Enhancing Patient Engagement with Machine Learning at a Novel Care Transition Clinic

Enhancing Patient Engagement with Machine Learning at a Novel Care Transition Clinic | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article Enhancing Patient Engagement with Machine Learning at a Novel Care Transition Clinic Seung-Yup Lee, Reid Eagleson, Larry Hearld, Madeline Gibson, Kristine Hearld, and 9 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-4477049/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Objective This study applies predictive analytics to identify patients at risk of missing appointments at a novel post-discharge clinic (PDC) in a large academic health system. Recognizing the critical role of appointment adherence in the success of new clinical ventures, this research aims to inform future targeted interventions to increase appointment adherence. Materials and Methods We analyzed electronic health records (EHR) capturing a wide array of demographic, socio-economic, and clinical variables from 2,168 patients with scheduled appointments at the PDC from September 2022 to August 2023. Logistic regression, decision trees, and XGBoost algorithms were employed to construct predictive models for appointment adherence. Results The XGBoost machine learning model outperformed logistic regression and decision trees with an area under the curve of 72% vs. 65% and 67%, respectively, in predicting missed appointments, despite limited availability of historical data. Key predictors included patient age, number of days between appointment scheduling and occurrence, insurance status, marital status, and mental health and cardiac disease conditions. Discussion Findings underscore the potential of machine learning predictive analytics to significantly enhance patient engagement and operational efficiency in emerging healthcare settings. Optimizing predictive models can help balance the early identification of patients at risk of non-adherence with the efficient allocation of resources. Conclusion The study highlights the potential value of employing machine learning techniques to inform interventions aimed at improving appointment adherence in a post-discharge transition clinic environment. Machine Learning Patient Compliance Post Discharge Care Health Services Research Appointments and Schedules Electronic Health Records Healthcare Disparities Risk Factors Figures Figure 1 Figure 2 Figure 3 Figure 4 Figure 5 Figure 6 1. Introduction Hospitals are responsible for ensuring effective care transitions from acute care to outpatient settings. Inadequate follow-up, particularly for patients with limited access to primary care and ongoing medical needs, can escalate hospital readmissions and emergency department (ED) visits, thereby straining hospital operations and finances [ 1 – 3 ]. Post-discharge care interventions can bridge this transition gap [ 4 ]. Research has shown that timely ED follow-up can reduce the likelihood of acute care readmissions [ 5 ] and is important for ensuring that patients adhere to treatment plans and recognize relevant symptoms [ 6 ]. Post discharge clinics (PDCs) are dedicated to providing follow-up care and have emerged as a key strategy for improving care transitions and reducing preventable health care utilization [ 7 – 10 ]. Their potential has garnered interest from many U.S. hospitals, particularly with initiatives like the Hospital Readmission Reduction Program, which penalizes hospitals with high readmission rates for selected conditions [ 11 , 12 ]. However, challenges in patient adherence to post-discharge transitional care services, such as attending outpatient appointments, persist, limiting the effectiveness of these interventions [ 13 , 14 ]. Because PDCs are conceptually new, evidenced-based protocols for supporting adherence to PDC appointments is limited. Novel predictive analytic techniques may improve patient prioritization and engagement by identifying patients at elevated risk of missing appointments. Although predictive analytics, such as machine learning, have shown promise in identifying patients at increased risk of adverse health outcomes and resource utilization (e.g., readmissions) [ 15 , 16 ], their application in supporting appointment adherence remains largely underexplored. Our study at a large academic medical center in the Southeastern United States employs predictive machine learning with electronic health record (EHR) data to identify patients at higher risk of appointment non-adherence at a new PDC. Such efforts are especially important during the early stages of a new clinic’s implementation when organizational routines are not yet ‘hardwired’ into social and technical norms, thus providing a unique opportunity to establish evidence-based interventions to support appointment adherence. By informing targeted adherence interventions, we anticipate enabling more efficient clinic operations and improved patient outcomes, thereby mitigating the ‘liability of newness’ challenge that new healthcare ventures often face [ 17 – 19 ]. In this analysis we employ three distinct analytical models: logistic regression, decision tree, and eXtreme Gradient Boosting (XGBoost). This multi-method approach allows us to evaluate the merits of applying predictive analytics with different levels of computational complexity, demonstrating the potential of advanced techniques to bolster patient appointment adherence. 2. Methods 2.1. Study setting Established in September 2022, the PDC at the University of Alabama at Birmingham (UAB) offers follow-up care to patients within 14 days of discharge from any UAB ED or Hospital. UAB Medicine is a comprehensive tertiary care academic medical center in Birmingham, AL with over 1,200 inpatient beds and approximately 135,000 ED visits and 409,000 inpatient days each year, spanning two hospitals and three EDs. 2.2. Data Our dataset included 2,168 appointments from 1,716 patients during the first operational year of the PDC (September 2022 to August 2023). Appointment data were anonymized for patient confidentiality and included dates and times when appointments were scheduled and set to occur, the type of appointment (following an ED visit or hospital discharge), medical history, and the status of the appointment (such as arrived, no show, canceled by patient, or bumped). A bumped appointment refers to one that the clinic rescheduled for operational reasons. The demographic and socio-economic data included city, state, zip code, age, gender, ethnicity, race, marital status, preferred language, religion, primary insurance provider, and specific needs (e.g., wheelchair and hearing impaired). These data are vital to assessing and predicting health care access and utilization after discharge and to pinpoint potential disparities in care, as highlighted in previous studies [ 12 , 20 ]. The clinical variables, including xxx types of medical history, were also tested as predictors. Appointment status (i.e., appointment adherence) was the primary outcome variable. We combined the ‘no show’ and ‘canceled’ categories to create a “missed” appointment category in our binary classification models. “Arrived” appointments served as the negative class in our models. Thirteen appointments were ‘bumped’ and excluded from the analysis, leaving 2,155 observations. 2.3. Predictive analysis methodologies We utilized logistic regression, decision trees, and XGBoost methods to analyze the dataset and test their effectiveness in predicting appointment adherence. We chose these three methods to represent a traditional statistical inference-based model, an interpretable machine learning model, and an advanced machine learning model, respectively. Logistic regression, a traditional statistical model, is simple, easily interpretable, and a staple in biomedical research [ 21 , 22 ]. Logistic regression assumes linear relationships between independent variables and the log odds of the dependent variable, unless the model explicitly accounts for non-linearity [ 23 , 24 ]. Decision Trees are machine learning algorithms generating inherently interpretable models, by recursively splitting the data into subsets based on the value of input features. The recursive split of the data is done such that it results in the highest information gain or the greatest reduction in impurity (i.e., splitting the data into homogeneous categories) [ 25 ]. XGBoost is an advanced machine learning algorithm that addresses complex data patterns through an ensemble model approach. It employs gradient boosting to sequentially construct a series of decision trees, each aiming to correct the errors of its predecessors [ 26 ]. Three models were trained (the first 85% of observations) and tested (the remaining 15%) on the same dataset, allowing for a comparative analysis of their performance. We report performance measure values from the testing data. This approach evaluates not only their predictive accuracy but also their ability to generalize findings to new, unseen data, which is crucial for practical applications in healthcare settings. We also explored how the three methods differently utilize the key variables by representing the importance (or significance) of the variables incorporated into the models. We utilized the R version 4.3.2 to conduct our modeling and analysis. 2.4. Performance measures We evaluated model performance by employing multiple performance measures, including accuracy, sensitivity, precision, specificity, and the area under the curve (AUC). Accuracy measures the proportion of true results (both true positives and true negatives) among the total number of tested observations [ 27 ]. However, accuracy does not distinguish between prediction error types, be they false positives or false negatives. Sensitivity evaluates the model’s ability to correctly identify true positives. Specificity assesses the model’s effectiveness in identifying true negatives. Precision measures the accuracy of positive predictions (i.e., true positives among all positives). Lastly, AUC reflects the trade-off between sensitivity and specificity by plotting the false positive rate (1 – specificity) against the true positive rate (sensitivity) at various prediction probability thresholds. An AUC value above 0.7 is considered acceptable for many applications, indicating that the model has a good measure of separability, while a value above 0.8 is deemed excellent [ 21 ]. 2.5. Feature importance In generalized linear models, such as logistic regression, the significance of a variable is often judged by the p-value of its coefficient. This provides an indicator of whether a variable significantly influences the model outcome. In contrast, feature importance in decision trees is computed based on the reduction in impurity that each variable contributes to the splits it participates in across the tree. To be specific, higher Gini importance values indicate a greater importance or contribution of a variable to the tree construction process [ 25 ]. Lastly, for XGBoost, the ‘Gain’ of a feature indicates the improvement in model performance it brings, measuring a feature’s contribution to each node across all trees, weighted by the number of observations passing through those nodes [ 26 ]. The ‘Cover’ metric reflects the average number of observations affected by a feature across all trees where it appears, indicating the breadth of a feature’s impact [ 25 ]. Meanwhile, ‘Frequency’ describes how often a feature is used to split the data across all boosting rounds [ 25 ], with a higher frequency suggesting a feature’s recurrent utility in making splits. 3. Results 3.1. Descriptive analysis results Figure 1 displays monthly appointment volumes (left y-axis) and attendance rates (right y-axis) by appointments status. Despite an increase in overall appointment volume, the PDC appointment attendance rate remained relatively constant around the 50% level. This observation underscores the critical need for an effective appointment adherence strategy to address the persistently high rate of missed appointments. Table 1 presents the univariate and bivariate analysis results, showing the distribution of demographic, socio-economic, and clinical variables across binary appointment status groups (arrived vs. missed). The average age of the 2,155 patients was 48.2 years (SD = 16.5). The majority of these patients were Black (52.6%), followed by White (39.5%), and Hispanic (3.9%). Most patients referred were female (53.9%), single (52.3%), and either covered by commercial insurance (47.7%) or uninsured/self-pay (18.8%). Table 1 PDC Patient Characteristics All Arrived Cancelled / No-show t -test / Chi-square Demographic and Socio-Economic Factor Age, mean (SD) 48.2 (16.5) 48.9 (16.4) 47.5 (16.6) t = 3.9 , p < .05 Race/ethnicity, N (% of total) χ 2 = 4.2, p = .52 Black 1,134 (52.6) 587 (52.0) 547 (53.3) White 852 (39.5) 456 (40.4) 396 (38.6) Hispanic 83 (3.9) 49 (4.3) 34 (3.3) Asian 39 (1.8) 18 (1.6) 21 (2.0) Unknown 35 (1.6) 12 (1.1) 23 (2.2) Gender, N (% of total) χ 2 = 0.2, p = .62 Female 1,161 (53.9) 602 (53.3) 559 (54.5) Male 994 (46.1) 527 (46.7) 467 (45.5) Marital status, N (% of total) χ 2 = 9.7, p = .14 Single 1,128 (52.3) 588 (52.1) 540 (52.6) Married/Life partner 645 (29.9) 363 (32.2) 282 (27.5) Divorced/Separated 227 (10.5) 103 (9.1) 124 (12.1) Widowed 103 (4.8) 51 (4.5) 52 (5.1) Unknown 52 (2.4) 24 (2.1) 28 (2.7) Preferred language χ 2 = 11.7 , p < .05 English 2,052 (95.2) 1,077 (95.4) 975 (95.0) Spanish 50 (2.3) 33 (2.9) 17 (1.7) Other 53 (2.5) 19 (1.7) 34 (3.3) Payer type χ 2 = 20.8 , p < .001 Commercial 1,028 (47.7) 585 (51.8) 443 (43.2) Medicare 336 (15.6) 172 (15.2) 164 (16.0) Medicaid 345 (16.0) 150 (13.3) 195 (19.0) Uninsured/Self-pay 405 (18.8) 201 (17.8) 204 (19.9) Specific needs description χ 2 = 7.9, p = .25 Not applicable 2,103 (97.6) 1,100 (97.4) 1,003 (97.8) Wheelchair 9 (0.4) 2 (0.2) 7 (0.7) Other 43 (2.0) 27 (2.4) 16 (1.5) Medical History Diabetes 622 (28.9) 319 (28.3) 303 (29.5) χ 2 = 1.3, p = .52 Myocardial infarction 264 (12.3) 123 (10.9) 141 (13.7) χ 2 = 4.9, p = .08 Coronary artery disease 367 (17.0) 189 (16.7) 178 (17.3) χ 2 = 1.0, p = .59 Chronic kidney disease 298 (13.8) 141 (12.5) 157 (15.3) χ 2 = 4.4 p = .11 Hypertension 972 (45.1) 544 (48.2) 428 (41.7) χ 2 = 10.1 , p < .01 COPD 259 (12.0) 121 (10.7) 138 (13.5) χ 2 = 4.7, p = .10 Venous thromboembolism pulmonary embolism 238 (11.0) 107 (9.5) 131 (12.8) χ 2 = 6.8 , p < .05 Asthma 357 (16.6) 169 (15.0) 188 (18.3) χ 2 = 5.2, p = .07 Anxiety 660 (30.6) 316 (28.0) 344 (33.5) χ 2 = 8.6 , p < .05 Autoimmune disorder 130 (6.0) 68 (6.0) 62 (6.0) χ 2 = 0.9, p = .63 Depression 479 (22.2) 227 (20.1) 252 (24.6) χ 2 = 7.0 , p < .05 Dyslipidemia 604 (28.0) 327 (29.0) 277 (27.0) χ 2 = 2.0, p < .37 Heart failure 347 (16.1) 161 (14.3) 186 (18.1) χ 2 = 6.8 , p < .05 Appointment Scheduling Factor Lag days, mean (SD) 6.6 (14.1) 4.5 (7.5) 8.8 (18.6) t = 52.5 , p < .001 The bivariate analysis (Table 1 , columns 3–5) reveals significant differences in patient characteristics based on appointment status. For instance, the average age of patients with “arrived” appointments was significantly greater than those with missed appointments ( t = 3.9, p < .05). Individuals using languages other than English and Spanish as their first language were more likely to miss appointments ( χ 2 = 11.7, p < .05), as were patients with Medicaid or uninsured/self-insured ( χ 2 = 20.8, p < .001). Additional statistical differences were identified through chi-square analysis. Patients with arrived appointments had a higher prevalence of hypertension (48.2%) than those who missed their appointments (41.7%) ( χ 2 = 10.1, p < .01). Conversely, anxiety and depression were more prevalent among patients who missed their appointments, suggesting these conditions as potential barriers to accessing PDC services ( χ 2 = 8.6, p < .05, χ 2 = 7.0, p < .05, respectively). Lastly, the average number of days between appointment scheduling and occurrence (lag days) was 6.6 days. A longer interval was significantly associated with a higher likelihood of missing appointments ( t = 52.5, p < .001). 3.2. Prediction outcomes As illustrated in Fig. 2 , the XGBoost model demonstrated superior performance in identifying cases prone to missed appointments. XGBoost demonstrated higher accuracy (68.4%), specificity (64.3%), precision (60.7%), and AUC (72.0%) than the other two models. The decision tree and XGBoost were equally as sensitive with 73.9% each. Nevertheless, XGBoost’s AUC value, serving as a summary measure of the models’ ability to balance sensitivity and specificity, indicates a stronger overall discriminative power in distinguishing between patients likely to attend or miss appointments. This underscores XGBoost’s potential impact on appointment adherence in a new clinic venture. 3.3. Logistic regression variable odds ratio Table 2 reports logistic regression results. Demographic and socio-economic variables were significantly associated with appointment status. Divorced/separated and widowed individuals were more likely to miss appointments (OR 1.57 ( p < .05) and 1.76 ( p < .05), respectively). Additionally, payer type emerged as a critical factor, with Medicaid (OR 1.67, p < .001) and uninsured/self-paying (OR 1.44, p < .05) persons having higher odds of non-attendance. In contrast, hypertension was significantly associated with lower odds of missing appointments (OR 0.58, p < .001), indicating that specific health conditions might differentially influence appointment adherence. The number of days between appointment scheduling and occurrence (lag days) remained highly significant ( p < .001) after controlling for all other factors in the model, underscoring its impact on appointment adherence. Table 2 Summary of logistic regression results Variable Odds ratio 95% CI p-value Intercept 8.7e-08 (0, 10.0) .99 Demographic and Socio-Economic Factor Age 0.99 (0.98, 1.01) .41 Race/ethnicity Black Reference White 1.03 (0.82, 1.30) .80 Hispanic 0.63 (0.21, 1.81) .39 Asian 1.22 (0.58, 2.56) .60 Unknown 4.21 (0.93, 29.81) .09 Gender Female Reference Male 1.01 (0.81, 1.25) .92 Marital status Married/Life partner Reference Single 1.09 (0.85, 1.41) .50 Divorced/Separated 1.57 * (1.06, 2.33) < .05 Widowed 1.76 * (1.04, 2.97) < .05 Unknown 1.02 (0.47, 2.18) 0.96 Preferred language English Reference Spanish 0.44 (0.21, 1.44) .23 Other 1.45 (0.88, 2.36) .32 Payer type Commercial Reference Medicare 1.31 (0.94, 1.84) .11 Medicaid 1.67 *** (1.23, 2.25) < .001 Uninsured/Self-pay 1.44 * (1.09, 1.91) < .05 Specific needs description Not applicable Reference Wheelchair 8.96 † (1.10, 48.0) .08 Other 1.61 (0.43, 6.94) .49 Medical History Diabetes 0.90 (0.34, 2.34) .83 Myocardial Infarction 1.12 (0.77, 1.63) .55 Coronary Artery Disease 0.90 (0.63, 1.26) .53 Chronic Kidney Disease 1.44 † (0.98, 2.11) .06 Hypertension 0.58 *** (0.45, 0.74) < .001 COPD 1.22 (0.86, 1.73) .27 Venous thromboembolism pulmonary embolism 1.63 (0.83, 3.22) .16 Asthma 1.12 (0.84, 1.50) .42 Anxiety 1.20 (0.92, 1.57) .17 Autoimmune Disorder 0.95 (0.61, 1.45) .80 Depression 0.86 (0.64, 1.17) .34 Dyslipidemia 1.09 (0.82, 1.44) .57 Heart Failure 1.30 (0.91, 1.86) .14 Appointment Scheduling Factor Lag days 1.05 *** (1.03, 1.07) < .001 3.3. Decision tree model feature importance Figure 3 presents the Gini importance value for the top 20 variables as identified by the decision tree method, ordered by the index values. This list encompasses a mix of scheduling-related, demographic, socio-economic, and clinical characteristics. Distinct from the findings in logistic regression analysis, age appeared to be highly important in the decision tree model. This indicates that age may exhibit a non-linear effect or interact with other variables. It is important to note that a higher feature importance signifies the variable's capacity to distinguish between the arrived and missed appointment classes, not directly suggesting a variable contributes to missed appointments. Figure 4 visualizes the decision tree, suggesting potential interactions among variables that contribute to appointment status. The numbers in the leaf nodes indicate the counts of arrived and missed appointments, respectively. The red leaf nodes represent the subgroups that have more missed appointments than arrived appointments. For effective visualization, the complexity parameter of the decision tree is set 1.0 x 10 − 4 , which produced the decision tree of depth five. Lag days and age appear to be the most influential variables. Demographic and socio-economic variables are associated with appointment adherence among patients with more than 2 but less than 4 lag days, but medical history variables are associated with adherence among those with more than 4 but less than 15 lag days. 3.4. XGBoost model feature importance Figure 5 presents the three feature metric values (Gain, Cover, and Frequency) for the top 20 variables as identified by the XGBoost method, ordered by their Gain measure values. Similar to the decision tree model, the list encompasses a mix of variable types. Lag days and age were identified as the most informative, followed by insurance status, marital status, and preferred language, highlighting the influence of socio-economic factors. 3.5. Sensitivity analysis on probability threshold values Figure 6 illustrates the trade-off relationship between sensitivity and precision in relation to the positive prediction threshold (i.e., the probability at which an appointment is predicted to be missed) in the XGBoost model. The detection prevalence represents the proportion of appointments predicted to be missed (i.e., the proportion of positive predictions). As the probability threshold increases, the model predicts fewer patients to miss their appointments (resulting in lower detection prevalence), leading to a decrease in sensitivity but an increase in precision. For example, setting the positive prediction threshold at 0.4 allows our model to detect 88.4% of positive cases with a precision of 49.0%. On the other hand, with the threshold set at 0.6, the model’s sensitivity drops to 34.8% while its precision increases to 70.6%. This describes that a lower threshold increases sensitivity but may include more false positives, while a higher threshold enhances precision but may miss identifying individuals at risk. 4. Discussion In this study, we applied a data-driven machine learning method to identify patients more likely to miss a scheduled appointment at a new PDC and compared this approach to more traditional prediction methods. XGBoost stands out for its ability to handle non-linear relationships and intricate interactions between variables [ 28 , 29 ], which are common in health-related data [ 30 , 31 ]. Our findings underscore the potential of machine learning to enhance clinic operations by better identifying those at higher risk of missing future appointments, for prioritizing patients for targeted interventions to improve health care operations and ultimately patient outcomes. Our findings shed light on the significant role that lag time, age, insurance status, marital status, and medical history have on predicting appointment adherence. Because these factors may impact patient engagement beyond appointment attendance, findings may have value in additional patient engagement strategies and across other clinic settings. Particularly, distinct from the findings in logistic regression analysis, the results from the XGBoost and decision tree models consistently identified age as a highly important variable. The significance of the age variable identified by both the decision tree and XGBoost method testifies the models’ advanced capacity for detecting intricate patterns. For health care clinical and operational leaders, these findings highlight the importance of minimizing the lag time and particular attention to older (≥ 81) or younger (< 62) patients who as groups were more likely to miss PDC appointments in our study. Our analysis also revealed the necessity of optimizing the sensitivity and precision trade-off in predictive models to enhance the effectiveness of appointment adherence interventions. A clinic can adjust the prediction threshold based on their financial resouces and physical and human resource capacity to follow up on predicted no-shows or their prioritization of minimizing missed appointments versus avoiding unnecessary follow-ups. This decision impacts resource allocation, the effectiveness of patient engagement efforts, and ultimately, the clinic's operational efficiency and patient care quality. This optimization ensures the detection of patients at increased risk of missing appointments while minimizing false positives that could lead to resource misallocation and alarm fatigue among healthcare professionals [ 32 , 33 ]. Fine-tuning the prediction threshold is paramount for deploying efficient and sustainable patient engagement strategies, ensuring that patient engagement interventions are both targeted and impactful. The dynamic nature of prediction thresholds demands continuous evaluation and adjustment of model parameters based on patient outcomes and new data. This adaptability is vital in health care practices. Integrating advanced predictive analytics into clinical and operational workflows marks a pivotal shift towards more proactive and personalized healthcare. However, realizing these benefits requires addressing potential optimization problems with these tools and enhancing their usability in partnership with healthcare professionals. Our findings have important implications for both post-discharge care operations as well as strategies to improve patient engagement such as appointment adherence. By demonstrating the value of machine learning models like XGBoost in predicting appointment adherence, our study contributes to the growing field of data-driven healthcare decision-making. Strategies such as these hold promise for allocating resources more judiciously, tailoring interventions to individual patient profiles, and ultimately elevating the quality and efficiency of care. 5. Limitations This study is subject to limitations. First, study findings are based on EHR data from a single institution, potentially limiting their applicability to broader post-discharge care settings. However, we note the variables included in prediction models are ubiquitous in health care settings. Second, the study’s sample size was relatively limited for predictive analytics research, a reflection of our intentional focus on supporting patient engagement for a new clinic venture, where patient engagement mechanisms were nascent and system-wide awareness was limited. This prompts consideration of how our findings may not be generalizable for more mature clinics with established patient engagement protocols and system-wide visibility, though the modeling approaches employed can easily be applied in other settings and compared to our findings. 6. Conclusion and Future Work This study has demonstrated the potential for predictive analytics to prioritize patients for enhanced engagement initiatives in transitional ambulatory care by identifying patients at increased risk of missing appointments. Highlighting key variables that influence appointment adherence, our findings offer actionable insights for developing focused interventions delivered to prioritized patients to improve post-discharge care. To the best of our knowledge, this work is among the first to apply machine learning to inform interventions that address post-discharge appointment adherence in new clinical ventures. An important next step involves integrating these models with feasibility studies (e.g., pilot implementation) to evaluate their potential impact on patient outcomes and care continuity. By advancing the application of predictive analytics in healthcare, our work aims to foster personalized and more effective post-discharge care, enhancing healthcare delivery and patient well-being. Declarations Acknowledgements : The authors thank the UAB Research and Informatics Service Center (RISC) and the Department of Biomedical Informatics and Data Science for their data extraction and transformation expertise. Competing Interest : The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article. Funding : This work was supported by the University of Alabama Health Services Foundation General Endowment Fund through a grant to establish the UAB Learning Health System Platform. Ethics Approval : This work was approved by the UAB Institutional Review Board (IRB-300009914). Availability of Data and Materials : The EHR datasets analyzed during the current study are not publicly available because they are governed by HIPAA regulations at the University of Alabama at Birmingham. Authors’ Contributions: S.L. designed the study, conducted analysis, and led the development of the manuscript. R.E. conducted quantitative data collection and handled correspondence with co-authors. L.H. contributed to the study design and Introduction, Results, and Discussion sections. M.G. contributed to project management tasks and the development of the manuscript. K.H. contributed to the study design, supervised the quantitative data collection, and reviewed the Results section. A.H. contributed to the study design and the Introduction and Discussion sections. G.B. contributed through her data extraction expertise to assist with acquiring quantitative data and reviewed the manuscript. J.M. contributed the literature review for the Introduction section and reviewed the manuscript. S.M. contributed to the study clinic implementation and reviewed the manuscript. C.S. provided information on the administrative and operation activities of the Post Discharge Clinic and provided feedback through draft review. T.B. assisted with project identification and study development and provided feedback through draft reviews. A.G. contributed to study recruitment and provided feedback on the manuscript through draft reviews. H.B. provided information related to care transition activities at the University of Alabama at Birmingham Health System and outpatient facilities. M.M. contributed to the study design the Discussion section and provided feedback and revisions through draft reviews. References Ragin, D.F., Hwang, U., Cydulka, R.K., Holson, D., Haley, L.L., Jr., Richards, C.F., Becker, B.M., Richardson, L.D., and Emergency Medicine Patients' Access To Healthcare Study, I.: ‘Reasons for using the emergency department: results of the EMPATH Study’, Acad Emerg Med, 2005, 12, (12), pp. 1158-1166 Brown, L.E., Burton, R., Hixon, B., Kakade, M., Bhagalia, P., Vick, C., Edwards, A., and Hawn, M.T.: ‘Factors influencing emergency department preference for access to healthcare’, West J Emerg Med, 2012, 13, (5), pp. 410-415 Moscelli, G., Siciliani, L., Gutacker, N., and Cookson, R.: ‘Socioeconomic inequality of access to healthcare: Does choice explain the gradient?’, J Health Econ, 2018, 57, pp. 290-314 Preen, D.B., Bailey, B.E., Wright, A., Kendall, P., Phillips, M., Hung, J., Hendriks, R., Mather, A., and Williams, E.: ‘Effects of a multidisciplinary, post-discharge continuance of care intervention on quality of life, discharge satisfaction, and hospital length of stay: a randomized controlled trial’, Int J Qual Health Care, 2005, 17, (1), pp. 43-51 Coppa, K., Kim, E.J., Oppenheim, M.I., Bock, K.R., Conigliaro, J., and Hirsch, J.S.: ‘Examination of Post-discharge Follow-up Appointment Status and 30-Day Readmission’, J Gen Intern Med, 2021, 36, (5), pp. 1214-1221 Elliott, K., Klein, J.W., Basu, A., and Sabbatini, A.K.: ‘Transitional care clinics for follow-up and primary care linkage for patients discharged from the ED’, Am J Emerg Med, 2016, 34, (7), pp. 1230-1235 Doctoroff, L., Nijhawan, A., McNally, D., Vanka, A., Yu, R., and Mukamal, K.J.: ‘The characteristics and impact of a hospitalist-staffed post-discharge clinic’, Am J Med, 2013, 126, (11), pp. 1016 e1019-1015 Lu, L.Y., Jackevicius, C., de Leon, N., Warner, A., Chang, D., and Mody, F.: ‘Impact of a Multi-Disciplinary Heart Failure Post-Discharge Management Clinic on Medication Adherence’, J Am Coll Cardiol, 2014, 63, (12), pp. A539-A539 Jackevicius, C.A., de Leon, N.K., Lu, L., Chang, D.S., Warner, A.L., and Mody, F.V.: ‘Impact of a Multidisciplinary Heart Failure Post-hospitalization Program on Heart Failure Readmission Rates’, Ann Pharmacother, 2015, 49, (11), pp. 1189-1196 Shah, M., Douglas, V., Scott, B., and Josephson, S.A.: ‘A Neurohospitalist Discharge Clinic Shortens the Transition From Inpatient to Outpatient Care’, Neurohospitalist, 2016, 6, (2), pp. 64-69 Burke, R.E., Whitfield, E., and Prochazka, A.V.: ‘Effect of a hospitalist-run postdischarge clinic on outcomes’, J Hosp Med, 2014, 9, (1), pp. 7-12 Baldino, M., Bonaguro, A.M., Burgwardt, S., Lombardi, A., Cristancho, C., Mann, C., Wright, D., Jackson, C., and Seth, A.: ‘Impact of a Novel Post-Discharge Transitions of Care Clinic on Hospital Readmissions’, J Natl Med Assoc, 2021, 113, (2), pp. 133-141 Distelhorst, K., Claussen, R., Dion, K., Bena, J.F., Morrison, S.L., Walker, D., Tai, H.L., and Albert, N.M.: ‘Factors Associated With Adherence to 14-Day Office Appointments After Heart Failure Discharge’, J Card Fail, 2018, 24, (6), pp. 407-411 Truong, E.I., DeMario, B.S., Hendrickson, S., Kalina, M.J., Jr., Vallier, H.A., Tseng, E.S., Claridge, J.A., and Ho, V.P.: ‘Factors Influencing Nonadherence to Recommended Postdischarge Follow-Up After Trauma’, J Surg Res, 2020, 256, pp. 143-148 Shamout, F., Zhu, T., and Clifton, D.A.: ‘Machine Learning for Clinical Outcome Prediction’, IEEE Rev Biomed Eng, 2021, 14, pp. 116-126 Brink, A., Alsma, J., van Attekum, L.A.A.M., Bramer, W.M., Zietse, R., Lingsma, H., and Schuit, S.C.E.: ‘Predicting inhospital admission at the emergency department: a systematic review’, Emerg Med J, 2022, 39, (3), pp. 191-198 Stinchcombe, A.L.: ‘Social structure and organizations’: ‘Hand of Organizations (RLE: Organizations)’ (Routledge, 2013), pp. 142-193 Aarons, G.A., and Palinkas, L.A.: ‘Implementation of evidence-based practice in child welfare: service provider perspectives’, Adm Policy Ment Health, 2007, 34, (4), pp. 411-419 Schell, S.F., Luke, D.A., Schooley, M.W., Elliott, M.B., Herbers, S.H., Mueller, N.B., and Bunger, A.C.: ‘Public health program capacity for sustainability: a new framework’, Implement Sci, 2013, 8, pp. 15 Kurtz Landy, C., Sword, W., and Ciliska, D.: ‘Urban women's socioeconomic status, health service needs and utilization in the four weeks after postpartum hospital discharge: findings of a Canadian cross-sectional survey’, BMC Health Serv Res, 2008, 8, pp. 203 Hosmer, D.W., Lemeshow, S., Sturdivant, R.X., and ProQuest: ‘Applied logistic regression’ (Wiley, 2013, Third edition edn. 2013) Boateng, E.Y.A., D. A.: ‘A Review of the Logistic Regression Model with Emphasis on Medical Research’, Journal of Data Analysis and Information Processing, 2019, 7, pp. 190 - 207 Omer, D.M., A. B.: ‘Modelling logistic regression using multivariable fractional polynomials’, Imperial Journal of Interdisciplinary Research, 2017, 3, (11), pp. 8-16 Bache-Mathiesen, L.K., Andersen, T.E., Dalen-Lorentsen, T., Clarsen, B., and Fagerland, M.W.: ‘Not straightforward: modelling non-linearity in training load and injury research’, BMJ Open Sport Exerc Med, 2021, 7, (3), pp. e001119 Hastie, T., Friedman, J.H., and Tibshirani, R.: ‘The elements of statistical learning : data mining, inference, and prediction’ (Springer, 2009. 2009) Chen, T.G., C. : ‘A scalable tree boosting system’, in Editor (Ed.)^(Eds.): ‘Book A scalable tree boosting system’ (Association for Computing Machinery, 2016, edn.), pp. 785-794 Bishop, C.M.: ‘Pattern Recognition and Machine Learning’ (Springer, 2006, Softcover reprint of the original 1st 2006. edn. 2006) Yang, C., Chen, M., and Yuan, Q.: ‘The application of XGBoost and SHAP to examining the factors in freight truck-related crashes: An exploratory analysis’, Accid Anal Prev, 2021, 158, pp. 106153 Liu, J.X., Wang, B., and Xiao, L.Z.: ‘Non-linear associations between built environment and active travel for working and shopping: An extreme gradient boosting approach’, J Transp Geogr, 2021, 92 Orfanoudaki, A.C., E.; Cadisch, C.; Stein, B.; Nouh, A.; Alberts, M. J.; Bertsimas, B. : ‘Machine learning provides evidence that stroke risk is not linear: The non-linear Framingham stroke risk score’, PLOs one, 2020, 21, (15) Rodgers, B., Korten, A.E., Jorm, A.F., Jacomb, P.A., Christensen, H., and Henderson, A.S.: ‘Non-linear relationships in associations of depression and anxiety with alcohol use’, Psychol Med, 2000, 30, (2), pp. 421-432 Cvach, M.: ‘Monitor alarm fatigue: an integrative review’, Biomed Instrum Technol, 2012, 46, (4), pp. 268-277 Hravnak, M., Pellathy, T., Chen, L., Dubrawski, A., Wertz, A., Clermont, G., and Pinsky, M.R.: ‘A call to alarms: Current state and future directions in the battle against alarm fatigue’, J Electrocardiol, 2018, 51, (6S), pp. S44-S48 Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-4477049","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":307180512,"identity":"93cb0527-a10a-4cbf-9eb3-c9d47199408e","order_by":0,"name":"Seung-Yup Lee","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAArklEQVRIiWNgGAWjYDACdiD+UMHAYACkJYjTwszAwDjjDKlamHnbSNEi38x88DHvvMPy5gzMB2/zEKPF4DBbsuHcbYcNdzawJVsTp4WZx0zi7bbDjBsO8JhJE6VFvpn/+w/eOYftNxzg/0acFobDPGyMvA2HE4G2sBGnBegXY8kZx9KTdzazGVvOIcph7c0PP3yosbbdzt788MYbohwGB8ykKR8Fo2AUjIJRgA8AAA7LLiBhszqvAAAAAElFTkSuQmCC","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":true,"prefix":"","firstName":"Seung-Yup","middleName":"","lastName":"Lee","suffix":""},{"id":307180513,"identity":"a2976e74-6fda-41ca-a61e-d90dd09e13dd","order_by":1,"name":"Reid Eagleson","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Reid","middleName":"","lastName":"Eagleson","suffix":""},{"id":307180514,"identity":"a9c6e94e-4e0d-4c61-a2bc-238d169ab5a7","order_by":2,"name":"Larry Hearld","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Larry","middleName":"","lastName":"Hearld","suffix":""},{"id":307180516,"identity":"4579a3cc-4085-4032-8dd9-af07de1c578d","order_by":3,"name":"Madeline Gibson","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Madeline","middleName":"","lastName":"Gibson","suffix":""},{"id":307180517,"identity":"5eba0575-051b-4e08-b205-230e0e8c9859","order_by":4,"name":"Kristine Hearld","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Kristine","middleName":"","lastName":"Hearld","suffix":""},{"id":307180518,"identity":"6c49c1df-b67d-4bd4-a2f7-016edc34cbea","order_by":5,"name":"Allyson Hall","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Allyson","middleName":"","lastName":"Hall","suffix":""},{"id":307180519,"identity":"6ee01783-1e16-4d0a-ad3e-b21bf563f3ff","order_by":6,"name":"Greer Burkholder","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Greer","middleName":"","lastName":"Burkholder","suffix":""},{"id":307180521,"identity":"3407a952-13f9-4f8a-b08b-92bccc542426","order_by":7,"name":"Jacob McMahon","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Jacob","middleName":"","lastName":"McMahon","suffix":""},{"id":307180523,"identity":"b4a1020e-f12d-4401-b3c3-b80c164109bc","order_by":8,"name":"Shoaib Mahmood","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Shoaib","middleName":"","lastName":"Mahmood","suffix":""},{"id":307180524,"identity":"4de00e3e-3349-4705-821e-6da558606133","order_by":9,"name":"Corey Spraberry","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Corey","middleName":"","lastName":"Spraberry","suffix":""},{"id":307180525,"identity":"b1cb38e4-1fd7-4fd2-a299-f3125f8c3f3e","order_by":10,"name":"Thalia Baker","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Thalia","middleName":"","lastName":"Baker","suffix":""},{"id":307180529,"identity":"3e80bc64-9eb3-47a8-906f-cca8063d18bd","order_by":11,"name":"Alison Garretson","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Alison","middleName":"","lastName":"Garretson","suffix":""},{"id":307180530,"identity":"8069ce9a-693e-4b10-bf5d-377e9abffead","order_by":12,"name":"Heather Bradley","email":"","orcid":"","institution":"Cooper Green Mercy Health Service Authority","correspondingAuthor":false,"prefix":"","firstName":"Heather","middleName":"","lastName":"Bradley","suffix":""},{"id":307180531,"identity":"3c2ab507-dea4-486f-9ae2-bdf040709fb9","order_by":13,"name":"Michael Mugavero","email":"","orcid":"","institution":"University of Alabama at Birmingham","correspondingAuthor":false,"prefix":"","firstName":"Michael","middleName":"","lastName":"Mugavero","suffix":""}],"badges":[],"createdAt":"2024-05-25 13:53:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-4477049/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-4477049/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":58154359,"identity":"f5ae9036-ea71-4231-a760-da233904f253","added_by":"auto","created_at":"2024-06-11 20:39:03","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":45531,"visible":true,"origin":"","legend":"\u003cp\u003eMonthly trend of PDC appointment status\u003c/p\u003e","description":"","filename":"Figure1.png","url":"https://assets-eu.researchsquare.com/files/rs-4477049/v1/26b9fae72cf629b8b460cec1.png"},{"id":58154362,"identity":"fbb765c3-94ce-4557-a7c1-b79da023b73d","added_by":"auto","created_at":"2024-06-11 20:39:04","extension":"png","order_by":2,"title":"Figure 2","display":"","copyAsset":false,"role":"figure","size":113266,"visible":true,"origin":"","legend":"\u003cp\u003eModel performance comparison\u003c/p\u003e","description":"","filename":"Figure2.png","url":"https://assets-eu.researchsquare.com/files/rs-4477049/v1/808e998ab4105201b0a61ee8.png"},{"id":58154360,"identity":"04e1d1f5-6133-45e6-b9cf-a31a0841f61b","added_by":"auto","created_at":"2024-06-11 20:39:04","extension":"png","order_by":3,"title":"Figure 3","display":"","copyAsset":false,"role":"figure","size":56487,"visible":true,"origin":"","legend":"\u003cp\u003eGini feature importance from the decision tree model\u003c/p\u003e","description":"","filename":"Figure3.png","url":"https://assets-eu.researchsquare.com/files/rs-4477049/v1/77689319aa0845801f27c0d1.png"},{"id":58154363,"identity":"dcc3c066-b03c-4744-99e6-bfe4569c60c0","added_by":"auto","created_at":"2024-06-11 20:39:04","extension":"png","order_by":4,"title":"Figure 4","display":"","copyAsset":false,"role":"figure","size":59008,"visible":true,"origin":"","legend":"\u003cp\u003eDecision tree structure\u003c/p\u003e","description":"","filename":"Figure4.png","url":"https://assets-eu.researchsquare.com/files/rs-4477049/v1/ed3f3531aa3c55a0ffdf189c.png"},{"id":58155362,"identity":"cf542c5a-790a-4279-8617-2f74e9bd8f90","added_by":"auto","created_at":"2024-06-11 20:47:04","extension":"png","order_by":5,"title":"Figure 5","display":"","copyAsset":false,"role":"figure","size":118200,"visible":true,"origin":"","legend":"\u003cp\u003eFeature importance from the XGBoost model\u003c/p\u003e","description":"","filename":"Figure5.png","url":"https://assets-eu.researchsquare.com/files/rs-4477049/v1/d6ce3e794231df0baf4dd317.png"},{"id":58154365,"identity":"fbce6ba2-8196-4c30-a16a-f488a3a04ead","added_by":"auto","created_at":"2024-06-11 20:39:04","extension":"png","order_by":6,"title":"Figure 6","display":"","copyAsset":false,"role":"figure","size":98848,"visible":true,"origin":"","legend":"\u003cp\u003eTrade-off between sensitivity, detection prevalence, and precision\u003c/p\u003e","description":"","filename":"Figure6.png","url":"https://assets-eu.researchsquare.com/files/rs-4477049/v1/222d886ae935bc8c8682ad5e.png"},{"id":58156056,"identity":"8c7aa41f-05c8-4fe5-b706-02db6adfe764","added_by":"auto","created_at":"2024-06-11 21:03:07","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1511914,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-4477049/v1/df80113c-6502-4437-a2ee-f4eac9d5314e.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Enhancing Patient Engagement with Machine Learning at a Novel Care Transition Clinic","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eHospitals are responsible for ensuring effective care transitions from acute care to outpatient settings. Inadequate follow-up, particularly for patients with limited access to primary care and ongoing medical needs, can escalate hospital readmissions and emergency department (ED) visits, thereby straining hospital operations and finances [\u003cspan additionalcitationids=\"CR2\" citationid=\"CR1\" class=\"CitationRef\"\u003e1\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e3\u003c/span\u003e]. Post-discharge care interventions can bridge this transition gap [\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e4\u003c/span\u003e]. Research has shown that timely ED follow-up can reduce the likelihood of acute care readmissions [\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e5\u003c/span\u003e] and is important for ensuring that patients adhere to treatment plans and recognize relevant symptoms [\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e6\u003c/span\u003e].\u003c/p\u003e \u003cp\u003ePost discharge clinics (PDCs) are dedicated to providing follow-up care and have emerged as a key strategy for improving care transitions and reducing preventable health care utilization [\u003cspan additionalcitationids=\"CR8 CR9\" citationid=\"CR7\" class=\"CitationRef\"\u003e7\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e10\u003c/span\u003e]. Their potential has garnered interest from many U.S. hospitals, particularly with initiatives like the Hospital Readmission Reduction Program, which penalizes hospitals with high readmission rates for selected conditions [\u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e11\u003c/span\u003e, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eHowever, challenges in patient adherence to post-discharge transitional care services, such as attending outpatient appointments, persist, limiting the effectiveness of these interventions [\u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e13\u003c/span\u003e, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e14\u003c/span\u003e]. Because PDCs are conceptually new, evidenced-based protocols for supporting adherence to PDC appointments is limited. Novel predictive analytic techniques may improve patient prioritization and engagement by identifying patients at elevated risk of missing appointments. Although predictive analytics, such as machine learning, have shown promise in identifying patients at increased risk of adverse health outcomes and resource utilization (e.g., readmissions) [\u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e15\u003c/span\u003e, \u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e16\u003c/span\u003e], their application in supporting appointment adherence remains largely underexplored.\u003c/p\u003e \u003cp\u003eOur study at a large academic medical center in the Southeastern United States employs predictive machine learning with electronic health record (EHR) data to identify patients at higher risk of appointment non-adherence at a new PDC. Such efforts are especially important during the early stages of a new clinic\u0026rsquo;s implementation when organizational routines are not yet \u0026lsquo;hardwired\u0026rsquo; into social and technical norms, thus providing a unique opportunity to establish evidence-based interventions to support appointment adherence. By informing targeted adherence interventions, we anticipate enabling more efficient clinic operations and improved patient outcomes, thereby mitigating the \u0026lsquo;liability of newness\u0026rsquo; challenge that new healthcare ventures often face [\u003cspan additionalcitationids=\"CR18\" citationid=\"CR17\" class=\"CitationRef\"\u003e17\u003c/span\u003e\u0026ndash;\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e19\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eIn this analysis we employ three distinct analytical models: logistic regression, decision tree, and eXtreme Gradient Boosting (XGBoost). This multi-method approach allows us to evaluate the merits of applying predictive analytics with different levels of computational complexity, demonstrating the potential of advanced techniques to bolster patient appointment adherence.\u003c/p\u003e"},{"header":"2. Methods","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1. Study setting\u003c/h2\u003e \u003cp\u003eEstablished in September 2022, the PDC at the University of Alabama at Birmingham (UAB) offers follow-up care to patients within 14 days of discharge from any UAB ED or Hospital. UAB Medicine is a comprehensive tertiary care academic medical center in Birmingham, AL with over 1,200 inpatient beds and approximately 135,000 ED visits and 409,000 inpatient days each year, spanning two hospitals and three EDs.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2. Data\u003c/h2\u003e \u003cp\u003eOur dataset included 2,168 appointments from 1,716 patients during the first operational year of the PDC (September 2022 to August 2023). Appointment data were anonymized for patient confidentiality and included dates and times when appointments were scheduled and set to occur, the type of appointment (following an ED visit or hospital discharge), medical history, and the status of the appointment (such as arrived, no show, canceled by patient, or bumped). A bumped appointment refers to one that the clinic rescheduled for operational reasons.\u003c/p\u003e \u003cp\u003eThe demographic and socio-economic data included city, state, zip code, age, gender, ethnicity, race, marital status, preferred language, religion, primary insurance provider, and specific needs (e.g., wheelchair and hearing impaired). These data are vital to assessing and predicting health care access and utilization after discharge and to pinpoint potential disparities in care, as highlighted in previous studies [\u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e12\u003c/span\u003e, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e20\u003c/span\u003e]. The clinical variables, including xxx types of medical history, were also tested as predictors.\u003c/p\u003e \u003cp\u003eAppointment status (i.e., appointment adherence) was the primary outcome variable. We combined the \u0026lsquo;no show\u0026rsquo; and \u0026lsquo;canceled\u0026rsquo; categories to create a \u0026ldquo;missed\u0026rdquo; appointment category in our binary classification models. \u0026ldquo;Arrived\u0026rdquo; appointments served as the negative class in our models. Thirteen appointments were \u0026lsquo;bumped\u0026rsquo; and excluded from the analysis, leaving 2,155 observations.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3. Predictive analysis methodologies\u003c/h2\u003e \u003cp\u003eWe utilized logistic regression, decision trees, and XGBoost methods to analyze the dataset and test their effectiveness in predicting appointment adherence. We chose these three methods to represent a traditional statistical inference-based model, an interpretable machine learning model, and an advanced machine learning model, respectively.\u003c/p\u003e \u003cp\u003eLogistic regression, a traditional statistical model, is simple, easily interpretable, and a staple in biomedical research [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e, \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e22\u003c/span\u003e]. Logistic regression assumes linear relationships between independent variables and the log odds of the dependent variable, unless the model explicitly accounts for non-linearity [\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e23\u003c/span\u003e, \u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e24\u003c/span\u003e]. Decision Trees are machine learning algorithms generating inherently interpretable models, by recursively splitting the data into subsets based on the value of input features. The recursive split of the data is done such that it results in the highest information gain or the greatest reduction in impurity (i.e., splitting the data into homogeneous categories) [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. XGBoost is an advanced machine learning algorithm that addresses complex data patterns through an ensemble model approach. It employs gradient boosting to sequentially construct a series of decision trees, each aiming to correct the errors of its predecessors [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eThree models were trained (the first 85% of observations) and tested (the remaining 15%) on the same dataset, allowing for a comparative analysis of their performance. We report performance measure values from the testing data. This approach evaluates not only their predictive accuracy but also their ability to generalize findings to new, unseen data, which is crucial for practical applications in healthcare settings. We also explored how the three methods differently utilize the key variables by representing the importance (or significance) of the variables incorporated into the models. We utilized the R version 4.3.2 to conduct our modeling and analysis.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4. Performance measures\u003c/h2\u003e \u003cp\u003eWe evaluated model performance by employing multiple performance measures, including accuracy, sensitivity, precision, specificity, and the area under the curve (AUC). Accuracy measures the proportion of true results (both true positives and true negatives) among the total number of tested observations [\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e27\u003c/span\u003e]. However, accuracy does not distinguish between prediction error types, be they false positives or false negatives. Sensitivity evaluates the model\u0026rsquo;s ability to correctly identify true positives. Specificity assesses the model\u0026rsquo;s effectiveness in identifying true negatives. Precision measures the accuracy of positive predictions (i.e., true positives among all positives). Lastly, AUC reflects the trade-off between sensitivity and specificity by plotting the false positive rate (1 \u0026ndash; specificity) against the true positive rate (sensitivity) at various prediction probability thresholds. An AUC value above 0.7 is considered acceptable for many applications, indicating that the model has a good measure of separability, while a value above 0.8 is deemed excellent [\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e21\u003c/span\u003e].\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5. Feature importance\u003c/h2\u003e \u003cp\u003eIn generalized linear models, such as logistic regression, the significance of a variable is often judged by the p-value of its coefficient. This provides an indicator of whether a variable significantly influences the model outcome.\u003c/p\u003e \u003cp\u003eIn contrast, feature importance in decision trees is computed based on the reduction in impurity that each variable contributes to the splits it participates in across the tree. To be specific, higher Gini importance values indicate a greater importance or contribution of a variable to the tree construction process [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e].\u003c/p\u003e \u003cp\u003eLastly, for XGBoost, the \u0026lsquo;Gain\u0026rsquo; of a feature indicates the improvement in model performance it brings, measuring a feature\u0026rsquo;s contribution to each node across all trees, weighted by the number of observations passing through those nodes [\u003cspan citationid=\"CR26\" class=\"CitationRef\"\u003e26\u003c/span\u003e]. The \u0026lsquo;Cover\u0026rsquo; metric reflects the average number of observations affected by a feature across all trees where it appears, indicating the breadth of a feature\u0026rsquo;s impact [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e]. Meanwhile, \u0026lsquo;Frequency\u0026rsquo; describes how often a feature is used to split the data across all boosting rounds [\u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e25\u003c/span\u003e], with a higher frequency suggesting a feature\u0026rsquo;s recurrent utility in making splits.\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Results","content":"\u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e3.1. Descriptive analysis results\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e displays monthly appointment volumes (left y-axis) and attendance rates (right y-axis) by appointments status. Despite an increase in overall appointment volume, the PDC appointment attendance rate remained relatively constant around the 50% level. This observation underscores the critical need for an effective appointment adherence strategy to address the persistently high rate of missed appointments.\u003c/p\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e presents the univariate and bivariate analysis results, showing the distribution of demographic, socio-economic, and clinical variables across binary appointment status groups (arrived vs. missed). The average age of the 2,155 patients was 48.2 years (SD\u0026thinsp;=\u0026thinsp;16.5). The majority of these patients were Black (52.6%), followed by White (39.5%), and Hispanic (3.9%). Most patients referred were female (53.9%), single (52.3%), and either covered by commercial insurance (47.7%) or uninsured/self-pay (18.8%).\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003ePDC Patient Characteristics\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"5\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eAll\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eArrived\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eCancelled / No-show\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cem\u003et\u003c/em\u003e-test / Chi-square\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDemographic and Socio-Economic Factor\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge, mean (SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e48.2 (16.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e48.9 (16.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e47.5 (16.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003et\u003c/b\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;3.9\u003c/b\u003e, \u003cb\u003ep\u003c/b\u003e\u0026thinsp;\u003cb\u003e\u0026lt;\u0026thinsp;.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRace/ethnicity, N (% of total)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;4.2, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.52\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBlack\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1,134 (52.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e587 (52.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e547 (53.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWhite\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e852 (39.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e456 (40.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e396 (38.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHispanic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e83 (3.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e49 (4.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e34 (3.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAsian\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e39 (1.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e18 (1.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e21 (2.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnknown\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e35 (1.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e12 (1.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e23 (2.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender, N (% of total)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.2, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.62\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1,161 (53.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e602 (53.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e559 (54.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e994 (46.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e527 (46.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e467 (45.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarital status, N (% of total)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;9.7, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSingle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1,128 (52.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e588 (52.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e540 (52.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarried/Life partner\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e645 (29.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e363 (32.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e282 (27.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDivorced/Separated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e227 (10.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e103 (9.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e124 (12.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWidowed\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e103 (4.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e51 (4.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e52 (5.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnknown\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e52 (2.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e24 (2.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e28 (2.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePreferred language\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eχ\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;11.7\u003c/b\u003e, \u003cb\u003ep\u003c/b\u003e\u0026thinsp;\u003cb\u003e\u0026lt;\u0026thinsp;.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEnglish\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2,052 (95.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,077 (95.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e975 (95.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpanish\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e50 (2.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e33 (2.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e17 (1.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e53 (2.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e19 (1.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e34 (3.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePayer type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eχ\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;20.8\u003c/b\u003e, \u003cb\u003ep\u003c/b\u003e\u0026thinsp;\u003cb\u003e\u0026lt;\u0026thinsp;.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCommercial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1,028 (47.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e585 (51.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e443 (43.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedicare\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e336 (15.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e172 (15.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e164 (16.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedicaid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e345 (16.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e150 (13.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e195 (19.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUninsured/Self-pay\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e405 (18.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e201 (17.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e204 (19.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpecific needs description\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;7.9, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.25\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNot applicable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e2,103 (97.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e1,100 (97.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e1,003 (97.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWheelchair\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9 (0.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e2 (0.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e7 (0.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e43 (2.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e27 (2.4)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e16 (1.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"3\" nameend=\"c3\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMedical History\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e622 (28.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e319 (28.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e303 (29.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;1.3, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.52\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMyocardial infarction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e264 (12.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e123 (10.9)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e141 (13.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;4.9, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoronary artery disease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e367 (17.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e189 (16.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e178 (17.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;1.0, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.59\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChronic kidney disease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e298 (13.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e141 (12.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e157 (15.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;4.4 \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHypertension\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e972 (45.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e544 (48.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e428 (41.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eχ\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;10.1\u003c/b\u003e, \u003cb\u003ep\u003c/b\u003e\u0026thinsp;\u003cb\u003e\u0026lt;\u0026thinsp;.01\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCOPD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e259 (12.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e121 (10.7)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e138 (13.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;4.7, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.10\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVenous thromboembolism\u003c/p\u003e \u003cp\u003epulmonary embolism\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e238 (11.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e107 (9.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e131 (12.8)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eχ\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;6.8\u003c/b\u003e, \u003cb\u003ep\u0026thinsp;\u0026lt;\u003c/b\u003e\u0026thinsp;\u003cb\u003e.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAsthma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e357 (16.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e169 (15.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e188 (18.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;5.2, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.07\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnxiety\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e660 (30.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e316 (28.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e344 (33.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eχ\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;8.6\u003c/b\u003e, \u003cb\u003ep\u0026thinsp;\u0026lt;\u003c/b\u003e\u0026thinsp;\u003cb\u003e.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAutoimmune disorder\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e130 (6.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e68 (6.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e62 (6.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;0.9, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;=\u0026thinsp;.63\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDepression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e479 (22.2)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e227 (20.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e252 (24.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eχ\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;7.0\u003c/b\u003e, \u003cb\u003ep\u003c/b\u003e\u0026thinsp;\u003cb\u003e\u0026lt;\u0026thinsp;.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDyslipidemia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e604 (28.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e327 (29.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e277 (27.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003eχ\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;2.0, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.37\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHeart failure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e347 (16.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e161 (14.3)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e186 (18.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003eχ\u003c/b\u003e\u003csup\u003e\u003cb\u003e2\u003c/b\u003e\u003c/sup\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;6.8\u003c/b\u003e, \u003cb\u003ep\u003c/b\u003e\u0026thinsp;\u003cb\u003e\u0026lt;\u0026thinsp;.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"5\" nameend=\"c5\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAppointment Scheduling Factor\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLag days, mean (SD)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e6.6 (14.1)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e4.5 (7.5)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e8.8 (18.6)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u003cb\u003et\u003c/b\u003e\u0026thinsp;\u003cb\u003e=\u0026thinsp;52.5\u003c/b\u003e, \u003cb\u003ep\u003c/b\u003e\u0026thinsp;\u003cb\u003e\u0026lt;\u0026thinsp;.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eThe bivariate analysis (Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, columns 3\u0026ndash;5) reveals significant differences in patient characteristics based on appointment status. For instance, the average age of patients with \u0026ldquo;arrived\u0026rdquo; appointments was significantly greater than those with missed appointments (\u003cem\u003et\u003c/em\u003e\u0026thinsp;=\u0026thinsp;3.9, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05). Individuals using languages other than English and Spanish as their first language were more likely to miss appointments (\u003cem\u003eχ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;11.7, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05), as were patients with Medicaid or uninsured/self-insured (\u003cem\u003eχ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;20.8, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001).\u003c/p\u003e \u003cp\u003eAdditional statistical differences were identified through chi-square analysis. Patients with arrived appointments had a higher prevalence of hypertension (48.2%) than those who missed their appointments (41.7%) (\u003cem\u003eχ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;10.1, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.01). Conversely, anxiety and depression were more prevalent among patients who missed their appointments, suggesting these conditions as potential barriers to accessing PDC services (\u003cem\u003eχ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;8.6, \u003cem\u003ep\u0026thinsp;\u0026lt;\u003c/em\u003e\u0026thinsp;.05, \u003cem\u003eχ\u003c/em\u003e\u003csup\u003e2\u003c/sup\u003e\u0026thinsp;=\u0026thinsp;7.0, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05, respectively). Lastly, the average number of days between appointment scheduling and occurrence (lag days) was 6.6 days. A longer interval was significantly associated with a higher likelihood of missing appointments (\u003cem\u003et\u003c/em\u003e\u0026thinsp;=\u0026thinsp;52.5, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec10\" class=\"Section2\"\u003e \u003ch2\u003e3.2. Prediction outcomes\u003c/h2\u003e \u003cp\u003eAs illustrated in Fig.\u0026nbsp;\u003cspan refid=\"Fig2\" class=\"InternalRef\"\u003e2\u003c/span\u003e, the XGBoost model demonstrated superior performance in identifying cases prone to missed appointments. XGBoost demonstrated higher accuracy (68.4%), specificity (64.3%), precision (60.7%), and AUC (72.0%) than the other two models. The decision tree and XGBoost were equally as sensitive with 73.9% each. Nevertheless, XGBoost\u0026rsquo;s AUC value, serving as a summary measure of the models\u0026rsquo; ability to balance sensitivity and specificity, indicates a stronger overall discriminative power in distinguishing between patients likely to attend or miss appointments. This underscores XGBoost\u0026rsquo;s potential impact on appointment adherence in a new clinic venture.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Logistic regression variable odds ratio\u003c/h2\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e reports logistic regression results. Demographic and socio-economic variables were significantly associated with appointment status. Divorced/separated and widowed individuals were more likely to miss appointments (OR 1.57 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05) and 1.76 (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05), respectively). Additionally, payer type emerged as a critical factor, with Medicaid (OR 1.67, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) and uninsured/self-paying (OR 1.44, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.05) persons having higher odds of non-attendance. In contrast, hypertension was significantly associated with lower odds of missing appointments (OR 0.58, \u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001), indicating that specific health conditions might differentially influence appointment adherence. The number of days between appointment scheduling and occurrence (lag days) remained highly significant (\u003cem\u003ep\u003c/em\u003e\u0026thinsp;\u0026lt;\u0026thinsp;.001) after controlling for all other factors in the model, underscoring its impact on appointment adherence.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSummary of logistic regression results\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"4\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOdds ratio\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003e95% CI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ep-value\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eIntercept\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8.7e-08\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0, 10.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.99\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"4\" nameend=\"c4\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003eDemographic and Socio-Economic Factor\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.99\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.98, 1.01)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.41\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eRace/ethnicity\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eBlack\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWhite\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.03\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.82, 1.30)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHispanic\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.21, 1.81)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.39\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAsian\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.58, 2.56)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.60\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnknown\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4.21\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.93, 29.81)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.09\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.01\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.81, 1.25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.92\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarital status\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMarried/Life partner\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSingle\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.85, 1.41)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.50\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDivorced/Separated\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.57\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.06, 2.33)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWidowed\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.76\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.04, 2.97)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUnknown\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.02\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.47, 2.18)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e0.96\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePreferred language\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eEnglish\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpanish\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.44\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.21, 1.44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.23\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.88, 2.36)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.32\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePayer type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCommercial\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedicare\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.31\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.94, 1.84)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.11\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMedicaid\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.67\u003csup\u003e***\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.23, 2.25)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eUninsured/Self-pay\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.44\u003csup\u003e*\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.09, 1.91)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.05\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eSpecific needs description\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eNot applicable\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eReference\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eWheelchair\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e8.96\u003csup\u003e\u0026dagger;\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.10, 48.0)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.61\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.43, 6.94)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.49\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eMedical History\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDiabetes\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.34, 2.34)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.83\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMyocardial Infarction\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.77, 1.63)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.55\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCoronary Artery Disease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.90\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.63, 1.26)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.53\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eChronic Kidney Disease\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.44\u003csup\u003e\u0026dagger;\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.98, 2.11)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.06\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHypertension\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.58\u003csup\u003e***\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.45, 0.74)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eCOPD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.22\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.86, 1.73)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.27\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVenous thromboembolism pulmonary embolism\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.63\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.83, 3.22)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.16\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAsthma\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.12\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.84, 1.50)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.42\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAnxiety\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.20\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.92, 1.57)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.17\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAutoimmune Disorder\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.61, 1.45)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.80\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDepression\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e0.86\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.64, 1.17)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.34\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eDyslipidemia\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.09\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.82, 1.44)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.57\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eHeart Failure\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(0.91, 1.86)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.14\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e\u003cb\u003eAppointment Scheduling Factor\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eLag days\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1.05\u003csup\u003e***\u003c/sup\u003e\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e(1.03, 1.07)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u003cb\u003e\u0026lt;\u0026thinsp;.001\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.3. Decision tree model feature importance\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig3\" class=\"InternalRef\"\u003e3\u003c/span\u003e presents the Gini importance value for the top 20 variables as identified by the decision tree method, ordered by the index values. This list encompasses a mix of scheduling-related, demographic, socio-economic, and clinical characteristics. Distinct from the findings in logistic regression analysis, age appeared to be highly important in the decision tree model. This indicates that age may exhibit a non-linear effect or interact with other variables. It is important to note that a higher feature importance signifies the variable's capacity to distinguish between the arrived and missed appointment classes, not directly suggesting a variable contributes to missed appointments.\u003c/p\u003e\u003cp\u003eFigure 4 visualizes the decision tree, suggesting potential interactions among variables that contribute to appointment status. The numbers in the leaf nodes indicate the counts of arrived and missed appointments, respectively. The red leaf nodes represent the subgroups that have more missed appointments than arrived appointments. For effective visualization, the complexity parameter of the decision tree is set 1.0 x 10\u003csup\u003e\u0026minus;\u0026thinsp;4\u003c/sup\u003e, which produced the decision tree of depth five. Lag days and age appear to be the most influential variables. Demographic and socio-economic variables are associated with appointment adherence among patients with more than 2 but less than 4 lag days, but medical history variables are associated with adherence among those with more than 4 but less than 15 lag days.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.4. XGBoost model feature importance\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig4\" class=\"InternalRef\"\u003e5\u003c/span\u003e presents the three feature metric values (Gain, Cover, and Frequency) for the top 20 variables as identified by the XGBoost method, ordered by their Gain measure values. Similar to the decision tree model, the list encompasses a mix of variable types. Lag days and age were identified as the most informative, followed by insurance status, marital status, and preferred language, highlighting the influence of socio-economic factors.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.5. Sensitivity analysis on probability threshold values\u003c/h2\u003e \u003cp\u003eFigure \u003cspan refid=\"Fig5\" class=\"InternalRef\"\u003e6\u003c/span\u003e illustrates the trade-off relationship between sensitivity and precision in relation to the positive prediction threshold (i.e., the probability at which an appointment is predicted to be missed) in the XGBoost model. The detection prevalence represents the proportion of appointments predicted to be missed (i.e., the proportion of positive predictions). As the probability threshold increases, the model predicts fewer patients to miss their appointments (resulting in lower detection prevalence), leading to a decrease in sensitivity but an increase in precision. For example, setting the positive prediction threshold at 0.4 allows our model to detect 88.4% of positive cases with a precision of 49.0%. On the other hand, with the threshold set at 0.6, the model\u0026rsquo;s sensitivity drops to 34.8% while its precision increases to 70.6%. This describes that a lower threshold increases sensitivity but may include more false positives, while a higher threshold enhances precision but may miss identifying individuals at risk.\u003c/p\u003e \u003cp\u003e \u003c/p\u003e \u003c/div\u003e"},{"header":"4. Discussion","content":"\u003cp\u003eIn this study, we applied a data-driven machine learning method to identify patients more likely to miss a scheduled appointment at a new PDC and compared this approach to more traditional prediction methods. XGBoost stands out for its ability to handle non-linear relationships and intricate interactions between variables [\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e28\u003c/span\u003e, \u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e29\u003c/span\u003e], which are common in health-related data [\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e30\u003c/span\u003e, \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e31\u003c/span\u003e]. Our findings underscore the potential of machine learning to enhance clinic operations by better identifying those at higher risk of missing future appointments, for prioritizing patients for targeted interventions to improve health care operations and ultimately patient outcomes.\u003c/p\u003e \u003cp\u003eOur findings shed light on the significant role that lag time, age, insurance status, marital status, and medical history have on predicting appointment adherence. Because these factors may impact patient engagement beyond appointment attendance, findings may have value in additional patient engagement strategies and across other clinic settings. Particularly, distinct from the findings in logistic regression analysis, the results from the XGBoost and decision tree models consistently identified age as a highly important variable. The significance of the age variable identified by both the decision tree and XGBoost method testifies the models\u0026rsquo; advanced capacity for detecting intricate patterns. For health care clinical and operational leaders, these findings highlight the importance of minimizing the lag time and particular attention to older (\u0026ge;\u0026thinsp;81) or younger (\u0026lt;\u0026thinsp;62) patients who as groups were more likely to miss PDC appointments in our study.\u003c/p\u003e \u003cp\u003eOur analysis also revealed the necessity of optimizing the sensitivity and precision trade-off in predictive models to enhance the effectiveness of appointment adherence interventions. A clinic can adjust the prediction threshold based on their financial resouces and physical and human resource capacity to follow up on predicted no-shows or their prioritization of minimizing missed appointments versus avoiding unnecessary follow-ups. This decision impacts resource allocation, the effectiveness of patient engagement efforts, and ultimately, the clinic's operational efficiency and patient care quality. This optimization ensures the detection of patients at increased risk of missing appointments while minimizing false positives that could lead to resource misallocation and alarm fatigue among healthcare professionals [\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e32\u003c/span\u003e, \u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e33\u003c/span\u003e]. Fine-tuning the prediction threshold is paramount for deploying efficient and sustainable patient engagement strategies, ensuring that patient engagement interventions are both targeted and impactful.\u003c/p\u003e \u003cp\u003eThe dynamic nature of prediction thresholds demands continuous evaluation and adjustment of model parameters based on patient outcomes and new data. This adaptability is vital in health care practices. Integrating advanced predictive analytics into clinical and operational workflows marks a pivotal shift towards more proactive and personalized healthcare. However, realizing these benefits requires addressing potential optimization problems with these tools and enhancing their usability in partnership with healthcare professionals.\u003c/p\u003e \u003cp\u003eOur findings have important implications for both post-discharge care operations as well as strategies to improve patient engagement such as appointment adherence. By demonstrating the value of machine learning models like XGBoost in predicting appointment adherence, our study contributes to the growing field of data-driven healthcare decision-making. Strategies such as these hold promise for allocating resources more judiciously, tailoring interventions to individual patient profiles, and ultimately elevating the quality and efficiency of care.\u003c/p\u003e"},{"header":"5. Limitations","content":"\u003cp\u003eThis study is subject to limitations. First, study findings are based on EHR data from a single institution, potentially limiting their applicability to broader post-discharge care settings. However, we note the variables included in prediction models are ubiquitous in health care settings. Second, the study\u0026rsquo;s sample size was relatively limited for predictive analytics research, a reflection of our intentional focus on supporting patient engagement for a new clinic venture, where patient engagement mechanisms were nascent and system-wide awareness was limited. This prompts consideration of how our findings may not be generalizable for more mature clinics with established patient engagement protocols and system-wide visibility, though the modeling approaches employed can easily be applied in other settings and compared to our findings.\u003c/p\u003e"},{"header":"6. Conclusion and Future Work","content":"\u003cp\u003eThis study has demonstrated the potential for predictive analytics to prioritize patients for enhanced engagement initiatives in transitional ambulatory care by identifying patients at increased risk of missing appointments. Highlighting key variables that influence appointment adherence, our findings offer actionable insights for developing focused interventions delivered to prioritized patients to improve post-discharge care. To the best of our knowledge, this work is among the first to apply machine learning to inform interventions that address post-discharge appointment adherence in new clinical ventures. An important next step involves integrating these models with feasibility studies (e.g., pilot implementation) to evaluate their potential impact on patient outcomes and care continuity. By advancing the application of predictive analytics in healthcare, our work aims to foster personalized and more effective post-discharge care, enhancing healthcare delivery and patient well-being.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eAcknowledgements\u003c/strong\u003e:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe authors thank the UAB Research and Informatics Service Center (RISC) and the Department of Biomedical Informatics and Data Science for their data extraction and transformation expertise.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting Interest\u003c/strong\u003e:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding\u003c/strong\u003e:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis work was supported by the University of Alabama Health Services Foundation General Endowment Fund through a grant to establish the UAB Learning Health System Platform.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics Approval\u003c/strong\u003e:\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThis work was approved by the UAB Institutional Review Board (IRB-300009914).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAvailability of Data and Materials\u003c/strong\u003e:\u003c/p\u003e\n\u003cp\u003eThe EHR datasets analyzed during the current study are not publicly available because they are governed by HIPAA regulations at the University of Alabama at Birmingham.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003e\u003cem\u003eAuthors\u0026rsquo; Contributions:\u0026nbsp;\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eS.L. designed the study, conducted analysis, and led the development of the manuscript.\u003c/p\u003e\n\u003cp\u003eR.E. conducted quantitative data collection and handled correspondence with co-authors.\u003c/p\u003e\n\u003cp\u003eL.H. contributed to the study design and Introduction, Results, and Discussion sections.\u003c/p\u003e\n\u003cp\u003eM.G. contributed to project management tasks and the development of the manuscript.\u003c/p\u003e\n\u003cp\u003eK.H. contributed to the study design, supervised the quantitative data collection, and reviewed the Results section.\u003c/p\u003e\n\u003cp\u003eA.H. contributed to the study design and the Introduction and Discussion sections.\u003c/p\u003e\n\u003cp\u003eG.B. contributed through her data extraction expertise to assist with acquiring quantitative data and reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003eJ.M. contributed the literature review for the Introduction section and reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003eS.M. contributed to the study clinic implementation and reviewed the manuscript.\u003c/p\u003e\n\u003cp\u003eC.S. provided information on the administrative and operation activities of the Post Discharge Clinic and provided feedback through draft review.\u003c/p\u003e\n\u003cp\u003eT.B. assisted with project identification and study development and provided feedback through draft reviews.\u003c/p\u003e\n\u003cp\u003eA.G. contributed to study recruitment and provided feedback on the manuscript through draft reviews.\u003c/p\u003e\n\u003cp\u003eH.B. provided information related to care transition activities at the University of Alabama at Birmingham Health System and outpatient facilities.\u003c/p\u003e\n\u003cp\u003eM.M. contributed to the study design the Discussion section and provided feedback and revisions through draft reviews.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eRagin, D.F., Hwang, U., Cydulka, R.K., Holson, D., Haley, L.L., Jr., Richards, C.F., Becker, B.M., Richardson, L.D., and Emergency Medicine Patients\u0026apos; Access To Healthcare Study, I.: \u0026lsquo;Reasons for using the emergency department: results of the EMPATH Study\u0026rsquo;, Acad Emerg Med, 2005, 12, (12), pp. 1158-1166\u003c/li\u003e\n\u003cli\u003eBrown, L.E., Burton, R., Hixon, B., Kakade, M., Bhagalia, P., Vick, C., Edwards, A., and Hawn, M.T.: \u0026lsquo;Factors influencing emergency department preference for access to healthcare\u0026rsquo;, West J Emerg Med, 2012, 13, (5), pp. 410-415\u003c/li\u003e\n\u003cli\u003eMoscelli, G., Siciliani, L., Gutacker, N., and Cookson, R.: \u0026lsquo;Socioeconomic inequality of access to healthcare: Does choice explain the gradient?\u0026rsquo;, J Health Econ, 2018, 57, pp. 290-314\u003c/li\u003e\n\u003cli\u003ePreen, D.B., Bailey, B.E., Wright, A., Kendall, P., Phillips, M., Hung, J., Hendriks, R., Mather, A., and Williams, E.: \u0026lsquo;Effects of a multidisciplinary, post-discharge continuance of care intervention on quality of life, discharge satisfaction, and hospital length of stay: a randomized controlled trial\u0026rsquo;, Int J Qual Health Care, 2005, 17, (1), pp. 43-51\u003c/li\u003e\n\u003cli\u003eCoppa, K., Kim, E.J., Oppenheim, M.I., Bock, K.R., Conigliaro, J., and Hirsch, J.S.: \u0026lsquo;Examination of Post-discharge Follow-up Appointment Status and 30-Day Readmission\u0026rsquo;, J Gen Intern Med, 2021, 36, (5), pp. 1214-1221\u003c/li\u003e\n\u003cli\u003eElliott, K., Klein, J.W., Basu, A., and Sabbatini, A.K.: \u0026lsquo;Transitional care clinics for follow-up and primary care linkage for patients discharged from the ED\u0026rsquo;, Am J Emerg Med, 2016, 34, (7), pp. 1230-1235\u003c/li\u003e\n\u003cli\u003eDoctoroff, L., Nijhawan, A., McNally, D., Vanka, A., Yu, R., and Mukamal, K.J.: \u0026lsquo;The characteristics and impact of a hospitalist-staffed post-discharge clinic\u0026rsquo;, Am J Med, 2013, 126, (11), pp. 1016 e1019-1015\u003c/li\u003e\n\u003cli\u003eLu, L.Y., Jackevicius, C., de Leon, N., Warner, A., Chang, D., and Mody, F.: \u0026lsquo;Impact of a Multi-Disciplinary Heart Failure Post-Discharge Management Clinic on Medication Adherence\u0026rsquo;, J Am Coll Cardiol, 2014, 63, (12), pp. A539-A539\u003c/li\u003e\n\u003cli\u003eJackevicius, C.A., de Leon, N.K., Lu, L., Chang, D.S., Warner, A.L., and Mody, F.V.: \u0026lsquo;Impact of a Multidisciplinary Heart Failure Post-hospitalization Program on Heart Failure Readmission Rates\u0026rsquo;, Ann Pharmacother, 2015, 49, (11), pp. 1189-1196\u003c/li\u003e\n\u003cli\u003eShah, M., Douglas, V., Scott, B., and Josephson, S.A.: \u0026lsquo;A Neurohospitalist Discharge Clinic Shortens the Transition From Inpatient to Outpatient Care\u0026rsquo;, Neurohospitalist, 2016, 6, (2), pp. 64-69\u003c/li\u003e\n\u003cli\u003eBurke, R.E., Whitfield, E., and Prochazka, A.V.: \u0026lsquo;Effect of a hospitalist-run postdischarge clinic on outcomes\u0026rsquo;, J Hosp Med, 2014, 9, (1), pp. 7-12\u003c/li\u003e\n\u003cli\u003eBaldino, M., Bonaguro, A.M., Burgwardt, S., Lombardi, A., Cristancho, C., Mann, C., Wright, D., Jackson, C., and Seth, A.: \u0026lsquo;Impact of a Novel Post-Discharge Transitions of Care Clinic on Hospital Readmissions\u0026rsquo;, J Natl Med Assoc, 2021, 113, (2), pp. 133-141\u003c/li\u003e\n\u003cli\u003eDistelhorst, K., Claussen, R., Dion, K., Bena, J.F., Morrison, S.L., Walker, D., Tai, H.L., and Albert, N.M.: \u0026lsquo;Factors Associated With Adherence to 14-Day Office Appointments After Heart Failure Discharge\u0026rsquo;, J Card Fail, 2018, 24, (6), pp. 407-411\u003c/li\u003e\n\u003cli\u003eTruong, E.I., DeMario, B.S., Hendrickson, S., Kalina, M.J., Jr., Vallier, H.A., Tseng, E.S., Claridge, J.A., and Ho, V.P.: \u0026lsquo;Factors Influencing Nonadherence to Recommended Postdischarge Follow-Up After Trauma\u0026rsquo;, J Surg Res, 2020, 256, pp. 143-148\u003c/li\u003e\n\u003cli\u003eShamout, F., Zhu, T., and Clifton, D.A.: \u0026lsquo;Machine Learning for Clinical Outcome Prediction\u0026rsquo;, IEEE Rev Biomed Eng, 2021, 14, pp. 116-126\u003c/li\u003e\n\u003cli\u003eBrink, A., Alsma, J., van Attekum, L.A.A.M., Bramer, W.M., Zietse, R., Lingsma, H., and Schuit, S.C.E.: \u0026lsquo;Predicting inhospital admission at the emergency department: a systematic review\u0026rsquo;, Emerg Med J, 2022, 39, (3), pp. 191-198\u003c/li\u003e\n\u003cli\u003eStinchcombe, A.L.: \u0026lsquo;Social structure and organizations\u0026rsquo;: \u0026lsquo;Hand of Organizations (RLE: Organizations)\u0026rsquo; (Routledge, 2013), pp. 142-193\u003c/li\u003e\n\u003cli\u003eAarons, G.A., and Palinkas, L.A.: \u0026lsquo;Implementation of evidence-based practice in child welfare: service provider perspectives\u0026rsquo;, Adm Policy Ment Health, 2007, 34, (4), pp. 411-419\u003c/li\u003e\n\u003cli\u003eSchell, S.F., Luke, D.A., Schooley, M.W., Elliott, M.B., Herbers, S.H., Mueller, N.B., and Bunger, A.C.: \u0026lsquo;Public health program capacity for sustainability: a new framework\u0026rsquo;, Implement Sci, 2013, 8, pp. 15\u003c/li\u003e\n\u003cli\u003eKurtz Landy, C., Sword, W., and Ciliska, D.: \u0026lsquo;Urban women\u0026apos;s socioeconomic status, health service needs and utilization in the four weeks after postpartum hospital discharge: findings of a Canadian cross-sectional survey\u0026rsquo;, BMC Health Serv Res, 2008, 8, pp. 203\u003c/li\u003e\n\u003cli\u003eHosmer, D.W., Lemeshow, S., Sturdivant, R.X., and ProQuest: \u0026lsquo;Applied logistic regression\u0026rsquo; (Wiley, 2013, Third edition edn. 2013)\u003c/li\u003e\n\u003cli\u003eBoateng, E.Y.A., D. A.: \u0026lsquo;A Review of the Logistic Regression Model with Emphasis on Medical Research\u0026rsquo;, Journal of Data Analysis and Information Processing, 2019, 7, pp. 190 - 207\u003c/li\u003e\n\u003cli\u003eOmer, D.M., A. B.: \u0026lsquo;Modelling logistic regression using multivariable fractional polynomials\u0026rsquo;, Imperial Journal of Interdisciplinary Research, 2017, 3, (11), pp. 8-16\u003c/li\u003e\n\u003cli\u003eBache-Mathiesen, L.K., Andersen, T.E., Dalen-Lorentsen, T., Clarsen, B., and Fagerland, M.W.: \u0026lsquo;Not straightforward: modelling non-linearity in training load and injury research\u0026rsquo;, BMJ Open Sport Exerc Med, 2021, 7, (3), pp. e001119\u003c/li\u003e\n\u003cli\u003eHastie, T., Friedman, J.H., and Tibshirani, R.: \u0026lsquo;The elements of statistical learning : data mining, inference, and prediction\u0026rsquo; (Springer, 2009. 2009)\u003c/li\u003e\n\u003cli\u003eChen, T.G., C. : \u0026lsquo;A scalable tree boosting system\u0026rsquo;, in Editor (Ed.)^(Eds.): \u0026lsquo;Book A scalable tree boosting system\u0026rsquo; (Association for Computing Machinery, 2016, edn.), pp. 785-794\u003c/li\u003e\n\u003cli\u003eBishop, C.M.: \u0026lsquo;Pattern Recognition and Machine Learning\u0026rsquo; (Springer, 2006, Softcover reprint of the original 1st 2006. edn. 2006)\u003c/li\u003e\n\u003cli\u003eYang, C., Chen, M., and Yuan, Q.: \u0026lsquo;The application of XGBoost and SHAP to examining the factors in freight truck-related crashes: An exploratory analysis\u0026rsquo;, Accid Anal Prev, 2021, 158, pp. 106153\u003c/li\u003e\n\u003cli\u003eLiu, J.X., Wang, B., and Xiao, L.Z.: \u0026lsquo;Non-linear associations between built environment and active travel for working and shopping: An extreme gradient boosting approach\u0026rsquo;, J Transp Geogr, 2021, 92\u003c/li\u003e\n\u003cli\u003eOrfanoudaki, A.C., E.; Cadisch, C.; Stein, B.; Nouh, A.; Alberts, M. J.; Bertsimas, B. : \u0026lsquo;Machine learning provides evidence that stroke risk is not linear: The non-linear Framingham stroke risk score\u0026rsquo;, PLOs one, 2020, 21, (15)\u003c/li\u003e\n\u003cli\u003eRodgers, B., Korten, A.E., Jorm, A.F., Jacomb, P.A., Christensen, H., and Henderson, A.S.: \u0026lsquo;Non-linear relationships in associations of depression and anxiety with alcohol use\u0026rsquo;, Psychol Med, 2000, 30, (2), pp. 421-432\u003c/li\u003e\n\u003cli\u003eCvach, M.: \u0026lsquo;Monitor alarm fatigue: an integrative review\u0026rsquo;, Biomed Instrum Technol, 2012, 46, (4), pp. 268-277\u003c/li\u003e\n\u003cli\u003eHravnak, M., Pellathy, T., Chen, L., Dubrawski, A., Wertz, A., Clermont, G., and Pinsky, M.R.: \u0026lsquo;A call to alarms: Current state and future directions in the battle against alarm fatigue\u0026rsquo;, J Electrocardiol, 2018, 51, (6S), pp. S44-S48\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Machine Learning, Patient Compliance, Post Discharge Care, Health Services Research, Appointments and Schedules, Electronic Health Records, Healthcare Disparities, Risk Factors","lastPublishedDoi":"10.21203/rs.3.rs-4477049/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-4477049/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003ch2\u003eObjective\u003c/h2\u003e \u003cp\u003eThis study applies predictive analytics to identify patients at risk of missing appointments at a novel post-discharge clinic (PDC) in a large academic health system. Recognizing the critical role of appointment adherence in the success of new clinical ventures, this research aims to inform future targeted interventions to increase appointment adherence.\u003c/p\u003e\u003ch2\u003eMaterials and Methods\u003c/h2\u003e \u003cp\u003eWe analyzed electronic health records (EHR) capturing a wide array of demographic, socio-economic, and clinical variables from 2,168 patients with scheduled appointments at the PDC from September 2022 to August 2023. Logistic regression, decision trees, and XGBoost algorithms were employed to construct predictive models for appointment adherence.\u003c/p\u003e\u003ch2\u003eResults\u003c/h2\u003e \u003cp\u003eThe XGBoost machine learning model outperformed logistic regression and decision trees with an area under the curve of 72% vs. 65% and 67%, respectively, in predicting missed appointments, despite limited availability of historical data. Key predictors included patient age, number of days between appointment scheduling and occurrence, insurance status, marital status, and mental health and cardiac disease conditions.\u003c/p\u003e\u003ch2\u003eDiscussion\u003c/h2\u003e \u003cp\u003eFindings underscore the potential of machine learning predictive analytics to significantly enhance patient engagement and operational efficiency in emerging healthcare settings. Optimizing predictive models can help balance the early identification of patients at risk of non-adherence with the efficient allocation of resources.\u003c/p\u003e\u003ch2\u003eConclusion\u003c/h2\u003e \u003cp\u003eThe study highlights the potential value of employing machine learning techniques to inform interventions aimed at improving appointment adherence in a post-discharge transition clinic environment.\u003c/p\u003e","manuscriptTitle":"Enhancing Patient Engagement with Machine Learning at a Novel Care Transition Clinic","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-06-11 20:38:59","doi":"10.21203/rs.3.rs-4477049/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"3cad213c-07d7-4bec-8c0f-3ccdca96fdc9","owner":[],"postedDate":"June 11th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2024-06-11T20:39:01+00:00","versionOfRecord":[],"versionCreatedAt":"2024-06-11 20:38:59","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-4477049","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-4477049","identity":"rs-4477049","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}