Readmission Prediction for Patients with Heterogeneous Medical History: A Trajectory-Based Deep Learning Approach

Hospital readmission refers to the situation where a patient is re-hospitalized with the same primary diagnosis within a specific time interval after discharge. Hospital readmission causes $26 billion preventable expenses to the U.S. health systems annually and often indicates suboptimal patient care. To alleviate those severe financial and health consequences, it is crucial to proactively predict patients’ readmission risk. Such prediction is challenging because the evolution of patients’ medical history is dynamic and complex. The state-of-the-art studies apply statistical models which use static predictors in a period, failing to consider patients’ heterogeneous medical history. Our approach – Trajectory-BAsed DEep Learning (TADEL) – is motivated to tackle the deficiencies of the existing approaches by capturing dynamic medical history. We evaluate TADEL on a five-year national Medicare claims dataset including 3.6 million patients per year over all hospitals in the United States, reaching an F1 score of 87.3% and an AUC of 88.4%. Our approach significantly outperforms all the state-of-the-art methods. Our findings suggest that health status factors and insurance coverage are important predictors for readmission. This study contributes to IS literature and analytical methodology by formulating the trajectory-based readmission prediction problem and developing a novel deep-learning-based readmission risk prediction framework. From a health IT perspective, this research delivers implementable methods to assess patients’ readmission risk and take early interventions to avoid potential negative consequences.

Development of a Readiness for Hospital Discharge assessment tool in Thai patients with stroke

Background: The transition from hospital to home among patients with stroke is quite challenging. If the patients are not ready for hospital discharge, their condition may worsen, which also causes a high rate of readmission. Although instruments to measure readiness for hospital discharge exist, none of them fit with the Thailand context. Objective: This study aimed to develop a Readiness for Hospital Discharge assessment tool in Thai patients with stroke. Methods: The study was conducted from February to September 2020, which consisted of several steps: 1) conducting an extensive literature review, 2) content validity with five experts, 3) pilot testing with 30 samples, and 4) field testing with 348 participants. Content validity index (CVI) was used to measure the content validity, Cronbach’s alpha and inter-item correlation to evaluate reliability, and multiple logistic regression analysis to measure the construct validity. Results: The findings showed good validity and reliability, with I-CVI of 0.85, Cronbach’s alpha of 0.94, and corrected item-total correlation ranging from 0.43 to 0.86. The construct validity was demonstrated through the results of regression analysis showing that the nine variables include level of consciousness (OR = 0.544; CI 95% = 0.311 - 0.951), verbal response (OR = 0.445; 95% CI 0.272- 0.729), motor power right leg (OR = 0.165; 95% CI 0.56- 0.485), visual field (OR = 0.188; 95% CI 0.60-0.587), dysphagia (OR = 0.618; 95% CI 0.410-0.932), mobility (OR = 0.376; 95% CI 0.190 - 0.741), self-feeding (OR = 0.098; 95% CI 0.036 -0.265), bathing (OR = 0.099; 95% CI 0.026-0.378), and bladder control (OR = 0.589; 95% CI 0.355-0.977) that significantly influenced the hospital readmission within 30 days in patients with stroke. Conclusion: The Readiness for Hospital Discharge assessment tool is valid and reliable. Healthcare providers, especially nurses, can use this tool to assess discharge conditions for patients with stroke with greater accuracy in predicting hospital readmission.

Hospital revisits after paediatric tonsillectomy: a cohort study

Background Tonsillectomy, with or without adenoidectomy, is the leading reason for paediatric unplanned hospital readmission, some of which are potentially avoidable. Reducing unplanned hospital revisits would improve patient safety and decrease use of healthcare resources. This study aimed to describe the incidence, timing and risk factors for any surgery-related hospital revisits (both emergency presentation and readmission) following paediatric tonsillectomy and adenotonsillectomy in a large state-wide cohort. Methods We conducted a population-based cohort study using linked administrative datasets capturing all paediatric tonsillectomy and adenotonsillectomy surgeries performed between 2010 and 2015 in the state of Victoria, Australia. The primary outcome was presentation to the emergency department or hospital readmission within 30-day post-surgery. Results Between 2010 and 2015, 46,583 patients underwent 47,054 surgeries. There was a total of 4758 emergency department presentations (10.11% total surgeries) and 2750 readmissions (5.84% total surgeries). Haemorrhage was the most common reason for both revisit types, associated with 33.02% of ED presentations (3.34% total surgeries) and 67.93% of readmissions (3.97% total surgeries). Day 5 post-surgery was the median revisit time for both ED presentations (IQR 3–7) and readmission (IQR 3–8). Predictors of revisit included older age, public and metropolitan hospitals and peri-operative complications during surgery. Conclusions Haemorrhage was the most common reason for both emergency department presentation and hospital readmission. The higher risk of revisits associated with older children, surgeries performed in public and metropolitan hospitals, and in patients experiencing peri-operative complications, suggest the need for improved education of postoperative care for caregivers, and avoidance of inappropriate early discharge. Graphical Abstract

Predicting Hospital Readmission for Campylobacteriosis from Electronic Health Records: A Machine Learning and Text Mining Perspective

(1) Background: This study investigates influential risk factors for predicting 30-day readmission to hospital for Campylobacter infections (CI). (2) Methods: We linked general practitioner and hospital admission records of 13,006 patients with CI in Wales (1990–2015). An approach called TF-zR (term frequency-zRelevance) technique was presented to evaluates how relevant a clinical term is to a patient in a cohort characterized by coded health records. The zR is a supervised term-weighting metric to assign weight to a term based on relative frequencies of the term across different classes. Cost-sensitive classifier with swarm optimization and weighted subset learning was integrated to identify influential clinical signals as predictors and optimal model for readmission prediction. (3) Results: From a pool of up to 17,506 variables, 33 most predictive factors were identified, including age, gender, Townsend deprivation quintiles, comorbidities, medications, and procedures. The predictive model predicted readmission with 73% sensitivity and 54% specificity. Variables associated with readmission included male gender, recurrent tonsillitis, non-healing open wounds, operation for in-gown toenails. Cystitis, paracetamol/codeine use, age (21–25), and heliclear triple pack use, were associated with a lower risk of readmission. (4) Conclusions: This study gives a profile of clustered variables that are predictive of readmission associated with campylobacteriosis.

Association between unmet medication needs after hospital discharge and readmission or death among acute respiratory failure survivors: the addressing post-intensive care syndrome (APICS-01) multicenter prospective cohort study

Introduction Survivors of acute respiratory failure (ARF) commonly experience long-lasting physical, cognitive, and/or mental health impairments. Unmet medication needs occurring immediately after hospital discharge may have an important effect on subsequent recovery. Methods and analysis In this multicenter prospective cohort study, we enrolled ARF survivors who were discharged directly home from their acute care hospitalization. The primary exposure was unmet medication needs. The primary outcome was hospital readmission or death within 3 months after discharge. We performed a propensity score analysis, using inverse probability weighting for the primary exposure, to evaluate the exposure–outcome association, with an a priori sample size of 200 ARF survivors. Results We enrolled 200 ARF survivors, of whom 107 (53%) were female and 77 (39%) were people of color. Median (IQR) age was 55 (43–66) years, APACHE II score 20 (15–26) points, and hospital length of stay 14 (9–21) days. Of the 200 participants, 195 (98%) were in the analytic cohort. One hundred fourteen (57%) patients had at least one unmet medication need; the proportion of medication needs that were unmet was 6% (0–15%). Fifty-six (29%) patients were readmitted or died by 3 months; 10 (5%) died within 3 months. Unmet needs were not associated (risk ratio 1.25; 95% CI 0.75–2.1) with hospital readmission or death, although a higher proportion of unmet needs may have been associated with increased hospital readmission (risk ratio 1.7; 95% CI 0.96–3.1) and decreased mortality (risk ratio 0.13; 95% CI 0.02–0.99). Discussion Unmet medication needs are common among survivors of acute respiratory failure shortly after discharge home. The association of unmet medication needs with 3-month readmission and mortality is complex and requires additional investigation to inform clinical trials of interventions to reduce unmet medication needs. Study registration number: NCT03738774. The study was prospectively registered before enrollment of the first patient.