| |
Psychological resources protect health: 5-year survival and immune function among HIV-infected women from four US cities
|
| |
| |
"....Psychological resources may protect against HIV-related mortality and immune system decline..."
AIDS: Volume 20(14) 11 September 2006 p 1851-1860
Ickovics, Jeannette Ra; Milan, Stephaniea; Boland, Robertb; Schoenbaum, Elliec; Schuman, Paulad; Vlahov, Davide; for the HIV Epidemiology Research Study (HERS) Group
From the aYale School of Public Health and Center for Interdisciplinary Research on AIDS, New Haven, Connecticut, USA
bDepartment of Psychiatry, Brown University, Providence, Rhode Island, USA
cDepartment of Medicine, Montefiore Medical Center, Bronx, New York, USA
dDepartment of Medicine, Virginia Commonwealth University, Richmond, Virginia, USA
eCenter for Urban Epidemiological Studies, New York Academy of Medicine, New York, New York, USA.
The HIV Epidemiologic Research Study (HERS): The cohort included 871 women who were HIV seropositive recruited from clinical sites in four US cities: Baltimore, Maryland; Bronx, New York; Providence Rhode Island; and Detroit, Michigan... 60% were black, 20% Latina, and 20% white or other race/ethnicity. The majority of women were unemployed (82%), reported monthly incomes below US$ 1000 (72%), and received public assistance (65%). Forty-five percent had no high school degree. Per design, biological markers of disease progression indicated that participants began this study in relatively early disease stages...
"....findings suggest that psychological resources play an important role in physical health of women with HIV beyond many relevant clinical and sociodemographic factors. Whether this role is direct or indirect is an area for future inquiry... The majority (58%; n = 429) of women reported substantial depressive symptoms.... odds of dying was 3.5 times higher in women with no versus three psychological resources.... Women with more psychological resources had slower rates of decline in CD4+ cell count...
As number of psychological resources increased from 0 to 3, HIV-related mortality decreased in a 'dose response manner': 17, 14, 10, and 6% respectively (Fig. 1a). Among women with baseline CD4+ cell counts < 200 cells/μl: 50% of those with no psychological resources died compared with only 16% of those with three resources (figure 1b).....
... Women with HIV are generally more vulnerable because they are disproportionately from racial/ethnic minority groups and face multiple socio-economic stressors [30]. This social vulnerability may diminish the extent to which women with HIV have and can utilize psychological resources. Findings have implications for understanding individual variability in HIV disease progression. Moreover, because psychological resources are potentially amenable to intervention, results can be applied to improve the health of women with HIV....
.... HIV disease progression is strongly driven by pathophysiological processes associated with HIV, including viral load and CD4+ cell count. Results of this study suggest that psychological processes also contribute independently to disease progression. Psychological resources appear to have protective effect against HIV-related mortality and immune decline. This effect was independent controlling for clinical/sociodemographic characteristics associated with HIV-related morbidity and mortality, and was notable for those with low baseline CD4+ cell count (< 200 cells/μl) where 50% of those with no psychological resources died. The consistency and dose-response effect of these results when examining survival rates and CD4+ trajectory lend confidence to these conclusions. Women with the greatest diversity of psychological resources (i.e., three resources) had greater longevity and slower immune decline...."
ABSTRACT
Objective: Guided by Cognitive Adaptation Theory, the aim was to determine whether psychological resources (positive affect, positive expectancy regarding health outcomes, finding meaning in challenging circumstances) protect against HIV-related mortality and decline in CD4 lymphocyte counts among women with HIV.
Design: The HIV Epidemiologic Research Study, a longitudinal prospective cohort study, with semi-annual interview, physical examination and laboratory assays.
Methods: A total of 773 HIV-seropositive women aged 16 to 55 years were recruited from four academic medical centers in Baltimore, Maryland; Bronx, New York; Providence, Rhode Island; and Detroit, Michigan. Semi-annually for up to 5 years, the women were interviewed, underwent physical examination, medical record abstraction, and venipuncture. Primary outcomes for these analyses included HIV-related mortality and CD4 cell count slope decline over 5 years.
Results: Psychological resources were inversely associated with HIV-related mortality and time to death, beyond the effects of potential confounding variables such as clinical status (e.g., HIV viral load, symptoms, antiretroviral therapy), sociodemographic characteristics (e.g. age, race), and depression at study entry (P < 0.05). Psychological resources also were inversely associated with CD4+ cell count decline (P < 0.01), serving as a possible mechanism linking resources to mortality.
Conclusions: Psychological resources may protect against HIV-related mortality and immune system decline. Findings have implications for understanding individual variability in HIV disease progression. Moreover, because psychological resources are potentially amenable to change, results can be applied to clinical interventions aimed at improving the health of women with HIV.
Results
Description of study participants
At baseline, participants ranged in age 19-55 years (mean = 35.5; SD = 6.7); 60% were black, 20% Latina, and 20% white or other race/ethnicity. The majority of women were unemployed (82%), reported monthly incomes below US$ 1000 (72%), and received public assistance (65%). Forty-five percent had no high school degree. Per design, biological markers of disease progression indicated that participants began this study in relatively early disease stages, with one-half asymptomatic, moderate CD4+ cell counts [mean = 430.9; SD = 270.8; median = 376.87; interquartile range (IQR) = 331.06] and HIV viral load (82% < 10 000 copies/ml).
Inter-correlations of psychological resources and depression
At study entry, most respondents had one (33%; n = 257) or more (49%; n = 380) psychological resources, nearly equally distributed among positive affect (mean = 3.00; SD = 0.93), positive expectancy (mean = 2.62; SD = 0.53), and finding meaning (mean = 3.02; SD = 1.57). The majority (58%; n = 429) of women reported substantial depressive symptoms (i.e., CES-D > 16, a priori clinical cutoff). Correlation between psychological resources and depressive symptoms (Spearman rho = -0.24) indicated that individual differences in psychological resources were not simply reflecting differences in depression.
Psychological resources and HIV-related mortality
Over the 5-year study period, the total number of HIV-related deaths was 106 (14.3%). As number of psychological resources increased from 0 to 3, HIV-related mortality decreased in a 'dose response manner': 17, 14, 10, and 6% respectively (linear trend, x2 = 8.93; P = 0.003) (Fig. 1a). Among women with baseline CD4+ cell counts < 200 cells/μl: 50% of those with no psychological resources died compared with only 16% of those with three resources (linear trend, x2 = 5.48; P = 0.02) (Fig. 1b).
Hierarchical regression tested whether psychological resources were predictive of HIV-related mortality beyond potential confounding factors (Table 1). The overall model was significant (model x2 = 210.22; P < 0.0001), with the psychological resources index incrementally predicting HIV-related mortality beyond effects associated with covariates (block x2 = 4.79; P = 0.03). Adjusting for differences in control variables (CD4+ cell count, viral load, antiretroviral use, opportunistic infections prophylaxis, symptoms, antidepressants, age), odds of dying was 3.5 times higher in women with no versus three psychological resources. Participants were enrolled in the study over a 2-year period and followed for 5 years (1993-2000); post-hoc analyses controlled for time of study entry to account for HAART availability, and results were unchanged.
Psychological resources and CD4+ cell count decline
Our next analyses examined whether psychological resources were associated with CD4+ trajectory, as a possible mechanism linking resources to survival. Linear growth model with serially correlated error terms was first fitted to the data; however, indices suggested that this could be improved (x2 = 317.33; df = 49; x2/df = 6.47; NFI = 0.97; RMSEA = 0.08). A quadratic term was added, which provided a better fit to the data (x2 = 145.87; df = 52; x2/df = 2.80; NFI = 0.99; RMSEA = 0.04) (Table 2). This model indicates that, on average, women began the study with CD4+ cell counts of 429.3 cells/μl. CD4+ cell count decreased at each assessment by -31.8 cells/μl; the quadratic effect indicates that rate of decline diminished over the study by 2.6*time2.
Significant beta values in the final model indicate that psychological resources were associated with CD4+ trajectory parameters with all other factors held constant (Table 2). Only clinical factors (e.g., viral load, medication use) and psychological resources were associated with CD4+ trajectory parameters. Covariates accounted for 37% of variability in intercept, 8% of variability in slope, and 13% of variability in quadratic factor. Beyond covariates, psychological resources did not account for additional intercept variance (i.e., women entered the study at approximately the same CD4+ cell count), but did account for an additional 2% of variability in slope and 3% of variability in quadratic effect. In other words, controlling for clinical and sociodemographic characteristics, differences in baseline psychological resources significantly predicted extent of CD4+ cell count decline over time. Women with more psychological resources had slower rates of decline in CD4+ cell count, as indicated by significant association between psychological resources and slope (β = -0.18) and quadratic (β = 0.20) factors of CD4+ trajectories (Table 2). The final model provided a good fit (x2 = 243.3; df = 103; NFI = 0.97; RMSEA = 0.04). Figure 2 illustrates estimated average CD4+ trajectory stratified by baseline psychological resources: women with more resources had slower CD4+ cell count decline over 5 years, controlling for relevant clinical factors (P < 0.20 in univariable associations, HIV viral load, HAART, and depressive symptoms).
Discussion
HIV disease progression is strongly driven by pathophysiological processes associated with HIV, including viral load and CD4+ cell count. Results of this study suggest that psychological processes also contribute independently to disease progression. Psychological resources appear to have protective effect against HIV-related mortality and immune decline. This effect was independent controlling for clinical/sociodemographic characteristics associated with HIV-related morbidity and mortality, and was notable for those with low baseline CD4+ cell count (< 200 cells/μl) where 50% of those with no psychological resources died. The consistency and dose-response effect of these results when examining survival rates and CD4+ trajectory lend confidence to these conclusions. Women with the greatest diversity of psychological resources (i.e., three resources) had greater longevity and slower immune decline.
Psychological resources may have direct effects on immunity through partial preservation of CD4+ lymphocytes, expected to decline over time among persons with HIV. This association may operate through 'muting' adverse effects of stress [45], which result in immunosuppression. Although exact mechanisms are not fully elucidated, empirical evidence suggests that stress downregulates immunity by: (a) activating autonomic nervous system fibers that descend from brain to lymphoid organs; (b) triggering secretion of hormones and neuropeptides that bind to white blood cells; and, (c) inducing immunomodulatory coping behaviors such as nicotine and alcohol consumption [46]. For women already immunocompromised with HIV, any physiological or environmental stress could heighten the risk for more rapid disease progression. Psychological resources may also have correlates that influence viral replication, resistance, treatment failure, morbidity and mortality [47].
This study was limited by failure to measure factors that might explain the association between psychological resources and HIV-related mortality such as medication adherence. We could not examine adherence for the entire sample because one-third received no antiretroviral treatment. Of those never on HAART, 82% had CD4+ cell counts of > 200 cells/μl and viral load < 10 000 copies/ml throughout the study, so there was probably no clinical indication. Post-hoc analyses of 282 women on two or more antiretrovirals after widespread availability (June-September 1999) and enrolled in an adherence sub-study indicated that psychological resources were not associated with adherence (skipped dose yesterday; number of doses skipped in last 3 days; frequency of drug holidays in last 6 months). Measures of association between adherence and the psychological resources index were all non-significant (P = 0.45 to 0.94). Adherence measures were limited to self-report; however, these were related to clinical outcomes such as viral failure [48-51]. Women in this cohort [52] and others from studies with similar socio-economically-vulnerable populations [53-55] indicate that overall adherence is variable and generally low.
Although not examined, we acknowledge potential indirect effects. For example, psychological resources could contribute to increased adherence, which could affect immunity and mortality. Psychological resources also could contribute to differences in other health behaviors that impact immunity, such as exercise, diet, and consistency of healthcare. The results do not eliminate the possibility of mediating factors; however, findings suggest that psychological resources play an important role in physical health of women with HIV beyond many relevant clinical and sociodemographic factors. Whether this role is direct or indirect is an area for future inquiry. Finally, although it is impossible to rule out reverse causation, we protect against many problems by being strict with temporal ordering and exposure definition. Further, post hoc testing indicated that there were no differences in baseline CD4+ cell count, viral load, and symptoms as a function of psychological resources; rather, the impact of psychological resources may be gradual across disease progression.
Baseline data were collected 1993-1995; although much has changed with regard to HIV epidemiology and treatment, we have no reason to believe that the association between psychological resources and health outcomes would vary over time. Our measurements appear to be sensitive, reflecting upswing in CD4+ cell counts after HAART availability. Limiting mortality analyses to HIV-related deaths that occurred 1996 or later (after HAART availability), significant linear trend by psychological resources was maintained despite fewer deaths (n = 44): 29% of those with no psychological resources died during the follow-up period, compared with only 4% of those with three psychological resources (linear trend, x2 = 5.40, P = 0.02). The cohort was healthy, based on study criteria that excluded women with previous AIDS diagnosis or opportunistic infections; the study should be replicated with women with more advanced HIV. Finally, although variance explained is moderate, maximum overall difference in CD4+ cell count between those with zero versus three resources was 61.3 cells/μl, a difference of > 14% based on CD4+ cell count at study entry.
There is no standardized approach to measure psychological resources, and measurement validity must be confirmed. Despite crude categorical condensations of data, association between psychological measures and disease progression were consistent. We included only three of many potential resources that people utilize when facing serious illness. Were other resources included (e.g., coping, support), effects of psychological resources on health would likely be strengthened. Identifying that psychosocial variations account for even a small proportion of variability in survival yields information about mechanisms of immunity and susceptibility to HIV-related illness. It is unclear whether clinical interventions could be designed to meaningfully enhance psychological resources. Fostering psychological resources either innate or acquired may affect treatment response and promote resilience to clinical setbacks likely as HIV disease progresses. The critical question is under what conditions, for which patients, might we have the capacity to enhance psychological resources and subsequently immune alterations and ultimately longevity [56]? One needs to tread carefully, however, to insure that patients with HIV do not perceive self-blame because of natural illness progression [57,58].
Positive psychosocial outcomes have been reported in response to diverse stressful health events, including cancer, transplantation, cardiovascular and autoimmune diseases [21]. Determining a causal relationship between psychological resources and HIV disease progression is complex because of the multi-factorial nature of HIV disease pathology. There are many biomedical and psychosocial factors that may result in more rapid disease progression or protect against expected decline. It is not a question of mind over matter, but it is important to recognize that mind does matter [59].
Introduction
Physician-scientist Sir William Osler (1849-1919) wrote: 'It is much more important to know what sort of patient has the disease than what sort of disease the patient has' - implying that individual characteristics can influence course and outcome of disease [1]. Longevity has been empirically associated with genetic influences [2,3]. However, non-genetic attributes, including psychological and behavioral factors, account for up to 80% of heterogeneity in longevity [2,4].
Cognitive Adaptation Theory suggests that psychological resources that enable individuals to deal effectively with major illness may result in better longevity and positive health outcomes [5-7]. Psychological resources include: (a) positive emotional disposition; (b) positive expectancy regarding health outcomes; and, (c) ability to find meaning in challenging circumstances. Although inter-related, these resources have distinct and cumulative health effects [6].
As far back as Hippocrates (460-400 BC), clinicians/investigators have posited that positive emotions are one psychological resource that has direct beneficial effects on health via immune function and indirect effects via health behaviors, elicitation of social support, and diminished perceptions of physical symptoms [8]. Second, positive expectancy, has been associated with better function and longer survival among patients with cardiac-related events and procedures [9-12], cancer [13,14], Parkinson's disease [15], and among men with HIV [16,17]. Positive expectancy appears to be salutary, even if unrealistic, given impairment associated with serious illness. Third, ability to find meaning despite threatening events, appears protective against adverse consequences of illness-related stressors [18-20]. Although it may seem counterintuitive, 83% of women with breast cancer found meaning from experience with disease (e.g., reassessment of priorities) [21]. Among a small sample of men with HIV experiencing bereavement, those who found meaning had less rapid decline in CD4+ lymphocytes and lower rates of AIDS-related death independent of health status at baseline, health behaviors, and other potential confounders [22].
Among individuals with HIV, considerable variability in disease progression exists that cannot be explained entirely by initial health status, biology, or medication [23]. Scientists have investigated the association of psychological factors to HIV-related mortality and disease progression. Depression has been linked to CD4+ lymphocyte decline in several studies [24,25], but unrelated in others [26,27]. Treatment advances have focused exclusively on the important goal of interrupting HIV pathogenesis. Relatively little attention has been paid to processes that could plausibly affect somatic resistance to HIV progression. To slow progression, we must understand many determinants of health and immune function for persons with HIV, including biomedical and psychosocial factors.
The primary objective of this study was to determine whether psychological resources protect against HIV-related mortality and decline in CD4+ cell counts among women with HIV. Using the HIV Epidemiologic Research Study (HERS), a large prospective cohort study of biological/psychosocial manifestations of HIV in women followed longitudinally [28], we hypothesized that women with more psychological resources would have lower mortality and less immune decline.
This study expands existing research by testing a model, derived from Cognitive Adaptation Theory, to examine whether psychological resources account for variability in HIV-related outcomes. Longitudinal studies linking psychological resources to biological markers of disease are scarce, and those among individuals with HIV include only men [29]. Women with HIV are generally more vulnerable because they are disproportionately from racial/ethnic minority groups and face multiple socio-economic stressors [30]. This social vulnerability may diminish the extent to which women with HIV have and can utilize psychological resources. Findings have implications for understanding individual variability in HIV disease progression. Moreover, because psychological resources are potentially amenable to intervention, results can be applied to improve the health of women with HIV.
Methods
Participants
Methods for the HIV Epidemiologic Research Study (HERS) have been described in detail [24,28]. The cohort included 871 women who were HIV seropositive recruited from clinical sites in four US cities: Baltimore, Maryland; Bronx, New York; Providence Rhode Island; and Detroit, Michigan. [Note: there were also 439 HIV-seronegative women recruited as a comparison group in the HERS, but the current analyses were restricted to the seropositive cohort only, given the focus on response to disease.] Eligibility included age 16-55 years, fluent in English or Spanish, documented HIV status within 60 days, and reported one or more HIV-related risk behavior. Women with an AIDS diagnosis or opportunistic infections (1987 Centers for Disease Control surveillance definition) were considered late in disease progression and were ineligible. The final sample for these analyses included 773 women; 98 women were excluded because they completed baseline assessment only (n = 72, 8.3%) or lacked data on psychological resources (n = 26, 3%). The excluded women had poorer physical health, with lower baseline CD4+ cell counts (P < 0.05) and greater likelihood of death (P < 0.001). Other measured characteristics were statistically similar.
Procedures
Recruitment was conducted 1993-1995; follow-up visits completed March 2000. Semi-annually, women were interviewed, underwent physical examination, medical record abstraction, and venipuncture. Serum plasma and cells were shipped for centralized processing and storage. For these analyses, we used up to 5 years of data, thus the total number of possible visits for each participant was 10 (mean = 7.42; SD = 2.50; median = 7; range, 2-10). Very few participants were lost to follow-up (n = 91; 12%) or withdrew (n = 21; 3%). Participation was voluntary, confidential, and did not influence provision of care. Procedures were approved by institutional review boards at each site and the Centers for Disease Control.
Measures
Primary outcomes
HIV-related mortality. HIV-related mortality was determined from systematic review of medical records, National Death Index, and individual death certificates using standard nosology. Participants who died of other causes (n = 79) were included, categorized as non-HIV-related death, and censored at time of death.
CD4+ cell count. Flow cytometry from whole blood was used to determine CD4+ cell counts at each assessment. All sites passed periodic proficiency panels [31].
Predictor variables
To protect against reverse causation, psychological resources, clinical and sociodemographic variables were assessed at the baseline interview, with the exception of medication use. Because medication use has a direct and profound effect on HIV-related disease progression, HAART and anti-depressant use were included in all models based on use over the study period.
Psychological resources (Appendix 1)
Positive Affect. Participants were asked about frequency of six positive emotions in the last 6 months (e.g., hopeful), adapted from the Profile of Mood States. Responses were on a Likert-type scale (1 = not at all to 5 = a lot). Averaged scores created a positive affect measure (Cronbach alpha = 0.88).
Positive HIV expectancy. Seven items assessing expectations about disease and sense of control were developed for this study. Responses were on a Likert-type scale (1 = strongly disagree to 4 = strongly agree). Higher average scores reflected more positive HIV expectancy (Cronbach alpha = 0.78).
Finding meaning. Five questions about positive life changes women attributed to HIV were developed (e.g., more time with family) (yes/no). Higher average scores reflected more positive changes attributed to HIV (Cronbach alpha = 0.72).
Psychological resources index. Principal component analysis determined whether items comprising the three measures assessed separate constructs. Examination of the scree plot indicated three factors with eigenvalues > 1.0. Items loaded on the expected factor, and accounted for 67% of total variance. Spearman rho correlations (0.11-0.24) suggested non-redundant measurement. To create a simple index, median splits based on scores for the cohort were performed on each psychological resource. Study participants received '1' each time they scored at/above median for each resource (range, 0-3).
Clinical/sociodemographic covariates
HIV viral load. Baseline HIV RNA viral load quantification was performed using branched-DNA signal amplification assay (Chiron, Emeryville, California, USA): categorized as < 500, 500-9999, or ≥ 10 000 copies/μl.
HIV-related constitutional symptoms. Baseline symptoms were documented: oral thrush; diarrhea; fever > 100 F; problems concentrating/remembering; numbness, tingling, or burning sensations in extremities; unexpected weight loss > 10 pounds lasting > 1 month. Unless specified, symptoms lasted >2 weeks and were not due to other, identifiable cause.
Highly active antiretroviral therapy. Participants were hierarchically classified based on percentage of study visits using HAART. Participants reporting no HAART were assigned 0. Participants reporting HAART at 1-32%, 33-66%, or > 66% of study visits were assigned values 1, 2, or 3, respectively. [Note: Protease inhibitors became available in December 1995 after initial recruitment was completed. This measure provides a crude but integrated indicator of HAART use across the entire study period. Post-hoc analyses also were conducted using the raw number of study visits on HAART, and results were unchanged. There was no systematic measure of antiretroviral adherence.]
Prophylaxsis for opportunistic infections. Because data collection began before widespread HAART availability, the use of opportunistic infection prophylaxis (i.e., Bactrim, Hoffmann, LaRoche, Nutley, New Jersey, USA; Pentamidine, US brand-Pentam 300, American Pharmaceutical Partners, Schaumburg, Illinois, USA) during study period was included (yes/no).
Illicit drug use. Participants reported non-injection crack/cocaine and any injection drugs (yes/no) in past 6 months, summed for a general indicator of recent drug use (0-2).
Alcohol use. Participants reported frequency of alcohol consumption in past 6 months, categorized as heavy (≥ 4 days/week), frequent (1-3 days/week) occasional (< 1 day per/week), or non-users.
Cigarette use. Participants reported frequency of cigarettes, categorized as > 1 pack/day, pack or less/day, or no smoking.
Depressive symptoms. Center for Epidemiological Studies-Depression (CES-D) was designed to measure major components of depression [32]. At baseline, respondents rate 20 symptoms (e.g., appetite change, hopelessness) over last 7 days, with responses from 0 (rarely/none of the time) to 3 (most of the time). CES-D has been used extensively in studies of the population [32-34], clinical samples [32,33,35], and persons with HIV [26,36]. Approximately 20% of the general population have scores > 16 and are classified as 'probable cases of depression' [32,34,37].
Antidepressant medication. Participants were dichotomized based on self-reports of prescribed antidepressants during study (yes/no).
Sociodemographic characteristics. Baseline demographic characteristics were assessed via questionnaire: age, race, marital status, education, employment, living conditions, and financial sources. Socio-economic risk was computed by summing frequency of four indicators of disadvantage: no high school degree, unemployed, public assistance, and monthly household income < US$ 1000.
Analytic plan
Hierarchical logistic regression was used to determine if baseline psychological resources predict the likelihood of a participant dying from HIV-related causes over 5 years, beyond effect of clinical factors, sociodemographic factors, and baseline depression [38]. The association between clinical/sociodemographic variables and mortality were first tested at the univariable level. Variables significant P ≦ 0.20 were maintained in the final model as potential confounders [39,40].
The structural equation modeling (SEM) approach to growth curve modeling (LGM) was used to examine the impact of psychological resources on CD4+ trajectories [41,42]. Observed variables (CD4+ from ten timepoints) were used as indicators of two latent factors: intercept and slope. Intercept reflects average level of initial CD4+ cell counts; slope reflects average change in CD4+ cell counts over the study period. Variance reflects individual differences in initial (intercept variance) and change in CD4+ cell counts (slope variance). If there is significant variability in slope, one can determine individual characteristics such as psychological resources that are predictive of different CD4+ trajectories. To predict trajectories, clinical and sociodemographic variables were first tested individually to determine which should be maintained in the final model as potential confounders (P ≦ 0.20) [39,40]. Then, the psychological resources index was added. Of interest are multiple squared correlations (R 2) and incremental multiple squared correlations (ΔR 2) at each block for the slope factor, and individual beta weights of predictor variables in the final model.
AMOS (SPSS Inc., Chicago, Illinois, USA) was used for analyses with full-information maximum likelihood methods of estimation to generate estimates based on available data, such that individuals with missing data were maintained with values at each missing timepoint estimated based on all available data using their existing trajectory. This approach provides largely unbiased parameter estimates in simulation studies of the effect of missing longitudinal data [43]. Indices of fit used to evaluate the unconditional and final model include the normed fit index (NFI, > 0.95 considered good fit) and root mean square error of approximation (RMSEA, < 0.05 considered good fit) [44].
|
|
| |
| |
|
|
|
