Generalized Propensity Score Matching with Multilevel Treatment Options
**Background:** Although conventional form of propensity score matching (PSM) is widely used in outcomes research field, its application on multilevel treatment is limited. **Objectives:** This article reviews PSM and illustrates their use when there are more than two treatment choices, which is v...
Saved in:
| Main Author: | |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Columbia Data Analytics, LLC
2013-03-01
|
| Series: | Journal of Health Economics and Outcomes Research |
| Online Access: | https://doi.org/10.36469/9847 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1823860322885173248 |
|---|---|
| author | Onur Baser |
| author_facet | Onur Baser |
| author_sort | Onur Baser |
| collection | DOAJ |
| description | **Background:** Although conventional form of propensity score matching (PSM) is widely used in outcomes research field, its application on multilevel treatment is limited.
**Objectives:** This article reviews PSM and illustrates their use when there are more than two treatment choices, which is very common in health services research.
**Methods:** Generalized PSM technique was applied to commercial claims data to estimate the treatment effect of reliever only, controller only and combination therapy of patients with asthma. The propensity score is estimated using multinomial logistic regression. The outcome variable was total annual health care costs. Inverse probability weighting was applied to calculate risk adjusted costs. Results are compared with multivariate regression analysis, where the generalized linear model is used with gamma family and log link function.
**Results:** Based on the study’s definitions of an asthma episode, we obtained a sample that included 25,124 patients in fee-for-service (FFS) plans and 6,603 patients in non-FFS plans. Under each plan type, patients who were prescribed three different treatment options were significantly different in terms of their demographic and clinical characteristics. Compared to combination therapy under FFS group, the difference of the total health care costs among reliever therapy and controller only group was significant ($728 and $1,216, respectively). Under non-FFS group, reliever only therapy totaled $1,266; controller only therapy was $1,959, and combination therapy totaled $1,933. Although the cost difference between reliever only and combination therapy was significant, there was no evidence that combination therapy cost more than controller only therapy. There were no significant differences in the multi-level propensity score adjusted results and multivariate regression results.
**Conclusion:** This analysis presents the potential value of generalized PSM methods in health services when there are multilevel treatment options. |
| format | Article |
| id | doaj-art-04a6373caa1f473ba866bbfaabee44a3 |
| institution | Kabale University |
| issn | 2327-2236 |
| language | English |
| publishDate | 2013-03-01 |
| publisher | Columbia Data Analytics, LLC |
| record_format | Article |
| series | Journal of Health Economics and Outcomes Research |
| spelling | doaj-art-04a6373caa1f473ba866bbfaabee44a32025-02-10T16:13:03ZengColumbia Data Analytics, LLCJournal of Health Economics and Outcomes Research2327-22362013-03-0111Generalized Propensity Score Matching with Multilevel Treatment OptionsOnur Baser**Background:** Although conventional form of propensity score matching (PSM) is widely used in outcomes research field, its application on multilevel treatment is limited. **Objectives:** This article reviews PSM and illustrates their use when there are more than two treatment choices, which is very common in health services research. **Methods:** Generalized PSM technique was applied to commercial claims data to estimate the treatment effect of reliever only, controller only and combination therapy of patients with asthma. The propensity score is estimated using multinomial logistic regression. The outcome variable was total annual health care costs. Inverse probability weighting was applied to calculate risk adjusted costs. Results are compared with multivariate regression analysis, where the generalized linear model is used with gamma family and log link function. **Results:** Based on the study’s definitions of an asthma episode, we obtained a sample that included 25,124 patients in fee-for-service (FFS) plans and 6,603 patients in non-FFS plans. Under each plan type, patients who were prescribed three different treatment options were significantly different in terms of their demographic and clinical characteristics. Compared to combination therapy under FFS group, the difference of the total health care costs among reliever therapy and controller only group was significant ($728 and $1,216, respectively). Under non-FFS group, reliever only therapy totaled $1,266; controller only therapy was $1,959, and combination therapy totaled $1,933. Although the cost difference between reliever only and combination therapy was significant, there was no evidence that combination therapy cost more than controller only therapy. There were no significant differences in the multi-level propensity score adjusted results and multivariate regression results. **Conclusion:** This analysis presents the potential value of generalized PSM methods in health services when there are multilevel treatment options.https://doi.org/10.36469/9847 |
| spellingShingle | Onur Baser Generalized Propensity Score Matching with Multilevel Treatment Options Journal of Health Economics and Outcomes Research |
| title | Generalized Propensity Score Matching with Multilevel Treatment Options |
| title_full | Generalized Propensity Score Matching with Multilevel Treatment Options |
| title_fullStr | Generalized Propensity Score Matching with Multilevel Treatment Options |
| title_full_unstemmed | Generalized Propensity Score Matching with Multilevel Treatment Options |
| title_short | Generalized Propensity Score Matching with Multilevel Treatment Options |
| title_sort | generalized propensity score matching with multilevel treatment options |
| url | https://doi.org/10.36469/9847 |
| work_keys_str_mv | AT onurbaser generalizedpropensityscorematchingwithmultileveltreatmentoptions |