An approach that uses predictive modeling may improve healthcare organizations' ability to meet the triple aim of healthcare reform — improved patient experience, improved population health and reduced costs, according to a commentary in Health Affairs.
There are currently two dominant approaches to meeting the triple aim: a population-wide approach that aims to affect everyone in a population, and a targeted approach that aims to affect subgroups of patients at high risk for poor health outcomes, according to the commentary.
The authors suggest a third, stratified approach to prevent "Triple Fail" events — events that result in a poor patient experience, a poor health outcome and high costs, according to the commentary. The approach identifies patients at high risk for "Triple Fail" events and targets patients who are predicted to respond positively to preventive programs.
The stratified approach consists of three phases:
1. Planning phase. Conduct an opportunity analysis and develop predictive and impactibility models. Impactibility (or intervenability) models "seek to identify subgroups of high-risk people who are most likely to engage with and respond to various preventive interventions," according to the authors.
2. Operational phase. Use the predictive and impactibility models to identify patients who are at risk and likely to respond to an intervention. Then, provide the intervention.
3. Ongoing feedback phase. Refine the predictive and impactibility models based on results from the interventions.
While the stratified approach has several benefits, it does present some ethical challenges. The authors suggest the following ethical criteria for stratifying patients by their risk of having a Triple Fail event:
• The predicted event is an important health problem.
• An intervention can reduce the risk of the event.
• Resources and systems can enable timely risk stratification and preventive interventions.
• There is adequate time to adopt the intervention after stratification and before the event occurs.
• The population at large accepts the predictive risk model, which should be relatively accurate, and the impactibility model.
• An accepted policy defines who should receive the preventive intervention.
• The organization offering the preventive intervention understands the common causes of the Triple Fail event.
• The cost of stratification does not exceed the cost of the overall program.
• Stratification is a continuous process.
The authors also offer four recommendations to "promote the appropriate use of the stratified approach to the triple aim:"
• Use pilots. Compare population, individual and stratified approaches in triple aim pilots.
• Reduce data lags. Reduce delays in the availability of information from the Medicare limited data set and specific-use files.
• Conduct an ethical review. Review ethics of the design and implementation of predictive and impactibility models.
• Record more events. Expand database indicators to cover more potentially adverse events.
3 Years of PPACA: The 5 Biggest Changes in Healthcare Since the Law's Passage
Improving Population Health — An ACO Tenet Undefined?
There are currently two dominant approaches to meeting the triple aim: a population-wide approach that aims to affect everyone in a population, and a targeted approach that aims to affect subgroups of patients at high risk for poor health outcomes, according to the commentary.
The authors suggest a third, stratified approach to prevent "Triple Fail" events — events that result in a poor patient experience, a poor health outcome and high costs, according to the commentary. The approach identifies patients at high risk for "Triple Fail" events and targets patients who are predicted to respond positively to preventive programs.
The stratified approach consists of three phases:
1. Planning phase. Conduct an opportunity analysis and develop predictive and impactibility models. Impactibility (or intervenability) models "seek to identify subgroups of high-risk people who are most likely to engage with and respond to various preventive interventions," according to the authors.
2. Operational phase. Use the predictive and impactibility models to identify patients who are at risk and likely to respond to an intervention. Then, provide the intervention.
3. Ongoing feedback phase. Refine the predictive and impactibility models based on results from the interventions.
While the stratified approach has several benefits, it does present some ethical challenges. The authors suggest the following ethical criteria for stratifying patients by their risk of having a Triple Fail event:
• The predicted event is an important health problem.
• An intervention can reduce the risk of the event.
• Resources and systems can enable timely risk stratification and preventive interventions.
• There is adequate time to adopt the intervention after stratification and before the event occurs.
• The population at large accepts the predictive risk model, which should be relatively accurate, and the impactibility model.
• An accepted policy defines who should receive the preventive intervention.
• The organization offering the preventive intervention understands the common causes of the Triple Fail event.
• The cost of stratification does not exceed the cost of the overall program.
• Stratification is a continuous process.
The authors also offer four recommendations to "promote the appropriate use of the stratified approach to the triple aim:"
• Use pilots. Compare population, individual and stratified approaches in triple aim pilots.
• Reduce data lags. Reduce delays in the availability of information from the Medicare limited data set and specific-use files.
• Conduct an ethical review. Review ethics of the design and implementation of predictive and impactibility models.
• Record more events. Expand database indicators to cover more potentially adverse events.
More Articles on the Triple Aim:
Superior Health Partners, BCBS of Michigan Partner for Accountable Care3 Years of PPACA: The 5 Biggest Changes in Healthcare Since the Law's Passage
Improving Population Health — An ACO Tenet Undefined?