Assessing healthcare providers' performance with and without risk adjustment
- Authors: Morales-Silva, Daniel
- Date: 2018
- Type: Text , Thesis , PhD
- Full Text:
- Description: This study focuses on how healthcare data can be used to draw comparisons between healthcare providers (surgeons or hospitals). Depending on the type of access to datasets, these comparisons can be done with or without risk adjustment. For us, risk adjustment refers to the use of patient-level information to explain variation in healthcare spending, resource utilisation and health outcomes. For unadjusted comparisons, we highlight the diagnostic potential that radar plots offer for reporting on outcome indicators. These outcome indicators were obtained from hospital admissions of patients undergoing certain surgical procedures. We address two drawbacks of radar plots: presence of missing information and order of indicators. By introducing a consolidated view at provider level, we define an uncomplicated ranking of providers which can be used to identify potential low and high performers. For risk adjusted comparisons, we introduce a novel and robust methodology that enables comparisons of healthcare providers across multiple hierarchies, namely, surgeons, teams, departments and hospitals, using a consistent approach. Our methodology puts the patient at the centre of the analysis, and thus, can be used for personalised predictions (e.g. expected length of stay, costs and probability of being transferred to intensive care unit). Our findings suggest that the observed variation in selected outcome indicators, such as length of stay and charges of healthcare providers, cannot be explained by patient characteristics alone. Importantly, we have also observed that the perceived performance, on selected outcome indicators, of providers can change substantially following risk adjustment. Healthcare is unique in that clinical expertise is essential in guiding decision making and in informing all statistical models that seek to describe patient outcomes. For future iterations of our models, we will seek greater clinical input.
- Description: Doctor of Philosophy
- Authors: Morales-Silva, Daniel
- Date: 2018
- Type: Text , Thesis , PhD
- Full Text:
- Description: This study focuses on how healthcare data can be used to draw comparisons between healthcare providers (surgeons or hospitals). Depending on the type of access to datasets, these comparisons can be done with or without risk adjustment. For us, risk adjustment refers to the use of patient-level information to explain variation in healthcare spending, resource utilisation and health outcomes. For unadjusted comparisons, we highlight the diagnostic potential that radar plots offer for reporting on outcome indicators. These outcome indicators were obtained from hospital admissions of patients undergoing certain surgical procedures. We address two drawbacks of radar plots: presence of missing information and order of indicators. By introducing a consolidated view at provider level, we define an uncomplicated ranking of providers which can be used to identify potential low and high performers. For risk adjusted comparisons, we introduce a novel and robust methodology that enables comparisons of healthcare providers across multiple hierarchies, namely, surgeons, teams, departments and hospitals, using a consistent approach. Our methodology puts the patient at the centre of the analysis, and thus, can be used for personalised predictions (e.g. expected length of stay, costs and probability of being transferred to intensive care unit). Our findings suggest that the observed variation in selected outcome indicators, such as length of stay and charges of healthcare providers, cannot be explained by patient characteristics alone. Importantly, we have also observed that the perceived performance, on selected outcome indicators, of providers can change substantially following risk adjustment. Healthcare is unique in that clinical expertise is essential in guiding decision making and in informing all statistical models that seek to describe patient outcomes. For future iterations of our models, we will seek greater clinical input.
- Description: Doctor of Philosophy
Using radar plots for performance benchmarking at patient and hospital levels using an Australian orthopaedics dataset
- Morales-Silva, Daniel, McPherson, Cameron, Pineda-Villavicencio, Guillermo, Atchison, Rory
- Authors: Morales-Silva, Daniel , McPherson, Cameron , Pineda-Villavicencio, Guillermo , Atchison, Rory
- Date: 2020
- Type: Text , Journal article
- Relation: Health Informatics Journal Vol. 26, no. 3 (2020), p. 2119-2137
- Full Text:
- Reviewed:
- Description: This study will highlight the diagnostic potential that radar plots display for reporting on performance benchmarking from patient admissions to hospital for surgical procedures. Two drawbacks of radar plots – the presence of missing information and ordering of indicators – are addressed. Ten different orthopaedic surgery procedures were considered in this study. Moreover, twelve outcome indicators were provided for each of the 10 surgeries of interest. These indicators were displayed using a radar plot, which we call a scorecard. At the hospital level, we propose a facile process by which to consolidate our 10 scorecards into one. We addressed the ordering of indicators in our scorecards by considering the national median of the indicators as a benchmark. Furthermore, our the consolidated scorecard facilitates concise visualisation and dissemination of complex data. It also enables the classification of providers into potential low and high performers that warrant further investigation. In conclusion, radar plots provide a clear and effective comparative tool for discerning multiple outcome indicators against the benchmarks of patient admission. A case study between two top and bottom performers on a consolidated scorecard (at hospital level) showed that medical provider charges varied more than other outcome indicators. © The Author(s) 2020.
- Authors: Morales-Silva, Daniel , McPherson, Cameron , Pineda-Villavicencio, Guillermo , Atchison, Rory
- Date: 2020
- Type: Text , Journal article
- Relation: Health Informatics Journal Vol. 26, no. 3 (2020), p. 2119-2137
- Full Text:
- Reviewed:
- Description: This study will highlight the diagnostic potential that radar plots display for reporting on performance benchmarking from patient admissions to hospital for surgical procedures. Two drawbacks of radar plots – the presence of missing information and ordering of indicators – are addressed. Ten different orthopaedic surgery procedures were considered in this study. Moreover, twelve outcome indicators were provided for each of the 10 surgeries of interest. These indicators were displayed using a radar plot, which we call a scorecard. At the hospital level, we propose a facile process by which to consolidate our 10 scorecards into one. We addressed the ordering of indicators in our scorecards by considering the national median of the indicators as a benchmark. Furthermore, our the consolidated scorecard facilitates concise visualisation and dissemination of complex data. It also enables the classification of providers into potential low and high performers that warrant further investigation. In conclusion, radar plots provide a clear and effective comparative tool for discerning multiple outcome indicators against the benchmarks of patient admission. A case study between two top and bottom performers on a consolidated scorecard (at hospital level) showed that medical provider charges varied more than other outcome indicators. © The Author(s) 2020.
- «
- ‹
- 1
- ›
- »