Early Epidemic Comparisons

[mmcleod89]

Direct comparisons between Python and SAS
From 1989 to 1995

• Population demographics: age / sex
• Non HIV death rates
• Number of short term partners (by age and sex)
• Partner balance

[pineapple-cat]

Hi @UCL/hiv-modelling, here are some early epidemic comparison graphs of mean averages between 25 HIVpy runs (population size: 100k) and the 100 core SAS runs kindly provided by Jenny.

Population

• Population (15-49) → s_alive1549
• Population (15-49, male) → s_alive1549_m
• Population (15-49, female) → s_alive1549_w

Deaths

• Deaths (tot) → s_dead_all
• Non-HIV deaths (tot) → s_dead_all - s_death_hiv
• Non-HIV deaths (ratio) → (s_dead_all - s_death_hiv) / (s_alive_m + s_alive_w)

Short Term Partners

• At least 1 short term partner (ratio) → s_newp_ge1 / (s_alive_m + s_alive_w)
• Short term partners (15-49, male) → (s_m_1524_newp + s_m_2534_newp + s_m_3544_newp) / s_alive1549_m
• Short term partners (15-49, female) → (s_w_1524_newp + s_w_2534_newp + s_w_3544_newp) / s_alive1549_w

Partner Balance

@mmcleod89 is currently in the process of making improvements to the partner balance regulation code so we don't need to take too much stock of these graphs, but having them here for later comparison won't hurt.

• Partner sex balance (15-24, male) → log(m15r, 10)
• Partner sex balance (15-24, female) → log(w15r, 10)
• Partner sex balance (25-34, male) → log(m25r, 10)
• Partner sex balance (25-34, female) → log(w25r, 10)
• Partner sex balance (35-44, male) → log(m35r, 10)
• Partner sex balance (35-44, female) → log(w35r, 10)

[andrew-phillips-1]

Thanks @pineapple-cat Good to start to compare these graphically. For the earlier graphs I would suggest having the y axis start at 0 to enable comparison visually.
I'm afraid I won't be on the call this coming Tuesday but hopefully you all can go ahead and discuss them.

[ValentinaCambiano]

Hi Emily, I’ve just noticed a couple of things, highlighted below From: Emily Dubrovska ***@***.***> Sent: Friday, February 16, 2024 8:33 PM To: UCL/hivpy ***@***.***> Cc: Cambiano, Valentina ***@***.***>; Team mention ***@***.***> Subject: Re: [UCL/hivpy] Early Epidemic Comparisons (Issue #175) ⚠ Caution: External sender Hi @UCL/hiv-modelling<https://github.com/orgs/UCL/teams/hiv-modelling>, here are some early epidemic comparison graphs of mean averages between 25 HIVpy runs (population size: 100k) and the 100 core SAS runs kindly provided by Jenny. Population * Population (15-49) → s_alive1549 * Population (15-49, male) → s_alive1549_m * Population (15-49, female) → s_alive1549_w Comparison.of.Population.15-49.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/15e2c5e7-d907-48b8-abff-1142d2fe4a89> Comparison.of.Population.15-49.Male.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/4c4b42d7-baf6-42ed-859a-259d0bfb4477> Comparison.of.Population.15-49.Female.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/0b16e4e3-00a8-47eb-9f3f-fcaaf9f08553> Deaths * Deaths (tot) → s_dead_all * Non-HIV deaths (tot) → s_dead_all - s_death_hiv * Non-HIV deaths (ratio) → (s_dead_all - s_death_hiv) / (s_alive_m + s_alive_w) Should this not be the total number of deaths? Or are you trying to calculate the death rate? Comparison.of.Deaths.Tot.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/c0298620-4ceb-4871-ac3c-6b72420a95fa> Comparison.of.Non-Hiv.Deaths.Tot.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/92445699-233c-4672-b333-549c0fecc2ab> Comparison.of.Non-Hiv.Deaths.Ratio.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/a2f2b636-1606-43ab-9097-18f5e0905f3e> Short Term Partners * At least 1 short term partner (ratio) → s_newp_ge1 / (s_alive_m + s_alive_w) * Short term partners (15-49, male) → (s_m_1524_newp + s_m_2534_newp + s_m_3544_newp) / s_alive1549_m * Short term partners (15-49, female) → (s_w_1524_newp + s_w_2534_newp + s_w_3544_newp) / s_alive1549_w Note that only people 15-64 have condomless sex partner, I think s_alive_m and s_alive_w are for the same age group. It seems like you are missing s_m45_49 from the numerator Comparison.of.At.Least.1.Short.Term.Partner.Ratio.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/3c46ddb1-5781-4367-bf31-b6f93678af7a> Comparison.of.Short.Term.Partners.15-49.Male.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/faa1856e-f2e6-457b-8c51-b9b4e88d6373> Comparison.of.Short.Term.Partners.15-49.Female.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/a7a39165-1652-4664-bc9e-8621665e1363> Partner Balance @mmcleod89<https://github.com/mmcleod89> is currently in the process of making improvements to the partner balance regulation code so we don't need to take too much stock of these graphs, but having them here for later comparison won't hurt. * Partner sex balance (15-24, male) → log(m15r, 10) * Partner sex balance (15-24, female) → log(w15r, 10) * Partner sex balance (25-34, male) → log(m25r, 10) * Partner sex balance (25-34, female) → log(w25r, 10) * Partner sex balance (35-44, male) → log(m35r, 10) * Partner sex balance (35-44, female) → log(w35r, 10) Comparison.of.Partner.Sex.Balance.15-24.Male.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/9cc67d2b-7904-41a3-90e8-3ea4c4e60369> Comparison.of.Partner.Sex.Balance.15-24.Female.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/17da86e9-ae46-4c76-b2b1-418c5b70cd28> Comparison.of.Partner.Sex.Balance.25-34.Male.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/9dd66384-d22b-40d1-aeee-ce5239ee9d35> Comparison.of.Partner.Sex.Balance.25-34.Female.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/28364db1-264b-40bb-869b-b94cf3f25043> Comparison.of.Partner.Sex.Balance.35-44.Male.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/0c426be2-4626-4e86-81db-a8ee58ec8505> Comparison.of.Partner.Sex.Balance.35-44.Female.Over.Time.png (view on web)<https://github.com/UCL/hivpy/assets/34184448/d2178626-f3b3-4e00-a996-f2f48e20cf04> — Reply to this email directly, view it on GitHub<#175 (comment)>, or unsubscribe<https://github.com/notifications/unsubscribe-auth/AJNUSMVAQA3RNWFQAZEMES3YT666RAVCNFSM6AAAAABCPMCLVGVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNBZGI4TEOBSGU>. You are receiving this because you are on a team that was mentioned.Message ID: ***@***.******@***.***>>

[pineapple-cat]

Hi @ValentinaCambiano, thanks for your comments.

• You are correct that Non-HIV deaths (ratio) is trying to calculate the death rate. I decided to add this metric when I noticed how different the population sizes were between the HIVpy and SAS models to see if that accounted for the increased number of deaths in the SAS runs.
• Short term partners (15-49, male) and Short term partners (15-49, female) are imperfectly calculated from the SAS runs because there are no s_m_4549_newp or s_w_4549_newp columns, there are only s_m_4554_newp or s_w_4554_newp, so I tried to get as close as I could without their inclusion. The bigger problem seems to be that I forgot to exclude sex workers from the HIVpy Short term partners (15-49, female) data, so I'll be sure to add that as a constraint in my next round of plots.