Drug and Alcohol Dependence 137 (2014) 55–61
Contents lists available at ScienceDirect
Drug and Alcohol Dependence journal homepage: www.elsevier.com/locate/drugalcdep
Alcohol use disorders among surgical patients: Unplanned 30-days readmissions, length of hospital stay, excessive costs and mortality Miguel Gili-Miner a,b,∗ , Luis Béjar-Prado b , Enrique Gili-Ortiz c , Gloria Ramírez-Ramírez a,b , Julio López-Méndez a,b , José-Manuel López-Millán c , Brett Sharp d a Unidad de Gestión Clínica de Medicina Preventiva, Vigilancia y Promoción de la Salud, Hospital Universitario Virgen Macarena, Av. Dr. Fedriani s/n, 41070 Seville, Spain b Departamento de Medicina Preventiva y Salud Pública, Universidad de Sevilla, Av. Sánchez Pizjuán s/n, 41007 Seville, Spain c Unidad de Gestión Clínica de Anestesiología y Bloque Quirúrgico, Hospital Universitario Virgen Macarena, Av. Dr. Fedriani s/n, 41070 Seville, Spain d Unidad de Gestión Clínica de Diagnóstico por la Imagen, Hospital Universitario Virgen de Rocío, Av. Manuel Siurot s/n, 41013 Seville, Spain
a r t i c l e
i n f o
Article history: Received 16 June 2013 Received in revised form 7 January 2014 Accepted 9 January 2014 Available online 30 January 2014 Keywords: Alcohol use disorders Unplanned hospital readmissions Mortality Prolonged hospital stay Excessive costs
a b s t r a c t Aims: Alcohol use disorders (AUD) have been associated with an increased risk of unplanned hospital readmissions (URA). We analyzed in a sample of 87 Spanish Hospitals if surgical patients with AUD had a higher risk of URA and if among patients with URA, those with AUD had an excess length of hospital stay, higher hospital expenses and increased risk of mortality. Method: We analyzed data of patients who underwent surgical operations during the period between 2008 and 2010. URA was defined as unplanned readmissions during the first 30 days after hospital departure. The primary outcome was risk of URA in patients with AUD. Secondary outcomes were mortality, excess length of stay and over expenditure. Results: A total of 2,076,958 patients who underwent surgical operations were identified: 68,135 (3.3%) had AUD, and 62,045 (3.0%) had at least one URA. Among patients with AUD 4212 (6.2%) had at least one URA and among patients without AUD 57,833 (2.9%) had at least one URA. Multivariable analysis demonstrated that AUD was an independent predictor of developing URA (Odds ratio: 1.56; 95% CI: 1.50–1.62). Among surgical patients with URA, those with AUD had longer lengths of hospital stay (2.9 days longer), higher hospital costs (2885.8 Euros or 3858.3 US Dollars), higher risk of death (OR: 2.16, 95% CI: 1.92–2.44) and higher attributable mortality (11.2%). Conclusions: Among surgical patients, AUD increase the risk of URA, and among patients with URA, AUD heighten the risk of in-hospital death, and cause longer hospital stays and over expenditures. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Readmissions have emerged internationally as an important outcome measure, as they occur frequently (all-cause readmissions range from 13% to 20%) and are very costly events (Kossovsky et al., 1999; Friedman and Basu, 2004; Jencks et al., 2009). The hospital readmission rate has been proposed as an outcome indicator of quality of health care computable from routine statistics because a link between poor quality of care and subsequent readmission seems intuitively reasonable, and because readmission rates are easy to monitor through administrative hospital databases
∗ Corresponding author at: Unidad de Gestión Clínica de Medicina Preventiva, Vigilancia y Promoción de la Salud, Hospital Universitario Virgen Macarena, Av. Dr. Fedriani s/n, 41070 Seville, Spain. Tel.: +34 955923386; fax: +34 955008930. E-mail address: [email protected] (M. Gili-Miner). 0376-8716/$ – see front matter © 2014 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.drugalcdep.2014.01.009
(Henderson et al., 1989; Milne and Clarke, 1990). Numerous studies have shown that premature discharge or substandard care during initial hospitalization increases the risk of readmission (Ashton et al., 1995; Wei et al., 1995). Hospital readmissions are considered a marker of general complications and potential substandard care. Reducing rates of rehospitalization has attracted attention from policymakers as a way to improve quality of care and reduce costs. Payment incentives to avoid readmissions have been cited in the Department of Health and Human Services’ strategic plan for 2010 through 2015 as an example of quality of care improvement (DHHS, 2012). Furthermore, several studies have shown the impact of patient related factors, such as case mix, severity of disease, comorbidities and chronic conditions on the risk of readmission (Henderson et al., 1989; Kossovsky et al., 1999; Librero et al., 1999; Halfon et al., 2002). Alcohol use disorders (AUD) are a risk factor for numerous comorbidities and may influence the risk of hospital admission and
56
M. Gili-Miner et al. / Drug and Alcohol Dependence 137 (2014) 55–61
subsequent readmissions, for both surgical and medical conditions. AUD are a risk factor for intensive-care readmission in patients undergoing elective vascular and thoracic surgical procedures (Maxson et al., 1999) and AUD are a risk factor for post-discharge surgical infections (Daneman et al., 2010). AUD affects the course of diabetes, leading to increased morbidity, hospital admissions and readmissions, and mortality (Engler et al., 2013). AUD are a risk factor for hospital discharge against medical advice, and patients discharged against medical advice have higher readmission rates and have higher in-hospital mortality (Southern et al., 2012). AUD have been recognized as a significant variable in predictive multivariate models of hospital unplanned readmissions (Billings et al., 2006; Howell et al., 2009). A recent systematic review on readmission prediction suggests that patient characteristics are strongly associated with readmission risk, but the quality of inpatient care is also an important predictor (Kansagara et al., 2011), and any analysis of AUD as a risk factor for unplanned readmissions (URA) should take into account hospital-level factors such as size and complexity of hospitals. With the increasing focus on containing health care costs, in the United States and elsewhere, a better understanding of factors contributing to the risk of URA is important. However, there is a notable lack of information in epidemiological literature about the impact of AUD in the risk of URA, in the risk of mortality among cases with URA and in the attributable mortality among patients with AUD and URA, if AUD influences the length of hospital stay among cases of URA and the magnitude of over expenditures caused by AUD among URA cases. To assess the impact of AUD on the risk of unplanned hospital readmissions among surgical patients we used a large administrative database. We sought to determine the incidence of URA in patients with AUD and without AUD, and the impact of AUD among patients with readmissions on length of hospital stay, excessive costs and in-hospital mortality, after controlling other individual and hospital-level factors. We analyzed these events in patients who underwent surgical operations in a sample of 87 Spanish Hospitals. 2. Method 2.1. Source of data Data for hospital admissions and readmissions were captured by the administrative minimal basic data set (MBDS) of 87 Spanish Hospitals during the period of 2008–2010. From written or digitalized information that is provided by the hospital physician who signs the clinical record, each patient’s diagnosis, external causes and procedures are codified according to International Classification of Diseases, 9th review (ICD-9-CM) codes. Codification and data entry in the electronic database are performed by dedicated administrative personnel who have completed in-depth training on medical data registration. This administrative database has demographic data, admission and discharge dates, surgical operation date, type of admission and type of discharge, diagnostic codes for principal cause and secondary diagnoses, external causes and procedures using ICD-9-CM codes. Diagnosis-related groups (DRG) and a classification of hospital groups based on size and complexity are included in this national database (Spanish Ministry of Health, 2012).
Table 1 International classification of diseases codes (9th review) used to identify patients with alcohol use disorders. Variables
ICD-9 codes
Alcohol use disorders Alcohol dependence Alcohol abuse Alcohol-induced mental disorders Alcoholic polyneuropathy Alcoholic cardiomyopathy Alcoholic gastritis Alcoholic liver diseases Excessive blood alcohol level Alcohol toxicity Alcohol poisoning
303.00–303.93 305.00–305.03 291.0–291.9 357.5 425.5 535.30–535.31 571.0–571.3 790.3 980.0–980.9 E860.0–E860.9
pulmonary circulation disorders, valvular disease, deficiency anemia, blood loss anemia, fluid and electrolyte disorders, weight loss, coagulopathy, myocardial infarction, congestive heart failure, peripheral vascular disorders, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatic disease, peptic ulcer disease, mild liver disease, diabetes without chronic complications, diabetes with chronic complications, hemiplegia or paraplegia, renal disease, moderate or severe liver disease, drugs use disorders, cancer, leukemia or lymphoma, metastatic cancer and AIDS. We used ICD-9 codes proposed for these comorbidities (Quan et al., 2005). We followed the classification of the Spanish Ministry of Health, which divides hospitals into five categories ranging from lower size and complexity to higher size and complexity, included in the MBDS (Spanish Ministry of Health, 2012), as depicted in Table 2. 2.3. Data analysis The primary outcome of interest was to compute the risk of unplanned readmissions (URA) in the first 30 days after hospital discharge in patients with AUD. Secondary analyses were used to determine risk of mortality and attributable mortality, length of hospital stay and costs in patients with and without AUD who had URA. Spanish Ministry Of Health calculates yearly the cost for every diagnosis related group (DRG) in a sample of Spanish Hospitals. These hospital costs include 12 items: nursing, standard daily care, structure, medical care, critical care, surgery, pharmacy, radiology, laboratory, medical supplies, therapies and clinical services. These costs are calculated for every DRG adjusted to the type of hospital, because costs of DRGs are higher in more complex hospitals. We applied these estimated national costs for every DRG and type of hospital in our multivariate estimation of costs (Spanish Ministry of Health, 2012). Charlson Comorbidity Index was calculated for every patient (Charlson et al., 1987). Univariate analysis examining association between AUD and URA, age, gender, tobacco dependence, type of hospital and comorbidity were performed (t test and Chi-square tests, or their equivalent non-parametric tests). Cohen’s d was calculated for continuous variables (age and Charlson Index) and because of large sample size
Table 2 Spanish Ministry of Health Classification of Hospitals according to complexity and size. Group
Characteristics
1
Small hospitals with an average of fewer than 150 beds, with nearly no high technology equipment, limited means, and low complexity in terms of care Basic general hospitals, mean size fewer than 200 beds, minimum of high technology equipment, certain teaching activity, and somewhat greater complexity Area hospitals, with a mean size of around 500 beds. More than 50 medical residents and an average of 269 physicians. Intermediate complexity (1.5 complex services and case-mix, 1.01) Group of large hospitals, but more heterogeneous in terms of equipment, size and activity. Highly intense teaching activity (more than 160 medical residents) and high complexity (an average of 4 complex services and case-mix >1.2). In this group, 81% of the hospitals have fewer than 1.000 beds Hospitals of great importance in the structural context and intense activity. Complete range of services. More than 680 physicians and around 300 medical residents. Includes the large hospital complexes. A hospital may consist of a single center or two or more that are organized and integrated into the hospital complex. The latter is identified by its unified administration and management. Thus, a hospital complex can consist of two or more hospitals, which can even be far apart, and one or several specialty centers. In this group, 88% of the hospitals have more than 1000 beds.
2
3 2.1.1. Inclusion and exclusion criteria. We included all patients aged 18 years or older who received a primary surgical procedure code and who underwent in-patient surgery. Patients who were transferred to another hospital were excluded.
4
2.2. Variables 2.2.1. Definition of alcohol use disorders. AUD were defined as alcohol dependence, alcohol abuse, alcohol-induced mental disorders, alcoholic polyneuropathy, alcoholic cardiomyopathy, alcoholic gastritis, alcoholic liver disorders, excessive blood alcohol level, alcohol toxicity and alcohol poisoning. Table 1 details the ICD codes used to diagnose AUD. We used principal diagnostic code and secondary diagnostic codes (13 fields). 2.2.2. Other variables. Other variables collected included age, gender, tobacco dependence (ICD-9 code 305.1), hospital group, and 29 comorbidities: obesity, uncomplicated hypertension, complicated hypertension, cardiac arrhythmias,
5
M. Gili-Miner et al. / Drug and Alcohol Dependence 137 (2014) 55–61 we considered results significant if p < 10−6 . Subsequently, multivariable models using unconditional logistic regression analysis were computed to determine the effects of AUD on development of URA and on the risk of in-hospital death among patients with URA. Multivariate analysis of covariance was performed to determine the effect of AUD on attributable mortality, length of hospital stay (LOS) and costs in patients with URA. Data were adjusted for age, gender, tobacco dependence, group of hospitals and severity of disease using comorbidities. Analysis was performed in STATA version MP 12.1.
3. Results 3.1. Patient characteristics Patient characteristics with AUD and without AUD are detailed in Table 3. A total of 2,076,958 patients who underwent surgical operations were identified: 68,135 (3.3%) had AUD, and 62,045 (3.0%) had at least one URA. Among patients with AUD 4212 (6.2%) had at least one URA and among patients without AUD 57,833 (2.9%) had at least one URA. Older age, male gender and tobacco dependence were associated with AUD. Patients with AUD were more commonly admitted to larger, more complex regional hospitals. Higher prevalence of comorbidity at the time of hospital admission was the norm among patients with AUD, with the exception of blood loss anemia, dementia and rheumatic disease. When comorbidity was measured by Charlson Comorbidity Index, it was associated with AUD. 3.2. Risk of unplanned readmissions Characteristics of patients with URA and without URA are shown in Table 4. AUD and tobacco dependence were associated with higher prevalence of URA. Higher prevalence of comorbidity at the time of hospital admission was standard among patients with URA, excepting for obesity, uncomplicated hypertension and peripheral vascular disorders. As a consequence of this, Charlson Comorbidity Index was associated with a higher prevalence of URA. Patients with AUD were significantly more likely to develop URA by multivariable analysis adjusted for age, gender, hospital group, tobacco dependence and comorbidities (OR: 1.56, 95% CI: 1.50–1.62). Size and complexity of hospitals were associated with a lower risk of URA (OR: 0.96, 95% CI: 0.95–0.96). 3.3. Mortality Crude in hospital mortality rate was 9.7% for those cases who simultaneously were admitted with AUD and URA. For those admitted with URA but without AUD in hospital mortality rate was 6.2%. The respective in hospital mortality rates for those with AUD but without URA and those without URA and without AUD were 4.1% and 1.8%. Multivariate analysis demonstrated that patients with AUD and URA were at increased risk of in hospital death. We found a significant association between AUD and death in patients with URA adjusting for age, gender, tobacco dependence and comorbidities (Odds ratio: 2.16, 95% CI: 1.92–2.44). We used multivariate analysis of covariance to calculate adjusted mortality rates and adjusted attributable mortality among these patients controlling for age, sex, group of hospital and comorbidities. Patients with both AUD and URA had an adjusted mortality rate of 5.40% (95% CI: 5.25–5.54, p < 0.0001) and patients with URA but without AUD had an adjusted mortality rate of 4.85% (95% CI: 4.75–4.96, p < 0.0001). The respective figures for patients with AUD but without URA were 2.40% (95% CI: 2.29–2.50, p < 0.0001) and for patients without AUD neither URA were 1.85% (95% CI: 1.83–1.87, p < 0.0001).
57
Adjusted attributable mortality of AUD among patients with URA compared to those patients with URA but without AUD was 11.2% (95% CI: 10.52–11.77, p < 0.0001). 3.4. Length of stay and costs AUD influenced LOS both in patients with URA and without URA. Multivariable model shows a mean LOS of 16.6 days for patients with AUD and URA and a mean LOS of 13.7 days for those with URA but without AUD. For patients with AUD but without URA mean LOS was 10.9 days and for patients without AUD neither URA mean LOS was 8.0 days. AUD influenced also over expenditures. Multivariable model found a mean cost of hospital stay of 16,476.6 Euros (22,029.2 US Dollars) for patients with AUD and URA, a mean cost of 13,590.7 Euros (18,170.8 US Dollars) for patients with URA but without AUD, a mean cost of 11,529.3 Euros (15,414.7 US Dollars) for patients with AUD but without URA, and a mean cost of 8643.5 Euros (11,556.4 US Dollars) for patients without AUD neither URA. Results of multivariable analysis including age, gender, hospital group, tobacco dependence and comorbidities for all hospitals and for every type of hospital are exposed in Table 5. Results demonstrated that among patients who had URA, patients with AUD had a mean 2.86-days longer LOS. Mean excess costs for stay among patients with AUD who had URA were 2885.8 Euros (3858.3 US Dollars). LOS and over expenditures were higher in hospitals of Group 5 (more complex hospitals). Hospitals of group 1 (less complex) ranked second in LOS but ranked fourth in over expenditures. 4. Discussion There is an abundance of evidence within the literature to support that comorbidities increase the risk of surgical site infections and other adverse events (Kassin et al., 2012; Sachdev and Napolitano, 2012) and it has been shown that AUD is a risk factor of surgical site infections, increased LOS and over-expenditures among surgical patients (de Wit et al., 2011, 2012). In some studies these postoperative complications increase risk of readmission by a factor of 4 (Kassin et al., 2012). Several groups have shown that patients with AUD have a higher risk of having URA. AUD has been identified as a risk factor for intensive-care readmissions in patients undergoing some elective surgical procedures (Maxson et al., 1999). Chronic alcoholics have up to five times more cardiac complications in the postoperative period, and AUD increases the risk for URA among patients with heart failure (Spies et al., 2001). Billings et al. (2006) found that alcohol related diagnoses formed one of the 21 variables that were significant predictors of URA during a period of twelve months. Howell et al. (2009) studied 17,699 URA during a period of twelve months and also found that alcohol was a risk factor in the predictive algorithm for URA (OR: 1.4; 95% CI: 1.2–1.7). In this study analyzing a national database, we evaluated more than 2 million patients. Our data demonstrate that AUD have a significant impact on the appearance of URA among surgical patients independently of age, gender, tobacco dependence, type of hospital and comorbidity. Considering the risk of URA in the multivariable model, surgical patients with AUD had a 56% excess of risk of having URA when compared to surgical patients without AUD. Among surgical patients with URA, those with AUD had longer LOS, higher hospital costs, increased risk of death and higher attributable mortality. As far as we know, this is the first report about the impact of AUD in the attributable mortality, prolonged LOS and overexpenditures among URA patients. The results of this study show that AUD have a notable impact on the risk of URA, the risk of in hospital mortality, the use of medical resources and hospital costs.
58
M. Gili-Miner et al. / Drug and Alcohol Dependence 137 (2014) 55–61
Table 3 Characteristics of surgical patients with AUD and without AUD included in the study.
Unplanned readmissions before 31 days a Age in years, mean (95% CI) Gender Female Male Tobacco dependence Type of hospital 1 2 3 4 5 Comorbidity Obesity Hypertension, uncomplicated Hypertension, complicated Cardiac arrhythmias Pulmonary circulation disorders Valvular disease Deficiency anemia Blood loss anemia Fluid and electrolyte disorders Weight loss Coagulopathy Myocardial infarction Congestive heart failure Peripheral vascular disorders Cerebrovascular disease Dementia Chronic pulmonary disease Rheumatic disease Peptic ulcer disease Mild liver disease Diabetes without chronic complications Diabetes with chronic complications Hemiplegia or paraplegia Renal disease Moderate or severe liver disease Drug use disorder Cancer, leukemia or lymphoma Metastatic cancer AIDS b Charlson Comorbidity Index, mean (95% CI)
With AUD (n = 68,135)
Without AUD (n = 2,008,823)
n
%
n
%
p
4212 58.7 (58.6–58.8)
6.2
57,833 55.7 (55.6–55.7)
2.9
Contents lists available at ScienceDirect
Drug and Alcohol Dependence journal homepage: www.elsevier.com/locate/drugalcdep
Alcohol use disorders among surgical patients: Unplanned 30-days readmissions, length of hospital stay, excessive costs and mortality Miguel Gili-Miner a,b,∗ , Luis Béjar-Prado b , Enrique Gili-Ortiz c , Gloria Ramírez-Ramírez a,b , Julio López-Méndez a,b , José-Manuel López-Millán c , Brett Sharp d a Unidad de Gestión Clínica de Medicina Preventiva, Vigilancia y Promoción de la Salud, Hospital Universitario Virgen Macarena, Av. Dr. Fedriani s/n, 41070 Seville, Spain b Departamento de Medicina Preventiva y Salud Pública, Universidad de Sevilla, Av. Sánchez Pizjuán s/n, 41007 Seville, Spain c Unidad de Gestión Clínica de Anestesiología y Bloque Quirúrgico, Hospital Universitario Virgen Macarena, Av. Dr. Fedriani s/n, 41070 Seville, Spain d Unidad de Gestión Clínica de Diagnóstico por la Imagen, Hospital Universitario Virgen de Rocío, Av. Manuel Siurot s/n, 41013 Seville, Spain
a r t i c l e
i n f o
Article history: Received 16 June 2013 Received in revised form 7 January 2014 Accepted 9 January 2014 Available online 30 January 2014 Keywords: Alcohol use disorders Unplanned hospital readmissions Mortality Prolonged hospital stay Excessive costs
a b s t r a c t Aims: Alcohol use disorders (AUD) have been associated with an increased risk of unplanned hospital readmissions (URA). We analyzed in a sample of 87 Spanish Hospitals if surgical patients with AUD had a higher risk of URA and if among patients with URA, those with AUD had an excess length of hospital stay, higher hospital expenses and increased risk of mortality. Method: We analyzed data of patients who underwent surgical operations during the period between 2008 and 2010. URA was defined as unplanned readmissions during the first 30 days after hospital departure. The primary outcome was risk of URA in patients with AUD. Secondary outcomes were mortality, excess length of stay and over expenditure. Results: A total of 2,076,958 patients who underwent surgical operations were identified: 68,135 (3.3%) had AUD, and 62,045 (3.0%) had at least one URA. Among patients with AUD 4212 (6.2%) had at least one URA and among patients without AUD 57,833 (2.9%) had at least one URA. Multivariable analysis demonstrated that AUD was an independent predictor of developing URA (Odds ratio: 1.56; 95% CI: 1.50–1.62). Among surgical patients with URA, those with AUD had longer lengths of hospital stay (2.9 days longer), higher hospital costs (2885.8 Euros or 3858.3 US Dollars), higher risk of death (OR: 2.16, 95% CI: 1.92–2.44) and higher attributable mortality (11.2%). Conclusions: Among surgical patients, AUD increase the risk of URA, and among patients with URA, AUD heighten the risk of in-hospital death, and cause longer hospital stays and over expenditures. © 2014 Elsevier Ireland Ltd. All rights reserved.
1. Introduction Readmissions have emerged internationally as an important outcome measure, as they occur frequently (all-cause readmissions range from 13% to 20%) and are very costly events (Kossovsky et al., 1999; Friedman and Basu, 2004; Jencks et al., 2009). The hospital readmission rate has been proposed as an outcome indicator of quality of health care computable from routine statistics because a link between poor quality of care and subsequent readmission seems intuitively reasonable, and because readmission rates are easy to monitor through administrative hospital databases
∗ Corresponding author at: Unidad de Gestión Clínica de Medicina Preventiva, Vigilancia y Promoción de la Salud, Hospital Universitario Virgen Macarena, Av. Dr. Fedriani s/n, 41070 Seville, Spain. Tel.: +34 955923386; fax: +34 955008930. E-mail address: [email protected] (M. Gili-Miner). 0376-8716/$ – see front matter © 2014 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.drugalcdep.2014.01.009
(Henderson et al., 1989; Milne and Clarke, 1990). Numerous studies have shown that premature discharge or substandard care during initial hospitalization increases the risk of readmission (Ashton et al., 1995; Wei et al., 1995). Hospital readmissions are considered a marker of general complications and potential substandard care. Reducing rates of rehospitalization has attracted attention from policymakers as a way to improve quality of care and reduce costs. Payment incentives to avoid readmissions have been cited in the Department of Health and Human Services’ strategic plan for 2010 through 2015 as an example of quality of care improvement (DHHS, 2012). Furthermore, several studies have shown the impact of patient related factors, such as case mix, severity of disease, comorbidities and chronic conditions on the risk of readmission (Henderson et al., 1989; Kossovsky et al., 1999; Librero et al., 1999; Halfon et al., 2002). Alcohol use disorders (AUD) are a risk factor for numerous comorbidities and may influence the risk of hospital admission and
56
M. Gili-Miner et al. / Drug and Alcohol Dependence 137 (2014) 55–61
subsequent readmissions, for both surgical and medical conditions. AUD are a risk factor for intensive-care readmission in patients undergoing elective vascular and thoracic surgical procedures (Maxson et al., 1999) and AUD are a risk factor for post-discharge surgical infections (Daneman et al., 2010). AUD affects the course of diabetes, leading to increased morbidity, hospital admissions and readmissions, and mortality (Engler et al., 2013). AUD are a risk factor for hospital discharge against medical advice, and patients discharged against medical advice have higher readmission rates and have higher in-hospital mortality (Southern et al., 2012). AUD have been recognized as a significant variable in predictive multivariate models of hospital unplanned readmissions (Billings et al., 2006; Howell et al., 2009). A recent systematic review on readmission prediction suggests that patient characteristics are strongly associated with readmission risk, but the quality of inpatient care is also an important predictor (Kansagara et al., 2011), and any analysis of AUD as a risk factor for unplanned readmissions (URA) should take into account hospital-level factors such as size and complexity of hospitals. With the increasing focus on containing health care costs, in the United States and elsewhere, a better understanding of factors contributing to the risk of URA is important. However, there is a notable lack of information in epidemiological literature about the impact of AUD in the risk of URA, in the risk of mortality among cases with URA and in the attributable mortality among patients with AUD and URA, if AUD influences the length of hospital stay among cases of URA and the magnitude of over expenditures caused by AUD among URA cases. To assess the impact of AUD on the risk of unplanned hospital readmissions among surgical patients we used a large administrative database. We sought to determine the incidence of URA in patients with AUD and without AUD, and the impact of AUD among patients with readmissions on length of hospital stay, excessive costs and in-hospital mortality, after controlling other individual and hospital-level factors. We analyzed these events in patients who underwent surgical operations in a sample of 87 Spanish Hospitals. 2. Method 2.1. Source of data Data for hospital admissions and readmissions were captured by the administrative minimal basic data set (MBDS) of 87 Spanish Hospitals during the period of 2008–2010. From written or digitalized information that is provided by the hospital physician who signs the clinical record, each patient’s diagnosis, external causes and procedures are codified according to International Classification of Diseases, 9th review (ICD-9-CM) codes. Codification and data entry in the electronic database are performed by dedicated administrative personnel who have completed in-depth training on medical data registration. This administrative database has demographic data, admission and discharge dates, surgical operation date, type of admission and type of discharge, diagnostic codes for principal cause and secondary diagnoses, external causes and procedures using ICD-9-CM codes. Diagnosis-related groups (DRG) and a classification of hospital groups based on size and complexity are included in this national database (Spanish Ministry of Health, 2012).
Table 1 International classification of diseases codes (9th review) used to identify patients with alcohol use disorders. Variables
ICD-9 codes
Alcohol use disorders Alcohol dependence Alcohol abuse Alcohol-induced mental disorders Alcoholic polyneuropathy Alcoholic cardiomyopathy Alcoholic gastritis Alcoholic liver diseases Excessive blood alcohol level Alcohol toxicity Alcohol poisoning
303.00–303.93 305.00–305.03 291.0–291.9 357.5 425.5 535.30–535.31 571.0–571.3 790.3 980.0–980.9 E860.0–E860.9
pulmonary circulation disorders, valvular disease, deficiency anemia, blood loss anemia, fluid and electrolyte disorders, weight loss, coagulopathy, myocardial infarction, congestive heart failure, peripheral vascular disorders, cerebrovascular disease, dementia, chronic pulmonary disease, rheumatic disease, peptic ulcer disease, mild liver disease, diabetes without chronic complications, diabetes with chronic complications, hemiplegia or paraplegia, renal disease, moderate or severe liver disease, drugs use disorders, cancer, leukemia or lymphoma, metastatic cancer and AIDS. We used ICD-9 codes proposed for these comorbidities (Quan et al., 2005). We followed the classification of the Spanish Ministry of Health, which divides hospitals into five categories ranging from lower size and complexity to higher size and complexity, included in the MBDS (Spanish Ministry of Health, 2012), as depicted in Table 2. 2.3. Data analysis The primary outcome of interest was to compute the risk of unplanned readmissions (URA) in the first 30 days after hospital discharge in patients with AUD. Secondary analyses were used to determine risk of mortality and attributable mortality, length of hospital stay and costs in patients with and without AUD who had URA. Spanish Ministry Of Health calculates yearly the cost for every diagnosis related group (DRG) in a sample of Spanish Hospitals. These hospital costs include 12 items: nursing, standard daily care, structure, medical care, critical care, surgery, pharmacy, radiology, laboratory, medical supplies, therapies and clinical services. These costs are calculated for every DRG adjusted to the type of hospital, because costs of DRGs are higher in more complex hospitals. We applied these estimated national costs for every DRG and type of hospital in our multivariate estimation of costs (Spanish Ministry of Health, 2012). Charlson Comorbidity Index was calculated for every patient (Charlson et al., 1987). Univariate analysis examining association between AUD and URA, age, gender, tobacco dependence, type of hospital and comorbidity were performed (t test and Chi-square tests, or their equivalent non-parametric tests). Cohen’s d was calculated for continuous variables (age and Charlson Index) and because of large sample size
Table 2 Spanish Ministry of Health Classification of Hospitals according to complexity and size. Group
Characteristics
1
Small hospitals with an average of fewer than 150 beds, with nearly no high technology equipment, limited means, and low complexity in terms of care Basic general hospitals, mean size fewer than 200 beds, minimum of high technology equipment, certain teaching activity, and somewhat greater complexity Area hospitals, with a mean size of around 500 beds. More than 50 medical residents and an average of 269 physicians. Intermediate complexity (1.5 complex services and case-mix, 1.01) Group of large hospitals, but more heterogeneous in terms of equipment, size and activity. Highly intense teaching activity (more than 160 medical residents) and high complexity (an average of 4 complex services and case-mix >1.2). In this group, 81% of the hospitals have fewer than 1.000 beds Hospitals of great importance in the structural context and intense activity. Complete range of services. More than 680 physicians and around 300 medical residents. Includes the large hospital complexes. A hospital may consist of a single center or two or more that are organized and integrated into the hospital complex. The latter is identified by its unified administration and management. Thus, a hospital complex can consist of two or more hospitals, which can even be far apart, and one or several specialty centers. In this group, 88% of the hospitals have more than 1000 beds.
2
3 2.1.1. Inclusion and exclusion criteria. We included all patients aged 18 years or older who received a primary surgical procedure code and who underwent in-patient surgery. Patients who were transferred to another hospital were excluded.
4
2.2. Variables 2.2.1. Definition of alcohol use disorders. AUD were defined as alcohol dependence, alcohol abuse, alcohol-induced mental disorders, alcoholic polyneuropathy, alcoholic cardiomyopathy, alcoholic gastritis, alcoholic liver disorders, excessive blood alcohol level, alcohol toxicity and alcohol poisoning. Table 1 details the ICD codes used to diagnose AUD. We used principal diagnostic code and secondary diagnostic codes (13 fields). 2.2.2. Other variables. Other variables collected included age, gender, tobacco dependence (ICD-9 code 305.1), hospital group, and 29 comorbidities: obesity, uncomplicated hypertension, complicated hypertension, cardiac arrhythmias,
5
M. Gili-Miner et al. / Drug and Alcohol Dependence 137 (2014) 55–61 we considered results significant if p < 10−6 . Subsequently, multivariable models using unconditional logistic regression analysis were computed to determine the effects of AUD on development of URA and on the risk of in-hospital death among patients with URA. Multivariate analysis of covariance was performed to determine the effect of AUD on attributable mortality, length of hospital stay (LOS) and costs in patients with URA. Data were adjusted for age, gender, tobacco dependence, group of hospitals and severity of disease using comorbidities. Analysis was performed in STATA version MP 12.1.
3. Results 3.1. Patient characteristics Patient characteristics with AUD and without AUD are detailed in Table 3. A total of 2,076,958 patients who underwent surgical operations were identified: 68,135 (3.3%) had AUD, and 62,045 (3.0%) had at least one URA. Among patients with AUD 4212 (6.2%) had at least one URA and among patients without AUD 57,833 (2.9%) had at least one URA. Older age, male gender and tobacco dependence were associated with AUD. Patients with AUD were more commonly admitted to larger, more complex regional hospitals. Higher prevalence of comorbidity at the time of hospital admission was the norm among patients with AUD, with the exception of blood loss anemia, dementia and rheumatic disease. When comorbidity was measured by Charlson Comorbidity Index, it was associated with AUD. 3.2. Risk of unplanned readmissions Characteristics of patients with URA and without URA are shown in Table 4. AUD and tobacco dependence were associated with higher prevalence of URA. Higher prevalence of comorbidity at the time of hospital admission was standard among patients with URA, excepting for obesity, uncomplicated hypertension and peripheral vascular disorders. As a consequence of this, Charlson Comorbidity Index was associated with a higher prevalence of URA. Patients with AUD were significantly more likely to develop URA by multivariable analysis adjusted for age, gender, hospital group, tobacco dependence and comorbidities (OR: 1.56, 95% CI: 1.50–1.62). Size and complexity of hospitals were associated with a lower risk of URA (OR: 0.96, 95% CI: 0.95–0.96). 3.3. Mortality Crude in hospital mortality rate was 9.7% for those cases who simultaneously were admitted with AUD and URA. For those admitted with URA but without AUD in hospital mortality rate was 6.2%. The respective in hospital mortality rates for those with AUD but without URA and those without URA and without AUD were 4.1% and 1.8%. Multivariate analysis demonstrated that patients with AUD and URA were at increased risk of in hospital death. We found a significant association between AUD and death in patients with URA adjusting for age, gender, tobacco dependence and comorbidities (Odds ratio: 2.16, 95% CI: 1.92–2.44). We used multivariate analysis of covariance to calculate adjusted mortality rates and adjusted attributable mortality among these patients controlling for age, sex, group of hospital and comorbidities. Patients with both AUD and URA had an adjusted mortality rate of 5.40% (95% CI: 5.25–5.54, p < 0.0001) and patients with URA but without AUD had an adjusted mortality rate of 4.85% (95% CI: 4.75–4.96, p < 0.0001). The respective figures for patients with AUD but without URA were 2.40% (95% CI: 2.29–2.50, p < 0.0001) and for patients without AUD neither URA were 1.85% (95% CI: 1.83–1.87, p < 0.0001).
57
Adjusted attributable mortality of AUD among patients with URA compared to those patients with URA but without AUD was 11.2% (95% CI: 10.52–11.77, p < 0.0001). 3.4. Length of stay and costs AUD influenced LOS both in patients with URA and without URA. Multivariable model shows a mean LOS of 16.6 days for patients with AUD and URA and a mean LOS of 13.7 days for those with URA but without AUD. For patients with AUD but without URA mean LOS was 10.9 days and for patients without AUD neither URA mean LOS was 8.0 days. AUD influenced also over expenditures. Multivariable model found a mean cost of hospital stay of 16,476.6 Euros (22,029.2 US Dollars) for patients with AUD and URA, a mean cost of 13,590.7 Euros (18,170.8 US Dollars) for patients with URA but without AUD, a mean cost of 11,529.3 Euros (15,414.7 US Dollars) for patients with AUD but without URA, and a mean cost of 8643.5 Euros (11,556.4 US Dollars) for patients without AUD neither URA. Results of multivariable analysis including age, gender, hospital group, tobacco dependence and comorbidities for all hospitals and for every type of hospital are exposed in Table 5. Results demonstrated that among patients who had URA, patients with AUD had a mean 2.86-days longer LOS. Mean excess costs for stay among patients with AUD who had URA were 2885.8 Euros (3858.3 US Dollars). LOS and over expenditures were higher in hospitals of Group 5 (more complex hospitals). Hospitals of group 1 (less complex) ranked second in LOS but ranked fourth in over expenditures. 4. Discussion There is an abundance of evidence within the literature to support that comorbidities increase the risk of surgical site infections and other adverse events (Kassin et al., 2012; Sachdev and Napolitano, 2012) and it has been shown that AUD is a risk factor of surgical site infections, increased LOS and over-expenditures among surgical patients (de Wit et al., 2011, 2012). In some studies these postoperative complications increase risk of readmission by a factor of 4 (Kassin et al., 2012). Several groups have shown that patients with AUD have a higher risk of having URA. AUD has been identified as a risk factor for intensive-care readmissions in patients undergoing some elective surgical procedures (Maxson et al., 1999). Chronic alcoholics have up to five times more cardiac complications in the postoperative period, and AUD increases the risk for URA among patients with heart failure (Spies et al., 2001). Billings et al. (2006) found that alcohol related diagnoses formed one of the 21 variables that were significant predictors of URA during a period of twelve months. Howell et al. (2009) studied 17,699 URA during a period of twelve months and also found that alcohol was a risk factor in the predictive algorithm for URA (OR: 1.4; 95% CI: 1.2–1.7). In this study analyzing a national database, we evaluated more than 2 million patients. Our data demonstrate that AUD have a significant impact on the appearance of URA among surgical patients independently of age, gender, tobacco dependence, type of hospital and comorbidity. Considering the risk of URA in the multivariable model, surgical patients with AUD had a 56% excess of risk of having URA when compared to surgical patients without AUD. Among surgical patients with URA, those with AUD had longer LOS, higher hospital costs, increased risk of death and higher attributable mortality. As far as we know, this is the first report about the impact of AUD in the attributable mortality, prolonged LOS and overexpenditures among URA patients. The results of this study show that AUD have a notable impact on the risk of URA, the risk of in hospital mortality, the use of medical resources and hospital costs.
58
M. Gili-Miner et al. / Drug and Alcohol Dependence 137 (2014) 55–61
Table 3 Characteristics of surgical patients with AUD and without AUD included in the study.
Unplanned readmissions before 31 days a Age in years, mean (95% CI) Gender Female Male Tobacco dependence Type of hospital 1 2 3 4 5 Comorbidity Obesity Hypertension, uncomplicated Hypertension, complicated Cardiac arrhythmias Pulmonary circulation disorders Valvular disease Deficiency anemia Blood loss anemia Fluid and electrolyte disorders Weight loss Coagulopathy Myocardial infarction Congestive heart failure Peripheral vascular disorders Cerebrovascular disease Dementia Chronic pulmonary disease Rheumatic disease Peptic ulcer disease Mild liver disease Diabetes without chronic complications Diabetes with chronic complications Hemiplegia or paraplegia Renal disease Moderate or severe liver disease Drug use disorder Cancer, leukemia or lymphoma Metastatic cancer AIDS b Charlson Comorbidity Index, mean (95% CI)
With AUD (n = 68,135)
Without AUD (n = 2,008,823)
n
%
n
%
p
4212 58.7 (58.6–58.8)
6.2
57,833 55.7 (55.6–55.7)
2.9
