Profile Matching for the Generalization and Personalization of Causal Inferences

Cohn, Eric R.; Zubizarreta, José R.

doi:10.1097/ede.0000000000001517

Cited by 5 publications

(4 citation statements)

References 63 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Finally, while we have focused on three estimands-the ATE, ATT, and ATO-balancing approaches have also been shown to perform well for other possible estimands. For instance, one can attempt to estimate the effect for a population with a particular target covariate profile 24 or to generalize or transport the effect for a population different from the sample. 25,26 These estimands have their own overlap assumptions; we leave a thorough investigation to future work.…”

Section: Discussionmentioning

confidence: 99%

Using Balancing Weights to Target the Treatment Effect on the Treated when Overlap is Poor

Ben‐Michael

Keele

2023

Epidemiology

View full text Add to dashboard Cite

Inverse probability weights are commonly used in epidemiology to estimate causal effects in observational studies. Researchers often focus on either the average treatment effect or the average treatment effect on the treated with inverse probability weighting estimators. However, poor overlap in the baseline covariates between the treated and control groups can produce extreme weights that can result in biased treatment effect estimates. One alternative to inverse probability weights are overlap weights, which target the population with the most overlap on observed covariates. Although estimates based on overlap weights produce less bias in such contexts, the causal estimand can be difficult to interpret. An alternative to model-based inverse probability weights are balancing weights, which directly target imbalances during the estimation process, rather than model fit. Here, we explore whether balancing weights allow analysts to target the average treatment effect on the treated in cases where inverse probability weights lead to biased estimates due to poor overlap. We conduct three simulation studies and an empirical application. We find that balancing weights often allow the analyst to still target the average treatment effect on the treated even when overlap is poor. We show that although overlap weights remain a key tool, more familiar estimands can sometimes be targeted by using balancing weights.

show abstract

Section: Discussionmentioning

confidence: 99%

Using Balancing Weights to Target the Treatment Effect on the Treated when Overlap is Poor

Ben‐Michael

Keele

2023

Epidemiology

View full text Add to dashboard Cite

show abstract

“…Given baseline differences among the 3 groups of practices, we used matching to balance observable practice characteristics. We used profile matching 32 to find the largest sample of practices in each group whereby characteristics were balanced compared with the overall practice characteristics in the full sample. This method optimizes the size of the matched sample and directly balances covariates across all 3 groups.…”

Section: Methodsmentioning

confidence: 99%

Use of Telemedicine and Quality of Care Among Medicare Enrollees With Serious Mental Illness

Wilcock,

Huskamp,

Busch

et al. 2023

JAMA Health Forum

View full text Add to dashboard Cite

ImportanceDuring the COVID-19 pandemic, a large fraction of mental health care was provided via telemedicine. The implications of this shift in care for use of mental health service and quality of care have not been characterized.ObjectiveTo compare changes in care patterns and quality during the first year of the pandemic among Medicare beneficiaries with serious mental illness (schizophrenia or bipolar I disorder) cared for at practices with higher vs lower telemedicine use.Design, Setting, and ParticipantsIn this cohort study, Medicare fee-for-service beneficiaries with schizophrenia or bipolar I disorder were attributed to specialty mental health practices that delivered the majority of their mental health care in 2019. Practices were categorized into 3 groups based on the proportion of telemental health visits provided during the first year of the pandemic (March 2020-February 2021): lowest use (0%-49%), middle use (50%-89%), or highest use (90%-100%). Across the 3 groups of practices, differential changes in patient outcomes were calculated from the year before the pandemic started to the year after. These changes were also compared with differential changes from a 2-year prepandemic period. Analyses were conducted in November 2022.ExposurePractice-level use of telemedicine during the first year of the COVID-19 pandemic.Main Outcomes and MeasuresThe primary outcome was the total number of mental health visits (telemedicine plus in-person) per person. Secondary outcomes included the number of acute hospital and emergency department encounters, all-cause mortality, and quality outcomes, including adherence to antipsychotic and mood-stabilizing medications (as measured by the number of months of medication fills) and 7- and 30-day outpatient follow-up rates after discharge for a mental health hospitalization.ResultsThe pandemic cohort included 120 050 Medicare beneficiaries (mean [SD] age, 56.5 [14.5] years; 66 638 females [55.5%]) with serious mental illness. Compared with prepandemic changes and relative to patients receiving care at practices with the lowest telemedicine use: patients receiving care at practices in the middle and highest telemedicine use groups had 1.11 (95% CI, 0.45-1.76) and 1.94 (95% CI, 1.28-2.59) more mental health visits per patient per year (or 7.5% [95% CI, 3.0%-11.9%] and 13.0% [95% CI, 8.6%-17.4%] more mental health visits per year, respectively). Among patients of practices with middle and highest telemedicine use, changes in adherence to antipsychotic and mood-stabilizing medications were −0.4% (95% CI, −1.3% to 0.5%) and −0.1% (95% CI, −1.0% to 0.8%), and hospital and emergency department use for any reason changed by 2.4% (95% CI, −1.5% to 6.2%) and 2.8% (95% CI, −1.2% to 6.8%), respectively. There were no significant differential changes in postdischarge follow-up or mortality rates according to the level of telemedicine use.Conclusions and RelevanceIn this cohort study of Medicare beneficiaries with serious mental illness, patients receiving care from practices that had a higher level of telemedicine use during the COVID-19 pandemic had more mental health visits per year compared with prepandemic levels, with no differential changes in other observed quality metrics over the same period.

show abstract

“…and . For recent optimization matching methods that leverage modern optimization using generic integer programming, see , , and Cohn and Zubizarreta (2022).…”

Section: Matchingmentioning

confidence: 99%

“…In our example, we implement two matching methods, optimal pair matching as proposed by and profile matching by Cohn and Zubizarreta (2022). The corresponding structures of the matched samples are depicted in Figures 6 and 7.…”

Section: Matchingmentioning

confidence: 99%

Causation, Comparison, and Regression

Chattopadhyay,

Zubizarreta

2024

Harvard Data Science Review

View full text Add to dashboard Cite

Comparison and contrast are the basic means to unveil causation and learn which treatments work. To build good comparison groups, randomized experimentation is key, yet often infeasible. In such non-experimental settings, we illustrate and discuss diagnostics to assess how well the common linear regression approach to causal inference approximates desirable features of randomized experiments, such as covariate balance, study representativeness, interpolated estimation, and unweighted analyses. We also discuss alternative regression modeling, weighting, and matching approaches and argue they should be given strong consideration in empirical work.

show abstract

Profile Matching for the Generalization and Personalization of Causal Inferences

Cited by 5 publications

References 63 publications

Using Balancing Weights to Target the Treatment Effect on the Treated when Overlap is Poor

Using Balancing Weights to Target the Treatment Effect on the Treated when Overlap is Poor

Use of Telemedicine and Quality of Care Among Medicare Enrollees With Serious Mental Illness

Causation, Comparison, and Regression

Contact Info

Product

Resources

About