Characterising adverse events in postdischarge high-risk patients with 6-month follow-up in a Korean tertiary hospital: a retrospective cohort study Authors: Jee-In Hwang, Yoon-Seok Chang, Ho Jun Chin Published: 2025 ABSTRACT Abstract AbstractObjectivesThis study aimed to examine adverse event occurrences in outpatient care settings and healthcare utilisation after discharge and to identify the characteristics of patients at high risk of safety issues during the care transition from the hospital to their homes. In addition, the performance of outpatient triggers for the detection of adverse events was explored.DesignA retrospective cohort study using hospital medical record reviews over 6 months postdischarge.SettingA tertiary teaching hospital in South Korea.ParticipantsA total of 746 adult patient medical records mainly including patients potentially at high risk of adverse events were analysed, comprising 5439 outpatient visits.Primary outcome measureOccurrence of adverse events.ResultsApproximately 14.2% of patients experienced at least one adverse event within 6 months postdischarge. Medication-related (84.2%) and surgical/procedural (13.2%) events were common. Most events were mild in severity and assessed as unpreventable. Patients who experienced adverse events had significantly more all-cause re-hospitalisations, emergency room visits and outpatient clinic visits. No difference was found in mortality. Controlling for other variables, multiple logistic regression analysis showed that patients who are overweight (OR=0.48, 95% CI 0.25 to 0.90), with circulatory system diseases (OR=0.38, 95% CI 0.15 to 0.96) and with a National Early Warning Score 2 (NEWS2) ≥2 at discharge (OR=0.26, 95% CI 0.12 to 0.54) were less likely to experience adverse events. Those with admissions via emergency rooms or day-surgery centres (OR=1.98, 95% CI 1.18 to 3.32), in medical departments (OR=1.93, 95% CI 1.15 to 3.23), with a diagnosis of cancer (OR=2.03, 95% CI 1.14 to 3.62) and longer hospital stays (OR=1.07, 95% CI 1.02 to 1.11) were more likely to experience adverse events postdischarge. Outpatient triggers with relatively high performance included ‘constipation with narcotics/calcium channel blockers’, ‘abrupt medication discontinuation’, ‘rash’ and ‘white blood cell count <3000/mm³’.ConclusionsApproximately 14% of patients, mainly comprising high-risk patients, experienced adverse events postdischarge, with more emergency room visits and all-cause re-hospitalisations. Patients experiencing adverse events were characterised by body mass index, type of admission route, clinical department, medical diagnosis, length of hospital stay and NEWS2. Patient and caregiver participation in the monitoring and reporting of adverse events should be encouraged to improve postdischarge patient safety. The use of triggers to detect potential patient safety problems is recommended.Keywords: Health policy, Quality in health care, Risk management STRENGTHS AND LIMITATIONS OF THIS STUDY This electronic medical record (HIMSS stage 7 certified)-based cohort study includes physicians’ documentation of adverse events and laboratory results, which may strengthen its findings. This electronic medical record (HIMSS stage 7 certified)-based cohort study includes physicians’ documentation of adverse events and laboratory results, which may strengthen its findings. This study examines the occurrence of adverse events in the outpatient setting during the 6 months after discharge, using a randomly sampled cohort of patients with or without a possible high risk of adverse events in a Korean tertiary teaching hospital. This study examines the occurrence of adverse events in the outpatient setting during the 6 months after discharge, using a randomly sampled cohort of patients with or without a possible high risk of adverse events in a Korean tertiary teaching hospital. This study uses previously validated outpatient triggers as screening tools for detecting adverse events. This study uses previously validated outpatient triggers as screening tools for detecting adverse events. However, this study has limitations owing to the use of patient data from a single tertiary teaching hospital and its reliance on retrospective medical record reviews. However, this study has limitations owing to the use of patient data from a single tertiary teaching hospital and its reliance on retrospective medical record reviews. Introduction Patient safety refers to the prevention and reduction of risk of unnecessary harm to patients due to healthcare management.1 Globally, approximately 1 in 10 patients experience a healthcare-related harm event, and the economic burden from these events is substantial, estimated to account for approximately 15% of the total hospital expenditure globally.2 In South Korea, continuous efforts to enhance patient safety have been underway.3 As of 2024, since 29 July 2016, approximately 109 906 patient safety events, including near misses, have been reported via the National Patient Safety Reporting and Learning System.4 Furthermore, studies based on medical record reviews indicate that approximately 7%–10.7% of hospitalised patients experience at least one adverse event during their hospital stay.5,7 Thus, there is an ongoing need for dedicated initiatives to improve patient safety. However, research on the occurrence of adverse events in outpatient care settings is lacking. Previous research has shown that 2%–25% of patients in primary and ambulatory care settings have experienced an adverse event.8,15 The frequency and type of adverse events in outpatient settings vary. For instance, a previous study, based on interviews with outpatients with no hospitalisation history in the previous 30 days across four Latin American countries, found that the incidence of adverse events was 2.4% during the 42 days of follow-up after the index ambulatory care visit, and 60% of these events were medication-related.8 Another study, using a retrospective telephone survey with outpatient clinic users in Germany, found that 14% of patients experienced patient safety problems, with most events related to patient history/diagnostic procedures and medication, mainly in general practice, orthopaedics and internal medicine care.11 Furthermore, studies have used medical records and triggers to facilitate the detection of adverse events in primary and ambulatory care settings.16,20 However, there is still a scarcity of research on the measurement and characteristics of adverse events in outpatient care settings.18,20 In particular, the care transition period, which includes the transition from the hospital to the patient’s home, is recognised as a high-risk time for patient safety.21 Previous studies found that 19%–28.7% of postdischarge patients experienced at least one adverse event, and approximately 6%–50% of the events were considered to be preventable.22,25 Among postdischarge events, medication-related events were common.22,2426 Although discharge planning and transitional care programmes to prevent adverse events during care transition and improve patient outcomes have been applied as routine practice in hospitals,27,29 additional studies including follow-up assessments after discharge are needed to capture the adverse events in outpatient settings and develop strategies for patient safety improvements. Furthermore, maintaining care continuity via timely postdischarge outpatient follow-up significantly impacts patient outcomes.30 Therefore, we aimed to (1) investigate the occurrence of adverse events in outpatient care settings and healthcare utilisation among patients over 6 months after hospital discharge and (2) characterise patients with a high risk of safety concerns during care transitions. Additionally, we explored the performance of outpatient triggers as screening tools to detect the occurrence of adverse events. The study findings could inform the development of strategies to enhance patient safety in postdischarge outpatient care. Methods Study design This retrospective cohort study involved medical record reviews with 6-month follow-up. Electronic medical records of patients admitted to a tertiary hospital (HIMSS Stage 7 certified) from 1 October 2021 to 31 January 2022 were reviewed. This study was part of a large project on patient safety resilience (2022–2025) funded by the National Research Foundation of Korea. Study participants and setting The target population included patients discharged from the hospital to their homes. The sample was randomly selected from the patient population, mainly comprising patients at potentially high risk of adverse events, who had been admitted to Seoul National University Bundang hospital, South Korea, between 1 October 2021 and 31 January 2022. The study hospital has 1223 beds. This tertiary hospital has fully electronic medical records and barcode systems for medications (HIMSS stage 7 certified). As of 2022, the annual number of inpatients was 425 395 and that of outpatients was 1 711 063, with the average daily census of outpatients being 6538. The inclusion criteria were as follows: (1) patients aged 19 years or older, (2) with a length of stay of 2 days or more and (3) who have had at least one follow-up visit at ambulatory clinics of the study hospital 6 months after discharge. The exclusion criteria included: (1) psychiatric, dental and paediatric patients (referring to the clinical department that patients were admitted to and discharged from) and (2) those discharged to other hospitals or healthcare facilities, including nursing homes at the time of index discharge. In addition, we excluded patients with a length of stay exceeding 31 days to obtain a more representative sample of the target population. To include patients at increased risk of patient safety issues during the care transition from the hospital to their homes (ie, potentially at-risk group), we primarily sampled patients (1) with a National Early Warning Score 2 (NEWS2) of 2 or higher, based on previous studies (ie, at-risk group 1) and (2) with medications at discharge falling under the following categories: anti-infective, potassium/electrolytes, insulin, narcotics/opioids/sedatives, chemotherapeutic agents and heparin/anticoagulants (APINCH); or with five or more medications at discharge (ie, at-risk group 2).31 32 In this study, specific lists of high-risk medications were identified through reviews of a pharmacist in a managerial position at the study hospital. The sample size was determined to be approximately 800 (500 for the potentially at-risk group and 300 for the general group) as a practical sample size based on similar studies.5 7 The final analysis included data from 746 patients, after excluding patients who were enrolled in clinical trials; those discharged to another hospital or nursing home; those under 19 years of age and those who had no outpatient visits after discharge, as identified based on further validation (figure 1). This sample size was appropriate for the final multiple logistic regression analysis with 14 variables. Flow chart of the study sample selection process. ED, emergency department; NEWS2, National Early Warning Score 2. Data collection Patient medical records at the hospital from index discharge up to 6 months postdischarge were reviewed in two phases. In phase one, the first author and two trained registered nurses reviewed electronic medical records using outpatient triggers to detect the occurrence of adverse events. To improve reliability among the reviewers, the first author trained them in a half-day session. One reviewer had prior experience of using inpatient triggers. We used 37 triggers based on the outpatient trigger tool of the Institute for Healthcare Improvement33 and medication-related and laboratory triggers.834,36 Among them, ‘complaint letters’ and ‘>3 calls/hospital visits in 1 week’ were reviewed only based on the notes on the medical records. We also used the Global Trigger Tool for inpatient care, when patients were readmitted.5 37 The first author re-reviewed cases flagged by trained nurse reviewers for additional evaluation, which represented approximately 10% of the medical records. Furthermore, the first author double-checked selected cases with adverse events to enhance the reliability of the review. Any inconsistencies were resolved through discussions with the reviewers. In phase two, two medical professors reviewed cases with positive triggers, along with the first author. The two medical professors had experience conducting similar studies. Adverse events were defined as injuries resulting from healthcare management rather than the patient’s underlying diseases.1 To characterise the adverse events detected in outpatient care, they were categorised using a classification system of adverse event types.38 It comprised six major categories: blood transfusion reaction (13 types), healthcare-associated infection (15 types), medication (12 types), patient care (6 types), perinatal/maternal (2 types) and surgical/procedural (16 types), after excluding ‘culminating event only’ types. The granularity of their subtypes varied depending on categories. Their severity was determined using five categories: (1) temporary patient harm, requiring intervention; (2) temporary patient harm, requiring initial or prolonged hospitalisation; (3) permanent patient harm; (4) intervention required to sustain life and (5) death.37 Preventability was measured using the following 6-point rating scale: ‘1’=no evidence of preventability, ‘2’=minimal possibility of preventability, ‘3’=slight possibility of preventability, ‘4’=moderate possibility of preventability, ‘5’=high possibility of preventability and ‘6’=total evidence of preventability.39 Preventability scores of 4 or higher were considered preventable.5 In addition, if patients received the standard of care (ie, no errors were reported), the event was judged to be unpreventable (‘1’ or ‘2’).12 All data on the patients’ general characteristics, healthcare utilisation and discharge nursing care plans were extracted from the Hospital Information System. Patients’ general characteristics included sex, birth year, marital status, education level, height, weight, residence, type of health insurance, previous admission history, admission route, admission method, discharge method, nursing care acuity classification and patient status at discharge. Healthcare utilisation characteristics included the clinical department, medical diagnoses, surgical procedures received, intensive care unit admission, cardiopulmonary resuscitation (CPR) events, transfers within hospital stay and length of stay for index hospitalisation. For the postdischarge period, data on emergency room visits and readmission dates were also extracted from the Hospital Information System. The level of care complexity was calculated using the Charlson Comorbidity Index (CCI).40 Levels of body mass index (BMI) were categorised as underweight (<18.5 kg/m2), normal weight (18.5–22.9 kg/m2), overweight (23.0–24.9 kg/m2) and obesity (≥25.0 kg/m2).41 The NEWS2 was calculated based on vital signs at the time of discharge.32 Statistical analysis Data were analysed using the SAS software (V.9.4; SAS Institute). We used the SAS SURVEYSELECT procedure with simple random sampling in selecting study sample. We sampled cases for interrater reliability assessment, using a random number generator of the Research Randomizer Programme.42 Moreover, if data on vital signs, height and weight at the time of discharge were unavailable, we used data closest to the discharge time to calculate NEWS2 and BMI. Since the study variables of interest were obtained from the Hospital Information System and were complete, no imputation was required. Patient characteristics, healthcare utilisation, occurrence of adverse events and patient outcomes were summarised using descriptive statistics. The inter-rater reliability among the first author and two medical professors for adverse event assessment, including severity and preventability, was calculated using Fleiss kappa with 60 random cases. To examine differences in the occurrence of adverse events according to patients’ general and healthcare utilisation characteristics, independent t-tests were conducted for continuous variables and χ2 tests were performed for categorical variables. The Wilcoxon rank-sum test was employed, when necessary. Adverse events detected in outpatient care were categorised using a classification system of adverse event types.38 Performance of each trigger was measured by calculating the positive predictive value. To identify the characteristics of patients at increased risk of patient safety issues during care transitions, multiple logistic regression analysis was performed based on univariate analysis results. Model fit was assessed using the Hosmer-Lemeshow goodness-of-fit test. The ORs with 95% CIs were calculated. Statistical significance was set at a two-tailed p value<0.05. Patient and public involvement Patients and the public were not involved in the design, conduct, reporting or dissemination plans of this research. Results Data from 746 patients (489 in the potentially at-risk group and 257 in the general group, respectively), encompassing 5439 outpatient care visits up to 6 months postdischarge, were analysed. In relation to our sampling strategy, we observed a significant difference in the occurrence of adverse events between patients with a NEWS2 ≥2 and those with NEWS2 <2 (χ2=5.33, p=0.021), and between patients taking five or more medications and those taking four or fewer medications (χ2=10.96, p<0.001). However, no significant difference was observed in the occurrence of adverse events between the potentially at-risk group (n=74/489, 15.1%) and the general group (n=32/257, 12.5%) (χ2=0.99, p=0.319). Additionally, no significant difference was found in the occurrence of adverse events between patients with high-risk medications and those without (χ2=1.52, p=0.218). Sample characteristics Sample characteristics are provided in table 1. Among the patients, 52.0% were women, with a median age of 64 years (IQR 53–74; range 19–98). Most (90.4%, n=538/594) were married, and 45.7% (n=271/593) had a university graduate or higher level of education. Most patients (90.2%) resided in smaller cities or provinces, and 95.7% were beneficiaries of the National Health Insurance. Of the patients, 79.5% had a previous admission history and 36.2% had a normal BMI. Table 1General characteristics of study sample by occurrence of adverse events (n=746)VariableCategoryn%Adverse eventχ2/ZP valueYes (n=106)No (n=640)n%n%SexMale35747.94845.330948.30.330.567Age19–45 years12516.81211.311317.74.850.18346–65 years28838.65047.223837.266–75 years17223.12220.815023.476–98 years16121.62220.813921.7Median, IQR (range 19–98)6453–746454–746452–740.14*0.886Marital status (n=594)Married53890.46864.246973.30.090.768Education (n=593)Elementary school or lower8614.51110.47511.70.360.949Middle school6511.098.5568.8High school17128.82321.714823.1University or higher27145.73230.223937.3ResidenceSmaller cities or provinces67390.29892.557589.80.700.402National health insurance(=Yes)71495.710195.361395.80.060.796†Body mass index<18.5466.2109.4365.69.950.01918.5 to <23.027036.24946.222134.523.0 to <25.017122.91615.115524.2≥25.025934.73129.222835.6Previous admission history(=Yes)59379.59286.850178.34.040.044Admission methodOn foot55374.17167.048275.33.790.150Wheelchair476.3109.4375.8Others14619.62523.612118.9Admission routeOutpatient clinic54272.76561.347774.57.990.005ED or day-surgery centre20427.44138.716325.5Clinical department Medical37350.07167.030247.214.25<0.001Surgical37350.03533.033852.8ICU stay(=Yes)7710.365.77111.12.900.089Ward transfer(=Yes)29139.04340.624838.80.130.723Surgical procedures(=Yes)30040.22220.827843.419.46<0.001Major medical diagnosisNeoplasms35147.16662.328544.514.390.006Circulatory system11916.076.611217.5Digestive system435.843.8396.1Genitourinary system435.865.7375.8Others19025.52321.716726.1No. of medical diagnosesMedian, IQR (range 1–13)21–221–221–20.83*0.406CCI scoreMedian, IQR (range 0–18)00–100–200–12.31*0.021Length of stayMedian, IQR (range 3–31)54–864–1154–83.25*0.001Patient acuity classificationGrade 1 or 241355.45249.136156.43.620.163Grade 330440.84744.325740.2Grade 4293.976.6223.4Discharge methodOn foot66689.39185.857589.81.620.444Wheelchair192.643.8152.3Others618.21110.4507.8Patient status at dischargeImproved68191.39286.858992.03.140.077Unchanged/maintained658.71413.2518.0Taking high-risk medication postdischarge (=Yes)38852.06157.632751.11.520.218Taking ≥5 medications at discharge (=Yes)27336.65450.921934.210.96<0.001NEWS2 at discharge (≥2) (=Yes)12917.3109.411918.65.330.021Follow-up care plan after dischargeLimited daily activity (=Yes)8010.7109.47010.90.210.643Special diet formula (=Yes)7610.21110.46510.20.000.944*Wilcoxon rank-sum test; Z includes a continuity correction of 0.5.†Fisher’s exact test.CCI, Charlson Comorbidity Index; ED, emergency department; ICU, intensive care unit; NEWS2, National Early Warning Score 2. General characteristics of study sample by occurrence of adverse events (n=746) Wilcoxon rank-sum test; Z includes a continuity correction of 0.5. Fisher’s exact test. CCI, Charlson Comorbidity Index; ED, emergency department; ICU, intensive care unit; NEWS2, National Early Warning Score 2. Most patients were admitted via outpatient clinics (72.7%) and arrived on foot (74.1%), reflecting their functional status. Overall, 50.0% of the patients were treated in medical departments. During their hospital stay, 10.3% of the patients were admitted to the intensive care unit, 39.0% underwent intra-hospital transfer and 40.2% underwent surgery. The median length of hospital stay was 5 days (IQR 4–8; range 3–31). The median number of medical diagnoses was 2 (IQR 1–2). Of the patients, 89.3% were discharged on foot, with an ‘improved’ status (91.3%) and five or more medications (36.6%) at discharge. The median CCI score was 0 (IQR 0–1), and 17.3% had a NEWS2 ≥2. None of the patients experienced a CPR event during their hospitalisation. Regarding follow-up care plans after discharge, 10.7% required limitations on their daily activities for stabilisation or rehabilitation treatments, and 10.2% needed a special diet formula. Four patients (two from the adverse event group and two from the non-adverse event group) were discharged with an intravenous line. Performance of outpatient triggers Fleiss’ kappa was 0.74 for presence of adverse events, 0.71 for their severity and 0.73 for their preventability. Results regarding the sensitivity of the 37 triggers used in this study are presented in table 2. The top five frequently observed triggers were as follows: ‘medications exceeding five in number’, ‘antiemetics use’, ‘diphenhydramine prescription’, ‘Serum glucose >110 mg/dL and systemic corticosteroids’ and ‘estimated glomerular filtration rate <35 mL/min/1.73 m²’. Among these, triggers demonstrating a positive predictive value exceeding 50% included ‘constipation with narcotics or calcium channel blockers’, ‘abrupt medication discontinuation’, ‘rash’ and ‘white blood cell (WBC) count <3000/mm³ and absolute neutrophil count (ANC) <1500/ mm³’. Table 2Performance of outpatient triggers (n=5439)Triggern%PositiveAEPPV95 % CIT1. New diagnosis of cancer100.200.00.0 to 0.0T2. Placement in nursing home or other hospitals250.500.00.0 to 0.0T3. Admission to and discharge from the hospital2184.010.50.4 to 0.5T4. ≥2 consultants90.200.00.0 to 0.0T5. Surgical procedure551.0610.910.8 to 11.0T6. Emergency department visit1562.9127.77.7 to 7.7T7. >5 medications265548.800.00.0 to 0.0 Drug-related D1. Diphenhydramine2524.6207.97.9 to 8.0 D2. Vitamin K00.00NA D3. Flumazenil00.00NA D4. Antiemetics3726.8154.04.0 to 4.1 D5. Naloxone00.00NA D6. Sodium polystyrene60.100.00.0 to 0.0 D7. Haloperidol00.00NA D8. Antidiarrhoeals571.01322.822.7 to 22.9 D9. Abrupt medication discontinuation370.72567.667.4 to 67.7T8. Physician change20.000.00.0 to 0.0T9. Complaint letter00.00NAT10. >3 calls/hospital visits in 1 week00.00NA Laboratory results L1. Glucose <50 mg/dL00.00NA L2. Serum glucose >110 mg/dL and systemic corticosteroids2374.4114.64.6 to 4.7 L3. Clostridium difficile positive stool culture00.00NA L4. Partial thromboplastin time >100 s00.00NA L5. International normalised ratio >600.00NA L6. WBC count <3000/mm3 and Absolute neutrophil count <1500/mm31783.39352.252.2 to 52.3 L7. Platelet <50 000/mm3581.11017.217.1 to 17.3 L8. Digoxin >2 ng/mL00.00NA L9. Two times or more than baseline BUN/creatinine60.100.00.0 to 0.0 L10. Estimated glomerular filtration rate <35 mL/min/1.73 m22374.410.40.41 to 0.43 L11. K >6.0 mEq/L140.300.00.0 to 0.0 L12. K <2.9 mEq/L20.0150.049.3 to 50.7 L13. Na <130 mEq/L230.400.00.0 to 0.0 L14. HbA1c >6% and glucocorticoid150.3640.039.8 to 40.2T12. Over-sedation/drowsiness/fall20.000.00.0 to 0.0T13. Rash150.3960.059.8 to 60.2T14. Constipation and narcotics/calcium channel blocker140.314100.0100.0 to 100.0T15. Recordings of an adverse event190.3947.447.1 to 47.6The number was counted based on each encounter at outpatient clinics. ‘Complaint letters’ and ‘>3 calls/hospital visits in one week’ were considered positive when they were noted on the medical records.NA, not applicable.AE, adverse event; BUN, blood urea nitrogen; HbA1c, Hemoglobin A1c; PPV, positive predictive value; WBC, white blood cell. Performance of outpatient triggers (n=5439) The number was counted based on each encounter at outpatient clinics. ‘Complaint letters’ and ‘>3 calls/hospital visits in one week’ were considered positive when they were noted on the medical records. NA, not applicable.AE, adverse event; BUN, blood urea nitrogen; HbA1c, Hemoglobin A1c; PPV, positive predictive value; WBC, white blood cell. Adverse event occurrences, healthcare utilisation and patient outcomes over 6 months postdischarge Of the 746 patients, 106 (14.2%) experienced at least one adverse event. There were significant differences in the occurrence of adverse events by patients’ BMI, previous admission history, admission route, clinical department, experiences of surgical procedures, medical diagnoses, CCI, length of stay, more than five medications postdischarge and NEWS2≥2 at discharge (table 1). Among the 5439 outpatient clinic visits, 225 were related to the occurrence of adverse events, after removing duplicate entries from different clinical departments for the same event on the same day. Additionally, there were three cases with two adverse events at one outpatient clinic visit. Thus, 228 adverse events were identified and considered in detailed analyses of adverse events (table 3). Of the 228 adverse events, medication-related events were the most frequent (84.2%). Other events included surgical and medical procedure-related events (13.2%), healthcare-associated infections (2.2%) and transfusion reactions (0.4%). Regarding the subtypes, medication-related abnormal blood counts represented the highest proportion (43.4%). In terms of severity, 92.5% were categorised as ‘1’ and the others were assessed as ‘2’. Most (99.6%) events were judged to be unpreventable (<4). Specifically, the preventability ratings were score 2 for 208 events (91.2%), score 3 for 19 events (8.3%) and score 4 for one event (0.4%). Table 3Types of adverse events detected in outpatient settings (n=228)TypeSubtypeExamplen%Blood transfusion reaction10.4Allergic reactionPruritus10.4Healthcare associated infection52.2Wound—surgical/surgical site infectionWound—surgical; surgical site infection41.8Other—specifyBiliary origin infection with PTBD10.4Medication19284.2Abnormal blood countAbnormal WBC; abnormal platelet; abnormal WBC and abnormal platelet9843.0GastrointestinalConstipation; nausea/vomiting resulting in need for intervention/treatment; diarrhoea resulting in intervention/treatment4720.6Allergic reactionHives; pruritus177.5Metabolic/endocrineHyperglycaemia177.5Others*Sensory or motor disturbance; dizziness, xerostomia; gynaecomastia135.7Surgical/procedural3013.2Other—specifyPTBD obstruction, PCD obstruction2310.1Radiation-related injuryAbnormal WBC; oronasal erosion; nausea/vomiting resulting in need for intervention/treatment41.8Others*Graft-versus-host disease rash after allogeneic stem cell transplantation; wound dehiscence31.3The number was counted based on each encounter at outpatient clinics.*It included the other subcategories.PCD, percutaneous catheter drainage.PTBD, percutaneous transhepatic biliary drainage; WBC, white blood cell. Types of adverse events detected in outpatient settings (n=228) The number was counted based on each encounter at outpatient clinics. It included the other subcategories. PCD, percutaneous catheter drainage.PTBD, percutaneous transhepatic biliary drainage; WBC, white blood cell. Results regarding healthcare utilisation and patient outcomes are shown in table 4. Patients had a median of 5 outpatient visits (IQR 2–10; range 1–37) over 6 months after discharge. Among the 746 patients, 23.2% had at least one emergency room visit and 35.0% were re-hospitalised, including planned admission within 6 months postdischarge. Among the 261 patients with all-cause readmission (IQR 0–1; range 0–11), 43.3% experienced at least one adverse event during their re-hospitalisation. There were significant differences in emergency room visits, number of outpatient clinic visits and number of all-cause re-hospitalisations after index discharge between patients who did and did not experience adverse events (table 4). A total of 28 patients had died at the 6-month follow-up. Table 4Postdischarge healthcare utilisation and patient outcomes (n=746)VariableCategoryn%Adverse eventχ2/ZP valueYes (n=106)No (n=640)n%n%Emergency department visitYes17323.24441.512920.223.28<0.001No57376.86258.551179.8No. of outpatient clinic visitsMedian (IQR)5(2–10)12(7–17)4(2–9)8.84*<0.001No. of hospitalisation048565.04643.443968.637.02<0.001111014.81817.09214.42527.0109.4426.63 or higher9913.33230.26710.5Median (IQR)00–110–300–15.66*<0.001Experience of adverse events during re-hospitalisation†Yes11343.32948.38441.80.810.369No14856.73151.711758.2Patient status at 6 months postdischargeDeath283.876.6213.32.780.101‡Survival71896.39993.461996.7*Wilcoxon rank-sum test; Z includes a continuity correction of 0.5.†n=261.‡Fisher’s exact test. Postdischarge healthcare utilisation and patient outcomes (n=746) Wilcoxon rank-sum test; Z includes a continuity correction of 0.5. n=261. Fisher’s exact test. Characteristics of patients at increased risk of transitional patient safety concerns To identify the characteristics associated with an increased risk of adverse events postdischarge, we conducted multiple logistic regression analysis based on the univariate results (table 1). Given its significant association with the variable of surgical operation, only clinical department type was included in the model. The results demonstrated that BMI, admission route, clinical department, medical diagnosis, NEWS2 at discharge and length of hospital stay were significant factors (Wald χ2=66.46, df=16, p<0.001) (table 5). Specifically, patients who were overweight (23.0–24.9 kg/m2; OR=0.48, 95% CI 0.25 to 0.90), had a circulatory system disease (OR=0.38, 95% CI 0.15 to 0.96) or those with a NEWS2 ≥2 at discharge (OR=0.26, 95% CI 0.12 to 0.54) were less likely to experience adverse events. Those admitted through emergency rooms or day-surgery centres rather than outpatient clinics (OR=1.98, 95% CI 1.18 to 3.32), those discharged from medical departments rather than surgical departments (OR=1.93, 95% CI 1.15 to 3.23), those diagnosed with cancer rather than other diseases (OR=2.03, 95% CI 1.14 to 3.62) and those with longer hospital stays (OR=1.07, 95% CI 1.02 to 1.11) were more likely to experience adverse events. Table 5Multiple logistic regression result for the occurrence of adverse eventsVariableOR95% CIBody mass index <18.50.890.37 to 2.16 23.0 to <25.00.480.25 to 0.90 ≥25.00.770.45 to 1.30 18.5 to <23.0ReferencePrevious admission history (=Yes)1.170.62 to 2.23Admission route (=emergency department or day-surgery centre)1.981.18 to 3.32 Ambulatory clinicsReferenceClinical department (=medical)1.931.15 to 3.23 SurgicalReferenceMajor medical diagnosis Neoplasm2.031.14 to 3.62 Circulatory system diseases0.380.15 to 0.96 Digestive diseases0.480.15 to 1.57 Genitourinary diseases1.080.39 to 3.02 OthersReferenceTaking ≥5 medications at discharge (=Yes)1.040.94 to 1.15NEWS2 ≥2 at discharge (=Yes)0.260.12 to 0.54Charlson Comorbidity Index score1.100.98 to 1.24Length of hospital stay1.071.02 to 1.11This result was adjusted additionally for patients’ age and sex. Hosmer and Lemeshow goodness-of-fit test: χ2=3.04, p=0.932.NEWS2, National Early Warning Score 2. Multiple logistic regression result for the occurrence of adverse events This result was adjusted additionally for patients’ age and sex. Hosmer and Lemeshow goodness-of-fit test: χ2=3.04, p=0.932. NEWS2, National Early Warning Score 2. Discussion To the best of our knowledge, this study is the first to examine the occurrence of adverse events through a follow-up of discharged patients, mainly including patients potentially at risk of adverse events, in the outpatient setting of a Korean hospital. Identifying outpatient triggers has important implications in monitoring adverse events in ambulatory care. Our findings indicated that approximately 14% of postdischarge patients at an outpatient setting experienced at least one adverse event, with medication-related events being the most frequent, followed by surgical/procedural events. Most adverse events were mild in severity and judged as unpreventable. Patients at increased risk of adverse events were characterised by admission type, clinical departments, medical diagnoses and length of hospital stays. Interestingly, patients with a lower NEWS2 at discharge were more likely to experience adverse events. In addition, patients who were overweight and those with circulatory diseases were less likely to experience adverse events. These findings indicate that a substantial number of patients experienced adverse events over 6 months after discharge and that patients at increased risk of adverse events should be appropriately managed to improve patient safety. The adverse event rate in this study was similar to the findings of a patient interview-based study conducted in Germany11 and lower than the rates of some studies.12 17 22 However, this rate was higher than that reported in other studies.8 15 16 These differences may be attributed to the sample characteristics. For instance, a previous study8 targeted patients without recent or planned hospital admissions, whereas we intentionally sampled patients with known high-risk characteristics in terms of medications and vital signs. In addition, our event rate was higher than the rates based on voluntary reporting systems or patient-reported data in primary care settings.13 14 Patients with adverse events had more visits to emergency departments and outpatient clinics and experienced more all-cause re-hospitalisation. This was consistent with the findings of previous studies.27,29 Regarding the types of adverse events, medication-related events were the most common, which was consistent with the findings of previous studies.8 15 16 The most frequent subtypes included abnormal platelets, abnormal WBC counts and gastrointestinal side effects. These may be related to side effects following chemotherapies, as 47.1% of patients in our sample had primary medical diagnosis of neoplasm. Additionally, surgical/procedural complications were frequently related to problems of drainage tube management, such as occlusion, dislocation and accidental extubation. Such situations often required visits to the emergency department. This highlights the need for developing coping strategies tailored to patient safety events in outpatient settings from the patient’s perspective. Interestingly, no events pertaining to the ‘patient care’ category, including falls or pressure ulcers, were identified in our study.40 This might reflect either absence of such events or a potential under-recognition of these events as patient safety problems, possibly leading to a lack of patient reporting in outpatient settings. These findings suggest areas for improvement: patients and caregivers at home must actively monitor for patient safety risks along with medication-related events. It is also noteworthy that patients often discontinued their medication when they perceived possible drug-related side effects (eg, vomiting, dizziness) according to data recorded in the medical records. Along with education on self-management after discharge, patient and caregiver participation in patient safety monitoring should be encouraged to prevent adverse events and avoid worsening conditions. Most adverse events in our study were mild in severity, with no severe harm. This contrasts with the findings of other studies.12 15 This may be attributed to the fact that outpatients were treated on the basis of ambulatory-based care without requiring hospitalisation. Thus, most were classified as having mild severity by definition of temporary harm without hospitalisation. Preventability ratings were consistently below 4, indicating low preventability. This was similar to the finding of a previous study with a rate of 1.5% preventable events (≥4).12 However, it was different from the findings of previous studies, suggesting that 42%–71.3% of adverse events were preventable.15 16 Several factors may contribute to this difference. First, prior studies often including patient interview-based methodologies captured diagnostic errors that were mainly considered to be preventable. In contrast, our study was based on medical records, and diagnostic errors were not identified. Second, the study hospital is a tertiary teaching hospital, which may represent quality care provision, thus resulting in fewer preventable events. Third, we focused on postdischarge adverse events in outpatient settings, primarily among relatively high-risk patients. These patients may be prone to unpreventable, rather than preventable, adverse events. Fourth, we considered adverse events to have occurred even when patient comments on the adverse events were recorded. These cases were generally considered to have low preventability. The risk of adverse events after discharge was associated with the admission route, clinical department, medical diagnosis, length of hospital stay, BMI and NEWS2 at discharge. These findings are consistent with those of previous studies. Specifically, patients admitted from the emergency department or day-surgery centres experienced more adverse events than those admitted from outpatient clinics. The former group may have a higher likelihood of being in critically ill conditions in comparison to the latter group. Such patient conditions might be linked to increased risk of adverse events in outpatient follow-up postdischarge. This result is in line with those of previous studies showing that patients admitted emergently or those with urgent or emergency surgeries experienced increased morbidity and mortality.43 44 Patients discharged from medical departments experienced more postdischarge adverse events than did those from surgical departments, consistent with the findings of a previous study based on patient reports showing that most events are related to certain clinical departments, including general practice and internal medicine.8 Furthermore, patients with cancer had a higher risk of experiencing adverse events than did those with other diagnoses, likely because of their chemotherapy treatments. Patients with circulatory diseases were less likely to experience safety issues. These findings indicate that disease characteristics should be considered in discharge planning to ensure safety during care transition. Patients with longer hospital stays experienced more adverse events after discharge. This is consistent with the findings of previous studies on adverse events experienced by inpatients.5 A longer hospital stay reflects more complex medical conditions, which may contribute to the occurrence of postdischarge adverse events. Interestingly, patients with a BMI of 23.0–24.9 kg/m2 were less likely to experience adverse events, compared with those with a BMI of 18.5–22.9 kg/m2. This result was different from the findings of a previous study, which found that BMI levels >25 kg/m2 in older patients were associated with reduced risks of postdischarge adverse outcomes of unplanned readmission and mortality within 30 days.45 This discrepancy might be due to the utilisation of different adverse outcomes. Meanwhile, another study found that BMI levels ≥25 kg/m2 among middle-aged and older adults were associated with higher frailty.46 Considering that frailty, referring to multidimensional construct of vulnerability,46 47 is tied to physical function, patients with slightly higher BMI levels of 23.0–24.9 kg/m2, which are also regarded as normal levels depending on the used criteria,48 likely have better physical function such as muscle strength outcomes and subsequently experience fewer adverse events than those with a BMI of 18.5–22.9 kg/m2, under condition of a lower risk of frailty. In addition, many patients in our study had a NEWS2 of 0–1 at discharge and were more likely to experience adverse events. This might be related to the fact that the score was based on vital signs at the time of discharge and that most patients were stable when discharged. This suggests that significant changes in patient status may occur after hospital discharge. However, our study did not analyse changes in NEWS2 during subsequent outpatient visits or re-hospitalisations. Future research is required to investigate relationships between NEWS2 changes and adverse events postdischarge. Outpatient triggers showed relatively low performance, except for a few triggers. Among them, constipation with narcotics or calcium channel blockers showed the highest performance. Rash and abrupt medication discontinuation also demonstrated a positive predictive value exceeding 60%. Although taking more than five medications was a frequently observed trigger, it showed very low positive predictive value. Although we used the previously validated triggers, this might indicate that additional factors—such as regular use or duration of medications—as well as different cut-off values for the numerical count of medications (eg, 10 or more) need to be considered.49 Furthermore, it may reflect advances in drug safety practices. Notably, certain triggers such as complaint letters or more than three phone calls/visits within a week were not identified, suggesting that these triggers may require detection through methods other than medical record reviews. These findings were different from those of a previous report.50 For certain triggers, we used additional guides, such as ‘ANC <1500/mm3’ along with ‘WBC count <3000/mm3. Regarding the triggers of ‘serum glucose level >110 mg/dL and the use of systemic corticosteroids’ and ‘glycated haemoglobin >6% and the use of glucocorticoids’, additional information on diabetes status, blood glucose trends, related values and medication changes was considered. This highlights the need for additional clinical considerations when using these triggers in daily clinical practice. Overall, these findings underscore the need for proactive efforts to prevent and reduce potential adverse events after discharge, with an emphasis on patient and caregiver participation in care management. The results suggest that patient and family engagement, information sharing and medication reconciliation should be increased to improve patient safety in primary and outpatient care settings.51 Specifically, in post-chemotherapy follow-up care, there is a need to strengthen education related to medication use, empower patients to respond appropriately, and report any safety problems or injuries that are suspected or occurred in a timely manner. Further research is required to ensure that these efforts are effectively implemented in transitional care settings. In particular, for patient safety problems linked to laboratory results, patients and caregivers should be educated on the importance of ongoing monitoring and encouraged to participate in related laboratory tests. Furthermore, monitoring related symptoms can help mitigate further negative effects. Study limitations This study has some limitations. First, the study population constituted patients discharged from a single tertiary teaching hospital, with a large proportion of patients who were primarily prescribed high-risk medications such as APINCH drugs, those who took multiple medications or had a NEWS2 ≥2 at discharge. We included only patients with at least one follow-up visit at ambulatory clinics of the study hospital within 6 months after discharge, excluding those transferred to other hospitals or healthcare facilities, including nursing homes. Consequently, the patient profile may differ from that of general discharge populations, which limits the generalisability of our findings. Second, there may be limitations linked to the recorded data. Although the study hospital, as a teaching hospital, generally maintains high-quality medical records, documenting information, including patient comments during outpatient visits, can be challenging owing to time constraints. This implies that the adverse events experienced by the patients may have been under-documented. Healthcare provider and patient characteristics can also influence the completeness of medical records and reporting of adverse events. Third, in this retrospective cohort study, not all adverse events experienced by patients were captured, especially when patients received follow-up care at other medical institutions before returning to the study hospital for subsequent treatment. In addition, adverse events were assessed exclusively within a 6-month follow-up period, irrespective of the time to the adverse event or readmission date because these were not the focus of our study. Furthermore, we did not include patient-reported outcomes such as quality of life or continuity of care when assessing patient safety. Despite these limitations, this study contributes to current knowledge as medical record-based research over 6 months, yielding distinct insights based on physicians’ documentation of adverse events and laboratory results. We suggest future studies including patient-reported follow-up conditions over time, along with consideration of time to readmission or the adverse event, to better understand postdischarge adverse event occurrences and patient outcomes. Conclusion Approximately 14% of patients—primarily those potentially at increased risk of adverse events—experienced at least one adverse event within 6 months after discharge, most of which were mild in severity and had low preventability. Patients with at least one adverse event were characterised by their BMI, clinical department, admission route, medical diagnosis, length of hospital stay and NEWS2 at discharge. These factors should be considered in transitional care to improve patient safety after discharge. Our findings indicate that the active engagement of patients and caregivers is essential for discharge planning and transitional care; and to monitor and report the occurrence of safety events. To mitigate further deterioration or harm from patient safety events, we recommend monitoring patients at increased risk of patient safety issues using the outpatient triggers employed in this study. Further studies are also needed to evaluate the sensitivity of outpatient triggers in different outpatient settings beyond hospitals, including primary care clinics. Acknowledgements The authors would like to thank the two nurses and one pharmacist who participated in the reviews. Footnotes Funding: This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Korea government (MSIT) (No. NRF-2022R1A2C1008479). Prepublication history for this paper is available online. To view these files, please visit the journal online (https://doi.org/10.1136/bmjopen-2025-108519). Patient consent for publication: Not applicable. Provenance and peer review: Not commissioned; externally peer reviewed. Patient and public involvement: Patients and/or the public were not involved in the design, or conduct, or reporting, or dissemination plans of this research. Ethics approval: The study was approved by the institutional review board of the Seoul National University Bundang hospital (IRB no: B-2205-758-101). The requirement for informed consent was waived due to the retrospective nature of the study. Data availability statement Data are available upon reasonable request to the authors. References Data Availability Statement Data are available upon reasonable request to the authors.