%PDF-1.7
%
131 0 obj
<>
endobj
121 0 obj
<>
endobj
120 0 obj
<>stream
Arbortext Advanced Print Publisher 9.1.440/W Unicode
2019-10-25T16:22:47-07:00
2019-10-21T19:17:03+08:00
2019-10-25T16:22:47-07:00
application/pdf
Dissecting racial bias in an algorithm used to manage the health of populations
Acrobat Distiller 8.1.0 (Windows)
uuid:9f2609a7-ad48-45a2-8f1c-76758acdbe42
uuid:1e75f1ae-1dd2-11b2-0a00-1e00d83deeff
endstream
endobj
117 0 obj
<>
endobj
2 0 obj
<>
endobj
119 0 obj
<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]/XObject<>>>/Rotate 0/Thumb 116 0 R/Type/Page>>
endobj
1 0 obj
<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]/XObject<>>>/Rotate 0/Thumb 3 0 R/Type/Page>>
endobj
21 0 obj
<>/ExtGState<>/Font<>/ProcSet[/PDF/Text/ImageC]/Properties<>/XObject<>>>/Rotate 0/Thumb 22 0 R/Type/Page>>
endobj
45 0 obj
<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]/Properties<>/XObject<>>>/Rotate 0/Thumb 46 0 R/Type/Page>>
endobj
57 0 obj
<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]/Properties<>/XObject<>>>/Rotate 0/Thumb 58 0 R/Type/Page>>
endobj
70 0 obj
<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]/XObject<>>>/Rotate 0/Thumb 71 0 R/Type/Page>>
endobj
77 0 obj
<>/ExtGState<>/Font<>/ProcSet[/PDF/Text]/XObject<>>>/Rotate 0/Thumb 78 0 R/Type/Page>>
endobj
201 0 obj
<>/ProcSet[/PDF/Text/ImageC]/XObject<>>>/Type/Page>>
endobj
226 0 obj
[231 0 R 232 0 R 233 0 R 234 0 R 235 0 R 236 0 R 237 0 R]
endobj
227 0 obj
<>stream
q
87.5 0 0 21.75 72 662.25 cm
/Im0 Do
Q
BT
/T1_0 1 Tf
10 0 0 10 72 622 Tm
(Dissecting racial bias in an algorithm used to manage the health of popu\
lations)Tj
ET
BT
/T1_1 1 Tf
8 0 0 8 72 609 Tm
(Ziad Obermeyer, Brian Powers, Christine Vogeli and Sendhil Mullainathan)Tj
ET
BT
/T1_1 1 Tf
8 0 0 8 72 562.99997 Tm
(DOI: 10.1126/science.aax2342)Tj
5.50301 1 Td
( \(6464\), 447-453.)Tj
/T1_0 1 Tf
-1.668 0 Td
(366)Tj
/T1_2 1 Tf
-3.83501 0 Td
(Science\240)Tj
ET
BT
/T1_1 1 Tf
8 0 0 8 145.45605 458.99973 Tm
(, this issue p. 447; see also p. 421)Tj
/T1_2 1 Tf
-3.55701 0 Td
(Science)Tj
/T1_1 1 Tf
-3.125 1 Td
(care.)Tj
0 1.00001 TD
(algorithm so that it no longer uses costs as a proxy for needs eliminate\
s the racial bias in predicting who needs extra)Tj
0 1 TD
(the algorithm thus falsely concludes that Black patients are healthier t\
han equally sick White patients. Reformulating the )Tj
0 1.00001 TD
(health costs as a proxy for health needs. Less money is spent on Black p\
atients who have the same level of need, and)Tj
0 1 TD
(reduces the number of Black patients identified for extra care by more t\
han half. Bias occurs because the algorithm uses )Tj
T*
(algorithm are sicker than White patients \(see the Perspective by Benjam\
in\). The authors estimated that this racial bias)Tj
0 1.00001 TD
(evidence of racial bias in one widely used algorithm, such that Black pa\
tients assigned the same level of risk by the )Tj
47.69692 1 Td
( find)Tj
/T1_2 1 Tf
-2.168 0 Td
(et al.)Tj
/T1_1 1 Tf
-42.40392 0 Td
(The U.S. health care system uses commercial algorithms to guide health d\
ecisions. Obermeyer )Tj
/T1_0 1 Tf
-3.125 1 Td
(Racial bias in health algorithms)Tj
ET
BT
/T1_0 1 Tf
6 0 0 6 79 393 Tm
(ARTICLE TOOLS)Tj
ET
BT
0 0 1 rg
/T1_1 1 Tf
8 0 0 8 187 391 Tm
(http://science.sciencemag.org/content/366/6464/447)Tj
ET
BT
0 g
/T1_0 1 Tf
6 0 0 6 79 350.99997 Tm
(MATERIALS)Tj
T*
(SUPPLEMENTARY )Tj
ET
BT
0 0 1 rg
/T1_1 1 Tf
8 0 0 8 187 355 Tm
(http://science.sciencemag.org/content/suppl/2019/10/23/366.6464.447.DC1)Tj
ET
BT
0 g
/T1_0 1 Tf
6 0 0 6 79 308.99997 Tm
(CONTENT)Tj
T*
(RELATED )Tj
ET
BT
0 0 1 rg
/T1_1 1 Tf
8 0 0 8 187 313 Tm
(http://science.sciencemag.org/content/sci/366/6464/421.full)Tj
ET
BT
0 g
/T1_0 1 Tf
6 0 0 6 79 273 Tm
(REFERENCES)Tj
ET
BT
0 0 1 rg
/T1_1 1 Tf
8 0 0 8 187 262.99997 Tm
(http://science.sciencemag.org/content/366/6464/447#BIBL)Tj
0 g
0 1 TD
(This article cites 34 articles, 7 of which you can access for free)Tj
ET
BT
/T1_0 1 Tf
6 0 0 6 79 237 Tm
(PERMISSIONS)Tj
ET
BT
0 0 1 rg
/T1_1 1 Tf
8 0 0 8 187 235 Tm
(http://www.sciencemag.org/help/reprints-and-permissions)Tj
ET
BT
/T1_1 1 Tf
8 0 0 8 193.38391 128 Tm
(Terms of Service)Tj
0 g
-15.17299 0 Td
(Use of this article is subject to the )Tj
ET
72 114 m
521 114 l
0 0 m
S
BT
ET
BT
/T1_1 1 Tf
8 0 0 8 357.45605 93.99997 Tm
( is a registered trademark of AAAS.)Tj
/T1_2 1 Tf
-3.55701 0 Td
(Science)Tj
/T1_1 1 Tf
-32.125 0 Td
(Science, 1200 New York Avenue NW, Washington, DC 20005. The title )Tj
3.83501 1 Td
(\(print ISSN 0036-8075; online ISSN 1095-9203\) is published by the Amer\
ican Association for the Advancement of)Tj
/T1_2 1 Tf
-3.83501 0 Td
(Science )Tj
ET
BT
/T1_1 1 Tf
8 0 0 8 72 71.99997 Tm
(Science. No claim to original U.S. Government Works)Tj
T*
(Copyright \251 2019 The Authors, some rights reserved; exclusive license\
e American Association for the Advancement of)Tj
ET
BT
/T1_1 1 Tf
0 -8 8 0 575 326.8761 Tm
(on October 25, 2019)Tj
-0.278 0 Td
( )Tj
0 0 1 rg
-13.50598 0 Td
(http://science.sciencemag.org/)Tj
0 g
-8.11398 0 Td
(Downloaded from )Tj
ET
endstream
endobj
230 0 obj
<>/Filter/FlateDecode/Height 87/Length 19684/Name/X/Subtype/Image/Type/XObject/Width 350>>stream
HyTS߽Ia
"6p\P縠x:Z.Suk:8ud,EԺУ:xtSVk9@na}/ Ҟ6Gy@{_rcL"bFSgjz"UH}>2Z
5f @a,uWP24
jѨ\i-~==Op(rw]EQڇMyƧH4E5(HQ_-*P1.#*kb"f_+XA¾ՆK$FxUFJeFq {vdcg8ZS#3_kX `9'p (3u_Od}uvn|TSI#
!y 3|].Fh{U_JA-M0bD22}[-(U~[yTk|H mL;ƶc 0X ?1Ar?oƀU]C:/^ّ,7(a"Gq?b#QUaS L6)$Hm[ANc tPĎvURl2!VV{^e(^I9Dã3l]]\3$!i;v{?.Ҕ}31,ER[^k4Nok#b7fIȐy`au^GUĉ
W&\y,$d<|rD;VӪn>Mј[Mi#DQ LY2Cme#BV~2`ls~s W&
=xAJE^dɡ>S=`_dgk+gqaنKgbK=E[> @AQbg(