Latent Class Profile Analysis Research Articles

Comparing symptom clusters in cancer survivors by cancer diagnosis: A latent class profile analysis

PurposeResearch on symptom clusters in oncology is progressing, but knowledge gaps remain. One question is whether the number and types of symptom subgroups (i.e., latent classes) differ based on cancer diagnosis. The purpose of this study was to: (1) identify and compare latent class subgroups based on four highly prevalent symptoms (pain, fatigue, sleep disturbance, and depression), and (2) examine the differences in sociodemographic and clinical factors in the identified latent classes across the seven cancer types (i.e., prostate, non-small cell lung, non-Hodgkin’s lymphoma, breast, uterine, cervical, and colorectal cancer).MethodsThis study is a cross-sectional secondary analysis of data obtained from the My-Health study in partnership with four Surveillance, Epidemiology, and End Results (SEER) cancer registries located in California (two), Louisiana, and New Jersey. The sample included 4,762 cancer survivors 6-13 months following diagnosis of one of the seven cancer types mentioned. Latent class profile analysis was used.ResultsSubjects were primarily young (59% age 21-64 years), Caucasian (41%), married/cohabitating (58%) and unemployed (55%). The number and types of symptom subgroups varied across these seven cancer populations: four-subgroups were the common in prostate, lung, non-Hodgkin’s lymphoma, and breast cancer survivors. Unmarried, low education, and unemployment status were associated with high risk of symptom burden across the cancer types.ConclusionIdentifying symptom subgroups by cancer diagnosis has the potential to develop innovative and effective targeted interventions in cancer survivors. Further research is needed to establish extensive knowledge in symptom clustering between treatment regimens, and short-term and long-term cancer survivors.

Latent class symptom profiles of colorectal cancer survivors with cancer-related cognitive impairment.

Colorectal cancer (CRC) survivors experience cancer-related cognitive impairment and co-occurring symptoms after cancer treatments. There has been little data to inform the risk factors of complex symptom phenotypes in CRC survivors. To determine if subgroups of CRC survivors after cancer treatments could be identified based on the cognitive impairment and common co-occurring symptoms (depression, anxiety, sleep disturbance, fatigue, and pain); and to explore risk factors (sociodemographic and clinical characteristics, perceived stress, and social support) of these subgroups. Latent class profile analysis (LCPA) was used to identify subgroups based on self-reported symptoms in 64 CRC survivors. Cognitive impairment was measured by assessing subjective cognitive function using the Patient-Reported Outcome Measurement Information System (PROMIS) measure. The Kruskal-Wallis test and regression analyses were performed. Three distinct latent classes were identified (Class 1: All Low '28.1%'; Class 2: High Psychological Symptoms (depression/anxiety) '25%'; Class 3: High Somatic Symptoms (fatigue, sleep disturbance, and pain) with High Cognitive Impairment'46.9%'). The pain was the most distinguishable symptom across the latent classes. The high symptom burden group was associated with less time since cancer diagnosis, higher perceived stress levels, and poor emotional social support. Our study adds to the information on interindividual variability in symptom experience of CRC survivors with cognitive impairment. Findings suggest a need for increased attention to screening for co-occurring symptoms (e.g., high pain) and future interventions focused on stress management and social support.

Risk Factors for a Higher Symptom Burden in Patients With Cancer During the COVID-19 Pandemic.

To evaluate for subgroups of patients with distinct symptom profiles and differences in demographic and clinical characteristics and stress and resilience among these subgroups. 1,145 patients with cancer aged 18 years or older completed a survey online. Data were collected between May 2020 and February 2021. Patients completed measures for depression, state anxiety, cognitive function, morning fatigue, evening fatigue, morning energy, evening energy, sleep disturbance, pain, stress, and resilience. Latent class profile analysis was used to identify subgroups of patients with distinct symptom profiles. Differences among the subgroups on study measures were evaluated using parametric and nonparametric tests. Four distinct profiles were identified (none, low, high, and very high). Patients in the high and very high classes reported clinically meaningful levels of all nine symptoms. Differences among the four profiles for stress and resilience exhibited a dose-response effect. Findings can serve as benchmark data of the symptom burden of patients with cancer following the COVID-19 pandemic.

Exploring latent profiles of disordered eating using an indicator of dietary restriction in an undergraduate sample of men and women.

Latent class or latent profile analysis (LPA) studies in patients with eating disorders consistently identify a low-weight, restrictive eating subgroup that does not endorse weight/shape concerns. To date, similar studies in samples unselected for disordered eating symptoms have not identified a high restriction-low weight/shape concerns group, which may be due to a lack of inclusion of measures of dietary restriction. We conducted an LPA using data from 1623 college students (54% female) recruited across three different studies. The Eating Pathology Symptoms Inventory Body Dissatisfaction, Cognitive Restraint, Restricting, and Binge Eating subscales were used as indicators, and body mass index, gender, and dataset were covaried. Purging, excessive exercise, emotion dysregulation, and harmful alcohol use were compared across resulting clusters. Fit indices supported a 10-class solution, including five disordered eating groups (largest to smallest): "Elevated General Disordered Eating", "Body Dissatisfied Binge Eating," "Most Severe General Disordered Eating," "Non-Body Dissatisfied Binge Eating," and "Non-Body Dissatisfied Restriction." The "Non-Body Dissatisfied Restriction" group scored as low on other measures of traditional eating pathology and harmful alcohol use as non-disordered eating groups but scored as high on a measure of emotion dysregulation as other disordered eating groups. This study is the first to identify a latent restrictive eating group that does not endorse traditional disordered eating cognitions in an unselected sample of undergraduate students. Results underscore the importance of using measures of disordered eating behaviors without implied motivation to capture overlooked problematic eating patterns in the population that are distinct from our "traditional" understanding of disordered eating. We identified a group of individuals with high levels of restrictive eating but low body dissatisfaction and intent to diet in an unselected adult sample of men and women. Results underscore the need to investigate restrictive eating outside of the traditional lens of body shape concerns. Findings also suggest that individuals with nontraditional eating difficulties may struggle with emotion dysregulation, putting them at risk of poor psychological and relational outcomes.

Patterns of health-promoting behaviors and associated factors in family caregivers of people receiving cancer treatment: A latent class profile analysis.

Family caregivers tend to neglect their health while prioritizing the needs of their care recipients. Identifying subgroups of caregivers based on the patterns of health-promoting behaviors (HPBs) may help develop tailored interventions for them, yet little is known. The purpose of this study was: (1) to identify latent classes with distinct patterns of HPBs in family caregivers of people with cancer; and (2) to investigate factors associated with the latent class membership. We performed a cross-sectional data analysis using the baseline dataset from a longitudinal survey study that assessed HPBs of family caregivers of individuals who received cancer treatment at a national research hospital (N=124). Latent class profile analysis was conducted to identify latent classes based on the subdomains of the Health-Promoting Lifestyle Profile II, followed by multinomial logistic regression analysis to investigate factors associated with the latent class membership. Three latent classes were identified: a high level of HPB (Class 1, 25.8%); a moderate level of HPB (Class 2, 53.2%); and a low level of HPB (Class 3, 21.0%) of HPBs. Controlling for caregiver age and sex, caregiver burden due to lack of family support, perceived stress, self-efficacy and body mass index were factors associated with the latent class membership. HPBs of our caregiver sample appeared in relatively stable patterns at different levels. Higher caregiver burden and perceived stress and lower self-efficacy were associated with the lower practice of HPBs overall. Our findings may serve as a reference for screening caregivers who need support and developing person-centered interventions.

Identification of Distinct Symptom Profiles in Cancer Patients Using a Pre-Specified Symptom Cluster.

Pain, fatigue, sleep disturbance, and depression often co-occur in oncology patients and negatively impact quality of life (QOL). Study purposes were to identify subgroups of patients with distinct symptom profiles based on their experiences with a pre-specified symptom cluster (i.e., pain, fatigue, depression, sleep disturbance) and to identify demographic, clinical, and symptom characteristics and QOL outcomes associated with each distinct profile. Patients with breast, lung, gastrointestinal, and gynecologic cancers (n=1340) were recruited from outpatient clinics during their first or second cycle of chemotherapy. They completed valid and reliable measures of pain, fatigue, sleep disturbance. depression, and QOL prior to their next dose of chemotherapy. Latent class profile analysis was used to identify the patient subgroups. Differences among the profiles were evaluated using parametric and non-parametric tests. Three distinct profiles were identified (i.e., Low (44.0%), Moderate (45.1%), High (10.8%). Compared to Low class, Moderate and High classes were younger and more likely to be female. Compared to the other two classes, High class was less likely to be married/partnered and employed, more likely to have a lower income and childcare responsibilities, had lower functional status, a higher body mass index, and exercised less. For both QOL scales, differences in subscale and total scores followed the same pattern (Low>Moderate>High). Over 55% of patients undergoing chemotherapy had a moderate to high symptom burden associated with these four common co-occurring symptoms. Multimodal interventions are needed to decrease symptom burden and improve QOL outcomes in these patients.

Латентные личностные профили стилевой регуляции принятия решений

Decision making (DM) generally assumes that the person is performing a choice between a multitude of alternatives under uncertainty and possible risk. According to the concept of dynamic regulative systems (Kornilova, 2016), preferred or most relied on DM strategies are linked in an integrative way with a variety of personality traits that can be at the top of the hierarchy. These include risk readiness, rationality, and Dark Triad traits as reflective of a generally unstable personality core. Decision-Making Tendency Inventory (DMTI; Misuraca et al., 2015) defined DM characteristics via maximization, satisficing and minimization. However, the relationships between DM characteristics captured by DMTI and the listed personality traits have not been explored before. The goal of the current study was establishing latent personality profiles in a person-centered approach that integrates DM “tendencies” and the listed personality traits by identifying relatively homogenous subgroups of individuals with similar profiles. Methods. 625 individuals in the age from 17 to 39 years (М = 20,17, SD = 3,02; 84% females) participated in the study. We used DMTI, Dirty Dozen, and LFR questionnaires to measure DM tendencies, Dark Triad traits, and risk readiness/rationality, respectively. Latent profile analysis was performed in VarSelLCM for R. Results. The results indicated the presence of three latent profiles in the data after adjustments for age and sex. Risk readiness and Dark Triad traits were positively related with maximizing and satisficing, forming one latent class. In another class lower rationality, on the other hand, was linked with minimization. In the third class higher rationality accompanied lower Dark Triad traits. Conclusions. The results provide evidence in favor of the general hypothesis that latent profiles of personality traits are associated with distinct preferences for specific DM tendencies. Higher levels of maximizing, satisficing, and minimizing were not related to subclinical psychopathy or Machiavellianism. Higher narcissism and risk readiness, generally unrelated, are nonetheless characteristic of the latent class that prefers maximizing and satisficing. Preference for minimization of effort during DM was associated with lower rationality. Latent class or latent profile analysis is a powerful technique that sheds new light on the relationships between personality and DM, beyond the contributions of variable-centered approaches such as correlational analysis.

Identification of Self-Management Behavior Clusters Among People Living with HIV in China: A Latent Class Profile Analysis.

BackgroundSelf-management directly affects the health outcomes and quality of life among people living with HIV (PLWH). A better understanding of self-management level will provide evidence for researchers to develop effective interventions.PurposeThis study aims to identify the latent classes among PLWH in their levels of self-management behavior, and to explore the sociodemographic and disease-related predictors within these classes.Materials and MethodsA total of 868 PLWH were recruited from August 2017 to January 2019 in Sichuan Province, China. A latent class profile analysis was used to identify participants’ self-management behavior, and multinomial logistic regression was used to explore the sociodemographic and disease-related predictors of the different latent classes.ResultsModel fit indices supported a three-class model. The mean self-management scores in the three classes were 23.56 (SD=6.02), 37.91 (SD=3.80), and 47.95 (SD=4.18), respectively. The latent classes were Class 1 (a poor level of self-management behavior, 12.1%, n=104), Class 2 (a moderate level of self-management behavior, 56.1%, n=491) and Class 3 (a good level of self-management behavior, 31.7%, n=273). Antiretroviral trerapy (ART) status, infection route, and educational level were the main predictors of self-management behavior.ConclusionThe findings indicated that the level of self-management behaviors among PLWH in China is inadequate. Those with a lower educational level, who were infected through blood/injecting drugs, and who were not receiving ART, showed a significantly lower level of self-management behavior. These results could help healthcare professionals to quickly recognize PLWH who are at a high risk of low-level self-management, using individual characteristics and could provide a scientific basis for the development of effective and targeted programs to improve self-management level in PLWH.

Profiles of HIV Risk, Sexual Power, and Decision-Making among Sexual Minority Men of Color Who Engage in Transactional Sex: A Latent Profile Analysis.

Though the transmission of HIV is preventable, there were still 37,968 new documented cases in the United States in 2018. HIV incidence is disproportionate in sexual minority men of color. The purpose of this study was to examine sexual relationship power risk profiles to identify distinct subgroups within the profiles who carry the highest HIV risk. Latent class profile analysis was used to identify subgroups of sexual minority men of color at the highest risk for contracting HIV based on their sexual power profiles. Among 322 sexual minority men, we identified four latent profiles. Profile 1: Low transactional sex and high power (n = 133; 14.3%); Profile 2: Transactional sex, high decision-making in sexual relationships, and low control in sexual relationship (n = 99; 30.7%); Profile 3: Low transactional sex, low decision-making, and moderate control (n = 43; 13.4%); Profile 4: High transactional sex and low power (n = 47; 14.6%). LPA was useful to identify distinct subgroups based on measures of sexual risk and relationship sexual power. Findings carry significant implications for developing tailored strategies to increase HIV knowledge and related HIV prevention and risk reduction services for sexual minority men of color who engage in transactional sex.

Bayesian multivariate latent class profile analysis: Exploring the developmental progression of youth depression and substance use

Multivariate latent class profile analysis (MLCPA) is a useful tool for exploring the stage-sequential process of multiple latent class variables, but the inference can be challenging due to the high-dimensional latent structure of the model. In this paper, a Bayesian approach via Markov chain Monte Carlo (MCMC) is proposed for MLCPA as an alternative to the maximum-likelihood (ML) method. Compared to the ML solution, Bayesian estimates are less sensitive to the set of initial values as well as easier to obtain standard errors. We also address issues in MCMC such as label-switching problem with a dynamic data-dependent prior and computational complexity with a recursive formula. Simulation studies revealed the validity and efficiency of the proposed algorithm. An empirical analysis of MLCPA using the National Longitudinal Survey of Youth 97 (NLSY97) identified a small number of representative developmental progressions of adolescent depression and substance use.

Psychological Allostatic Load: the Cost of Persistence in STEM Disciplines

The purpose of this study is to explore the impact of persistence within STEM learning environments as a function of cumulative stress and latent trauma. The biopsychosocial impact of prolonged stressors due to hostile environments and academic demands has deleterious health effects on historically underrepresented students who enter STEM disciplines. The Trauma Symptoms Checklist for Children, clinical histories, and psychophysiomeasurement tools were used to measure the effects of cumulative stress and latent trauma as historically underrepresented students persisted through high school STEM discipline classes. Elevated responses on the inventory and history were triangulated through measures of biological markers for cumulative stress and developed into a profile combination of traits to identify those students likely to show symptomology consistent with the negative effects of cumulative stress and latent trauma. Examination of these outcomes using a latent class profile analysis model suggested the presence of cumulative stress resulting from program participation was significant.

Multilevel Latent Class Profile Analysis: An Application to Stage-Sequential Patterns of Alcohol Use in a Sample of Canadian Youth.

Recently, latent class analysis (LCA) and its variants have been proposed to identify subgroups of individuals who follow similar sequential patterns of latent class membership for longitudinal study. A primary assumption underlying the family of LCA is that individual observations are independent. In many applications, however, particularly in research on adolescent substance use, individuals are often dependent because of multilevel data structure, where the unit of observation (e.g., students) is nested in higher level units (e.g., schools). In this study, we propose multilevel latent class profile analysis (MLCPA), which will allow us to analyze the longitudinal data with a multilevel structure under the framework of LCA. We apply an MLCPA using data from the COMPASS study, a 9-year study funded by the Canadian Institutes of Health Research and Health Canada, in order to identify representative sequential drinking patterns of Canadian youth and investigate whether these sequential patterns vary across schools. The MLCPA identified three common student-level drinking behaviors: non-drinker, ever lifetime, and binge drinker. The sequence of drinking behaviors can be classified into one of three longitudinal sequential patterns: non-drinking stayer, light drinking advancer, and heavy drinking advancer. In addition, MLCPA uncovered two latent clusters (low-use school and high-use school) out of 64 schools in Ontario and Alberta based on the prevalences of sequential drinking patterns.

Latent Class Analysis for Repeatedly Measured Multiple Latent Class Variables

Research on stage-sequential shifts across multiple latent classes can be challenging in part because it may not be possible to observe the particular stage-sequential pattern of a single latent class variable directly. In addition, one latent class variable may affect or be affected by other latent class variables and the associations among multiple latent class variables are not likely to be directly observed either. To address this difficulty, we propose a multivariate latent class analysis for longitudinal data, joint latent class profile analysis (JLCPA), which provides a principle for the systematic identification of not only associations among multiple discrete latent variables but sequential patterns of those associations. We also propose the recursive formula to the EM algorithm to overcome the computational burden in estimating the model parameters, and our simulation study shows that the proposed algorithm is much faster in computing estimates than the standard EM method. In this work, we apply a JLCPA using data from the National Longitudinal Survey of Youth 1997 in order to investigate the multiple drug-taking behavior of early-onset drinkers from their adolescence, via young adulthood, to adulthood.

Cooking Frequency and Perception of Diet among US Adults Are Associated with US Healthy and Healthy Mediterranean-Style Dietary Related Classes: A Latent Class Profile Analysis.

Background: Meal habits are associated with overall dietary quality and favorable dietary patterns determined by the Healthy Eating Index (HEI). However, within dietary patterns, complexities of food combinations that are not apparent through composite score determination may occur. Also, explorations of these food combinations with cooking and perceived diet quality (PDQ) remain unknown. Methods: Data from the National Health and Nutrition Examination Survey (NHANES) 2007–2010 were utilized to determine the frequency of cooking at home and PDQ, along with sociodemographic variables. Latent class profile analysis was performed to determine person-centered data-driven analysis using the dietary index, HEI-2010, at both the daily and dinner meal-time levels. Multinomial logistic regression analysis was utilized to evaluate the association of dietary patterns with all covariates. Results: For daily HEI, five distinct dietary classes were identified. For dinner HEI, six classes were identified. In comparison to the standard American diet classes, home cooking was positively associated with daily (p < 0.05) and dinner (p < 0.001) dietary classes that had the highest amounts of total vegetable and greens/beans intake. PDQ was positively associated with these classes at the daily level (p < 0.001), but negatively associated with healthier classes at the dinner level (p < 0.001). Conclusion: The use of latent class profile analysis at the daily and dinner meal-time levels identified that food choices coalesce into diverse intakes, as shown by identified dietary classes. Home cooking frequency could be considered a positive factor associated with higher vegetable intake, particularly greens/beans, at the daily and dinner levels. At the same time, the perception of diet quality has a positive association only with daily choices.

Identification of health behaviour clusters among people at high risk of stroke: A latent class profile analysis

To identify the possible latent classes of health behaviour reported by people at high risk of stroke and to explore the predictors of these different classes of health behaviour. A cross-sectional survey study. A stratified cluster random sampling method was used to collect data from 2,500 individuals at high risk of stroke who were from Henan Province, China, from January 2018-January 2019. A latent class profile analysis was used to identify the health behaviour clusters and multinomial logistic regression was used to determine which factors predicted the emergent latent classes of health behaviour. High-risk individuals (N=2,236) at high risk of stroke replied to the survey (89.44% response rate). Model fit indices (AIC=257,509.610, BIC=260,228.733, Entropy=0.956) supported a three-class model of health behaviours. The latent classes were Class 1 (a good level of adaptive health behaviour, 31%, N=693), Class 2 (a moderate level of adaptive health behaviour, 36%, N=805) and Class 3 (a poor level of adaptive health behaviour, 33%, N=738); Based on physical and belief, behaviour and clinical profiles, the three classes were further labelled self-realization deficiency subgroup, social contact anxiety subgroup and health responsibility absence subgroup respectively. Older age, male gender, no spouse, lower education and household income were risk factors associated with good health behaviour. After controlling these socio-demographic variables, high levels of health-related knowledge and attitude were the main positive predictors of health behaviour. This study has identified three different latent classes of health behaviour and their predictive factors in people at high risk of stroke in the Chinese setting. This study has significance for the promotion of adaptive health behaviour in individuals at high risk of stroke. It has allowed the identification of specific clusters of health behaviour that vary in terms of their adaptiveness and forms the basis for the development of a targeted intervention to promote health behaviour for each different subgroup.

A multivariate latent class profile analysis for longitudinal data with a latent group variable

In research on behavioral studies, significant attention has been paid to the stage-sequential process for multiple latent class variables. We now explore the stage-sequential process of multiple latent class variables using the multivariate latent class profile analysis (MLCPA). A latent profile variable, representing the stage-sequential process in MLCPA, is formed by a set of repeatedly measured categorical response variables. This paper proposes the extended MLCPA in order to explain an association between the latent profile variable and the latent group variable as a form of a two-dimensional contingency table. We applied the extended MLCPA to the National Longitudinal Survey on Youth 1997 (NLSY97) data to investigate the association between of developmental progression of depression and substance use behaviors among adolescents who experienced Authoritarian parental styles in their youth.

The adult outcome of children referred for autism: typology and prediction from childhood.

BackgroundAutism Spectrum Disorder is highly heterogeneous, no more so than in the complex world of adult life. Being able to summarize that complexity and have some notion of the confidence with which we could predict outcome from childhood would be helpful for clinical practice and planning.MethodsLatent class profile analysis is applied to data from 123 participants from the Early Diagnosis Study (Lord et al., Archives of General Psychiatry, 2006, 63, 694) to summarize in a typology the multifacetted early adult outcome of children referred for autism around age 2. The form of the classes and their predictability from childhood is described.ResultsDefined over 15 measures, the adult outcomes were reduced to four latent classes, accounting for much of the variation in cognitive and functional measures but little in the affective measures. The classes could be well and progressively more accurately predicted from childhood IQ and symptom severity measurement taken at age 2 years to age 9 years. Removing verbal and nonverbal IQ and autism symptom severity measurement from the profile of adult measures did not change the number of the latent classes; however, there was some change in the class composition and they were more difficult to predict.ConclusionsWhile an empirical summary of adult outcome is possible, careful consideration needs to be given to the aspects that should be given priority. An outcome typology that gives weight to cognitive outcomes is well predicted from corresponding measures taken in childhood, even after account for prediction bias from fitting a complex model to a small sample. However, subjective well‐being and affective aspects of adult outcome were weakly related to functional outcomes and poorly predicted from childhood.

Symptom Clusters in Breast Cancer Survivors: A Latent Class Profile Analysis.

To identify symptom clusters in breast cancer survivors and to determine sociodemographic and clinical characteristics influencing symptom cluster membership. The authors performed a cross-sectional secondary analysis of data obtained from a community-based cancer registry-linked survey with 1,500 breast cancer survivors 6-13 months following a breast cancer diagnosis. Symptom clusters were identified using latent class profile analysis of four patient-reported symptoms (pain, fatigue, sleep disturbance, and depression) with custom PROMIS® short forms. Four distinct classes were identified. Common symptom clusters may lead to better prevention and treatment strategies that target a group of symptoms. Results also suggest that certain factors place patients at high risk for symptom burden, which can guide tailored interventions.

RESILIENCY PHENOTYPES FOLLOWING HIP FRACTURE IN OLDER ADULTS

Defining common patterns of recovery after an acute health stressor (resiliency phenotypes) has clinical and research implications. We examined groups of patients with similar recovery patterns across 10 outcomes following hip fracture to determine the most important predictors of resiliency group membership. This study is a secondary analysis of three prospective cohort studies. Participants, community-dwelling adults aged >65 with recent surgical repair of a hip fracture (n=541), were recruited from eight hospitals near Baltimore and followed for up to one year. Self-reported function and activity measures were collected using validated scales at baseline, 2, 6, and 12 months. Physical performance tests were administered at all follow-up visits. Stressor characteristics, co-morbidities, psychosocial and environmental factors were collected at baseline, and latent class profile analysis was used to identify resiliency phenotypes and logistic regression models to identify associated factors. Three resiliency phenotypes had similar patterns across the 10 outcome measures and were defined as “high resilience” (n=163, 30.1%), “medium resilience” (n=242, 44.7%), and “low resilience” (n=136, 25.2%). Recovery trajectories for outcome measures were plotted for each resiliency group. Self-reported pre-fracture function was by far the strongest predictor of resilience group membership (AUC 0.84). Demographic factors, co-morbidities, stressor characteristics, environmental factors, and psychosocial characteristics were less predictive, but several factors remained significant in a fully adjusted model (AUC 0.88). These three resiliency phenotypes have immediate utility for clinical decision-making. They can be measured in future studies with a more parsimonious set of variables, and may prove useful for understanding mediators of physical resilience.

Resiliency Groups Following Hip Fracture in Older Adults.

Defining common patterns of recovery after an acute health stressor (resiliency groups) has both clinical and research implications. We sought to identify groups of patients with similar recovery patterns across 10 outcomes following hip fracture (stressor) and to determine the most important predictors of resiliency group membership. Secondary analysis of three prospective cohort studies. Participants were recruited from various hospitals in the Baltimore Hip Studies network and followed for up to 1 year in their residence (home or facility). Community-dwelling adults aged 65 years or older with recent surgical repair of a hip fracture (n = 541). Self-reported physical function and activity measures using validated scales were collected at baseline (within 15-22 d of fracture), 2, 6, and 12 months. Physical performance tests were administered at all follow-up visits. Stressor characteristics, comorbidities, and psychosocial and environmental factors were collected at baseline via participant report and chart abstraction. Latent class profile analysis was used to identify resiliency groups based on recovery trajectories across 10 outcome measures and logistic regression models to identify factors associated with those groups. Latent profile analysis identified three resiliency groups that had similar patterns across the 10 outcome measures and were defined as "high resilience" (n = 163 [30.1%]), "medium resilience" (n = 242 [44.7%]), and "low resilience" (n = 136 [25.2%]). Recovery trajectories for the outcome measures are presented for each resiliency group. Comparing highest with the medium- and low-resilience groups, self-reported pre-fracture function was by far the strongest predictor of high-resilience group membership with area under the curve (AUC) of .84. Demographic factors, comorbidities, stressor characteristics, environmental factors, and psychosocial characteristics were less predictive, but several factors remained significant in a multivariable model (AUC = .88). These three resiliency groups following hip fracture may be useful for understanding mediators of physical resilience. They may provide a more detailed description of recovery patterns in multiple outcomes for use in clinical decision making. J Am Geriatr Soc 67:2519-2527, 2019.