1080 — Identification of Latent Groups using Veterans’ Responses on the HIV Symptom Inventory
Kudel I (Cincinnati VAMC, University of Cincinnati), Tsevat J
(Cincinnati VAMC, University of Cincinnati), Justice A
(VA Connecticut Healthcare System)
Objectives:
The HIV Symptoms Index (HSI) is a 20-item symptom inventory that has been used in research and is being introduced into clinical care. It is unclear whether patients’ responses on this measure are highly individualistic or whether they can be categorized. In this study we tested whether a latent profile analyses can be used to extract distinct, latent subgroups using veterans’ responses on the HSI.
Methods:
Veterans' ratings of their HIV symptoms were dichotomized as non-bothersome vs. bothersome, and then the sample was randomly split into an exploratory (N=1,311) and a validation group (N=1,339). We assessed whether the groups were similar (SPSS 14.0.1) and then performed latent profile analysis (LPA; Mplus 4.2). Model fit was based on 3 criteria: the Bayesian information criterion, the Lo-Mendell-Rubin likelihood ratio test of model fit, and the average posterior probability. We then examined whether groups differed with regard to CD4 counts and common comorbidities (those reported by at least 10% of the sample).
Results:
The exploratory and validation groups were similar with regard to age, race, CD4 count, and SF-12 mental and physical component summary scale scores (P=NS for all comparisons). LPAs on both groups indicated that a 4-class solution was optimal and yielded similar symptom profiles. Of the 4 classes, the optimal functioning class (exploratory sample=50.1%, validation sample=53.2%) reported the lowest proportion of bothersome HIV symptoms (0.0%-7.8%) and the multi-symptom class (exploratory sample=14.3%, validation group=11.1%) reported the highest proportion of bothersome symptoms (42.3%-87.3%). Two intermediate classes reported similar proportions on 17 of the symptoms (0.0%-44.9% and 5.1%-50.6%, respectively), but differed on items associated with depressed mood: a high-distress class (exploratory=11.4%, validation=11.9%) reported high levels of sadness (87.4%), nervousness (75.4%), and poor sleep (66.9%), whereas a low-distress class (exploratory=24.3%, validation=23.8%) reported almost no sadness (8.0%), no nervousness (0.0%), and half the prevalence of bothersome sleep problems (33.2%). Classes differed significantly by CD4 counts (F=114.18, df=3, p < 0.01) and number of self-reported comorbid conditions (F=5.21, df=3, p < 0.01).
Implications:
Future research will determine whether these classification groups are associated with different clinical outcomes including mortality and quality of life.
Impacts:
Four distinct classes of symptom patterns in patients with HIV can be identified by using LPA.
