Penchant Scoring and Past: Why? also, How?

Slide1 l.jpg
1 / 8
1363 days ago, 373 views
PowerPoint PPT Presentation
... with just ~1K patients (Lindner Center: Kereiakes et al. Amer Heart J. ... Lingling Li, Harvard Medical School and Harvard Pilgrim Health Care ...

Presentation Transcript

Slide 1

Affinity Scoring and Beyond: Why? also, How? Midwest Biopharmaceutical Statistics Workshop, 2009

Slide 2

Notation for Variables y = watched result variable(s) x = watched standard covariate(s) t = watched treatment task (for the most part non-arbitrary) z = in secret logical variable(s)

Slide 3

Confounding An essential trouble in observational research is that the probability of treatment task, t , is NOT autonomous of the watched gauge x - covariates. Additionally, these gauge x - covariates are frequently not unimportant/subordinate. They may of themselves be prescient of y - result.

Slide 4

Propensity Score "Calculating" Joint dissemination of x and t given p : Pr( x , t | p ) = Pr( x | p ) Pr( t | p ) i.e. x and t are restrictively autonomous given the penchant for "new" treatment, p = Pr( t = 1 | x ).

Slide 5

Why Go Beyond PS Methods? Genuine PS as often as possible UNKNOWN. Assessed PS can without much of a stretch neglect to work the same as genuine PS. Approving PS Estimates can be Tedious and Frustrating. Options merit investigating!

Slide 6

Numerical Example for this Session. The information in the unreservedly conveyed "logical documents," Lsim10K and Lsim5K, utilized as a part of this session were recreated to be "like" that in a real OS with just ~1K patients (Lindner Center: Kereiakes et al. Amer Heart J. 2000.) Unfortunately, numerous creators and study supports don't perceive that sharing their information improves the believability of both their study and their dissects!

Slide 7

The "LSIM10K" dataset contains 10 reproduced estimations on 10,325 theoretical patients. [1] mort6mo : Binary 6-month mortality marker. [2] cardcost : Cumulative 6-month heart related charges. [3] trtm : Binary marker (1 => treated, 0 => untreated). [4] stent : Binary pointer (1 => coronary stent arrangement.) [5] stature : Patient tallness adjusted to the closest centimeter. [6] female : Binary sex pointer (1 => yes, 0 => male.) [7] diabetic : Binary marker (1 => diabetes mellitus, 0 => no.) [8] acutemi : Binary pointer (1 => intense myocardial dead tissue inside the past 7 days, 0 => no.) [9] ejecfrac : Left discharge portion % adjusted to whole number. [10] ves1proc : Number of vessels required in introductory PCI.

Slide 8

Confounding change: ideas and heuristic thoughts. Lingling Li, Harvard Medical School and Harvard Pilgrim Health Care Confounding change: thoughts in real life – a contextual investigation. Xiaochun Li, Div. Biostatistics, IU School of Medicine BREAK The "Neighborhood Control" Approach. Bounce Obenchain, Risk Benefit Statistics LLC DISCUSSION Gerhardt Pohl, Research Advisor, Lilly USA