Overview

To ensure the fidelity of estimated parameters and mathematical functions in PEARL, we compared individual fits between projected outcomes (such as age distribution, CD4 count, prevalence of risk factors and comorbidities) from simulation data and corresponding data from NAACORD (“in-sample” approach) among ART-users in 2009 and among the ART-initiators from 2010–2022.

Furthermore, to ensure the fidelity of model projections, we compared the projected age distribution and prevalence of comorbidities of ART-users in PEARL to the observed data from NA-ACCORD from 2010-2022 that they’re not directly used to inform the model (“out-of-sample” approach).

To account for outliers, we present the median value across 200 simulations. Agents and NAACORD members that initate ART or reengage are included in calculations for that given year. Likewise, agents and members that disengage or die are also counted in that given year.

Initial Population in Year 2009

Age Distribution of Initial Population in Year 2009

The age distribution of the simulated population on ART in year 2009 is shown, modeled via a two-component mixed normal distribution (black line) based on data from NA-ACCORD (blue bars).

Figure 1a: HET Black Female

Figure 1b: HET White Female

Figure 1c: HET Hispanic Female

Figure 1d: HET Black Male

Figure 1e: HET White Male

Figure 1f: HET Hispanic Male

Figure 1g: IDU Black Female

Figure 1h: IDU White Female

Figure 1i: IDU Hispanic Female

Figure 1j: IDU Black Male

Figure 1k: IDU White Male

Figure 1l: IDU Hispanic Male

Figure 1m: MSM Black Male

Figure 1n: MSM White Male

Figure 1o: MSM Hispanic Male

CD4 Count Distribution at ART Initiation of Initial Population in Year 2009

The CD4 count distribution of the simulated population on ART in year 2009 is shown. The black line represents the simulated CD4 count values against NA-ACCORD data (blue bars). CD4 count is drawn from a normal distribution with parameters fit to a linear function of ART initiation year.

Figure 2a: HET Black Female

Figure 2b: HET White Female

Figure 2c: HET Hispanic Female

Figure 2d: HET Black Male

Figure 2e: HET White Male

Figure 2f: HET Hispanic Male

Figure 2g: IDU Black Female

Figure 2h: IDU White Female

Figure 2i: IDU Hispanic Female

Figure 2j: IDU Black Male

Figure 2k: IDU White Male

Figure 2l: IDU Hispanic Male

Figure 2m: MSM Black Male

Figure 2n: MSM White Male

Figure 2o: MSM Hispanic Male

Population Initiating ART from 2010 - 2022

Age Distribution of ART Initiators

The age distribution of the simulated population initiating ART from 2010 - 2022 is shown. The black line represents the simulated age distribution against NA-ACCORD data (blue bars). Age at ART initiation is modeled by a two-component mixed normal distribution with parameters fit to a linear function of ART initiation year.

Figure 3a: HET Black Female

Figure 3b: HET White Female

Figure 3c: HET Hispanic Female

Figure 3d: HET Black Male

Figure 3e: HET White Male

Figure 3f: HET Hispanic Male

Figure 3g: IDU Black Female

Figure 3h: IDU White Female

Figure 3i: IDU Hispanic Female

Figure 3j: IDU Black Male

Figure 3k: IDU White Male

Figure 3l: IDU Hispanic Male

Figure 3m: MSM Black Male

Figure 3n: MSM White Male

Figure 3o: MSM Hispanic Male

CD4 Distribution of ART Initiators

The CD4 count distribution of the simulated population initiating ART from 2010 - 2022 is shown. The black line represents the simulated CD4 count values against NA-ACCORD data (blue bars). CD4 count is drawn from a normal distribution with parameters fit to a linear function of ART initiation year.

Figure 4a: HET Black Female

Figure 4b: HET White Female

Figure 4c: HET Hispanic Female

Figure 4d: HET Black Male

Figure 4e: HET White Male

Figure 4f: HET Hispanic Male

Figure 4g: IDU Black Female

Figure 4h: IDU White Female

Figure 4i: IDU Hispanic Female

Figure 4j: IDU Black Male

Figure 4k: IDU White Male

Figure 4l: IDU Hispanic Male

Figure 4m: MSM Black Male

Figure 4n: MSM White Male

Figure 4o: MSM Hispanic Male

Population Disengaging from Care

Age Distribution of ART Users Disengaging from ART

The age distribution of the simulated population disengaging from ART from 2010 - 2021 is shown. The black line represents the simulated age distribution against NA-ACCORD data (blue bars). Age at disengagement from ART is an emergent property of the model.

Figure 5a: HET Black Female

Figure 5b: HET White Female

Figure 5c: HET Hispanic Female

Figure 5d: HET Black Male

Figure 5e: HET White Male

Figure 5f: HET Hispanic Male

Figure 5g: IDU Black Female

Figure 5h: IDU White Female

Figure 5i: IDU Hispanic Female

Figure 5j: IDU Black Male

Figure 5k: IDU White Male

Figure 5l: IDU Hispanic Male

Figure 5m: MSM Black Male

Figure 5n: MSM White Male

Figure 5o: MSM Hispanic Male

CD4 Count Distribution of ART Users Disengaging from ART

The cd4 distribution of the simulated population disengaging from ART from 2010 - 2021 is shown. The black line represents the simulated cd4 distribution against NA-ACCORD data (blue bars). cd4 at disengagement from ART is an emergent property of the model.

Figure 5a: HET Black Female

Figure 5b: HET White Female

Figure 5c: HET Hispanic Female

Figure 5d: HET Black Male

Figure 5e: HET White Male

Figure 5f: HET Hispanic Male

Figure 5g: IDU Black Female

Figure 5h: IDU White Female

Figure 5i: IDU Hispanic Female

Figure 5j: IDU Black Male

Figure 5k: IDU White Male

Figure 5l: IDU Hispanic Male

Figure 5m: MSM Black Male

Figure 5n: MSM White Male

Figure 5o: MSM Hispanic Male

Mortality Rate of ART Users

The mortality rate of the simulated population using ART from 2010 - 2020 is shown. The blue line represents the simulated mortality rate against NA-ACCORD data (black dots). Mortality rate is an emergent property of the model.

Figure 6a: HET Black Female

Figure 6b: HET White Female

Figure 6c: HET Hispanic Female

Figure 6d: HET Black Male

Figure 6e: HET White Male

Figure 6f: HET Hispanic Male

Figure 6g: IDU Black Female

Figure 6h: IDU White Female

Figure 6i: IDU Hispanic Female

Figure 6j: IDU Black Male

Figure 6k: IDU White Male

Figure 6l: IDU Hispanic Male

Figure 6m: MSM Black Male

Figure 6n: MSM White Male

Figure 6o: MSM Hispanic Male

Collapsed Mortality Rates

Due to limitations in subgroup sample size, some groups were combined for final analysis. The collapsed group mortality rates are shown below:

Figure 6p: HET White Female + Het Hisp Female

Figure 6q: HET White Male + HET Hisp Male

Figure 6r: IDU Female

Figure 6s: IDU White Male + IDU Hisp Male

Figure 6t: MSM White Male + MSM Hisp Male

Age Distribution of ART Users

The age distribution of the simulated population using ART from 2010 - 2017 is shown. The black line represents the simulated age distribution against NA-ACCORD data (blue bars). Age distribution of ART users is an emergent property of the model.

Figure 7a: Black HET Female

Figure 7b: Black HET Male

Figure 7c: Hispanic HET Female

Figure 7d: Hispanic HET Male

Figure 7e: White HET Female

Figure 7f: White HET Male

Figure 7g: Black IDU Female

Figure 7h: Black IDU Male

Figure 7i: Hispanic IDU Female

Figure 7j: Hispanic IDU Male

Figure 7k: White IDU Female

Figure 7l: White IDU Male

Figure 7m: Black MSM

Figure 7n: Hispanic MSM

Figure 7o: White MSM

Reengagement with HIV Care

The number of years spent out of care following disengagement for the simulated population is modeling by a normalized Poisson distribution. The probability of spending a certain number of years out of care was fit such that the probability of staying disengaged for more than 12 years was zero. The resulting fit is shown below with the blue line representing the probability of being assigned a number of years out of care and the black dots represented the observed NAACORD data.

Figure 8: Number of Years Spent Out of Care

CD4 at Reengagement

The distribution of CD4 count at reengagement of the simulated population using ART from 2011 - 2021 is shown. The black line represents the simulated CD4 count distribution against NA-ACCORD data (blue bars). CD4 Count distribution of ART users is an emergent property of the model.

Figure 9a: Black HET Female

Figure 9b: Black HET Male

Figure 9c: Hispanic HET Female

Figure 9d: Hispanic HET Male

Figure 9e: White HET Female

Figure 9f: White HET Male

Figure 9g: Black IDU Female

Figure 9h: Black IDU Male

Figure 9i: Hispanic IDU Female

Figure 9j: Hispanic IDU Male

Figure 9k: White IDU Female

Figure 9l: White IDU Male

Figure 9m: Black MSM

Figure 9n: Hispanic MSM

Figure 9o: White MSM

Comorbidities

The PEARL model allows for subgroups specific projection of future comorbidity burden. To ensure that our model aligns with real-world data, we compare the comorbidity incidence, prevalence, and prevalence at initiation of ART users across all modeled comorbidities.

Incidence

Below we present the incidence comparison of ART users between simulated comorbidities from the PEARL model and comorbity data from the NAACORD. For each comorbidity plot, we show all subgroups and their alignment with acceptable deviation defined by the maximum of 95% confidence interval and 0.05 absolute value. The fit is defined as the number of years that fall within acceptable deviation and is listed below the subgroup title for each plot. Subgroups that have more than 25% of years outside deviation are shaded red. Due to sample size constraints, some groups are combined in the PEARL model. In these cases, the individual subgroup plots are presented in addition to the combined subgroup plots. The rules for collapsing are shown in the legend to the right.

Figure 10a: Anxiety

Figure 10b: Chronic Kidney Disease

Figure 10c: Diabetes

Figure 10d: Depression

Figure 10e: ESLD

Figure 10f: Hypertension

Figure 10g: Hypolipidemia

Figure 10h: Cancer

Figure 10i: MI

Prevalence

Below we present the prevalence comparison of ART users between simulated comorbidities from the PEARL model and comorbity data from the NAACORD. For each comorbidity plot, we show all subgroups and their alignment with acceptable deviation defined by the maximum of 95% confidence interval and 0.05 absolute value. The fit is defined as the number of years that fall within acceptable deviation and is listed below the subgroup title for each plot. Subgroups that have more than 25% of years outside deviation are shaded red. Due to sample size constraints, some groups are combined in the PEARL model. In these cases, the individual subgroup plots are presented in addition to the combined subgroup plots. The rules for collapsing are shown in the legend to the right.

Figure 11a: Anxiety

Figure 11b: Chronic Kidney Disease

Figure 11c: Diabetes

Figure 11d: Depression

Figure 11e: ESLD

Figure 11f: Hypertension

Figure 11g: Hypolipidemia

Figure 11h: Cancer

Figure 11i: MI

Prevalence at ART Initiation

Below we present the prevalence comparison of ART users at initiation between simulated comorbidities from the PEARL model and comorbity data from the NAACORD. For each comorbidity plot, we show all subgroups and their alignment with acceptable deviation defined by the maximum of 95% confidence interval and 0.05 absolute value. The fit is defined as the number of years that fall within acceptable deviation and is listed below the subgroup title for each plot. Subgroups that have more than 25% of years outside deviation are shaded red. Due to sample size constraints, some groups are combined in the PEARL model. In these cases, the individual subgroup plots are presented in addition to the combined subgroup plots. The rules for collapsing are shown in the legend to the right.

Figure 12a: Anxiety

Figure 12b: Chronic Kidney Disease

Figure 12c: Diabetes

Figure 12d: Depression

Figure 12e: ESLD

Figure 12f: Hypertension

Figure 12g: Hypolipidemia

Figure 12h: Cancer

Figure 12i: MI