Risk Factors and Impact of Patterns of Co-Occurring Comorbidities in People Living With HIV

Davide De Francesco; Jonathan Underwood; Emmanouil Bagkeris; Jane Anderson; Ian Williams; Jaime H. Vera; Frank A. Post; Marta Boffito; Margaret Johnson; Patrick W.G. Mallon; Alan Winston; Caroline A. Sabin; on behalf of the Pharmacokinetic and Clinical Observations in PeoPle Over fiftY (POPPY) study


AIDS. 2019;33(12):1871-1880. 

In This Article


Study Participants

The Pharmacokinetic and Clinical Observations in People Over Fifty (POPPY) study is an observational study that aims to examine the effects of ageing on the clinical outcomes of PLWH in the United Kingdom and Ireland. Full details have been described previously.[14] Briefly, two cohorts of PLWH were recruited from eight HIV outpatient clinics in London, Brighton (United Kingdom) and Dublin (Ireland) between April 2013 and January 2016: an 'older' group of PLWH aged at least 50 years and a 'younger' group of PLWH aged between 18 and 50 years. Inclusion criteria were: documented presence of HIV infection, white or black-African ethnicity, likely route of HIV acquisition via sexual exposure and ability to comprehend the study information leaflet. The younger group of PLWH was frequency matched on sex, ethnicity, sexual orientation and location (in or out of London) to the older PLWH. In addition, the study recruited a group of HIV-negative individuals aged at least 50 years which was not included in the present analysis. These inclusion criteria were selected to ensure that study participants were representative of the wider population of PLWH over the age of 50 in the United Kingdom and Ireland, as has previously been shown.[14] The study was approved by the UK National Research Ethics Service (Fulham, London, United Kingdom number 12/LO/1409). All participants provided written informed consent.

Patterns of Comorbidities

We considered a list of 65 individual comorbidities (Supplementary Table 1, http://links.lww.com/QAD/B496) obtained as follows. Full clinical history, medications and healthcare resources used over the year preceding the study visit were obtained via a structured interview with trained staff who, where possible, also reviewed hospital notes to validate the presence of comorbidities. Participants were asked whether they ever experienced any of the comorbidities or medical conditions from a detailed list and to report any other relevant comorbidity not included in the initial list. Answers to these free-text questions, reasons for any healthcare utilization over the previous year, and use of (non-antiretroviral) medication in the previous year were also examined to update the existing list of comorbidities or to include additional ones. Congenital diseases and conditions with a prevalence less than 1.5% in the study population were subsequently excluded.

As previously described,[5] principal component analysis (PCA) with oblimin rotation was applied to the matrix containing the pairwise associations (as measured by the Somers' D) between the 65 comorbidities. Six components (i.e. patterns) were extracted as suggested by very simple structure criterion[15] and comorbidities associated with each pattern (i.e. with a correlation >0.40) were reported. For each participant and each pattern, a severity score for that pattern was obtained using data on the presence/absence of comorbidities and coefficients returned by the PCA (rescaled so that the lowest score is 0). Severity scores were proportional to the number of comorbidities included in the pattern that were present in an individual, with higher scores indicating a greater number of comorbidities characterizing the pattern.

Risk Factors

We considered a range of potential risk factors for each comorbidity pattern, including sociodemographic, lifestyle and HIV-specific characteristics. Age, sex, ethnicity, sexual orientation (MSM or heterosexual), previous/current smoking and alcohol consumption, current recreational drug use (within the 6 months preceding study visit) and history of injection drug use were self-reported by study participants via a structured interview. BMI was also measured at study visit and historical data on HIV-specific parameters were obtained via linkage with the UK CHIC study[16] and the Mater Misericordiae University Hospital Infectious Diseases cohort for participants recruited in Ireland.[17] Years since HIV diagnosis, nadir CD4+ T-cell count and prior AIDS (defined as a previous report of a category C event[18]) were considered as potential risk factors.

Health Outcomes

Five health-related outcomes were considered: patient-reported physical and mental health, number of general practitioner (GP) visits, hospitalization and functional impairment. Physical and mental health summary scores were obtained from the Short Form Health Survey (SF-36) questionnaire.[19] Briefly, scores for eight subscales (physical functioning, physical limitation, emotional limitation, energy/fatigue, emotional well being, social functioning, pain and general health) were derived as recommended by the developers of the SF-36[20] and then standardized into z-scores (with a mean of 0 and a standard deviation of 1) using sex-specific and age-specific means and standard deviations obtained from the 1996 Health Survey for England dataset.[21] Standardized scores were then combined to obtain the two summary scores of physical and mental health and rescaled to have a mean of 50 and a standard deviation of 10.

Functional impairment was assessed using the Lawton IADL scale[22] and defined as impairment in at least one activity of daily living. The number of GP visits and hospitalization (yes or no) in the year preceding study visit were derived from information collected on healthcare resource use.

Statistical Analysis

For each pattern's severity score, a multivariable median regression model was fitted to assess the independent associations with risk factors previously described. Associations of sociodemographic, lifestyle and HIV-specific factors with each pattern's severity score are reported as associated average change [with 95% confidence interval (CI)] in the median severity score.

The independent associations of each pattern severity score and each health-related outcome considered were assessed using a different multivariable model, depending on the type and distribution of the outcome. In each model, all pattern severity scores were added simultaneously to account for the correlations between the scores. For physical and mental health summary scores, linear regression models were used and regression coefficients (β: average increase/decrease in the physical/mental score associated with a one-unit increase in the severity score of a pattern, while holding all other severity scores constant) are reported. Logistic regression was used for hospitalization and functional impairment and odds ratios (ORs) are reported. The number of GP visits was analysed using Poisson regression and incident risk ratios (IRRs) associated with a one-unit increase in severity scores are reported.

All analyses were performed using the statistical software R v3.3.3; (R Core Team, Vienna, Austria) P values less than 0.05 were considered statistically significant.