Toward a Prognostic Model for Mortality Risk in Older People Living With HIV: A Prospective Cohort Study From Southwestern China

Abstract

ObjectiveThe existing prognostic models for mortality risk in people living with HIV (PLWH) may not be applicable for older PLWH because the risk factors were confined to biomarkers and clinical variables. We developed and validated a nomogram for the prognosis of all-cause mortality in older PLWH based on comprehensive predictors. DesignProspective cohort study. Setting and ParticipantsWe included 824 participants aged ≥50 years (mean age, 64.0 ± 7.6 years) from 30 study sites in Sichuan, China, and followed up from Nov 2018 to Mar 2021. MethodsData on demographics, biomarkers, and clinical indicators were extracted from the registry; mental and social factors were assessed by a survey. Elastic net was used to select predictors. A nomogram was developed based on Cox proportional hazards regression model to visualize the relative effect size (points) of the selected predictors. The prognostic index (PI) was calculated by summing points of all predictors to quantify mortality risk. ResultsPredictive performance of PI from the nomogram was good, with area under the curve of 0.76 for the training set, and 0.77 for the validation set. Change in CD4 count, virological failure in antiretroviral therapy, and living with comorbidities were robust predictors. Depressive symptoms were an important predictor in men, those aged ≥65 years, and those with time of diagnosis <1 year; low social capital was an additional predictor in people aged <65. Mortality risk increased approximately 10-fold among participants whose PI was in the fourth quartile compared with those in the first quartile (hazard ratio, 9.5; 95% CI, 2.9–31.5). Conclusion and ImplicationsAlthough biological and clinical factors are crucial predictors, mental and social predictors are essential for specific groups. The developed nomogram is useful for identifying risk factors and groups at risk of mortality in older PLWH.