Validation of coding algorithms for the identification of patients hospitalized for alcoholic hepatitis using administrative data

The Association Between Homelessness and Key Liver-Related Outcomes in Veterans With Cirrhosis

INTRODUCTION

Homelessness adversely affects patient outcomes in broad cohort studies; however, its impact on key liver-related outcomes in patients with cirrhosis is understudied. We aimed to address this knowledge gap using data from the Veterans Health Administration, a cohort disproportionately affected by homelessness.

METHODS

This was a retrospective cohort study of the Veterans Health Administration patients with incident cirrhosis diagnosis between January 2008 and February 2022. Homeless status was classified at baseline and as time-updating variable during follow-up. Inverse probability treatment weighted Cox regression was performed to evaluate the association between homelessness and outcomes of all-cause mortality, cirrhosis decompensation, and hepatocellular carcinoma.

RESULTS

A total of 117,698 patients were included in the cohort, of whom 14,243 (12.1%) were homeless at baseline. In inverse probability treatment weighted Cox regression, homelessness was associated with a 24% higher hazard of all-cause mortality (hazard ratio [HR] 1.24, 95% confidence interval [CI] 1.22-1.26, P < 0.001). However, in competing risk regression models, homelessness was associated with a reduced subhazard of decompensation (subhazard ratio 0.86, 95% CI 0.84-0.88, P < 0.001) and hepatocellular carcinoma (subhazard ratio 0.86, 95% CI 0.83-0.89, P < 0.001). In cause-specific mortality analysis, homeless patients had significantly increased non-liver-related and liver-related mortality; however, the magnitude of effect size was greater for non-liver-related mortality (csHR 1.38, 95% CI 1.35-1.40, P < 0.001).

DISCUSSION

Homelessness in veterans with cirrhosis is associated with increased all-cause mortality; however, this is likely mediated primarily through non-liver-related factors. Future studies are needed to explore drivers of mortality and improve mitigation strategies in these patients.

Emergency services utilization by patients with alcohol-associated hepatitis: An analysis of national trends

BACKGROUND

Hospitalization and mortality in patients with alcohol-associated hepatitis (AH), a severe form of liver disease, continue to increase over time. Given the severity of the illness, most hospitalized patients with AH are admitted from the emergency department (ED). However, there are no data on ED utilization by patients with AH. Thus, the Nationwide Emergency Department Sample (NEDS) dataset was analyzed to determine the ED utilization for AH.

METHODS

Temporal trends (2016-2019) and outcomes of ED visits for AH were determined. Primary or secondary AH diagnoses were based on coding priority. Numbers of patients evaluated in the ED, severity of disease, complications of liver disease, and discharge disposition were analyzed. Crude and adjusted rates were examined, and temporal trends evaluated using logistic regression with orthogonal polynomial contrasts for each year.

RESULTS

There were 466,014,370 ED visits during 2016-2019, of which 448,984 (0.096%) were for AH, 85.0% of which required hospitalization. The rate of visits for AH (primary and secondary) between 2016 and 2019 increased from 85 to 106.8/100,000 ED visits. The rate of secondary AH increased more than the rate of primary AH (from 68.6 to 86.5 vs. from 16.4 to 20.3/100,000 ED visits). Patients aged 45-64 years had the highest rate of ED visits for AH, which decreased during the study period, while the rate of ED visits for AH increased in those aged 25-44 years (from 38.5% to 42.9%). The severity of disease (ascites, hepatic encephalopathy, and acute kidney injury) also increased over time. Medicaid and private insurance were the most common payors for patients seeking care in the ED for AH.

CONCLUSIONS

Temporal trends show an overall increase in ED utilization rates for AH, more patients requiring hospitalization, and an increase in the proportion of younger patients presenting to the ED with AH.

Incidence, clinical characteristics, and risk factors associated with recurrent alcohol-associated hepatitis

BACKGROUND

Alcohol relapse occurs frequently in alcohol-associated hepatitis (AH) survivors, but data on the frequency and course of recurrent alcohol-associated hepatitis (rAH) are sparse. We investigated the incidence, risk factors, and outcomes of rAH.

METHODS

Hospitalized patients with AH from 2010 to 2020 at a large health care system were followed until death/liver transplant, last follow-up, or end of study (December 31, 2021). AH was defined by NIAAA Alcoholic Hepatitis Consortium criteria; rAH was defined a priori as a discrete AH episode >6 months from index AH hospitalization with interim >50% improvement or normalization of total bilirubin. Multivariable competing risk analysis was performed to identify factors associated with rAH. Landmark Kaplan-Meier analysis was performed to compare survival between patients who did versus those who did not develop rAH.

RESULTS

Of 1504 hospitalized patients with AH, 1317 (87.6%) survived and were analyzed. During a 3055 person-year follow-up, 116 (8.8%) developed rAH at an annual incidence rate of 3.8% (95% CI: 2.8-4.8). On multivariable competing risk analysis, marital status [sub-HR 0.54 (95% CI: 0.34, 0.92), p=0.01] and medications for alcohol use disorder [sub-HR 0.56 (95% CI: 0.34, 0.91), p=0.02] were associated with a lower risk for rAH. On landmark Kaplan-Meier analysis, the cumulative proportion surviving at 1 year (75% vs. 90%) and 3 years (50% vs. 78%) was significantly lower in patients who developed rAH compared to those who did not develop rAH (log-rank p<0.001).

CONCLUSIONS

rAH develops in ~1 in 10 AH survivors and is associated with lower long-term survival. Medications for alcohol use disorder lower the risk for rAH and, therefore, could be a key preventative strategy to improve outcomes.

Lower survival and higher rates of cirrhosis in patients with ROUX-EN-Y gastric bypass hospitalised with alcohol-associated hepatitis

OBJECTIVE

The incidence of alcohol-associated liver disease (ALD) is increasing, and weight loss surgery is more common due to the obesity epidemic. Roux-en-Y gastric bypass (RYGB) is associated with alcohol use disorder and ALD; however, its impact on outcomes in patients hospitalised for alcohol-associated hepatitis (AH) is unclear.

DESIGN

We performed a single-centre, retrospective study of patients with AH from June 2011 to December 2019. Primary exposure was the presence of RYGB. The primary outcome was inpatient mortality. Secondary outcomes included overall mortality, readmissions and cirrhosis progression.

RESULTS

2634 patients with AH met the inclusion criteria; 153 patients had RYGB. Median age of the entire cohort was 47.3 years; median Model for End Stage Liver Disease - Sodium (MELD-Na) was 15.1 in the study group versus 10.9 in the control group. There was no difference in inpatient mortality between the two groups. On logistic regression, increased age, elevated body mass index, MELD-Na >20 and haemodialysis were all associated with higher inpatient mortality. RYGB status was associated with increased 30-day readmission (20.3% vs 11.7%, p<0.01), development of cirrhosis (37.5% vs 20.9%, p<0.01) and overall mortality (31.4% vs 24%, p=0.03).

CONCLUSIONS

Patients with RYGB have higher rates of readmissions, cirrhosis and overall mortality after discharge from hospital for AH. Allocation of additional resources on discharge may improve clinical outcomes and reduce healthcare expenditure in this unique patient population.

The Liverpool alcohol-related liver disease algorithm identifies twice as many emergency admissions compared to standard methods when applied to Hospital Episode Statistics for England

BACKGROUND

Emergency admissions in England for alcohol-related liver disease (ArLD) have increased steadily for decades. Statistics based on administrative data typically focus on the ArLD-specific code as the primary diagnosis and are therefore at risk of excluding ArLD admissions defined by other coding combinations.

AIM

To deploy the Liverpool ArLD Algorithm (LAA), which accounts for alternative coding patterns (e.g., ArLD secondary diagnosis with alcohol/liver-related primary diagnosis), to national and local datasets in the context of studying trends in ArLD admissions before and during the COVID-19 pandemic.

METHODS

We applied the standard approach and LAA to Hospital Episode Statistics for England (2013-21). The algorithm was also deployed at 28 hospitals to discharge coding for emergency admissions during a common 7-day period in 2019 and 2020, in which eligible patient records were reviewed manually to verify the diagnosis and extract data.

RESULTS

Nationally, LAA identified approximately 100% more monthly emergency admissions from 2013 to 2021 than the standard method. The annual number of ArLD-specific admissions increased by 30.4%. Of 39,667 admissions in 2020/21, only 19,949 were identified with standard approach, an estimated admission cost of £70 million in under-recorded cases. Within 28 local hospital datasets, 233 admissions were identified using the standard approach and a further 250 locally verified cases using the LAA (107% uplift). There was an 18% absolute increase in ArLD admissions in the seven-day evaluation period in 2020 versus 2019. There were no differences in disease severity or mortality, or in the proportion of admissions with decompensation of cirrhosis or alcoholic hepatitis.

CONCLUSIONS

The LAA can be applied successfully to local and national datasets. It consistently identifies approximately 100% more cases than the standard coding approach. The algorithm has revealed the true extent of ArLD admissions. The pandemic has compounded a long-term rise in ArLD admissions and mortality.

Concordance of ICD-10 Codes and the Clinical Diagnosis of Alcoholic Hepatitis

INTRODUCTION

Many studies on alcoholic hepatitis (AH) use the International Classification of Disease (ICD) coding to identify patients. Data regarding the diagnostic accuracy of ICD codes for AH are limited.

METHODS

A total of 151 patients with ICD-10 codes for AH were reviewed for the presence or absence of AH using standardized diagnostic criteria.

RESULTS

Sixty-eight of the 151 patients met AH criteria, corresponding to a positive predictive value of 45%. Patients with AH experienced higher model for end-stage liver disease and mortality than those who did not ( P < 0.05).

DISCUSSION

Our results suggest ICD-10 codes are not reliable for identifying AH. Studies using the ICD codes should be interpreted cautiously.

Algorithms to Identify Alcoholic Hepatitis Hospitalizations in Patients with Cirrhosis

BACKGROUND

Alcoholic hepatitis (AH) is a clinically diagnosed syndrome with high short-term mortality for which liver transplantation may be curative. A lack of validated algorithms to identify AH hospitalizations has hindered clinical epidemiology research.

METHODS

This was a retrospective cohort study of patients with cirrhosis using Veterans Health Administration (VHA) data from 2008 to 2015. We randomly sampled hospitalizations based upon abnormal liver tests and administrative codes for acute hepatitis or alcohol-associated liver disease (ALD). Hospitalizations were manually adjudicated for AH per society guidelines. A priori algorithms were evaluated to compute positive predicted value (PPV) and positive likelihood ratio (LR+), and were tested in an external University of Pennsylvania Health System (UPHS) cohort.

RESULTS

Of 368 hospitalizations, 142 (38.6%) were adjudicated as AH. AH patients were younger (55 vs. 58 years, p < 0.001), less likely to have prior cirrhosis decompensation (57% vs. 73.9%, p < 0.001), and had higher AST-to-ALT ratios (median 2.9 vs. 1.9 mg/dL, p < 0.001) and higher bilirubin levels (median 2.9 vs. 1.9 mg/dL, p < 0.001). Algorithms combining clinical laboratory criteria (AST > 85 U/L but < 450 U/L, AST-to-ALT ratio > 2, total bilirubin > 5 mg/dL) and administrative coding criteria yielded the highest PPV (96.4%, 95% CI 87.7-99.6) and the highest LR+ (43.0, 95% CI 10.6-173.5). Several algorithms demonstrated 100% PPV for definite AH in the UPHS external cohort.

CONCLUSION

We have identified algorithms for AH hospitalizations with excellent PPV and LR+. These high-specificity algorithms may be used in VHA datasets to identify patients with high likelihood of AH, but should not be used to study AH incidence.

Impact of the COVID-19 Pandemic on Hospitalizations for Alcoholic Hepatitis or Cirrhosis in Alberta, Canada

BACKGROUND & AIMS

Coronavirus disease 2019 (COVID-19) pandemic lockdown and restrictions had significant disruption to patient care. We aimed to evaluate the impact of COVID-19 restrictions on hospitalizations of patients with alcoholic and nonalcoholic cirrhosis as well as alcoholic hepatitis (AH) in Alberta, Canada.

METHODS

We used validated International Classification of Diseases (ICD-9 and ICD-10) coding algorithms to identify liver-related hospitalizations for nonalcoholic cirrhosis, alcoholic cirrhosis, and AH in the province of Alberta between March 2018 and September 2020. We used the provincial inpatient discharge and laboratory databases to identify our cohorts. We used elevated alanine aminotransferase or aspartate aminotransferase, elevated international normalized ratio, or bilirubin to identify AH patients. We compared COVID-19 restrictions (April-September 2020) with prior study periods. Joinpoint regression was used to evaluate the temporal trends among the 3 cohorts.

RESULTS

We identified 2916 hospitalizations for nonalcoholic cirrhosis, 2318 hospitalizations for alcoholic cirrhosis, and 1408 AH hospitalizations during our study time. The in-hospital mortality rate was stable in relation to the pandemic for alcoholic cirrhosis and AH. However, nonalcoholic cirrhosis patients had lower in-hospital mortality rate after March 2020 (8.5% vs 11.5%; P = .033). There was a significant increase in average monthly admissions in the AH cohort (22.1/10,000 admissions during the pandemic vs 11.6/10,000 admissions before March 2020; P < .001).

CONCLUSIONS

Before and during COVID-19 monthly admission rates were stable for nonalcoholic and alcoholic cirrhosis; however, there was a significant increase in AH admissions. Because alcohol sales surged during the pandemic, future impact on alcoholic liver disease could be detrimental.

Is There a Futility Discriminant Function Score for Alcoholic Hepatitis?

The treatment for patients with alcoholic hepatitis (AH) who have a discriminant function (DF) score greater than 32 has been steroids. A prior study reported that mortality approaches 100% when the DF score is greater than 54, despite the use of prednisolone. Our aim was to determine if a DF score greater than 54 is associated with high mortality despite steroids. We conducted a retrospective study of 531 first-time inpatient encounters of AH. We compared 30-day mortality between patients with DF 54 or less to those greater than 54 treated with steroids, as well as a matched group not treated with steroids. A total of 531 inpatients diagnosed with AH were identified, of which 124 had a DF greater than 32 and 52 were treated with steroids. Among patients treated with steroids, 30-day mortality for patients with DF greater than 54 (n = 27) and 54 or below (n = 25) was 22% and 12%, respectively (p = 0.47). Among patients with DF greater than 54, the 30-day mortality for those who did (n = 27) and did not (n = 29) receive steroids was 22% and 24%, respectively (p = 0.87). In our study population, steroids were not futile in patients with a DF score of greater than 54.

The Changing Burden of Alcoholic Hepatitis: Rising Incidence and Associations with Age, Gender, Race, and Geography

Alcoholic hepatitis is a major cause of morbidity and mortality. However, there are limited population-based data on its incidence, demographics, and temporal trends. We performed a retrospective cohort study using the State Inpatient Databases from Florida, Massachusetts, New York, and Washington from 2010 to 2014. We included patients aged 20-79 years admitted with alcoholic hepatitis and calculated incidence using population denominators obtained from the Centers for Disease Control and Prevention WONDER database. We fit multivariable Poisson regression models to explore interactions between alcoholic hepatitis incidence rates and several predictors including state, calendar year, age, race/ethnicity, and gender. Among 56,973 unique individuals with alcoholic hepatitis, the majority were male (39,702; 69.7%) and white non-Hispanic (40,934; 72.0%). In multivariable Poisson models, there was a significant interaction between calendar year and age group (p < 0.001), with the highest incidence rates in those ages 40-49 and 50-59 across all years. The absolute increase in incidence rate across calendar years was highest in the 20-29 and 30-39 age groups in every state. Female gender was associated with a lower rate (incidence rate ratio (IRR) 0.42, 95% confidence interval (CI) 0.41-0.42, p < 0.001). Compared to white non-Hispanics, black non-Hispanics (IRR 0.79, CI 0.77-0.81, p < 0.001) and Hispanics (0.66, CI 0.65-0.68, p < 0.001) had lower incidence rates. The incidence of alcoholic hepatitis in the USA varies by age, gender, race/ethnicity, and state of residence. The group with the fastest rising incidence is those aged 20-39. More work is needed to evaluate the reasons for the temporal trends for admissions for alcoholic hepatitis.

Reducing variation in hospital mortality for alcohol-related liver disease in North West England

BACKGROUND

Variations in emergency care quality for alcohol-related liver disease (ARLD) have been highlighted.

AIM

To determine whether introduction of a regional quality improvement (QI) programme was associated with a reduction in potentially avoidable inpatient mortality.

METHOD

Retrospective observational cohort study using hospital administrative data spanning a 1-year period before (2014/2015) and 3 years after a QI initiative at seven acute hospitals in North West England. The intervention included serial audit of a bundle of process metrics. An algorithm was developed to identify index ("first") emergency admissions for ARLD (n = 3887). We created a standardised mortality ratio (SMR) to compare relative mortality and regression models to examine risk-adjusted odds of death.

RESULTS

In 2014/2015, three of seven hospitals had an SMR above the upper control limit ("outliers"). Adjusted odds of death for patients admitted to outlier hospitals was higher than non-outliers (OR 2.13, 95% CI 1.32-3.44, P = 0.002). Following the QI programme there was a step-wise reduction in outliers (none in 2017/2018). Odds of death was 67% lower in 2017/2018 compared to 2014/2015 at original outlier hospitals, but unchanged at other hospitals. Process audit performance of outliers was worse than non-outliers at baseline, but improved after intervention.

CONCLUSIONS

There was a reduction in unexplained variation in hospital mortality following the QI intervention. This challenges the pessimism that is prevalent for achieving better outcomes for patients with ARLD. Notwithstanding the limitations of an uncontrolled observational study, these data provide hope that co-ordinated efforts to drive adoption of evidence-based practice can save lives.

Readmission Rates and Associated Outcomes for Alcoholic Hepatitis: A Nationwide Cohort Study

BACKGROUND/AIMS

Alcoholic hepatitis (AH) can lead to sudden and severe hepatic decompensation necessitating recurrent hospitalizations. We evaluated the trends, predictors, and healthcare cost burden of AH-related readmissions in the USA.

METHODS

Utilizing the National Readmissions Database 2010-2014, we performed a retrospective longitudinal analysis to identify the index readmission with AH for up to 90 days after discharge. Annual trends of 30- and 90-day AH-related readmissions were calculated. Predictors of 30- and 90-day readmission were determined by multivariate logistic regression. Annual healthcare cost burden associated with AH-linked readmissions was estimated.

RESULTS

Of the 21,572 (unweighted: 50,769) AH-related hospitalizations, 4917 (22.8%) and 7890 (36.6%) were readmitted in 30 and 90 day, respectively, with rates that were statistically unchanged from 2010 to 2014. Predictors of 30-day readmissions included female gender, hepatitis C virus infection, cirrhosis, ascites, acute kidney injury, urinary tract infection, history of bariatric surgery, chronic pancreatitis, and high medical comorbidity index. Acute pancreatitis and palliative care consultation were associated with a lower risk of 30-day readmission. Predictors of 90-day readmission were similar to risk factors for 30-day readmission. From 2010 to 2014, the annual cost (and total hospitalization days) burden increased in 2014 to $164 million (22,244 days) and $321 million (42,772 days) for 30- and 90-day AH-related readmissions, respectively.

CONCLUSION

Despite relatively stable trends in AH-related readmission, the total LOS and cost has been rising. A target-directed approach with a focus on high-risk subpopulations may help understand the unique challenges associated with the rising cost of AH-related readmissions.

Direct Health Care Costs, Health Services Utilization, and Outcomes of Biliary Atresia: A Population-based Cohort Study

OBJECTIVES

Biliary atresia (BA) is the most common reason for liver transplant in childhood, and outcomes worsen with older age at hepatoportoenterostomy (HPE). We determined direct health care costs in children with BA, compared to controls in a population-based cohort of children in Ontario, Canada.

METHODS

We used health administrative data to identify all children diagnosed with BA between 2002 and 2016 (n = 121) and matched controls (n = 602). We determined annual direct healthcare costs, and rates of health services utilization, liver transplantation, death, portal hypertension, cirrhosis, esophageal varices, and major upper gastrointestinal bleeding requiring hospitalization. Multivariable regression models determined the association between age at HPE, risk of liver transplant, and direct costs.

RESULTS

Incidence of BA was 6.07 (4.99-7.15) per 100,000 live births. The annual median (interquartile range) direct health care costs were higher in BA cases ($4210; interquartile range $1091-$16,765) compared to controls ($283; $112-$634). Compared to age at HPE <45 days, there was no significant association between direct costs and HPE ≥90 days (rate ratio 1.24, 95% confidence interval [CI] 0.78-1.97) or 45 to 90 days (rate ratio 1.05, 95% CI 0.73-1.50). Age at HPE ≥90 days was significantly associated with risk of undergoing liver transplant compared to age <45 days (hazard ratio 5.27, 95% CI 2.45-11.34). Direct costs were higher in patients with BA who underwent liver transplantation compared to those who did not ($39,476±$84,367 vs $22,579 ± $67,913).

CONCLUSIONS

Direct ealth care costs were high in patients with BA, especially in those who underwent liver transplantation. Age at HPE was associated with risk of liver transplantation, but not direct health care costs, utilization, or other risk outcomes.

The North American Consortium for the Study of End-Stage Liver Disease-Acute-on-Chronic Liver Failure Score Accurately Predicts Survival: An External Validation Using a National Cohort

Acute-on-chronic liver failure (ACLF) carries high short-term mortality. The North American Consortium for the Study of End-Stage Liver Disease (NACSELD)-ACLF score, positive if ≥2 organ failures are present, is a bedside tool that predicts short-term mortality in patients with cirrhosis. However, it was created using major liver referral centers, where a minority of patients with cirrhosis are hospitalized. Therefore, this study used the Nationwide Inpatient Sample, a nationally representative database, from 2005 to 2014 to externally validate the NACSELD-ACLF score in a cohort of patients with decompensated cirrhosis who were identified by a validated algorithm. Organ failures were identified using diagnosis codes. The primary objective was to evaluate the association between the NACSELD-ACLF score and inpatient mortality, whereas secondary objectives compared outcomes depending on presence of infection or hospitalization at a transplant center. Multivariate logistic regression was used to compare outcomes, and area under the curve was calculated. There were 1,523,478 discharges that were included with 106,634 (7.0%) having a positive NACSELD-ACLF score. Patients were a mean 58 years old, and a majority were white men. Infection was present in 33.7% of the sample. Inpatient survival decreased with each organ failure and if infection was present. Patients with the NACSELD-ACLF score had significantly lower inpatient survival on crude (94% versus 48%; P < 0.001) and multivariate analysis (odds ratio [OR], 0.08; 95% confidence interval [CI], 0.07-0.08) and area under the receiver operating characteristic curve 0.77 (95% CI, 0.77-0.78). Liver transplant centers had clinically similar but significantly better survival at each organ failure, in patients with the NACSELD-ACLF score, and on multivariate analysis (OR, 1.17; 95% CI, 1.13-1.22). Using a national cohort, our study validated the NACSELD-ACLF score as an excellent, simple bedside tool to predict short-term survival in patients with decompensated cirrhosis.

The high burden of alcoholic cirrhosis in privately insured persons in the United States

Alcoholic cirrhosis (AC) is a major cause of liver-related morbidity and mortality in the United States. Rising rates of alcohol use disorders in the United States will likely result in more alcoholic liver disease. Our aim was to determine the prevalence, health care use, and costs of AC among privately insured persons in the United States. We collected data from persons aged 18-64 with AC (identified by codes from the International Classification of Diseases, Ninth and Tenth Revisions) enrolled in the Truven MarketScan Commercial Claims and Encounters database (2009-2015). We determined yearly prevalence, weighted to the national employer-sponsored, privately insured population. Using competing risk analysis, we estimated event rates for portal hypertensive complications and estimated the association between AC and costs as well as admissions and readmissions. In 2015, 294,215 people had cirrhosis and 105,871 (36%) had AC. Mean age at AC diagnosis was 53.5 years, and 32% were women. Over the 7 years queried, estimated national cirrhosis prevalence rose from 0.19% to 0.27% (P < 0.001) and for AC from 0.07% to 0.10% (P < 0.001). Compared to non-AC, AC enrollees were significantly more likely to have portal hypertensive complications at diagnosis and higher yearly cirrhosis and alcohol-related admissions (25 excess cirrhosis admissions and 6.3 excess alcohol-related admissions per 100 enrollees) as well as all-cause readmissions. Per-person costs in the first year after diagnosis nearly doubled for AC versus non-AC persons (US$ 44,835 versus 23,319).

CONCLUSION

In a nationally representative cohort of privately insured persons, AC enrollees were disproportionately sicker at presentation, were admitted and readmitted more often, and incurred nearly double the per-person health care costs compared to those with non-AC. (Hepatology 2018).

Discharge Against Medical Advice in the United States, 2002-2011

OBJECTIVE

To describe the national frequency, prevalence, and trends of discharge against medical advice (DAMA) among inpatient hospitalizations in the United States and identify differences across patient- and hospital-level characteristics, overall and in clinically distinct diagnostic subgroups.

PATIENTS AND METHODS

We conducted a retrospective, cross-sectional analysis of inpatient hospitalizations (≥18 years), discharged between January 1, 2002, and December 31, 2011, using the Nationwide Inpatient Sample. Descriptive statistics, multivariable logistic, and joinpoint regression were used for statistical analyses.

RESULTS

Between January 1, 2002, and December 31, 2011, more than 338,000 inpatient hospitalizations were discharged against medical advice each year, with a 1.9% average annual increase in prevalence over the decade (95% CI, 0.8%-3.0%). Temporal trends in DAMA varied by principal diagnosis. Among patients hospitalized for mental health- or substance abuse-related disorders, there was a -2.3% (95% CI, -3.8% to -0.8%) average annual decrease in the rate of DAMA. A statistically significant temporal rate change was not observed among hospitalizations for pregnancy-related disorders. Multivariable regression revealed several patient and hospital characteristics as predictors of DAMA, including lack of health insurance (odds ratio [OR], 3.78; 95% CI, 3.62-3.94), male sex (OR, 2.40; 95% CI, 2.36-2.45), and northeast region (OR, 1.91; 95% CI, 1.72-2.11). Other predictors included age, race/ethnicity, income, primary diagnosis, severity of illness, and hospital location/type and size.

CONCLUSION

Rates for DAMA have increased in the United States, and key differences exist across patient and hospital characteristics. Early identification of vulnerable patients and preventive measures such as improved patient-provider communication may reduce DAMA.