bs_bs_banner
Journal of Evaluation in Clinical Practice ISSN 1365-2753
Risk factors for unplanned hospital re-admissions: a secondary data analysis of hospital discharge summaries Anja Braet MD,1,2 Caroline Weltens MD PhD,3 Walter Sermeus PhD4 and Arthur Vleugels MD PhD4 1
Doctoral Candidate, Department of Public Health and Primary Care, KU Leuven-University of Leuven, Leuven, Belgium Assistant of Medical Director, Medical Department, az Sint-Blasius, Dendermonde, Belgium 3 Professor, Flemish Hospital Network, University Hospitals Leuven, KU Leuven, Leuven, Belgium 4 Professor, Public Health, KU Leuven-University of Leuven, Leuven, Belgium 2
Keywords care transition, continuity of care, discharge, hospital re-admission, logistic regression, quality of care Correspondence Dr Anja Braet az Sint-Blasius Kroonveldlaan 50 9200 Dendermonde Belgium E-mail: [email protected] Accepted for publication: 9 December 2014 doi:10.1111/jep.12320
Abstract Rationale, aims and objectives To identify patient groups at risk for unplanned hospital re-admissions and risk factors for re-admission. Method We analysed the Belgian Hospital Discharge Dataset including data from 1 130 491 patients discharged in 2008. Patient and hospital factors contributing to re-admission rate were analysed using a multivariable model for logistic regression. Results The overall unplanned re-admission rate was 5.2%. Cardiovascular and pulmonary diagnoses were the most common reasons for re-admission. We found that 10.4% of all re-admissions were due to complications. A high number of previous emergency department (ED) visits proved to be a predictor for re-admission [odds ratio (OR) for patients with at least four ED visits in the past 6 months 4.65; 95% confidence interval (CI) 4.25–5.08]. Patients discharged on Friday (OR 1.05; 95% CI 1.01–1.08) and patients with a long length of stay (OR 1.19; 95% CI 1.15–1.23) also had a higher risk for re-admission. Patients with short lengths of stay were not at risk for re-admission (OR 0.99; 95% CI 0.95–1.02). Conclusions Actions to reduce re-admissions can be targeted to patient groups at risk, and should be aimed at the caring for chronic cardiovascular or pulmonary diseases, preventing complications and multiple ED visits, and ensuring continuity of care after discharge, especially for patients discharged on Friday.
Introduction Unplanned hospital re-admissions occur frequently and are expensive. In 2004, almost one-fifth of US Medicare patients were re-admitted within 30 days of discharge [1]. The cost of these re-admissions was $17.4 billion, out of $102.6 billion in total hospital payments. To reduce hospital re-admissions, national programmes are introduced in many countries. The best known example is the Medicare Hospital Re-admission Reduction Program (HRRP) as part of the Affordable Care Act, penalizing US hospitals with high re-admission rates. In 2013, two-thirds of the US hospitals were affected and $280 million was charged in re-admission penalties [2]. A shorter length of stay might be associated with a higher probability for re-admission because patients tend to be sicker when they leave the hospital, and the time available to prepare patients and caregivers for discharge becomes shorter. However 560
the effect of length of stay on re-admission rates is not yet clear. Some studies have shown that re-admission rates rise with length of stay [3–6]. In other studies, an association is found between short lengths of stay and re-admission rates [7,8] and one study had to conclude that there was no association with length of stay and re-admissions [9]. Severity of illness might be a mediator effect, which is rarely corrected for in these studies. The day of discharge is another important factor that may affect re-admission rates. The risk of a lapse in continuity is assumed to be greater for patients discharged on Friday [10]. Studies searching for associations between re-admission rate and day of discharge show inconclusive results [11,12]. Since discharges on Friday are common, this parameter as it relates to re-admissions will be studied. The research questions of this study are (i) which patient groups are most frequently re-admitted; (ii) what patient characteristics are determining the risk for re-admission; (iii) is length of
Journal of Evaluation in Clinical Practice 21 (2015) 560–566 © 2015 John Wiley & Sons, Ltd.
A. Braet et al.
Risk factors for hospital re-admissions
in-hospital stay associated with re-admission; and (iv) is discharge on Friday associated with a higher risk for re-admission?
identifier, which allows calculating readmissions in the same hospitals, but not between hospitals.
Methods Data selection and definitions Study type and data source We conducted a cross-sectional study using the 2008 Belgian Hospital Discharge Dataset (Be-HDDS) which is similar to international administrative data and includes data for all inpatients in acute hospitals. The 110 acute hospitals consist of general hospitals (80%), seven university hospitals (6%) and general hospitals with a university character (14%). In Belgium, no patient groups are excluded from admission to an acute hospital. The collection of hospital discharge data has been compulsory in Belgium since 1990 for all inpatients in all acute hospitals. The Be-HDDS was commissioned by the Belgian Ministry of Public Health via the Royal Decree of 6 December 1994. The quality of the data is audited by the Ministry of Public Health in two ways. Firstly, a software program checks the data for missings, outliers and inconsistent data. Secondly, by regular hospital audits, a random selection of patient records is reviewed to evaluate the accuracy of the records [13]. The Be-HDDS contains patient demographics, data about the hospital stay (date and type of admission and discharge, referral data, admitting department and destination after discharge) and clinical data [primary and secondary diagnoses as described in the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), diagnostic and therapeutic procedures as described in the ICD-9CM]. The Be-HDDS is used for hospital financing, epidemiology and surveys of national quality. Patient conditions are categorized into 25 Major Diagnosis Categories (MDCs), and patients are further classified into All Patient Refined Diagnostic Related Groups (APR-DRGs) (2008:version 15). APR-DRGs are the subgroups of patients with similar clinical conditions and utilization patterns [14]. Next to APR-DRGs, four classes (minor, moderate, major and extreme) of severity of illness and risk of mortality are calculated. The Be-HDDS does contain a unique hospital patient
We analysed patients re-admitted to the same hospital within 30 days after discharge. The 30-day interval is generally accepted as the optimal balance between a high rate of re-admissions and a low rate of unrelated or ‘false positive’ re-admissions [1,15–17]. A re-admission was classified as ‘unplanned’ when it was coded as an urgent admission in the Be-HDDS. An initial admission was defined as the admission preceding a re-admission. An index admission was defined as any admission that can be followed by a re-admission. By this definition, an admission ending with the patient’s death was not considered an index admission. Patients discharged to another hospital were also excluded from analysis because these discharges are transfers between hospitals and cannot be seen as a patient discharge from hospital. Patients could have more than one index admission and more than one re-admission, but an initial admission could only be followed by one re-admission. We defined the re-admission rate as the number of patients discharged from an acute hospital and urgently re-admitted to the same hospital within 30 days, divided by the number of index admissions. For this study, we sampled all medical and surgical patients >17 years of age discharged in 2008 from all 110 acute general hospitals in Belgium. Two hospitals were excluded from the analysis, because of too small numbers for adult admissions. Outpatients, 1-day clinics and patients staying in the hospital for more than 6 months were excluded. Because of the chronic nature of certain conditions with expected or unavoidable re-admissions, patients with burns (MDC 022), multiple significant trauma (MDC 025), myeloproliferative diseases (MDC 017), HIV (MDC 024), obstetric patients (MDC 014) and psychiatric ward patients were excluded from analysis. The number of hospitals and selected stays for each step in the selection process is presented in Table 1.
Table 1 Different steps in the selection of admissions
Actions 1. Selection of all stays 2. Selection of type of stays
3. Selection of pathology 4. Selection of age 5. Selection of hospitals
6. Selection of stays with no 30-day follow-up 7. Selection of index stays
– Exclusion of outpatient emergency stays – Exclusion of 1-day stays – Exclusion of newborns – Exclusion of psychiatric stays – Exclusion of in-hospital stays of more than 6 months Exclusion of MDCs 14, 17, 22, 24 and 25 Exclusion of patients with birth year > 1990 – Exclusion of non-acute hospitals – Exclusion of one hospital for children – Exclusion of one hospital with < 1000 admissions Exclusion of stays with discharge date > 1 December 2008 – Exclusion of stays ending with patient’s decease – Exclusion of stays with discharge to another hospital
Number of hospitals
Number of admissions
139 139
6 104 474 1 737 985
139 139 110
1 543 113 1 363 876 1 341 337
110 110
1 230 616 1 130 491
MDC, Major Diagnostic Category.
© 2015 John Wiley & Sons, Ltd.
561
Risk factors for hospital re-admissions
Analysis To determine the factors that influence risk for re-admission, we constructed a model consisting of two levels: patient and hospital. The patient variables included gender, age, discharge with or against medical advice, severity of illness, Charlson co-morbidity index, length of stay, previous visits to the emergency department (ED), acuity at admission and discharge destination. For length of stay, we did not use the absolute number of days spent in hospital because this is strongly linked to the reason for admission, severity of illness and age. Instead, we classified each hospital stay as a short, intermediate or long stay by comparing the observed length of stay to the expected length of stay for patients with the same APR-DRG, age category (
Journal of Evaluation in Clinical Practice ISSN 1365-2753
Risk factors for unplanned hospital re-admissions: a secondary data analysis of hospital discharge summaries Anja Braet MD,1,2 Caroline Weltens MD PhD,3 Walter Sermeus PhD4 and Arthur Vleugels MD PhD4 1
Doctoral Candidate, Department of Public Health and Primary Care, KU Leuven-University of Leuven, Leuven, Belgium Assistant of Medical Director, Medical Department, az Sint-Blasius, Dendermonde, Belgium 3 Professor, Flemish Hospital Network, University Hospitals Leuven, KU Leuven, Leuven, Belgium 4 Professor, Public Health, KU Leuven-University of Leuven, Leuven, Belgium 2
Keywords care transition, continuity of care, discharge, hospital re-admission, logistic regression, quality of care Correspondence Dr Anja Braet az Sint-Blasius Kroonveldlaan 50 9200 Dendermonde Belgium E-mail: [email protected] Accepted for publication: 9 December 2014 doi:10.1111/jep.12320
Abstract Rationale, aims and objectives To identify patient groups at risk for unplanned hospital re-admissions and risk factors for re-admission. Method We analysed the Belgian Hospital Discharge Dataset including data from 1 130 491 patients discharged in 2008. Patient and hospital factors contributing to re-admission rate were analysed using a multivariable model for logistic regression. Results The overall unplanned re-admission rate was 5.2%. Cardiovascular and pulmonary diagnoses were the most common reasons for re-admission. We found that 10.4% of all re-admissions were due to complications. A high number of previous emergency department (ED) visits proved to be a predictor for re-admission [odds ratio (OR) for patients with at least four ED visits in the past 6 months 4.65; 95% confidence interval (CI) 4.25–5.08]. Patients discharged on Friday (OR 1.05; 95% CI 1.01–1.08) and patients with a long length of stay (OR 1.19; 95% CI 1.15–1.23) also had a higher risk for re-admission. Patients with short lengths of stay were not at risk for re-admission (OR 0.99; 95% CI 0.95–1.02). Conclusions Actions to reduce re-admissions can be targeted to patient groups at risk, and should be aimed at the caring for chronic cardiovascular or pulmonary diseases, preventing complications and multiple ED visits, and ensuring continuity of care after discharge, especially for patients discharged on Friday.
Introduction Unplanned hospital re-admissions occur frequently and are expensive. In 2004, almost one-fifth of US Medicare patients were re-admitted within 30 days of discharge [1]. The cost of these re-admissions was $17.4 billion, out of $102.6 billion in total hospital payments. To reduce hospital re-admissions, national programmes are introduced in many countries. The best known example is the Medicare Hospital Re-admission Reduction Program (HRRP) as part of the Affordable Care Act, penalizing US hospitals with high re-admission rates. In 2013, two-thirds of the US hospitals were affected and $280 million was charged in re-admission penalties [2]. A shorter length of stay might be associated with a higher probability for re-admission because patients tend to be sicker when they leave the hospital, and the time available to prepare patients and caregivers for discharge becomes shorter. However 560
the effect of length of stay on re-admission rates is not yet clear. Some studies have shown that re-admission rates rise with length of stay [3–6]. In other studies, an association is found between short lengths of stay and re-admission rates [7,8] and one study had to conclude that there was no association with length of stay and re-admissions [9]. Severity of illness might be a mediator effect, which is rarely corrected for in these studies. The day of discharge is another important factor that may affect re-admission rates. The risk of a lapse in continuity is assumed to be greater for patients discharged on Friday [10]. Studies searching for associations between re-admission rate and day of discharge show inconclusive results [11,12]. Since discharges on Friday are common, this parameter as it relates to re-admissions will be studied. The research questions of this study are (i) which patient groups are most frequently re-admitted; (ii) what patient characteristics are determining the risk for re-admission; (iii) is length of
Journal of Evaluation in Clinical Practice 21 (2015) 560–566 © 2015 John Wiley & Sons, Ltd.
A. Braet et al.
Risk factors for hospital re-admissions
in-hospital stay associated with re-admission; and (iv) is discharge on Friday associated with a higher risk for re-admission?
identifier, which allows calculating readmissions in the same hospitals, but not between hospitals.
Methods Data selection and definitions Study type and data source We conducted a cross-sectional study using the 2008 Belgian Hospital Discharge Dataset (Be-HDDS) which is similar to international administrative data and includes data for all inpatients in acute hospitals. The 110 acute hospitals consist of general hospitals (80%), seven university hospitals (6%) and general hospitals with a university character (14%). In Belgium, no patient groups are excluded from admission to an acute hospital. The collection of hospital discharge data has been compulsory in Belgium since 1990 for all inpatients in all acute hospitals. The Be-HDDS was commissioned by the Belgian Ministry of Public Health via the Royal Decree of 6 December 1994. The quality of the data is audited by the Ministry of Public Health in two ways. Firstly, a software program checks the data for missings, outliers and inconsistent data. Secondly, by regular hospital audits, a random selection of patient records is reviewed to evaluate the accuracy of the records [13]. The Be-HDDS contains patient demographics, data about the hospital stay (date and type of admission and discharge, referral data, admitting department and destination after discharge) and clinical data [primary and secondary diagnoses as described in the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM), diagnostic and therapeutic procedures as described in the ICD-9CM]. The Be-HDDS is used for hospital financing, epidemiology and surveys of national quality. Patient conditions are categorized into 25 Major Diagnosis Categories (MDCs), and patients are further classified into All Patient Refined Diagnostic Related Groups (APR-DRGs) (2008:version 15). APR-DRGs are the subgroups of patients with similar clinical conditions and utilization patterns [14]. Next to APR-DRGs, four classes (minor, moderate, major and extreme) of severity of illness and risk of mortality are calculated. The Be-HDDS does contain a unique hospital patient
We analysed patients re-admitted to the same hospital within 30 days after discharge. The 30-day interval is generally accepted as the optimal balance between a high rate of re-admissions and a low rate of unrelated or ‘false positive’ re-admissions [1,15–17]. A re-admission was classified as ‘unplanned’ when it was coded as an urgent admission in the Be-HDDS. An initial admission was defined as the admission preceding a re-admission. An index admission was defined as any admission that can be followed by a re-admission. By this definition, an admission ending with the patient’s death was not considered an index admission. Patients discharged to another hospital were also excluded from analysis because these discharges are transfers between hospitals and cannot be seen as a patient discharge from hospital. Patients could have more than one index admission and more than one re-admission, but an initial admission could only be followed by one re-admission. We defined the re-admission rate as the number of patients discharged from an acute hospital and urgently re-admitted to the same hospital within 30 days, divided by the number of index admissions. For this study, we sampled all medical and surgical patients >17 years of age discharged in 2008 from all 110 acute general hospitals in Belgium. Two hospitals were excluded from the analysis, because of too small numbers for adult admissions. Outpatients, 1-day clinics and patients staying in the hospital for more than 6 months were excluded. Because of the chronic nature of certain conditions with expected or unavoidable re-admissions, patients with burns (MDC 022), multiple significant trauma (MDC 025), myeloproliferative diseases (MDC 017), HIV (MDC 024), obstetric patients (MDC 014) and psychiatric ward patients were excluded from analysis. The number of hospitals and selected stays for each step in the selection process is presented in Table 1.
Table 1 Different steps in the selection of admissions
Actions 1. Selection of all stays 2. Selection of type of stays
3. Selection of pathology 4. Selection of age 5. Selection of hospitals
6. Selection of stays with no 30-day follow-up 7. Selection of index stays
– Exclusion of outpatient emergency stays – Exclusion of 1-day stays – Exclusion of newborns – Exclusion of psychiatric stays – Exclusion of in-hospital stays of more than 6 months Exclusion of MDCs 14, 17, 22, 24 and 25 Exclusion of patients with birth year > 1990 – Exclusion of non-acute hospitals – Exclusion of one hospital for children – Exclusion of one hospital with < 1000 admissions Exclusion of stays with discharge date > 1 December 2008 – Exclusion of stays ending with patient’s decease – Exclusion of stays with discharge to another hospital
Number of hospitals
Number of admissions
139 139
6 104 474 1 737 985
139 139 110
1 543 113 1 363 876 1 341 337
110 110
1 230 616 1 130 491
MDC, Major Diagnostic Category.
© 2015 John Wiley & Sons, Ltd.
561
Risk factors for hospital re-admissions
Analysis To determine the factors that influence risk for re-admission, we constructed a model consisting of two levels: patient and hospital. The patient variables included gender, age, discharge with or against medical advice, severity of illness, Charlson co-morbidity index, length of stay, previous visits to the emergency department (ED), acuity at admission and discharge destination. For length of stay, we did not use the absolute number of days spent in hospital because this is strongly linked to the reason for admission, severity of illness and age. Instead, we classified each hospital stay as a short, intermediate or long stay by comparing the observed length of stay to the expected length of stay for patients with the same APR-DRG, age category (
