Se nsitivity Analysis in the Conte xt of Ne twor k Me ta - Analysis: De monstr ation for Ac ade myHe alth Annual Re se ar c h Me e ting June 27, 2017 Sa ndra Ze lma n L e wis, PhD Cra ig Whitting to n, PhD T o dd F e inma n, MD Roshan Shah, MS
L e ar ning Obje c tive s Sta te of the a rt dig ita l te c hnolog ie s c a n be le ve ra g e d to • Pe rfo rm se nsitivity a na lyse s in the c o nte xt o f ne two rk me ta -a na lysis • I mple me nt se ve ra l me tho ds o f se nsitivity a na lyse s • T o c ritic a lly a sse ss the a dva nta g e s o f using e ffic ie nt se nsitivity a na lyse s to g e ne ra te info rma tive re sults
Ne twor k Me ta- Analysis (NMA) Analytic al me thod to e xamine both dir e c t and indir e c t e vide nc e Inc r e ase s c omple xity of managing data c onfigur ation and pe r for ming se nsitivity analysis
Re se ar c h Obje c tive : F ind, synthe size , and analyze data fr om public sour c e s into fit- to- pur pose ac tionable knowle dge Se a rc h Se a r c he d, sc r e e ne d a nd inde xing of a ll r e le va nt a r tic le s in the the r a pe utic a r e a T e c hno lo g y a ssiste d pro c e ss se a rc h a nd sc re e n a rtic le s a ssig ning the m pe r inc lusio n/ e xc lusio n c rite rio n with PRISMA d ia g ra ms d yna mic a lly g e ne ra te d a nd c o ntinuo us mo nito ring Mine Ne a r a utoma te d digitization of lists a nd fr e que nc y tr e e s of a ll te r ms (inc luding AE s) in e a c h tic le s with hig hly va lida te d transparent & a c c ur a bstr a c t a nd full-te xt a r a c y T e c hno lo g y a ssiste d e vid e nc e a na lysts b uild fre q ue nc y tre e s o f te rms, Ad ve rse E ve nts a nd pa tie nt c ha ra c te ristic s, o utc o me s a nd e nd po ints in e a c h a rtic le De live r Re que ste d da ta tha t is ne e de d for a ny submission is a va ila ble on de ma nd a nd de live r e d in DRE pla tfor m a va ila ble for a na lysis, ta ble s, a nd othe r E BM outputs & fing e r tip a na lysis Any a na lysis c a pa b le o f b e ing run in a n R mo d ule a re d ire c tly ha rd c o d e d into the e vid e nc e d a ta b a se with a ll the sta tistic a l c o nso le s g e ne ra te d a nd a uto sa ve d
Data Configur ation Pr oc e ss • Study L e ve l Va ria b le s • I nte rve ntio n va ria b le s Da ta • Pa tie nt c ha ra c te ristic s a nd o utc o me s Co nfig ura tio n Pro to c o l • Da ta a nd me ta -da ta fro m te xt, ta b le s, fig ure s, c ha rts • Qua lity c o ntro l Da ta E xtra c tio n • Pre se rva tio n o f e xa c t te rm • Co nsiste nt na ming c o nve ntio ns Onto lo g y • Co de ma pping to he a lth la ng ua g e dic tio na rie s Ma na g e me nt
Se nsitivity Analysis Why r un se nsitivity analyse s? E xplo ra tio n E xa mina tio n o f ro b ustne ss a nd re le va nc e o f re sults Che c king a ssumptio ns Mo r e info r me d de c isio n-making
Conduct a Network Meta-Analysis De mo nstra tio n
Se nsitivity Analysis F ac tor s to e xplor e whe n c r e ating se nsitivity analyse s • Study de sig n • Po pula tio n (I / E c rite ria , b a se line c ha ra c te ristic s) • Study b ia s • F o llo w-up time • He te ro g e ne ity / o utlie rs • Da te o f pub lic a tio n • L e ve ra g e sa ve d wo rkflo ws to q uic kly upda te a na lyse s • Wha t-if a na lysis
Implic ations for HSR E mploying today’s te c hnologie s, the multivar iate c omple xitie s of HSR c an r e adily be ne fit fr om se nsitivity analyse s e asily pe r for me d in the c onte xt of NMAs Inc r e ase d tr anspar e nc y allows inde pe nde nt audits and/ or ve r ific ation of binding and inc lusion c r ite r ia, the r e for e inc r e asing c onfide nc e in r e sults and suppor ting r e plic ations of the se analyse s, a sc ie ntific te ne t ofte n impossible be c ause tr aditional pr oc e sse s and public ation r e sour c e s do not doc ume nt de tails (e .g., te r m manage me nt). F ur the r , lac k of te c hnology inhibits r e plic ation due to the high labor and time r e quir e me nts. Captur ing all te r ms (i.e . patie nt c har ac te r istic s and outc ome s) in publishe d studie s e nable s the c r e ation of a we b- base d dic tionar y of te r ms - - an e sse ntial tool for HSR topic s that do not have e xisting ontologie s.
T hank you! Sandr a L e wis, PhD: sle wis@doc tor e vide nc e .c om Roshan Shah, MS: r shah@doc tor e vide nc e .c om
Recommend
More recommend