A comparison of different methods to handle missing data in the context of propensity score analysis

Choi, Jung-Yeon; Dekkers, Olaf M.; Cessie, Saskia le

doi:10.1007/s10654-018-0447-z

Cited by 121 publications

(104 citation statements)

References 44 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…In these articles, the focus is on handling missing data in covariates used in propensity scores, whereas here we consider missing data in the exposure of interest. Nevertheless, [3] in particular noted that the use of missing indicators can partly adjust for unmeasured confounding, similar to our findings.…”

Section: Implications For Causal Estimationsupporting

confidence: 91%

“…With such application, missing indicator is known to lead to biased estimation even under MCAR [4,5]. The idea of combining the missing indicator approach with multiple imputation was first proposed by [6], and has been further explored by [7] and [3]. In these articles, the focus is on handling missing data in covariates used in propensity scores, whereas here we consider missing data in the exposure of interest.…”

Section: Implications For Causal Estimationmentioning

confidence: 99%

“…In these cases, the missingness mechanism (i.e. presence or absence of the laboratory test) may act as a proxy for unmeasured confounding, allowing for partial adjustment [3].…”

Section: Introductionmentioning

confidence: 99%

“…Although in MCAR and MAR scenarios, use of missing indicators can introduce bias [4,5], use of missing indicators may reduce bias where the missingness is informative, and particularly when the missing indicator is used in conjunction with multiple imputation (MIMI) [3,[6][7][8].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Use of Missing Indicators as Proxies for Unmeasured Variables: Simulation Study

Sperrin¹,

Martin²

2020

Preprint

View full text Add to dashboard Cite

Background Within routinely collected health data, missing data for an individual might provide useful information in itself. This occurs, for example, in the case of electronic health records, where the presence or absence of data is informative. While the naive use of missing indicators to try to exploit such information can introduce bias when used inappropriately, its use in conjunction with other imputation approaches may unlock the potential value of missingness to reduce bias and improve prediction.Methods We conducted a simulation study to determine when the use of a missing indicator, combined with an imputation approach, such as multiple imputation, would lead to improved model performance, in terms of minimising bias for causal effect estimation, and improving predictive accuracy, under a range of scenarios with unmeasured variables. We use directed acyclic graphs and structural models to elucidate causal structures of interest. We consider a variety of missingness mechanisms, then handle these using complete case analysis, unconditional mean imputation, regression imputation and multiple imputation. In each case we evaluate supplementing these approaches with missing indicator terms. Results For estimating causal effects, we find that multiple imputation combined with a missing indicator gives minimal bias in most scenarios. For prediction, we find that regression imputation combined with a missing indicator minimises mean squared error.Conclusion In the presence of missing data, careful use of missing indicators, combined with appropriate imputation, can improve both causal estimation and prediction accuracy.

show abstract

Section: Implications For Causal Estimationsupporting

confidence: 91%

Section: Implications For Causal Estimationmentioning

confidence: 99%

“…In these cases, the missingness mechanism (i.e. presence or absence of the laboratory test) may act as a proxy for unmeasured confounding, allowing for partial adjustment [3].…”

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Use of Missing Indicators as Proxies for Unmeasured Variables: Simulation Study

Sperrin¹,

Martin²

2020

Preprint

View full text Add to dashboard Cite

show abstract

“…With such application, missing indicator is known to lead to biased estimation even under MCAR [9][10][11][12]. The idea of combining the missing indicator approach with multiple imputation was first proposed by [6], and has been further explored by [7] and [4]. In those articles, the focus is on handling missing data in covariates used in propensity scores, whereas here we consider missing data in the exposure of interest.…”

Section: Discussionmentioning

confidence: 99%

Multiple Imputation with Missing Indicators as Proxies for Unmeasured Variables: Simulation Study

Sperrin

Martin

2020

Preprint

View full text Add to dashboard Cite

Background : Within routinely collected health data, missing data for an individual might provide useful information in itself. This occurs, for example, in the case of electronic health records, where the presence or absence of data is informative. While the naive use of missing indicators to try to exploit such information can introduce bias, its use in conjunction with multiple imputation may unlock the potential value of missingness to reduce bias in causal effect estimation, particularly in missing not at random scenarios and where missingness might be associated with unmeasured confounders. Methods: We conducted a simulation study to determine when the use of a missing indicator, combined with multiple imputation, would reduce bias for causal effect estimation, under a range of scenarios including unmeasured variables, missing not at random, and missing at random mechanisms. We use directed acyclic graphs and structural models to elucidate a variety of causal structures of interest. We handled missing data using complete case analysis, and multiple imputation with and without missing indicator terms. Results: We find that multiple imputation combined with a missing indicator gives minimal bias for causal effect estimation in most scenarios. In particular the approach: 1) does not introduce bias in missing (completely) at random scenarios; 2)reduces bias in missing not at random scenarios where the missing mechanism depends on the missing variable itself; and 3) may reduce or increase bias when unmeasured confounding is present. Conclusion : In the presence of missing data, careful use of missing indicators, combined with multiple imputation, can improve causal effect estimation when missingness is informative, and is not detrimental when missingness is at random.

show abstract

Mental Health Conditions in Partners and Adult Children of Stroke Survivors

Skajaa,

Farkas,

Laugesen

et al. 2024

JAMA Netw Open

View full text Add to dashboard Cite

ImportanceFamily caregiving after critical illness has been associated with several adverse health outcomes, including various aspects of mental health, but research focusing specifically on family members of stroke survivors is limited.ObjectivesTo examine the associations of stroke in a partner or parent with the risk of depression, substance use disorders, anxiety disorders, and self-harm or suicide.Design, Setting, and ParticipantsThis nationwide, population-based cohort study used data from Danish nationwide administrative and clinical registries (2004-2021). Participants included partners and adult children of survivors of stroke. Data analysis was performed from March to December 2023.ExposureHaving a partner or parent who survived stroke.Main Outcomes and MeasuresThe Aalen-Johansen estimator was used to compute propensity score–weighted 3-year absolute risks, risk differences, and risk ratios for depression, substance use disorders, anxiety disorders, and self-harm or suicide among partners or children of survivors of stroke compared with partners or children of survivors of myocardial infarction (MI) and matched individuals from the general population.ResultsThe study included a total of 1 923 732 individuals: 70 917 partners of stroke survivors (median [IQR] age, 68 [59-76] years; 46 369 women [65%]), 70 664 partners of MI survivors (median [IQR] age, 65 [55-73] years; 51 849 women [73%]), 354 570 partners of individuals from the general population (median [IQR] age, 68 [59-76] years; 231 833 women [65%]), 207 386 adult children of stroke survivors (median [IQR] age, 45 [36-52] years; 99 382 women [48%]), 183 309 adult children of MI survivors (median [IQR] age, 42 [33-49] years; 88 078 women [48%]), and 1 036 886 adult children of individuals from the general population (median [IQR] age, 45 [36-52] years; 496 875 women [48%]). Baseline characteristics were well balanced across cohorts after propensity score weighting. Among partners of stroke survivors, the 3-year absolute risk was 1.0% for depression, 0.7% for substance use disorders, 0.3% for anxiety disorders, and 0.04% for self-harm or suicide. Risk ratio point estimates for the assessed outcomes ranged from 1.14 to 1.42 compared with the general population and from 1.04 to 1.09 compared with partners of MI survivors. The elevated risk of depression in partners of stroke survivors was more pronounced after severe or moderate stroke than after mild stroke. Among adult children of stroke survivors, the 3-year absolute risk was 0.6% for depression, 0.6% for substance use disorders, 0.2% for anxiety disorders, and 0.05% for self-harm or suicide. Both absolute risks and risk ratios for adult children of stroke survivors were smaller than those reported in the partner analyses.Conclusions and RelevanceIn this cohort study of partners and adult children of stroke survivors, risks of several mental health conditions and self-harm or suicide were moderately higher compared with the general population and, to a lesser extent, partners and adult children of MI survivors. These findings highlight the potential consequences of stroke among family members, particularly partners, and its findings may possibly serve as a quantitative foundation for the development of future stroke rehabilitation services.

show abstract

A comparison of different methods to handle missing data in the context of propensity score analysis

Cited by 121 publications

References 44 publications

Use of Missing Indicators as Proxies for Unmeasured Variables: Simulation Study

Use of Missing Indicators as Proxies for Unmeasured Variables: Simulation Study

Multiple Imputation with Missing Indicators as Proxies for Unmeasured Variables: Simulation Study

Mental Health Conditions in Partners and Adult Children of Stroke Survivors

Contact Info

Product

Resources

About