Optimal compliance prediction models for estimating causal effects
| dc.creator | Odondi, Lang'o | |
| dc.creator | Bekolle, David | |
| dc.date | 02/24/2015 | |
| dc.date | Tue, 24 Feb 2015 | |
| dc.date | Tue, 24 Feb 2015 20:09:12 | |
| dc.date | Year: 2014 | |
| dc.date | Tue, 24 Feb 2015 20:09:12 | |
| dc.date.accessioned | 2015-03-18T11:29:15Z | |
| dc.date.available | 2015-03-18T11:29:15Z | |
| dc.description | Paper presented at the 2nd Strathmore International Mathematics Conference (SIMC 2013), 12 - 16 August 2013, Strathmore University, Nairobi, Kenya. | |
| dc.description | Understanding the causal relationship between intervention and outcome is at the heart of most research in the health sciences, and a variety of statistical methods have been developed to address causality. However, noncompliance with treatment assignment is a key source of complication for causal inference. Estimation of causal effects is likely to be compounded by the presence of noncompliance in both treatment arms of clinical trials where the intention-to-treat (ITT) analysis produces a biased estimator for the true causal estimate even under homogeneous treatment effects assumption. Principal stratification method has been developed to address such posttreatment complications by stratifying the population into partially latent classes (principal strata) based on potential values observed after randomization (e.g. noncompliance) under each of the levels of randomized intervention. The present work combines the two strategies of model selection and principal stratification with a novel application to a real data from a trial conducted to ascertain whether or not unopposed oestrogen (hormone replacement therapy - HRT) reduced the risk of further cardiac events in postmenopausal women who survive a first myocardial infarction. The causal model links the resulting two marginal prediction models with a user-defined sensitivity parameter which is a function of the correlation between the two compliance behaviours. The method's key assumption of conditional prediction is verified for our data via sensitivity analysis comparing results of causal estimates using different sets of predictors of compliance. We adjust for noncompliance in both treatment arms under a Bayesian framework to produce causal risk ratio estimates for each principal stratum. The results suggested better efficacy for HRT among those who would comply with it compared to those who would comply with either HRT or placebo: compliance with HRT treatment only and with either treatment allocation would reduce the risk for death (reinfarction) by 47%(25%) and 13%(60%) respectively. | |
| dc.description.abstract | Understanding the causal relationship between intervention and outcome is at the heart of most research in the health sciences, and a variety of statistical methods have been developed to address causality. However, noncompliance with treatment assignment is a key source of complication for causal inference. Estimation of causal effects is likely to be compounded by the presence of noncompliance in both treatment arms of clinical trials where the intention-to-treat (ITT) analysis produces a biased estimator for the true causal estimate even under homogeneous treatment effects assumption. Principal stratification method has been developed to address such posttreatment complications by stratifying the population into partially latent classes (principal strata) based on potential values observed after randomization (e.g. noncompliance) under each of the levels of randomized intervention. The present work combines the two strategies of model selection and principal stratification with a novel application to a real data from a trial conducted to ascertain whether or not unopposed oestrogen (hormone replacement therapy - HRT) reduced the risk of further cardiac events in postmenopausal women who survive a first myocardial infarction. The causal model links the resulting two marginal prediction models with a user-defined sensitivity parameter which is a function of the correlation between the two compliance behaviours. The method's key assumption of conditional prediction is verified for our data via sensitivity analysis comparing results of causal estimates using different sets of predictors of compliance. We adjust for noncompliance in both treatment arms under a Bayesian framework to produce causal risk ratio estimates for each principal stratum. The results suggested better efficacy for HRT among those who would comply with it compared to those who would comply with either HRT or placebo: compliance with HRT treatment only and with either treatment allocation would reduce the risk for death (reinfarction) by 47%(25%) and 13%(60%) respectively. | |
| dc.format | Pages:1 - 24 | |
| dc.format | Volume Number:Volume 1, Number 1, (2014) | |
| dc.identifier | 13.1007/s00009-003-0000 | |
| dc.identifier | ||
| dc.identifier.uri | http://hdl.handle.net/11071/3829 | |
| dc.language | eng | |
| dc.publisher | IMHOTEP Mathematical Proceedings | |
| dc.rights | By agreeing with and accepting this license, I (the author(s), copyright owner or nominated agent) agree to the conditions, as stated below, for deposit of the item (referred to as .the Work.) in the digital repository maintained by Strathmore University, or any other repository authorized for use by Strathmore University. Non-exclusive Rights Rights granted to the digital repository through this agreement are entirely non-exclusive. I understand that depositing the Work in the repository does not affect my rights to publish the Work elsewhere, either in present or future versions. I agree that Strathmore University may electronically store, copy or translate the Work to any approved medium or format for the purpose of future preservation and accessibility. Strathmore University is not under any obligation to reproduce or display the Work in the same formats or resolutions in which it was originally deposited. SU Digital Repository I understand that work deposited in the digital repository will be accessible to a wide variety of people and institutions, including automated agents and search engines via the World Wide Web. I understand that once the Work is deposited, metadata may be incorporated into public access catalogues. I agree as follows: 1.That I am the author or have the authority of the author/s to make this agreement and do hereby give Strathmore University the right to make the Work available in the way described above. 2.That I have exercised reasonable care to ensure that the Work is original, and to the best of my knowledge, does not breach any laws including those relating to defamation, libel and copyright. 3.That I have, in instances where the intellectual property of other authors or copyright holders is included in the Work, gained explicit permission for the inclusion of that material in the Work, and in the electronic form of the Work as accessed through the open access digital repository, or that I have identified that material for which adequate permission has not been obtained and which will be inaccessible via the digital repository. 4.That Strathmore University does not hold any obligation to take legal action on behalf of the Depositor, or other rights holders, in the event of a breach of intellectual property rights, or any other right, in the material deposited. 5.That if, as a result of my having knowingly or recklessly given a false statement at points 1, 2 or 3 above, the University suffers loss, I will make good that loss and indemnify Strathmore University for all action, suits, proceedings, claims, demands and costs occasioned by the University in consequence of my false statement. | |
| dc.subject | prediction models | |
| dc.subject | causal effects | |
| dc.subject | causal relationship | |
| dc.subject | statistical methods | |
| dc.subject | health sciences | |
| dc.title | Optimal compliance prediction models for estimating causal effects | |
| dc.type | Article | |
| dc.type | Conference Paper |