The other morning I was sitting in a meeting and a question was raised. That question was "do we really know that understanding our data better, and using tools such as data visualization, will really improve healthcare?"
Only in healthcare would this question still be asked in 2011.
Industries of all types have learned over the last 20 years the power of developing knowledge based on their actions and the actions of their customers. Many of these industries are not as complicated as healthcare. But I believe that the idea that healthcare is the most complex industry is erroneous; however, it's been my experience that this idea is deeply rooted in the culture of medicine. It seems to be felt that since there are so many uncontrollable factors such as patient complexity, patient compliance, variations in disease presentation, and the intricate web of payers and delivery systems, it is not possible to understand the "healthcare system" with data. I do not believe this is true, and in support of my position, I'd like to present a case from an industry that is at least as complex as the healthcare system: the airline industry.
Within the airline industry, the uncontrollable variable range from mechanical problems to passenger behavior to natural disasters to terrorist attacks... to the most unpredictable of all- the weather! Despite all of these variables, there are many examples of how the airline industry has transformed data into knowledge, improved service, and remained profitable.
Take, for example, the recent case study reported by American Airlines. American Airlines had identified fraud as a major cost to their business. However, they had no data warehouse technology or knowledge management plan for addressing fraud in their system. It was originally estimated that an effective data analytics system would save the company $150,000 per year. Using an "off-the-shelf" data warehouse solution, great gains were immediately seen, and ultimately saved the company $5 million over 5 years. The success was credited to the new system's ability to identify forms of fraud that the company never knew existed and giving the company the ability to make changes to eliminate those causes.
While it is true that we may be many decades away from being truly knowledgeable about how the US healthcare system works, it is also true that there are many technologies available today that, if utilized in healthcare, could have immediate and meaningful impact. Targeted solutions can quickly exceed expectations when we focus on creating new actionable knowledge with current technologies.
Monday, April 25, 2011
Thursday, April 14, 2011
CKM and Spaghetti Sauce
One of the key components to clinical knowledge management is the discovery of actionable knowledge. The process of discovering knowledge is often more of an art than a science. A complex part of the art is asking the right question. The knowledge discovered and developed through data driven techniques such as data mining and statistical hypothesis testing are always framed by the questions being asked. Ask the wrong question, generate the wrong knowledge. Unfortunately, unless you know you are asking the wrong question you assume you are working with the right knowledge.
Gains can be made when making decisions with the “wrong knowledge,” but they will be less than the gains made if decisions were made from the right knowledge. An excellent example of how knowledge is improved when you ask the right question is described by Malcolm Gladwell in his TED talk about the food industry and a spaghetti sauce breakthrough.
Gladwell describes how chunky spaghetti sauce revolutionized the food industry… because companies stopped asking their research teams to find the perfect food and started to ask them to find the best food for a cluster of people. The right question was not “what is the perfect spaghetti sauce?”(or mustard or soda.) The right question was “which varieties of spaghetti sauce greatly appealed to large groups of people?” The result was more food options, happier customers and increased revenue.
Monday, April 11, 2011
Communication: Changing Behaviors
I just received my copy of the Journal of the American Medical Informatics Association. In it, I found an article that is very revealing with regard to communication: Actionable reminders did not improve performance over passive reminders for overdue tests in the primary care setting (abstract). In the article, El-Kareh et al. investigated the impact of altering a passive reminder to be a passive reminder that facilitated direct ordering of recommended tests. The reminders created highly sensitive and specific actionable knowledge from the clinic's electronic health records. The new "enhanced" reminders had absolutely no affect on the screening rates of bone density exams, HgA1C, and LDL monitoring; in fact, they may have decreased the rate of screening mammograms.
From a health IT standpoint, this outcome makes no sense whatsoever. The software provided streamlined functionality to place orders and reminders were based on reliable knowledge. However, a psychology major would have predicted these results. So, why didn't the enhanced alerts work?
They did not work because they did not effect the physician's intention to perform the preventative services. But in order to understand this, you must first understand the theory of planned behavior.
The theory of planned behavior, like most psychological theories, is fairly complex. I like how Wikipedia explains it here, if you are interested, but basically, what it comes down to is this: when a person has the option of whether or not to do an action, their choice to act or not act is dependent upon their intention to act.
The intention to perform a behavior is the summation of three components: a person's attitude towards the behavior; the subjective norm; perceived behavioral control. Increasing someone's intention to perform the behavior increases the frequency of the behavior. So how do you increase a person's intention to perform a behavior? By changing their attitude towards the behavior, their perception of how others view the behavior, and the beliefs about their ability to complete the action.
The "enhanced" reminder studied by El-Kareh et al. did not alter any of the three components needed to increase intention. Both the basic reminder and the enhanced reminder were passive meaning the physician's perception of how colleagues felt about him/her completing preventative screening would not have changed. Streamlining the ability to place orders would not effect the physician's attitude towards providing screening to his patients. It is possible that the perceived ability to complete screening would have been increased due to increased ease of placing orders; however, the authors report that 79% of the physicians almost never used the system or were unaware of the functionality, despite receiving training on the new reminders. Thus, it is not surprising that the enhanced reminders did not result in improved care.
When you are sharing actionable knowledge with the intention of effecting behavior, you will be most effective when you keep the three tenets of intention in mind. For instance, in our VTE project, we purposely addressed each of these tenets in our interventions. We instituted a major, mandatory education program to change the attitude of providers. We used forcing functions and pop-up alerts to change the perception of VTE prophylaxis and reinforce the perceived importance of stratifying and prophylaxing patients. We embedded guidelines for risk stratification in the risk assessment tool and allowed order entry with a single click from the pop-up alerts to impact beliefs about the ability to complete the behavior.
Using the theory of planned behavior to optimize communication will only be effective if the health information system is designed to be functional for the provider. However, as El-Karah et al. demonstrated, improved functionality without improved communication is often ineffective.
From a health IT standpoint, this outcome makes no sense whatsoever. The software provided streamlined functionality to place orders and reminders were based on reliable knowledge. However, a psychology major would have predicted these results. So, why didn't the enhanced alerts work?
They did not work because they did not effect the physician's intention to perform the preventative services. But in order to understand this, you must first understand the theory of planned behavior.
The theory of planned behavior, like most psychological theories, is fairly complex. I like how Wikipedia explains it here, if you are interested, but basically, what it comes down to is this: when a person has the option of whether or not to do an action, their choice to act or not act is dependent upon their intention to act.
The intention to perform a behavior is the summation of three components: a person's attitude towards the behavior; the subjective norm; perceived behavioral control. Increasing someone's intention to perform the behavior increases the frequency of the behavior. So how do you increase a person's intention to perform a behavior? By changing their attitude towards the behavior, their perception of how others view the behavior, and the beliefs about their ability to complete the action.
The "enhanced" reminder studied by El-Kareh et al. did not alter any of the three components needed to increase intention. Both the basic reminder and the enhanced reminder were passive meaning the physician's perception of how colleagues felt about him/her completing preventative screening would not have changed. Streamlining the ability to place orders would not effect the physician's attitude towards providing screening to his patients. It is possible that the perceived ability to complete screening would have been increased due to increased ease of placing orders; however, the authors report that 79% of the physicians almost never used the system or were unaware of the functionality, despite receiving training on the new reminders. Thus, it is not surprising that the enhanced reminders did not result in improved care.
When you are sharing actionable knowledge with the intention of effecting behavior, you will be most effective when you keep the three tenets of intention in mind. For instance, in our VTE project, we purposely addressed each of these tenets in our interventions. We instituted a major, mandatory education program to change the attitude of providers. We used forcing functions and pop-up alerts to change the perception of VTE prophylaxis and reinforce the perceived importance of stratifying and prophylaxing patients. We embedded guidelines for risk stratification in the risk assessment tool and allowed order entry with a single click from the pop-up alerts to impact beliefs about the ability to complete the behavior.
Using the theory of planned behavior to optimize communication will only be effective if the health information system is designed to be functional for the provider. However, as El-Karah et al. demonstrated, improved functionality without improved communication is often ineffective.
Thursday, April 7, 2011
Communication: The Basics
One of the most fundamental elements of communication is using the same vocabulary so that each participant can understand the others. The best example I have of this comes from my 3 year old daughter.
She was walking all kinds of funky across the living room. I called to her, "Anna, do you have a wedgie?" She turned around, looked me square in the eye, and with all the will in her little body corrected me, "No! My panties are stuck in my butt!"
Clearly, this was a situation where we were not communicating well because we did not have a shared vocabulary. Whether evaluating information to make knowledge, sharing knowledge to effect behavior, or developing understanding in the AURI cycle, all parties involved must have a shared vocabulary. In my experience, this is a common pitfall in HIS implementations and data-driven quality improvement.
So, how do you establish this shared vocabulary in teams utilizing clinical knowledge management processes or quality improvement initiatives, such as with the AURI cycle? Defining terms and establishing metrics must be your first order of business in any of these projects; there's really no point in participating in these processes if the participants are not able to equally engage and speak with a common vocabulary.
As clinicians, we all understood what a VTE was, but in the first meetings of the VTE team, we had to define VTE clearly and establish which VTEs would be included in our metrics. For instance, would we count hospital-acquired VTEs in upper extremities that were associated with PICCs or other central lines as nosocomial VTEs and include them as a target of our interventions? We also had to decide if we would include VTEs discovered as outpatients and during readmissions or only those discovered during a single hospital admission as part of our intervention. Ultimately, we chose to include all VTEs, regardless of physical location or the clinical setting in which it was discovered.
As you can see, making another choice would have changed our evaluation of metrics, knowledge, and understanding. The interventions described in the VTE project were the result of the definitions and metrics we agreed upon as a team. If we had not established a common vocabulary, we would have experienced significant delays in the progress of our project, and possibly have been significantly less successful than we were.
Keep in mind that communication is dynamic and anytime a new member enters, you need to ensure you re-establish the shared vocabulary. For instance, when writing out the above story, my 8 year old daughter read it. Then she turned to her mother and asked, "Momma, what is a weed-ghee?"
In my next post, I will outline the science behind communicating with the goal of effecting volitional behavior.
She was walking all kinds of funky across the living room. I called to her, "Anna, do you have a wedgie?" She turned around, looked me square in the eye, and with all the will in her little body corrected me, "No! My panties are stuck in my butt!"
Clearly, this was a situation where we were not communicating well because we did not have a shared vocabulary. Whether evaluating information to make knowledge, sharing knowledge to effect behavior, or developing understanding in the AURI cycle, all parties involved must have a shared vocabulary. In my experience, this is a common pitfall in HIS implementations and data-driven quality improvement.
So, how do you establish this shared vocabulary in teams utilizing clinical knowledge management processes or quality improvement initiatives, such as with the AURI cycle? Defining terms and establishing metrics must be your first order of business in any of these projects; there's really no point in participating in these processes if the participants are not able to equally engage and speak with a common vocabulary.
As clinicians, we all understood what a VTE was, but in the first meetings of the VTE team, we had to define VTE clearly and establish which VTEs would be included in our metrics. For instance, would we count hospital-acquired VTEs in upper extremities that were associated with PICCs or other central lines as nosocomial VTEs and include them as a target of our interventions? We also had to decide if we would include VTEs discovered as outpatients and during readmissions or only those discovered during a single hospital admission as part of our intervention. Ultimately, we chose to include all VTEs, regardless of physical location or the clinical setting in which it was discovered.
As you can see, making another choice would have changed our evaluation of metrics, knowledge, and understanding. The interventions described in the VTE project were the result of the definitions and metrics we agreed upon as a team. If we had not established a common vocabulary, we would have experienced significant delays in the progress of our project, and possibly have been significantly less successful than we were.
Keep in mind that communication is dynamic and anytime a new member enters, you need to ensure you re-establish the shared vocabulary. For instance, when writing out the above story, my 8 year old daughter read it. Then she turned to her mother and asked, "Momma, what is a weed-ghee?"
In my next post, I will outline the science behind communicating with the goal of effecting volitional behavior.
Monday, April 4, 2011
Success Story: VTE Reduction
As part of the QualityBLUE Pay for Performance partnership between Highmark Blue Cross Blue Shield and Penn State Hershey Medical Center, hospital-acquired venous thromboembolism (VTE) was identified as an area for quality improvement. This coincided with the release of the latest ACCP Guidelines on Antithrombolytic and Thrombolytic Therapy (8th Ed.) in the summer of 2008.
An interdisciplinary team of nurses, physicians, pharmacists, quality improvement specialists, and informatics specialists was assembled to determine how to implement the new guidelines at HMC. Through the use of Clinical Knowledge Management and the Analyze-Understand-Redesign-Implement cycle, we made tremendous gains in this QI project, including a sustained 25% reduction in nosocomial VTEs, reduced mortality associated with nosocomial VTEs, and cost avoidance estimated at $2-4 million annually. Let me walk you through the use of the CKM process and the AURI cycle in this QI initiative.
The CKM Process
Capturing & Classifying: At the time of the project initiation, HMC had been using an electronic medical record (EMR) with computerized physician order entry (CPOE) for greater than two years. This mean that in addition to the best practices outlined in the ACCP Guidelines, we also had data from greater than 50,000 inpatient visits available to us. This included: risk stratification data like demographics and clinical conditions; use of pharmocologic and non-pharmocologic prophylaxis; time elapsed from admission to first prophylaxis dose; rate of occurrence of nosocomial VTE.
Retrieving: Queries were developed to gather information about current VTE prophylaxis behavior from HMC's clinical database. Results from the queries were transferred into Excel spreadsheets.
Evaluating: Rates and timing of appropriate prophylaxis and rates of development of nosocomial VTEs were determined to identify gaps between current recommendations and current HMC practice. Underutilization of risk-scoring at admission as well as underutilization of pharmacologic and mechanical prophylaxis were identified. Inappropriate risk stratification was common as was inappropriate use of prophylaxis.
Concurrently, the pharmacists and clinicians from both Medical and Surgical services condensed the ACCP Guidelines to an easy-reference pocket card that contained risk stratification guidelines and appropriate treatment options.
Sharing: Required education for all pharmacists, physicians, and nurses was provided along with the quick-reference pocket cards. The education reviewed the new ACCP Guidelines as well as required changes to HMC practice.
Action: Based on the education and availability of the pocket cards, there was a modest improvement in guideline compliance and a slight decrease in hospital acquired VTEs.
A second cycle of the CKM process was then initiated. The data that was captured and classified after the roll-out of the education and pocket cards was retrieved and evaluated. It was determined that significant opportunities for improvement were as of yet untapped. It was also determined that a more structured and standardized approach was needed to accommodate the resident learning curve.
The sharing step was multifaceted. A clinical decision support (CDS) tool was created that included forcing functions at the time of admission that required VTE risk stratification on all patients. Residents were provided with the stratification criteria at the time of the risk assessment. Interactive alerts were developed to present providers with prophylaxis guidelines based on patient risk stratification at the time of order entry. Providers were required to either place appropriate prophylaxis orders or document contraindications.
The resulting actions from the providers were immediate. There was an increase in appropriate prophylaxis and a decrease in nosocomial VTEs. This reduction has been sustained for two years and through two intern classes. Partial results were presented at the Society of Medical Decision Making Annual Conference in October, 2010.
The AURI Cycle
As you can see from the above, the traditional Plan-Do-Study-Act (PDSA) cycle was not followed during this initiative. Most notably, no system changes were made until the data was carefully analyzed and understood by all team members. However, the AURI cycle was extremely effective in producing sustained behavioral change and improved outcomes. Let's examine how the steps of the CKM process fit into the AURI cycle.
Analyze: The retrieval and evaluation of both internal and external data to create new knowledge comprises this part of the AURI cycle. In this QI project, retrieving and evaluating data from HMC's EMR and studying and condensing the ACCP Guidelines represents the analyze component. The new knowledge gained from the analyze phase was three-fold: the correct prophylaxis choices were often clear to experienced clinicians but not to inexperienced residents; ideally, a single drug would be suggested for pharmacoprophylaxis; risk stratification was rarely performed at admission.
Understand: Sharing the new knowledge derived from the analysis phase with the quality improvement team comprises this part of the AURI cycle. The team identified barriers such as baseline resident knowledge, inability to use a single low-molecular weight heparin, insufficient availability of mechanical prophylaxis devices, and concerns for the feasibility of improvement with voluntary compliance.
Redesign: The redesign phase is a result of the CKM process rather than a step in the CKM process. The new standard of practice at HMC that involved risk stratification of all patients at the time of admission based on a standard risk stratification system as well as the testing of multiple models of mechanical prophylaxis devices by the Department of Nursing comprises this part of the AURI cycle.
Implementation: Sharing new knowledge derived from the analysis phase with all clinicians and other stakeholders comprises this part of the AURI cycle. The first barrier to success, baseline resident knowledge, was tackled during the roll-out of required education. Additionally, the Operations Department purchased an adequate number of machines and made them easily accessible. The EHR was modified to capture the use of mechanical prophylaxis devices and the use of pharmacoprophylaxis continued to be captured. We were not able to capture the timing of the risk assessment, which unfortunately, had to be done manually on a small sample of the patients.
Despite seeing modest gains, a second AURI cycle was needed to address the remaining two barriers. As noted above, the analysis phase identified the need for a more structured, standardized approach. The understanding phase led to the new knowledge that risk assessment needed to be required instead of voluntary, prophylaxis needed to be simplified, and appropriate guidelines needed to be shared in real-time with residents. The redesign phase resulted in the clinical decision support tool described above as well as a policy change that Pharmacy would substitute appropriate low-molecular weight heparin for patients with renal failure. The implement phase was the education and go-live of the CDS tool.
The time elapsed from the beginning of the first AURI cycle to the implementation of the second AURI cycle was only 9 months. As you can see, the CKM process and the AURI cycle can allow for rapid institutional improvement and identification of barriers to improvement. The success of this project highlights the way that efficient quality improvement methodologies result in significant financial gains as well as reduced morbidity and mortality.
An interdisciplinary team of nurses, physicians, pharmacists, quality improvement specialists, and informatics specialists was assembled to determine how to implement the new guidelines at HMC. Through the use of Clinical Knowledge Management and the Analyze-Understand-Redesign-Implement cycle, we made tremendous gains in this QI project, including a sustained 25% reduction in nosocomial VTEs, reduced mortality associated with nosocomial VTEs, and cost avoidance estimated at $2-4 million annually. Let me walk you through the use of the CKM process and the AURI cycle in this QI initiative.
The CKM Process
Capturing & Classifying: At the time of the project initiation, HMC had been using an electronic medical record (EMR) with computerized physician order entry (CPOE) for greater than two years. This mean that in addition to the best practices outlined in the ACCP Guidelines, we also had data from greater than 50,000 inpatient visits available to us. This included: risk stratification data like demographics and clinical conditions; use of pharmocologic and non-pharmocologic prophylaxis; time elapsed from admission to first prophylaxis dose; rate of occurrence of nosocomial VTE.
Retrieving: Queries were developed to gather information about current VTE prophylaxis behavior from HMC's clinical database. Results from the queries were transferred into Excel spreadsheets.
Evaluating: Rates and timing of appropriate prophylaxis and rates of development of nosocomial VTEs were determined to identify gaps between current recommendations and current HMC practice. Underutilization of risk-scoring at admission as well as underutilization of pharmacologic and mechanical prophylaxis were identified. Inappropriate risk stratification was common as was inappropriate use of prophylaxis.
Concurrently, the pharmacists and clinicians from both Medical and Surgical services condensed the ACCP Guidelines to an easy-reference pocket card that contained risk stratification guidelines and appropriate treatment options.
Sharing: Required education for all pharmacists, physicians, and nurses was provided along with the quick-reference pocket cards. The education reviewed the new ACCP Guidelines as well as required changes to HMC practice.
Action: Based on the education and availability of the pocket cards, there was a modest improvement in guideline compliance and a slight decrease in hospital acquired VTEs.
A second cycle of the CKM process was then initiated. The data that was captured and classified after the roll-out of the education and pocket cards was retrieved and evaluated. It was determined that significant opportunities for improvement were as of yet untapped. It was also determined that a more structured and standardized approach was needed to accommodate the resident learning curve.
The sharing step was multifaceted. A clinical decision support (CDS) tool was created that included forcing functions at the time of admission that required VTE risk stratification on all patients. Residents were provided with the stratification criteria at the time of the risk assessment. Interactive alerts were developed to present providers with prophylaxis guidelines based on patient risk stratification at the time of order entry. Providers were required to either place appropriate prophylaxis orders or document contraindications.
The resulting actions from the providers were immediate. There was an increase in appropriate prophylaxis and a decrease in nosocomial VTEs. This reduction has been sustained for two years and through two intern classes. Partial results were presented at the Society of Medical Decision Making Annual Conference in October, 2010.
The AURI Cycle
As you can see from the above, the traditional Plan-Do-Study-Act (PDSA) cycle was not followed during this initiative. Most notably, no system changes were made until the data was carefully analyzed and understood by all team members. However, the AURI cycle was extremely effective in producing sustained behavioral change and improved outcomes. Let's examine how the steps of the CKM process fit into the AURI cycle.
Analyze: The retrieval and evaluation of both internal and external data to create new knowledge comprises this part of the AURI cycle. In this QI project, retrieving and evaluating data from HMC's EMR and studying and condensing the ACCP Guidelines represents the analyze component. The new knowledge gained from the analyze phase was three-fold: the correct prophylaxis choices were often clear to experienced clinicians but not to inexperienced residents; ideally, a single drug would be suggested for pharmacoprophylaxis; risk stratification was rarely performed at admission.
Understand: Sharing the new knowledge derived from the analysis phase with the quality improvement team comprises this part of the AURI cycle. The team identified barriers such as baseline resident knowledge, inability to use a single low-molecular weight heparin, insufficient availability of mechanical prophylaxis devices, and concerns for the feasibility of improvement with voluntary compliance.
Redesign: The redesign phase is a result of the CKM process rather than a step in the CKM process. The new standard of practice at HMC that involved risk stratification of all patients at the time of admission based on a standard risk stratification system as well as the testing of multiple models of mechanical prophylaxis devices by the Department of Nursing comprises this part of the AURI cycle.
Implementation: Sharing new knowledge derived from the analysis phase with all clinicians and other stakeholders comprises this part of the AURI cycle. The first barrier to success, baseline resident knowledge, was tackled during the roll-out of required education. Additionally, the Operations Department purchased an adequate number of machines and made them easily accessible. The EHR was modified to capture the use of mechanical prophylaxis devices and the use of pharmacoprophylaxis continued to be captured. We were not able to capture the timing of the risk assessment, which unfortunately, had to be done manually on a small sample of the patients.
Despite seeing modest gains, a second AURI cycle was needed to address the remaining two barriers. As noted above, the analysis phase identified the need for a more structured, standardized approach. The understanding phase led to the new knowledge that risk assessment needed to be required instead of voluntary, prophylaxis needed to be simplified, and appropriate guidelines needed to be shared in real-time with residents. The redesign phase resulted in the clinical decision support tool described above as well as a policy change that Pharmacy would substitute appropriate low-molecular weight heparin for patients with renal failure. The implement phase was the education and go-live of the CDS tool.
The time elapsed from the beginning of the first AURI cycle to the implementation of the second AURI cycle was only 9 months. As you can see, the CKM process and the AURI cycle can allow for rapid institutional improvement and identification of barriers to improvement. The success of this project highlights the way that efficient quality improvement methodologies result in significant financial gains as well as reduced morbidity and mortality.
Subscribe to:
Posts (Atom)