Predicting readmission risk following coronary artery bypass surgery at the time of admission

Leveraging Machine Learning to Predict 30-Day Hospital Readmission after Cardiac Surgery

BACKGROUND

Hospital readmission within 30 days of discharge is a well-studied outcome. Predicting readmission after cardiac surgery, however, is notoriously challenging; the best-performing models in the literature have AUCs around .65. A reliable predictive model would enable clinicians to identify patients at-risk for readmission and develop prevention strategies.

METHODS

We analyzed our institution's STS Adult Cardiac Surgery Database (STS), augmented with electronic medical record (EMR) data. Predictors included demographics, pre-operative comorbidities, proxies for intra-operative risk, indicators of post-operative complications, and timeseries-derived variables. We trained several machine learning models, evaluating each on a held-out test set.

RESULTS

Our analysis cohort consisted of 4,924 cases from 2011-2016. 723 (14.7%) were readmitted within 30 days of discharge. Our models included 141 STS-derived and 24 EMR-derived variables. A random forest model performed best, with test AUC .76 (95% CI: (.73, .79)). Using exclusively pre-operative variables, as in STS calculated risk scores, degraded the AUC: .64 (95% CI: .60, .68). Key predictors included length of stay (12.5x more important than the average variable) and whether the patient was discharged to a rehab facility (11.2x).

CONCLUSIONS

Our approach, augmenting STS variables with EMR data and employing flexible machine learning modeling, yielded state-of-the-art performance for predicting 30-day readmission. Separately, the importance of variables not directly related to in-patient care, such as discharge location, amplifies questions about the efficacy of assessing care quality via readmissions.

Patient and hospital factors associated with 30-day readmissions after coronary artery bypass graft (CABG) surgery: a systematic review and meta-analysis

BACKGROUND

Readmission after coronary artery bypass graft (CABG) surgery is associated with adverse outcomes and significant healthcare costs, and 30-day readmission rate is considered as a key indicator of the quality of care. This study aims to: quantify rates of readmission within 30 days of CABG surgery; explore the causes of readmissions; and investigate how patient- and hospital-level factors influence readmission.

METHODS

We conducted systematic searches (until June 2020) of PubMed and Embase databases to retrieve observational studies that investigated readmission after CABG. Random effect meta-analysis was used to estimate rates and predictors of 30-day post-CABG readmission.

RESULTS

In total, 53 studies meeting inclusion criteria were identified, including 8,937,457 CABG patients. The pooled 30-day readmission rate was 12.9% (95% CI: 11.3-14.4%). The most frequently reported underlying causes of 30-day readmissions were infection and sepsis (range: 6.9-28.6%), cardiac arrythmia (4.5-26.7%), congestive heart failure (5.8-15.7%), respiratory complications (1-20%) and pleural effusion (0.4-22.5%). Individual factors including age (OR per 10-year increase 1.12 [95% CI: 1.04-1.20]), female sex (OR 1.29 [1.25-1.34]), non-White race (OR 1.15 [1.10-1.21]), not having private insurance (OR 1.39 [1.27-1.51]) and various comorbidities were strongly associated with 30-day readmission rates, whereas associations with hospital factors including hospital CABG volume, surgeon CABG volume, hospital size, hospital quality and teaching status were inconsistent.

CONCLUSIONS

Nearly 1 in 8 CABG patients are readmitted within 30 days and the majority of these are readmitted for noncardiac causes. Readmission rates are strongly influenced by patients' demographic and clinical characteristics, but not by broadly defined hospital characteristics.

Predicting patient outcomes after far lateral lumbar discectomy

INTRODUCTION

The LACE+ (Length of Stay, Acuity of Admission, Charlson Comorbidity Index (CCI) Score, Emergency Department (ED) visits within the previous 6 months) index has never been tested in a purely spine surgery population. This study assesses the ability of LACE + to predict adverse patient outcomes following discectomy for far lateral disc herniation (FLDH).

PATIENTS AND METHODS

Data were obtained for patients (n = 144) who underwent far lateral lumbar discectomy at a single, multi-hospital academic medical center (2013-2020). LACE + scores were calculated for all patients with complete information (n = 100). The influence of confounding variables was assessed and controlled with stepwise regression. Logistic regression was used to test the ability of LACE + to predict risk of unplanned hospital readmission, ED visits, outpatient office visits, and reoperation after surgery.

RESULTS

Mean age of the population was 61.72 ± 11.55 years, 69 (47.9 %) were female, and 126 (87.5 %) were non-Hispanic white. Patients underwent either open (n = 92) or endoscopic (n = 52) surgery. Each point increase in LACE + score significantly predicted, in the 30-day (30D) and 30-90-day (30-90D) post-discharge window, higher risk of readmission (p = 0.005, p = 0.009; respectively) and ED visits (p = 0.045). Increasing LACE + also predicted, in the 30D and 90-day (90D) post-discharge window, risk of reoperation (p = 0.022, p = 0.016; respectively), and repeat neurosurgical intervention (p = 0.026, p = 0.026; respectively). Increasing LACE + score also predicted risk of reoperation (p = 0.011) within 30 days of initial surgery.

CONCLUSIONS

LACE + may be suitable for characterizing risk of adverse perioperative events for patients undergoing far lateral discectomy.

Predictors of Hospital Readmission within 30 Days after Coronary Artery Bypass Grafting: Data Analysis of 2,272 Brazilian Patients

Introduction

In order to reduce readmission rates after coronary artery bypass grafting (CABG), its predictors should be known in different contexts. The objective of this study was to identify predictive factors of hospital readmission within 30 days after CABG in a Brazilian center.

Methods

A secondary analysis of an electronic database of patients submitted to isolated CABG was performed. The relationship between readmission within 30 days and demographic, anthropometric, clinical, and surgery-related characteristics was investigated by univariate analyses. Predictors were identified by multiple logistic regression.

Results

Data from 2,272 patients were included, with an incidence of readmission of 8.6%. The predictors of readmission were brown skin color (Beta=1.613; 95% confidence interval [CI] 1.047-2.458; P=0.030), African-American ethnicity (Beta=0.136; 95% CI 0.019-0.988; P=0.049), chronic kidney disease (Beta=2.214; 95% CI 1.269-3.865; P=0.005), postoperative use of blood products (Beta=1.515; 95% CI 1.101-2.086; P=0.011), chronic obstructive pulmonary disease (Beta=2.095; 95% CI 1.284-3.419; P=0.003), and use of acetylsalicylic acid (Beta=1.418; 95% CI 1.000-2.011; P=0.05). Preoperative antibiotic prophylaxis (Beta=0.742; 95% CI 0.5471.007; P=0.055) was marginally significant.

Conclusion

The predictors identified may support a closer postoperative follow-up and individualized planning for a safe discharge.

Development of a Risk Score to Predict 90-Day Readmission After Coronary Artery Bypass Graft

BACKGROUND

Readmission after coronary artery bypass grafting (CABG) is used for quality metrics and may negatively affect hospital reimbursement. Our objective was to develop a risk score system from a national cohort that can predict 90-day readmission risk for CABG patients.

METHODS

Using the National Readmission Database between 2013 and 2014, we identified 104,930 patients discharged after CABG, for a total of 234,483 patients after weighted analysis. Using structured random sampling, patients were divided into a training set (60%) and test data set (40%). In the training data set, we used multivariable analysis to identify risk factors. A point system risk score was developed based on the odds ratios. Variables with odds ratio less than 1.3 were excluded from the final model to reduce noise. Performance was assessed in the test data set using receiver operator characteristics and accuracy.

RESULTS

In the United States, overall 90-day readmission rate after CABG was 19% (n = 44,559 of 234,483). Nine demographic and clinical variables were identified as important in the training data set. The final risk score ranged from 0 to 52; the 2 largest risks were associated with length of stay greater than 10 days (score = +10) and Medicaid insurance (score = +7). The final model's C-statistic was 0.67. Using an optimal cutoff of 18 points, the accuracy of the risk score was 77%.

CONCLUSIONS

Ninety-day readmission after CABG surgery is frequent. A readmission risk score higher than 18 points predicts readmission in 77% of patients. Based on 9 demographic and clinical factors, this risk score can be used to target high-risk patients for additional postdischarge resources to reduce readmission.

The LACE+ Index as a Predictor of 90-Day Supratentorial Tumor Surgery Outcomes

BACKGROUND

The LACE+ (Length of stay, Acuity of admission, Charlson Comorbidity Index [CCI] score, and Emergency department [ED] visits in the past 6 mo) index risk-prediction tool has never been successfully tested in a neurosurgery population.

OBJECTIVE

To assess the ability of LACE+ to predict adverse outcomes after supratentorial brain tumor surgery.

METHODS

LACE+ scores were retrospectively calculated for all patients (n = 624) who underwent surgery for supratentorial tumors at the University of Pennsylvania Health System (2017-2019). Confounding variables were controlled with coarsened exact matching. The frequency of unplanned hospital readmission, ED visits, and death was compared for patients with different LACE+ score quartiles (Q1, Q2, Q3, and Q4).

RESULTS

A total of 134 patients were matched between Q1 and Q4; 152 patients were matched between Q2 and Q4; and 192 patients were matched between Q3 and Q4. Patients with higher LACE+ scores were significantly more likely to be readmitted within 90 d (90D) of discharge for Q1 vs Q4 (21.88% vs 46.88%, P = .005) and Q2 vs Q4 (27.03% vs 55.41%, P = .001). Patients with larger LACE+ scores also had significantly increased risk of 90D ED visits for Q1 vs Q4 (13.33% vs 30.00%, P = .027) and Q2 vs Q4 (22.54% vs 39.44%, P = .039). LACE+ score also correlated with death within 90D of surgery for Q2 vs Q4 (2.63% vs 15.79%, P = .003) and with death at any point after surgery/during follow-up for Q1 vs Q4 (7.46% vs 28.36%, P = .002), Q2 vs Q4 (15.79% vs 31.58%, P = .011), and Q3 vs Q4 (18.75% vs 31.25%, P = .047).

CONCLUSION

LACE+ may be suitable for characterizing risk of certain perioperative events in a patient population undergoing supratentorial brain tumor resection.

Coronary artery bypass graft readmission rates and risk factors - A retrospective cohort study

BACKGROUND

Hospital readmissions contribute substantially to the overall healthcare cost. Coronary artery bypass graft (CABG) is of particular interest due to its relatively high short-term readmission rates and mean hospital charges.

METHODS

A retrospective review was performed on 2007-2011 data from California, Florida, and New York from the State Inpatient Databases, Healthcare Cost and Utilization Project. All patients ≥18 years of age who underwent isolated CABG and met inclusion/exclusion criteria were included. Insurance status was categorized by Medicaid, Medicare, Private Insurance, Uninsured, and Other. Primary outcomes were unadjusted rates and adjusted odds of readmission at 30- and 90-days. Secondary outcomes included diagnosis at readmission.

RESULTS

A total of 177,229 were included in the analyses after assessing for exclusion criteria. Overall 30-day readmission rate was 16.1%; rates were highest within Medicare (18.4%) and Medicaid (20.2%) groups and lowest in the private insurance group (11.7%; p < 0.0001). Similarly, 90-day rates were highest in Medicare (27.3%) and Medicaid (29.8%) groups and lowest in the private insurance group (17.6%), with an overall 90-day rate of 24.0% (p < 0.0001). The most common 30-day readmission diagnoses were atrial fibrillation (26.7%), pleural effusion (22.5%), and wound infection (17.7%). Medicare patients had the highest proportion of readmissions with atrial fibrillation (31.7%) and pleural effusions (23.3%), while Medicaid patients had the highest proportion of readmissions with wound infections (21.8%). Similar results were found at 90 days. Risk factors for readmission included non-private insurance, age, female sex, non-white race, low median household income, non-routine discharge, length of stay, and certain comorbidities and complications.

CONCLUSIONS

CABG readmission rates remain high and are associated with insurance status and racial and socioeconomic markers. Further investigation is necessary to better delineate the underlying factors that relate racial and socioeconomic disparities to CABG readmissions. Understanding these factors will be key to improving healthcare outcomes and expenditure.

The impact of care management information technology model on quality of care after Coronary Artery Bypass Surgery: "Bridging the Divides"

BACKGROUND

Reducing readmissions and improving metrics of care are a national priority. Supplementing traditional care with care management may improve outcomes. The Bridges program was an initial evaluation of a care management platform (CareLinkHub), supported by information technology (IT) developed to improve the quality and transition of care from hospital to home after Coronary Artery Bypass Surgery (CABG) and reduce readmissions.

METHODS

CareLink is comprised of care managers, patient navigators, pharmacists and physicians. Information to guide care management is guided by a middleware layer to gather information, PLR (ColdLight Solutions, LLC) and presented to CareLink staff on a care management platform, Aerial™ (Medecision). In addition there is an analytic engine to help evaluate and guide care, Neuron™ (Coldlight Solutions, LLC).

RESULTS

The "Bridges" program enrolled a total of 716 CABG patients with 850 admissions from April 2013 through March 2015. The data of the program was compared with those of 1111 CABG patients with 1203 admissions in the 3years prior to the program. No impact was seen with respect to readmissions, Blood Pressure or LDL control. There was no significant improvement in patients' reported outcomes using either the CTM-3 or any of the SAQ-7 scores. Patient follow-up with physicians within 1week of discharge improved during the Bridges years.

CONCLUSIONS

The CareLink hub platform was successfully implemented. Little or no impact on outcome metrics was seen in the short follow-up time.

Surgical data science: The new knowledge domain

Healthcare in general, and surgery/interventional care in particular, is evolving through rapid advances in technology and increasing complexity of care, with the goal of maximizing the quality and value of care. Whereas innovations in diagnostic and therapeutic technologies have driven past improvements in the quality of surgical care, future transformation in care will be enabled by data. Conventional methodologies, such as registry studies, are limited in their scope for discovery and research, extent and complexity of data, breadth of analytical techniques, and translation or integration of research findings into patient care. We foresee the emergence of surgical/interventional data science (SDS) as a key element to addressing these limitations and creating a sustainable path toward evidence-based improvement of interventional healthcare pathways. SDS will create tools to measure, model, and quantify the pathways or processes within the context of patient health states or outcomes and use information gained to inform healthcare decisions, guidelines, best practices, policy, and training, thereby improving the safety and quality of healthcare and its value. Data are pervasive throughout the surgical care pathway; thus, SDS can impact various aspects of care, including prevention, diagnosis, intervention, or postoperative recovery. The existing literature already provides preliminary results, suggesting how a data science approach to surgical decision-making could more accurately predict severe complications using complex data from preoperative, intraoperative, and postoperative contexts, how it could support intraoperative decision-making using both existing knowledge and continuous data streams throughout the surgical care pathway, and how it could enable effective collaboration between human care providers and intelligent technologies. In addition, SDS is poised to play a central role in surgical education, for example, through objective assessments, automated virtual coaching, and robot-assisted active learning of surgical skill. However, the potential for transforming surgical care and training through SDS may only be realized through a cultural shift that not only institutionalizes technology to seamlessly capture data but also assimilates individuals with expertise in data science into clinical research teams. Furthermore, collaboration with industry partners from the inception of the discovery process promotes optimal design of data products as well as their efficient translation and commercialization. As surgery continues to evolve through advances in technology that enhance delivery of care, SDS represents a new knowledge domain to engineer surgical care of the future.