Improving Flow, Quality of Care, and Decision Making: A Data Driven Analysis of LPCH Surgical Heart Center
Surgical demand at Lucile Packard Children’s Hospital heart center has been steadily increasing over the past several years, in large part due to its cutting-edge pediatric cardiothoracic surgery program that is known for performing complex unifocalization and pulmonary artery reconstruction procedures for patients. To keep up with demand, the heart center is operating at nearly full capacity at all times, creating a systematic bottleneck in the cardiovascular intensive care unit (CVICU). More than other intensive care units, utilization of beds in the CVICU can fluctuate wildly from day-to-day, due to the high variability in the recovery time for a pediatric cardiovascular surgical patient, which can range from spending one night in the CVICU, to spending hundreds of nights. Compounding this issue, the surgical schedule is currently not optimized to account for patient complexity, so the case-mix of low, medium, and high complexity patients in the CVICU could have high variability from week-to-week. Our project aims to address some of these issues, identify ways to improve the quality of care for patients, and provide insights for better decision making regarding scheduling.
To this end, we first conducted a historical analysis of the frequency and volume of various patient populations flowing into the CVICU according to the current way leadership at the heart center tracks them: transfers, orbiters, cardiac surgical cases, and medical inpatients/transplants. Many of these inflow sources are highly variable making them hard to predict. Last minute or “urgent” surgical cases are the most difficult to accommodate, so for our historical analysis, we filtered our data to examine cases that were entered in the system as needing surgery within the next 24 hours. Using this filtered data we discovered that on average, the heart center can expect ~2 “urgent transfer” surgical cases per week, ~2 “urgent orbiter” surgical cases per week, ~1 heart transplant case every other week, and ~6 cardiovascular surgical cases (excluding CATH/EP cases) per week. By studying the statistical properties of patient arrivals to the Heart Center, we created a reference graphic for hospital leadership to use, that depicts the frequency and volume the heart center can expect per week, for each of the different patient populations.
Next, our team began to analyze the relationship between occasional “upticks” in emergent cases that happen throughout the year, and an increase in the number of cancellations in the heart center. Specifically, we noted that in early February of 2021, the heart center had 10 cardiovascular OR / CATH cancellations in one week, 6x the weekly average (1.6), and nearly 12 percent of the total number of cancellations they had in all of fiscal year 2020 (86). Interestingly, we noticed that the spike in cancellations between February 1 and February 7 coincided with the CVICU receiving 7 “urgent transfer” surgical cases that same week, an event that according to our earlier analysis (based on the statistical properties of patient arrivals to the Heart Center) should only occur about once a year.
Conducting a deeper dive into the association between the number of “urgent transfer” surgical cases every week, and the number of cancellations, we found that according to the data available from 2019-09-01 to 2021-04-05, many (greater than 4) “urgent transfer” surgical cases per week is associated with more cardiovascular OR / CATH surgical cancellations. In short, we were able to identify a new risk factor that is strongly associated with an increase in CV OR / CATH surgical cancellations, and the hope is that Heart Center leadership can use this information as a planning tool, so when they have already reached a certain amount of “urgent transfer surgical cases” in a given week, they can start to explore ways to mitigate future “urgent transfer surgical case” arrivals either by diverting future transfer requests to other capable hospitals, increasing staffing, or possibly creating extra capacity in the schedule.
This association between many “urgent transfer surgical case” arrivals and an increase in cancellations is particularly troubling, given that over time, there has been an increase in the share of beds in the CVICU that are occupied by emergent patients. If this trend of having more and more beds occupied by emergent surgical patients persists, the Heart Center may see a corresponding increase in the number of cancellations, if no mitigating measures are taken.
Finally, following multiple interviews with project sponsors and healthcare providers, our team started to notice a common idea manifest: that individual patient complexity, along with the overall unit complexity of the heart center might be associated with an increase or decrease in quality of care for patients. With this, we formulated a hypothesis that served to study whether the quality of care for low risk patients is affected by the number of high complexity patients in the unit. We hypothesized that the quality of care will decrease when there is a high number of complex patients in the unit. To test this hypothesis, our team analyzed various proxies for quality of care and patient complexity, ultimately choosing to use STAT category scores (which are assigned based on the risk for mortality associated with a particular heart surgery procedure) as a proxy for patient and unit complexity, and hospital acquired conditions (HACs) and patient discharge status as proxies for quality of care. We defined a patient with an associated STAT score of 1 or 2 as a “low complexity” patient, and a patient with an associated STAT score of 3, 4, or 5 as a “high complexity” patient. We used HAC and patient discharge data to formulate a risk score for each patient, for each week they were in the CVICU following surgery, and after thorough investigation and data analysis, we found that “high complexity” patients are more likely to have negative quality of care outcomes. This plot shows that under average unit complexity conditions, the predicted probability of a “low complexity” patient getting a HAC or passing away is about 3.8 percent, and the predicted probability of a “high complexity” patient getting a HAC or passing away is about 7.2 percent. Our statistical testing was inconclusive regarding the impact unit complexity has on negative quality of care outcomes, so further testing is required to determine if an association exists. These results suggest that based on (1) our chosen proxies for complexity and quality of care; (2) the data we had access to; and (3) our chosen methodology; the most significant predictor for negative quality of care outcomes is the individual complexity of the patients themselves.
We would like to thank Dr. David Scheinker, Dr. Andy Shin, Blake Gentile, Yuan Shi, Saied Mehdian, Jonaelle Lucas, Andrew Nowak, the MS&E 463 teaching team, Heart Center “Flow” Team, and the leadership, administration, clinicians, and staff of the Lucile Packard Children’s Hospital who provided us with data and extensive feedback through interviews.