HVI Analytics: Leveraging Big Data to Identify Opportunities to Improve Patient Care

January 30, 2020

UPMC is home to one of the nation’s largest cardiology divisions, with more than 100 cardiologists working at the 40 hospitals in the UPMC system and various office locations. These clinicians perform more than 23,000 diagnostic and interventional procedures each year. In the process, they generate a massive quantity of clinical data, which is stored in the electronic health record (EHR). Suresh Mulukutla, MD, an interventional cardiologist in the UPMC Heart and Vascular Institute (HVI), is the Director of Analytics for the HVI-Cardiology. Dr. Mulukutla and his team integrate inpatient and outpatient data from across the UPMC system, aggregate this data, and use it to identify opportunities to improve care.

When Dr. Mulukutla first participated in the National Heart, Lung, and Blood Institute (NHLBI) Dynamic Registry in the early 2000s, he began what he describes as a “long-standing relationship with data.” He and Oscar Marroquin, MD, a cardiologist in the HVI and chief clinical analytics officer, UPMC Health Services Division, recognized the opportunity to replicate the power of the NHLBI multicenter registry within UPMC. According to Dr. Mulukutla, clinical analytics allows us to “integrate and harmonize all of the data across the health system,” and provides a host of opportunities to understand cardiac pathologies and improve care. In the HVI analytics division, Dr. Mulukutla oversees clinical data specialists and data managers who develop programs to draw out information that the clinicians request and help them visualize it. The HVI analytics division partners closely with UPMC Clinical Analytics, which is headed by Dr. Marroquin. The HVI also employs a team of statisticians proficient in the sophisticated analyses required to decipher big data sets. 

Harnessing the Power of the EHR

Although the EHR was not specifically created for physicians to use to learn and improve, it houses a wealth of information.¹ UPMC has invested in transforming the utility of the EHR through the clinical analytics program. Business application tools sit on the EHR and allow Dr. Mulukutla to visualize inpatient, outpatient, pharmacy, and laboratory data and toggle between different ways of viewing this data in real time. Dr. Mulukutla and his analytics team can stratify patients by condition, by patient characteristics, or by operational aspects. With such a large system and so many different providers, there is inherent variability in care. However, using clinical analytics, this variability can be used to identify opportunities. When differences are identified that lead to better outcomes, they can then be applied systemically at all UPMC facilities for heart and vascular care.

There are several challenges to large-scale data mining from the EHR. The sheer volume of data is one challenge; this data volume also is a tool. The incredible volume of data that Dr. Mulukutla and his colleagues can access is key to their success. Variability in care is a challenge but allows them to identify opportunities. Additionally, knowing which measures are the most helpful and how to tease out subtle differences is essential. UPMC is unique in its support of helping physicians use data from the EHR. Additionally, many other hospitals lack either the integrated data or the patient volume available at UPMC.

The HVI analytics division can select specific patient populations with conditions of interest for analysis. They can quickly determine how many patients are seen by UPMC physicians during a given period for the condition in question and then examine differences in hospitalization rates, mortality, and other outcomes measures. 

For example, in the past, it would take weeks to compile a list of UPMC patients who had undergone a coronary intervention, such as stent placement, who also have diabetes and heart failure. Now this information can be compiled within a day. Translating published data from clinical trials into real-world practice normally can take months or years. The use of clinical analytics and large EHR data sets has the potential to speed up this process. For example, the HVI clinical analytics division can narrow down the data to identify groups of patients with high cardio-vascular risk who also may meet specific criteria to be eligible for novel therapies.

Projects in MV-CAD, Atrial Fibrillation (AFib), and Valvular Intervention

One population that HVI analytics has examined in detail is patients with multivessel coronary artery disease (MV-CAD). Coronary artery bypass grafting (CABG) and percutaneous coronary intervention (PCI) are both safe and established treatment options for MV-CAD, and the optimal treatment is still a matter of debate. Multiple factors drive the decision of which therapy to use, but there have been few studies directly comparing the outcomes of CABG and PCI in routine clinical practice in patients with cardiac disease of varying complexity and with differing comorbidities. 
Dr. Mulukutla and his team, in collaboration with Arman Kilic, MD, from the Department of Cardiothoracic Surgery, leveraged the ability to dissect the vast data available from the EHR at UPMC and studied outcomes after treatment of MV-CAD in more than 6,000 patients at five UPMC hospitals. Mortality, hospital readmissions, and major adverse cardiovascular events (MACE) were compared after CABG and PCI in two propensity-matched groups of more than 800 patients who would be eligible for either therapy. Dr. Mulukutla and his colleagues found that CABG had a mortality benefit in the overall population (estimated one-year mortality of 11.5% for PCI and 7.2% for CABG, p < 0.001) and in every subset population examined independent of the comorbidities of diabetes and heart failure.² CABG also resulted in fewer readmissions (38.4% vs. 28.1%, p < 0.001), fewer patients who required revascularization (6.7% vs. 1.0%, p < 0.001), and fewer adverse cardiovascular events. 

Prior data supported directing patients with MV-CAD and diabetes or heart failure to CABG. Dr. Mulukutla’s study provided an important analysis of outcomes under real-world conditions with contemporary treatment options and suggests that CABG should be considered more frequently for patients with MV-CAD. A multidisciplinary approach from a revascularization heart team may help patients make the best decisions regarding their treatment options.³ 

Dr. Mulukutla is working with Sandeep Jain, MD, director of the Atrial Fibrillation Center of Excellence, in using clinical analytics to ensure that UPMC patients are treated according to evidence-based guidelines that have been put forth by professional societies in cardiovascular medicine. One example is using large-scale data mining to increase the proportion of patients with AFib who are on anticoagulation therapy, which is recommended by several clinical guidelines to reduce the risk of a cerebrovascular accident (CVA, stroke).^4,5 The EHR is being queried for all patients with AFib who are not being treated with anticoagulation therapy. Although some patients may have a valid reason to abstain from anticoagulation therapy, an opportunity is present to evaluate all patients to see if they are receiving the correct care. Dr. Mulukutla can identify these patients and then access scheduling information to know when they will next be seen by a primary care doctor or specialist in the UPMC system. His team then will contact physicians and ask them to evaluate the patient’s care in light of their AFib and anticoagulation therapy status.

Similarly, the HVI analytics division is working on identifying patients who would likely benefit from a valvular intervention or a defibrillator based on evidence-based guidelines. Historically, these patients present to the cardiology clinic after a referral from another physician. This is not a proactive approach to delivering care that will likely improve the patient’s health; it requires the cardiologist to wait for actions both by the patient and by another physician. If the EHR can be used to identify patients meeting guideline criteria for a valvular intervention, the cardiologist can then contact the primary care providers and facilitate an evaluation of the patients. The HVI analytics team currently is determining the best ways to collaborate with primary care physicians and help them efficiently care for their patients. HVI analytics is developing a process that does not overload busy physicians but allows them to have directed information for these patients.

Combatting Unplanned Readmissions

Another example of how Dr. Mulukutla and his colleagues have used data to care for patients in a smarter way is by developing a tool to assess the risk of unplanned hospital readmission at the outset of a hospital stay. Unplanned readmissions are a big challenge for every hospital, and the UPMC team has focused on lowering seven-day and 30-day readmission rates.

Dr. Mulukutla, Dr. Marroquin, and their colleagues found that if patients are seen in the clinic within 30 days of discharge, there is a 50% reduction in rehospitalization within 30 days.6 Using data from the EHR, Dr. Mulukutla and colleagues can identify patients at high-risk for readmission when they are first admitted and intervene before they are discharged. At several UPMC hospitals, discharge planning has become more efficient. Follow-up appointments in the clinic can be scheduled as early as the first day of a hospital admission. Readmission rates have since declined for cardiology patients. Additionally, these efforts allow UPMC to better use its resources. Larger implementation, further analysis, and publication of these results are in progress. 

Conclusion

Currently, Dr. Mulukutla is using the power of HVI analytics to address the important issue of improving patient access to UPMC cardiologists. UPMC is looking for opportunities at an operational level to serve more patients. Using big data from the EHR, Dr. Mulukutla is addressing ways that UPMC can run outpatient clinics to accommodate more patients and locating subpopulations within the community who may not have easy access to the UPMC outpatient clinics. The results of these analyses will allow UPMC to strategically grow and provide cardiology care to more patients. 
HVI analytics and UPMC Clinical Analytics reflect the combined commitment of UPMC to developing cutting-edge technology that harnesses patient-derived EHR data to improve clinical care and outcomes. 

References and Further Reading

1. Masri A, Althouse AD, McKibben J, Lee JS, Mulukutla SR. Limitations of Administrative Data for Studying Patients Hospitalized With Heart Failure. Ann Intern Med. 2017; 166(12): 916-917.
2. Mulukutla SR, Gleason TG, Sharbaugh M, Sultan I, Marroquin OC, Thoma F, et al. Coronary Bypass Versus Percutaneous Revascularization in Multivessel Coronary Artery Disease. Ann Thorac Surg. 2019; 108(2): 474-480.
3. Sanchez CE, Dota A, Badhwar V, Kliner D, Smith AJ, Chu D, et al. Revascularization Heart Team Recommendations as an Adjunct to Appropriate Use Criteria for Coronary Revascularization in Patients With Complex Coronary Artery Disease. Catheter Cardiovasc Interv. 2016; 88(4): E103-E12.
4. January CT, Wann LS, Calkins H, Chen LY, Cigarroa JE, Cleveland Jr JC, et al. 2019 AHA/ACC/HRS Focused Update of the 2014 AHA/ACC/HRS Guideline for the Management of Patients With Atrial Fibrillation: A Report of the American College of Cardiology/American Heart Association Task Force on Clinical Practice Guidelines and the Heart Rhythm Society. J Am Coll Cardiol. 2019; 74(1): 104-132.
5. January CT, Wann LS, Alpert JS, Calkins H, Cigarroa JE, Cleveland Jr JC, et al. 2014 AHA/ACC/HRS Guideline for the Management of Patients With Atrial Fibrillation: A Report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines and the Heart Rhythm Society. J Am Coll Cardiol. 2014; 64(21): e1-76.
6. Hagland M. At UPMC, Turbo-Charging Quality Improvement Efforts Through Data Analytics. Healthcare Innovations. 2018.
7. Galetsi P, Katsaliaki K, Kumar S. Values, Challenges and Future Directions of Big Data Analytics in Healthcare: A Systematic Review. Soc Sci Med. 2019; 241: 112533.