (JGSDF) personnel dispatched to the GEJE were enrolled in this 7-year longitudinal cohort study. PTSD symptom severity was measured using the Impact of Event Scale-Revised. Trajectories were identified using latent growth mixture models (LGMM). Nine potential risk factors for the symptom severity trajectories were analysed using multinomial logistic regression. Five symptom severity trajectories were identified: 'resilient' (54.8%), 'recovery' (24.6%), 'incomplete recovery' (10.7%), 'late-onset' (5.7%), and 'chronic' (4.3%). The main risk factors for the four non-resilient trajectories were older age, personal disaster experiences and working conditions. These working conditions included duties involving body recovery or radiation exposure risk, longer deployment length, later or no post-deployment leave and longer post-deployment overtime. The majority of first responders to GEJE were resilient and developed few or no PTSD symptoms. A substantial minority experienced late-onset and chronic symptom severity trajectories. The identified risk factors can inform policies for prevention, early detection and intervention in individuals at risk of developing symptomatic trajectories.

The course of symptoms in the first 27 months following bereavement: A latent trajectory analysis of prolonged grief, posttraumatic stress, and depression

BackgroundMuch remains unknown about the course of grief in the early months following bereavement, including the prevalence and timing of a recovery trajectory, whether specific symptoms mark a failure to recover, and the co-occurrence of chronic prolonged grief (PG), posttraumatic stress (PTS) and depression symptoms. MethodsTwo hundred fifty-nine participants completed PG, PTS and depression questionnaires up to eleven times every six weeks during the two years post-bereavement. We used Latent Class Growth Mixture Modeling (LCGMM) to identify subgroups of bereaved individuals sharing similar trajectories for each disorder. We used repeated measures ANOVA to evaluate differences in individual symptoms between trajectories. Finally, we investigated to what extent chronic trajectories of these three disorders co-occurred. ResultsThree trajectories of PG symptoms emerged: resilient (66.4%), chronic (25.1%) and acute recovery (8.4%). The overall severity and symptom profile of the acute recovery group were indistinguishable from that of the chronic group through 6 months post-bereavement, followed by reduction in PG from 6 to 18 months post-bereavement. Chronic PTS in the first-year post-bereavement tended to co-occur with chronic PG and/or chronic depression. ConclusionsTwenty five percent of those with initial elevations in grief recovered in the period of 6 to 12 months post-bereavement. These findings highlight the clinical importance of severe grief in the initial months following loss, but also suggests caution in diagnosing a grief disorder within the first-year post-bereavement.

Sex-differential PTSD symptom trajectories across one year following suspected serious injury

ABSTRACT Background Recent years have shown an increased application of prospective trajectory-oriented approaches to posttraumatic stress disorder (PTSD). Although women are generally considered at increased PTSD risk, sex and gender differences in PTSD symptom trajectories have not yet been extensively studied. Objective To perform an in-depth investigation of differences in PTSD symptom trajectories across one-year post-trauma between men and women, by interpreting the general trends of trajectories observed in sex-disaggregated samples, and comparing within-trajectory symptom course and prevalence rates. Method We included N = 554 participants (62.5% men, 37.5% women) from a multi-centre prospective cohort of emergency department patients with suspected severe injury. PTSD symptom severity was assessed at 1, 3, 6, and 12 months post-trauma, using the Clinician-Administered PTSD Scale for DSM-IV. Latent growth mixture modelling on longitudinal PTSD symptoms was performed within the sex-disaggregated and whole samples. Bayesian modelling with informative priors was applied for reliable model estimation, considering the imbalanced prevalence of the expected latent trajectories. Results In terms of general trends, the same trajectories were observed for men and women, i.e. resilient, recovery, chronic symptoms and delayed onset. Within-trajectory symptom courses were largely comparable, but resilient women had higher symptoms than resilient men. Sex differences in prevalence rates were observed for the recovery (higher in women) and delayed onset (higher in men) trajectories. Model fit for the sex-disaggregated samples was better than for the whole sample, indicating preferred application of sex-disaggregation. Analyses within the whole sample led to biased estimates of overall and sex-specific trajectory prevalence rates. Conclusions Sex-disaggregated trajectory analyses revealed limited sex differences in PTSD symptom trajectories within one-year post-trauma in terms of general trends, courses and prevalence rates. The observed biased trajectory prevalence rates in the whole sample emphasize the necessity to apply appropriate statistical techniques when conducting sex-sensitive research.

Suicide Ideation and Social Support Trajectories in National Guard and Reserve Servicemembers

Objective: Since 2004 increased rates of suicide have been noted in the US Armed Forces. We examined the association of social support (SS) trajectories and suicide ideation (SI) over a four-year period in Reserve Component (RC) servicemembers (National Guard and Reserve). We also examined baseline mental health measures, as predictors of the identified trajectories. Methods: Structured interviews were conducted with a nationally representative sample of 1,582 RC servicemembers at baseline and three follow-up waves. Latent growth mixture modeling identified SS trajectories and the association with follow-up SI. Multinomial logistic regression analyses were used to predict SS trajectories using baseline measures of demographics and mental health. Results: We identified four trajectories of SS and their associated prevalence of follow-up SI: low (n = 60, 3.8%; SI = 30.5%), medium (n = 229, 14.5%; SI = 14.1%), high–low (n = 66, 4.2%; SI = 13.6%), and high–high (n = 1,227, 77.5%; SI = 4.2%). There were significant differences in follow-up SI prevalence between each pair of SS trajectories except between the medium-SS and high-low-SS trajectories. Baseline SI, post-traumatic stress disorder (PTSD), depression, binge drinking, and mental health diagnosis were associated with increased likelihood of being on a low-SS or medium-SS trajectory. Baseline PTSD discriminated being on the high-high-SS and high-low-SS trajectories. Conclusion: Results support four trajectories of social support and that individuals with low or decreasing SS are likely to have greater follow-up SI. Baseline mental health assessments can identify these risk trajectories.

Patterns and predictors of dyspnoea following left ventricular assist device implantation.

Dyspnoea is a common symptom of heart failure (HF) that often prompts patients to seek treatment. Implantation of a left ventricular assist device (LVAD) has been associated with reduced dyspnoea but it is unclear if all patients experience similar improvements in dyspnoea over time following LVAD implantation. Our aim was to identify distinct trajectories of dyspnoea symptoms over time following LVAD implantation and predictors of dyspnoea trajectory. We hypothesized that at least two, distinct trajectories of dyspnoea would be observed following LVAD implantation. This was a secondary analysis of data from the Profiling Biobehavioral Responses to Mechanical Support in Advanced Heart Failure study. In the parent study, sociodemographic and clinical data were collected prior to LVAD implantation and at 1, 3, and 6 months following LVAD implantation from a sample (n = 101) of patients with advanced HF. Latent growth mixture modelling was performed to identify distinct trajectories of dyspnoea symptoms. Backwards stepwise logistic regression was used to identify predictors of dyspnoea trajectory. Two, distinct trajectories of dyspnoea symptoms were identified: sustained improvement and unsustained improvement. Participants who experienced sustained improvement (86.7% of sample) demonstrated large, significant improvement in dyspnoea from pre-implantation to 3 months post-implant followed by smaller, non-significant improvement from 3 to 6 months. Participants who experienced unsustained improvement (13.3% of sample) demonstrated initial improvement from pre-implantation to 3 months post-implantation followed by worsening of dyspnoea from 3 to 6 months. Greater depressive symptoms at baseline and living alone were significant predictors of unsustained improvement. Patients experience different patterns of dyspnoea over time following LVAD implantation. Clinicians should inquire about living arrangements and depressive symptoms at each visit to determine risk of unsustained improvement in dyspnoea.

Short-term pain trajectories in patients with knee osteoarthritis.

It is unknown if pain in knee osteoarthritis (KOA) follows distinct patterns over the short term. Therefore, the aim of this study was to identify whether persons with a previous history of KOA pain fluctuations have distinct trajectories of pain over 90days and to examine associations between baseline characteristics and pain trajectories. People with a previous history of KOA were selected from a web-based longitudinal study. Baseline variables were sex, age, being obese/overweight, years of KOA, knee injury, knee buckling, satisfactory Lubben Social Support Score, pain and stress scales, Intermittent Constant Osteoarthritis Pain Score (ICOAP), medication use, and physical activity. Participants completed a Knee Injury and Osteoarthritis Outcomes Score (KOOS) pain subscale (KOOS-p, rated 0=extreme to 100=no knee problems) at 10-day intervals for 90days. Short-term KOOS-p trajectories were identified using latent growth mixture modeling and the baseline risk factors for these pain trajectories were examined. Participants (n=313) had a mean age of 62.2 (SD ± 8.1) years and and a body mass index of 29.8 (SD ± 6.6) kg/m2 . The three-class latent growth mixture modeling quadratic model with best fit indices was chosen (based on lowest sample-size-adjusted Bayesian Information Criterion, high probability of belonging, interpretability). Three distinct pain trajectory clusters (over 90days) were identified: low-moderate pain at baseline with large improvement (n=11), minimal change in pain over 90days (n=248), and moderate-high pain with worsening (n=46). Higher ICOAP (intermittent scale), perceived stress, negative affect score, and knee buckling at baseline were associated with a worse knee pain trajectory (P<0.05). Persons with KOA showed unique short-term pain trajectories over 90days, with distinct characteristics at baseline associated with each trajectory.

Prediction of depressive symptoms onset and long-term trajectories in home-based older adults using machine learning techniques

Objectives Our aim was to explore the possibility of using machine learning (ML) in predicting the onset and trajectories of depressive symptom in home-based older adults over a 7-year period. Methods Depressive symptom data (collected in the year 2011, 2013, 2015 and 2018) of home-based older Chinese (n = 2650) recruited in the China Health and Retirement Longitudinal Study (CHARLS) were included in the current analysis. The latent class growth modeling (LCGM) and growth mixture modeling (GMM) were used to classify different trajectory classes. Based on the identified trajectory patterns, three ML classification algorithms (i.e. gradient boosting decision tree, support vector machine and random forest) were evaluated with a 10-fold cross-validation procedure and a metric of the area under the receiver operating characteristic curve (AUC). Results Four trajectories were identified for the depressive symptoms: no symptoms (63.9%), depressive symptoms onset {incident increasing symptoms [new-onset increasing (16.8%)], chronic symptoms [slowly decreasing (12.5%), persistent high (6.8%)]}. Among the analyzed baseline variables, the 10-item Center for Epidemiologic Studies Depression Scale (CESD-10) score, cognition, sleep time, self-reported memory were the top five important predictors across all trajectories. The mean AUCs of the three predictive models had a range from 0.661 to 0.892. Conclusions ML techniques can be robust in predicting depressive symptom onset and trajectories over a 7-year period with easily accessible sociodemographic and health information. Supplemental data for this article is available online at http://dx.doi.org/10.1080/13607863.2022.2031868

Anxiety Trajectories in Adolescents and the Impact of Social Support and Peer Victimization.

This paper examines whether adolescents can be reliably categorized into subgroups based on their patterns of anxiety levels over time and whether low levels of social support from parents, peers, and their school, and high levels of peer victimization, predict a pattern of increasing anxiety. Participants were 3392 youth from the Longitudinal Study of Australia's Children (LSAC). Youth-reported anxiety was measured at three occasions at ages 12/13years, 14/15years, and 16/17years, with social support and victimization assessed at age 12/13years. Anxiety trajectories were identified using latent class growth mixture modelling, and predictors of class membership were examined using multinomial logistic regression analyses. Three discrete classes of anxiety trajectories were identified. Most youth fell within a stable-low anxiety symptom class (89.5% males; 78.2% females), with smaller percentages in low-increasing (5.6% males; 14.4% females) or high-decreasing (4.9% males; 7.4% females) classes. Low support from parents and teachers, low sense of school belonging, and high peer victimization predicted membership of the low-increasing anxiety trajectory class, irrespective of gender. Social support did not moderate the effect of peer victimization upon the risk of developing anxiety, with peer victimization remaining a risk factor even when adolescents experienced good social support from parents, peers, and school. The findings highlight the need for screening in early adolescence to identify those who are experiencing low social support and high peer victimization and are thus at increased risk of developing anxiety problems. These youth could then be offered targeted intervention to reduce the likelihood of anxiety development.

Distinct trajectories of perinatal depression in Chinese women: application of latent growth mixture modelling

BackgroundCurrent research on perinatal depression rarely pays attention to the continuity and volatility of depression symptoms over time, which is very important for the early prediction and prognostic evaluation of perinatal depression. This study investigated the trajectories of perinatal depression symptoms and aimed to explore the factors related to these trajectories.MethodsThe study recruited 550 women during late pregnancy (32 ± 4 weeks of gestation) and followed them up 1 and 6 weeks postpartum. Depressive symptoms were measured using the Edinburgh Postnatal Depression Scale (EPDS). Latent growth mixture modelling (LGMM) was used to identify trajectories of depressive symptoms during pregnancy.ResultsTwo trajectories of perinatal depressive symptoms were identified: “decreasing” (n = 524, 95.3%) and “increasing” (n = 26, 4.7%). History of smoking, alcohol use and gestational hypertension increased the chance of belonging to the increasing trajectories, and a high level of social support was a protective factor for maintaining a decreasing trajectory.ConclusionsThis study identified two trajectories of perinatal depression and the factors associated with each trajectory. Paying attention to these factors and providing necessary psychological support services during pregnancy would effectively reduce the incidence of perinatal depression and improve patient prognosis.

Developmental Trend of Subjective Well-Being of Weibo Users During COVID-19: Online Text Analysis Based on Machine Learning Method.

Currently, the coronavirus disease 2019 (COVID-19) pandemic experienced by the international community has increased the usage frequency of borderless, highly personalized social media platforms of all age groups. Analyzing and modeling texts sent through social media online can reveal the characteristics of the psychological dynamic state and living conditions of social media users during the pandemic more extensively and comprehensively. This study selects the Sina Weibo platform, which is highly popular in China and analyzes the subjective well-being (SWB) of Weibo users during the COVID-19 pandemic in combination with the machine learning classification algorithm. The study first invokes the SWB classification model to classify the SWB level of original texts released by 1,322 Weibo active users during the COVID-19 pandemic and then combines the latent growth curve model (LGCM) and the latent growth mixture model (LGMM) to investigate the developmental trend and heterogeneity characteristics of the SWB of Weibo users after the COVID-19 outbreak. The results present a downward trend and then an upward trend of the SWB of Weibo users during the pandemic as a whole. There was a significant correlation between the initial state and the development rate of the SWB after the COVID-19 outbreak (r = 0.36, p < 0.001). LGMM results show that there were two heterogeneous classes of the SWB after the COVID-19 outbreak, and the development rate of the SWB of the two classes was significantly different. The larger class (normal growth group; n = 1,229, 93.7%) showed a slow growth, while the smaller class (high growth group; n = 93, 6.3%) showed a rapid growth. Furthermore, the slope means across the two classes were significantly different (p < 0.001). Therefore, the individuals with a higher growth rate of SWB exhibited stronger adaptability to the changes in their living environments. These results could help to formulate effective interventions on the mental health level of the public after the public health emergency outbreak.

Latent trajectories of symptom change during cognitive-behavior therapy predict post-treatment worsening of symptoms: a preliminary examination among outpatients with bulimia-spectrum eating disorders.

PurposeUp to 44% of individuals with bulimia nervosa (BN) experience worsening of symptoms after cognitive behavior therapy (CBT). Identifying risk for post-treatment worsening of symptoms using latent trajectories of change in eating disorder (ED) symptoms during treatment could allow for personalization of treatment to improve long-term outcomesMethodsParticipants (N = 56) with BN-spectrum EDs received 16 sessions of CBT and completed digital self-monitoring of eating episodes and ED behaviors. The Eating Disorder Examination was used to measured ED symptoms at post-treatment and 3-month follow-up. Latent growth mixture modeling of digital self-monitoring data identified latent growth classes. Kruskal–Wallis H tests examined effect of trajectory of change in ED symptoms on post-treatment to follow-up symptom change.ResultsMulti-class models of change in binge eating, compensatory behaviors, and regular eating improved fit over one-class models. Individuals with high frequency-rapid response in binge eating (H(1) = 10.68, p =0 .001, η2 = 0.24) had greater recurrence of compensatory behaviors compared to individuals with low frequency-static response. Individuals with static change in regular eating exhibited greater recurrence of binge eating than individuals with moderate response (H(1) = 8.99, p = 0.003, η2 = 0.20).ConclusionTrajectories of change in ED symptoms predict post-treatment worsening of symptoms. Personalized treatment approaches should be evaluated among individuals at risk of poor long-term outcomes.Level of evidenceIV, evidence obtained from multiple time series.Trial registrationClinicalTrials.gov registration number NCT03673540, registration date: September 17, 2018.Supplementary InformationThe online version contains supplementary material available at 10.1007/s40519-021-01348-5.