Matching Algorithms for Blood Donation Donor blood is a scarce - - PowerPoint PPT Presentation

▶
matching algorithms for blood donation
SMART_READER_LITE
LIVE PREVIEW

Matching Algorithms for Blood Donation Donor blood is a scarce - - PowerPoint PPT Presentation

Apr 28, 2023 128 likes •178 views

Matching Algorithms for Blood Donation (EC20 Poster) Matching Algorithms for Blood Donation Donor blood is a scarce resource: every 2 seconds, someone in the US needs blood shortages especially impact developing countries, and in

slide-1
SLIDE 1

Donor blood is a scarce resource:

  • every 2 seconds, someone in the US needs blood
  • shortages especially impact developing countries, and in

particular, children and women facing complications during childbirth

  • donation rates are correlated with a country’s wealth;

high-income countries have median donation rate of 31.5 donations per 1000 people, while low-income countries have a median of 5.0 donations per 1000 people.

1

Statistics and quotes from the American Red Cross (ARC) and the World Health Organization (WHO)

* : median donation rates

Matching Algorithms for Blood Donation

Matching Algorithms for Blood Donation (EC20 Poster)

slide-2
SLIDE 2

Donor Coordination & Recruitment

2

Donation opportunities (“recipients”) Web app

  • Browse opportunities
  • receive notifications
  • State a need
  • Reach out to donors

Notify donors using a notification policy Potential blood donors

In our setting, blood donors and recipients use a web application to connect with one another. The web app can send notifications to donors, about a particular donation

  • pportunity. These notifications are sent

automatically, by a notification policy.

Blood donors can find donation opportunities, and can choose to receive notifications about opportunities Blood recipients can state their need & availability

  • Individuals
  • Hospitals
  • Blood drives

* : As of June 2020 https://socialgood.fb.com/health/blood-donations/

Taking the perspective of the web app, we study the Facebook Blood Donation Tool, which connects donors with opportunities to donate in several countries around the world, with ~70 million registered donors.*

Matching Algorithms for Blood Donation (EC20 Poster)

slide-3
SLIDE 3

Donors

  • offline, but rate limited: donors can only be

matched once every K days

u ∈ U

<latexit sha1_base64="3d9oLNhISrRhdgH1oiYT/wRCxc=">AB/nicbVDLSsNAFL2pr1pfVZduBovgqiQ+0GXBje4qmLbQhjKZTtqhM5MwMxFKPgLbnXvTtz6K279EidtFtp64MLhnHs5lxMmnGnjul9OaWV1bX2jvFnZ2t7Z3avuH7R0nCpCfRLzWHVCrClnkvqGU47iaJYhJy2w/FN7rcfqdIslg9mktBA4KFkESPYWKmdoh6TyO9Xa27dnQEtE68gNSjQ7Fe/e4OYpIJKQzjWu5iQkyrAwjnE4rvVTBJMxHtKupRILqoNs9u4UnVhlgKJY2ZEGzdTfFxkWk9EaDcFNiO96OXiv14oFpJNdB1kTCapoZLMg6OUIxOjvAs0YIoSwyeWYKY/R2REVaYGNtYxZbiLVawTFpnde+8fnl/UWvcFfWU4QiO4RQ8uIG3EITfCAwhmd4gVfnyXlz3p2P+WrJKW4O4Q+czx/Su5Wx</latexit>

3

Finite time horizon (days): t ∈ {1, . . . , T}

<latexit sha1_base64="pZdwSN3gKoiPRqat3S9aM68rAwE=">ACDnicbVDLSsNAFJ3UV62vqLhyM1gEF6UkPtBlwY3uKvQFTSiT6aQdOpmEmRuhP6Dv+BW9+7Erb/g1i9x2mahrQcuHM65h3s5QSK4Bsf5sgorq2vrG8XN0tb2zu6evX/Q0nGqKGvSWMSqExDNBJesCRwE6ySKkSgQrB2Mbqd+5EpzWPZgHC/IgMJA85JWCkn0E2OMSe5lbwV4/Bl3BDW/Ss8tO1ZkBLxM3J2WUo96zv02YphGTQAXRus6CfgZUcCpYJOSl2qWEDoiA9Y1VJKIaT+bvT/Bp0bp4zBWZiTgmfo7kZFI63EUmM2IwFAvelPxXy+IFi5DeONnXCYpMEnh8NUYIjxtBvc54pREGNDCFXc/I7pkChCwTRYMqW4ixUsk9Z51b2oXj1clmv3eT1FdIxO0Bly0TWqoTtUR01EUYae0Qt6tZ6sN+vd+pivFqw8c4j+wPr8Ab6tmvs=</latexit>

Recipients

  • Both offline recipients (always available), and
  • nline recipients (sometimes available)
  • Probability

that v is available at time t

v ∈ V

<latexit sha1_base64="hla6shafeQgXxdDl61HBdG6wQc=">AB/nicbVC7SgNBFL0bXzG+opY2g0GwCrs+0DJgo10E84BkCbOT2WTIPJaZ2UBYAv6CrfZ2Yuv2PolTpItNHrgwuGcezmXEyWcGev7n15hZXVtfaO4Wdra3tndK+8fNI1KNaENorjS7QgbypmkDcsp+1EUywiTlvR6Gbmt8ZUG6bkg50kNBR4IFnMCLZOao1Rl0nU7JUrftWfA/0lQU4qkKPeK391+4qkgkpLODamE/iJDTOsLSOcTkvd1NAEkxEe0I6jEgtqwmz+7hSdOKWPYqXdSIvm6s+LDAtjJiJymwLboVn2ZuK/XiSWkm18HWZMJqmlkiyC45Qjq9CsC9RnmhLJ45gopn7HZEh1phY1jJlRIsV/CXNM+qwXn18v6iUrvL6ynCERzDKQRwBTW4hTo0gMAInuAZXrxH79V7894XqwUvzmEX/A+vgHV6ZWz</latexit>

pvt ∈ [0, 1]

<latexit sha1_base64="MLWDe9oPZvfUOnOHxSOA1JLZMCw=">AC3icbVDLSsNAFL2pr1pfsS7dDBbBhZTEB7osuNFdBfuANoTJdNoOnUzCzKRYQj/BX3Cre3fi1o9w65c4abPQ1gMXDufcy7mcIOZMacf5sgorq2vrG8XN0tb2zu6evV9uqiRhDZIxCPZDrCinAna0Exz2o4lxWHAaSsY3WR+a0ylYpF40JOYeiEeCNZnBGsj+XY59tOxnqIuE6jnCLXK/l2xak6M6Bl4uakAjnqv3d7UkCanQhGOlOq4Tay/FUjPC6bTUTRSNMRnhAe0YKnBIlZfOfp+iY6P0UD+SZoRGM/X3RYpDpSZhYDZDrIdq0cvEf70gXEjW/WsvZSJONBVkHtxPONIRyopBPSYp0XxiCaSmd8RGWKJiTb1ZaW4ixUsk+Z1T2vXt5fVGp3eT1FOIQjOAEXrqAGt1CHBhB4hGd4gVfryXqz3q2P+WrBym8O4A+szx9Z+Zm</latexit>

Edges (potential recipients)

  • potential notifications, which may depend on

distance or donor/recipient preferences

  • Edge weights

(likelihood of MA)

e = (u, v) ∈ E

<latexit sha1_base64="AwIlAF0i9XI5w4ws2M0GIS6Df4w=">ACBXicbVDLSgMxFL3js9ZX1aWbYBEqSJnxgW6Egi6q2Af0I4lk2ba0CQzJlCGbr2F9zq3p249Tvc+iWmj4W2HrhwOdezuUEMWfauO6Xs7C4tLymlnLrm9sbm3ndnarOkoUoRUS8UjVA6wpZ5JWDOc1mNFsQg4rQW965Ff61OlWSQfzCmvsAdyUJGsLHSI71CheQY9Y+aTKbVi7vFt0x0DzxpiQPU5Rbue9mOyKJoNIQjrVueG5s/BQrwinw2wz0TGpIc7tGpxIJqPx1/PUSHVmjMFJ2pEFj9fdFioXWAxHYTYFNV896I/FfLxAzySa89FMm48RQSbBYcKRidCoEtRmihLDB5Zgopj9HZEuVpgYW1zWluLNVjBPqidF7R4fn+WL91N68nAPhxATy4gBLcQhkqQEDBM7zAq/PkvDnvzsdkdcGZ3uzBHzifP4ACl5w=</latexit>

wet ∈ [0, 1]

<latexit sha1_base64="UuoaHY5PqFitslp+IdpMw6nQM1A=">ACnicbVDLSsNAFJ3UV62vVJduBovgQkriA10W3Oiugn1AG8JketMOnUzCzMRSQv/AX3Cre3fi1p9w65c4bPQ1gMXDufcy7mcIOFMacf5sgorq2vrG8XN0tb2zu6eXd5vqjiVFBo05rFsB0QBZwIamkO7UQCiQIOrWB4M/VbjyAVi8WDHifgRaQvWMgo0Uby7fLIz0BPcJcJ3HFOsev5dsWpOjPgZeLmpIJy1H37u9uLaRqB0JQTpTquk2gvI1IzymFS6qYKEkKHpA8dQwWJQHnZ7PUJPjZKD4exNCM0nqm/LzISKTWOArMZET1Qi95U/NcLoVkHV57GRNJqkHQeXCYcqxjPO0F95gEqvnYEIlM79jOiCSUG3aK5lS3MUKlknzrOqeVy/vLyq1u7yeIjpER+gEuegK1dAtqMGomiEntELerWerDfr3fqYrxas/OYA/YH1+QMPQJmI</latexit>

We formalize an online matching model for this setting, which differs from standard

  • nline matching in several key ways

Matching Algorithms for Blood Donation (EC20 Poster)

Standard Online Matching / Adwords Blood Donor Matching Same? U entirely offline U entirely offline Fixed time horizon T Fixed time horizon T U removed when matched U return after K days V have capacity 1 V have unlimited capacity V arrive one-by-one V arrive in batches, and some are offline |V| >> |U| |V| << |U|

Matching Model

slide-4
SLIDE 4

4

Greedy Myopic Policy Random Policy LP-based Policy

Total matched Weight (expected # donations) Median recipient weight* * : normalized by expected weight assigned by the random policy

Matching Algorithms for Blood Donation (EC20 Poster) In simulations we find that a greedy policy (which maximizes edge weight) increases

  • verall matching weight by 5-20%.

This comes at a cost of ignoring some recipients, which are not well-connected, or have low edge weights. In online experiments, using 1.3 million donors, we find that notifications which maximize (estimated) edge weight also increase overall donor action rate by about 5% (p<<0.001).

Initial Results