Mortality and costs associated with alcoholic hepatitis: A claims analysis of a commercially insured population

Accepted Manuscript Mortality and Costs Associated with Alcoholic Hepatitis: A Claims Analysis of a Commercially Insured Population Julie A. Thompson, Noel Martinson, Melissa Martinson PII:

S0741-8329(17)30801-7

DOI:

10.1016/j.alcohol.2018.02.003

Reference:

ALC 6778

To appear in:

Alcohol

Received Date: 10 June 2017 Revised Date:

12 February 2018

Accepted Date: 22 February 2018

Please cite this article as: Thompson J.A., Martinson N. & Martinson M., Mortality and Costs Associated with Alcoholic Hepatitis: A Claims Analysis of a Commercially Insured Population, Alcohol (2018), doi: 10.1016/j.alcohol.2018.02.003. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Fig Mortality and Costs Associated with Alcoholic Hepatitis: A Claims Analysis of a Commercially Insured Population

RI PT

Julie A. Thompson(1) Noel Martinson(2) Melissa Martinson(3) (1)

Division of Gastroenterology & Hepatology, University of Minnesota, Minneapolis MN Acclaim Data Analytics, Minneapolis MN (3) Technomics Research, Minneapolis MN

SC

(2)

Abstract

EP

TE D

M AN U

Rising mortality in the United States due to alcoholic liver disease (ALD) and the dearth of effective treatments for ALD have led to increased research in this area, particularly in alcoholic hepatitis. To understand the burden of illness and potential economic value of effective treatments, we conducted a healthcare claims analysis of over 15,000 commercially insured adults who were hospitalized with alcoholic hepatitis (AH) between 2006 and 2013 and followed for up to 5 years. Their average age was 54 years and 68% were male. Over 5 years, about two-thirds of these adults died (44% in the first year), and fewer than 500 received liver transplants. There were nearly 40,000 re-hospitalizations, with over 50% of the survivors rehospitalized within a year and nearly 75% through the second year. The total costs were nearly $145,000 per patient, with costs decreasing over time from over $50,000 in the first year (including the index hospitalization) to about $10,000 per year in the later years. Total costs for the cohort over 5 years were $2.2 billion. Patients who received a liver transplant averaged about $300,000 in transplanted-related costs and over $1,000,000 in total healthcare costs over 5 years. Average costs in years following the index hospitalization were similar to diabetes. AH has a high mortality and is a high cost condition.

AC C

Introduction

Rising mortality among middle- and working-class whites in the United States has garnered a great deal of attention recently. This increase was largely due to increasing death rates from drug and alcohol poisonings, suicide, chronic liver diseases and cirrhosis.1,2 Other reports have shown increasing rates of hospitalizations and emergency department use due to alcoholrelated conditions.3 Not surprisingly, health care expenditures for alcohol-related diagnoses are increasing. Costs associated with alcohol disorders are reported to account for more than 1% of the gross national product in high- and middle-income countries.4 Prevalence of alcohol abuse/dependence is not completely known but is estimated to effect around 8% of adult Americans.5

1

ACCEPTED MANUSCRIPT

RI PT

Acute alcoholic hepatitis (AH), a result of heavy, prolonged alcohol use, is one of these causes of hospitalizations and emergency department visits. One study reported that approximately 1% of hospital admissions in the U.S. in 2010 were due to AH, which represented a 30% increase over 8 years.3 Guidelines recommend prednisolone in eligible patients,6 but this therapy has limited benefit and was not shown to improved 90-day mortality in a large, randomized, controlled trial.7 Severely affected patients face prolonged hospitalizations, significant morbidity, and high mortality: 20-30% at 1 month and 40-70% at 6 months.8

SC

AH most often occurs in persons ages 40 to 60 years:9 individuals who are in the prime of their working lives. The productivity loss to the economy and to employers is likely very high. Because individuals of this age have employer-based health insurance coverage, the healthcare burden of this highly morbid condition falls largely upon employers, payers and society at large, through increased insurance premiums.

M AN U

There are publications reporting on costs of hospitalization due to AH10,11 but an assessment of healthcare costs after the initial hospitalization remains unreported. The cost of AH has not been studied outside of acute inpatient care. Those costs alone are high enough to indicate that further study is warranted. There are no reports quantifying the total burden of illness due to AH. Understanding the total economic burden of AH can help to focus efforts on developing effective treatments for this growing population. Quantifying the cost burden of AH is important in comprehending the impact of this serious condition and in understanding how to weigh potential treatment options.

Claims Data

EP

Methods

TE D

We sought to evaluate the costs due to AH after index hospitalization for this diagnosis from a commercial insurance claims data set over 5 years. We also looked at death rates, hospital utilization and frequency of liver transplantation.

AC C

Study Sample. For the analysis, we utilized a data set of commercial insurance claims from IMS Health Inc., which has one of the largest U.S. health plan claims databases. The data covered the period from 2006 to 2013 with over 40 million enrollees per year, aged 0-85 years. The dataset has a diverse representation of geography, employers, payers and providers. It has claims from 90% of U.S. hospitals, 80% of all doctors and 85% of the Fortune 500 companies. Patients with commercial and self-insured plans from all parts of the U.S. make up over 96% of the claims in this data set with Medicaid and Medicare Advantage comprising the remainder. Original Medicare plans such as Part A and Part B are not included in this commercial dataset. Allowed charges and reimbursements for acute care hospital inpatient and outpatient visits, long-term skilled nursing facilities, office and clinic visits, and pharmaceuticals were included. Mortality data were available if death occurred during an inpatient stay, but deaths were 2

ACCEPTED MANUSCRIPT

otherwise indistinguishable from coverage terminated for other reasons with a particular plan. (Adjustments for this limitation are discussed below.)

RI PT

Subjects were initially identified by an inpatient hospitalization only for liver decompensation due to recent alcohol ingestion (i.e., no other major comorbidities were present at the time of admission). They had a hospital stay of at least three days with a primary diagnosis of alcoholrelated liver disease, or liver disease with a diagnosis of alcohol abuse at the time of admission or within the past year. Based on those criteria, 30,373 patients aged 18 and older in this dataset were identified as having AH. Adjusting this population for enrollment levels for each year in 2006 thru 2013 and further adjusting for age and population with respect to U.S census estimates produced an estimated annual U.S. case estimate of 29,716.

M AN U

SC

The index hospital stay for a patient was defined as the index event. A “clean period” of one year prior to the index event and continuous enrollment during the prior year were required. A “clean period” is a period of time during which there are no claims with codes associated with the condition of interest. Therefore, the subjects included in this study had no AAH-related claims for at least one year prior to the index hospitalization. Additionally, patients were excluded if they had conditions unrelated to AH that could add significantly to on-going hospital costs. Patients were excluded if they had a diagnosis of metastatic cancer, pregnancy, pneumonia or sepsis at the time of the index event. These criteria reduced the number of patients in this study to 15,496.

EP

TE D

Death Adjustments. The commercial claims data includes death dates when patients die in the hospital, but out-of-hospital deaths are not recorded as deaths. Patients who appear to leave the health plan are a mix of those who lose insurance, transition to another insurance company, or die out-of-hospital (collectively, these are known as the “administratively censored”). As a result, the observed mortality of the claims population was about 25% lower than that reported in the medical literature. If all patients who were administratively censored were also considered to have died, the death rate was about double that in the medical literature. Clearly, neither of these ways of counting deaths seemed appropriate.

AC C

Yet if costs were to be accurately estimated, it was important that the real death rate be reflected in the data. This is because post-mortem, deaths are true $0 costs from the perspective of the heath care payers, while plan members who dis-enroll incur healthcare costs elsewhere and should not be treated as true $0 costs.12 Fortunately, Medicare includes death information regardless of where the death occurred. Therefore, the mortality of the commercial claims patients could be adjusted to more accurately reflect true deaths by taking advantage of the death data in the Medicare. To adjust the death rate of the commercial claims patients, we used a propensity score analysis of the Medicare patients. The purpose of a propensity score analysis is to estimate the probability that a patient with a given set of characteristics ends up in one category versus another category. It is commonly used to compare two medical treatments from observational

3

ACCEPTED MANUSCRIPT

data to adjust the comparison for imbalances in the characteristics of patients in the two treatment groups, and it is the preferred method by FDA for adjusting observational data. The adjustment has been shown to approximate the results of randomized clinical trials more closely than an unadjusted comparison.13

M AN U

SC

RI PT

For example, suppose some researchers wanted to compare observational data on the incidence of heart attacks from patients with coronary artery disease who are treated with drugs compared to those who get cardiovascular interventions. In many observational data sets, the latter patients tend to be older and have a higher percentage of male patients. A propensity score analysis would first use logistic regression on all of the data to estimate the probability that a patient got an intervention (versus drug therapy) based on the patient’s known covariates of age and sex. Then, using the probability and the actual treatment of each patient, the effect of the intervention (versus drug therapy) would be estimated in a time-toevent analysis of the time to a heart attack. The propensity score is a stand-in for a potentially large number of covariates.

TE D

In the case of this study, patients in the Medicare data either died or did not, so we estimated their probability of dying based on a large set of demographic covariates and diagnosis codes common to both the Medicare and commercial claims data. The resulting equation was then applied to patients in the commercial claims sample who were administratively censored to get their predicted probability of death. Then these enrollees were changed to status of “dead” starting with the highest probability of death and moving down the probabilities until the mortality of the group approximately matched that of the Medicare AH population.

EP

Payer-perspective Costs. Allowed charges were used in the payer-perspective analysis because they represented negotiated amounts that providers could be paid for a service by the insurer including patient copays and payments by other insurers; they most closely represented the costs from all payers. Total allowed charges (TAC) were separated into treatments for mental health/addiction, AH-related, liver-transplant related, and other.

AC C

Mental health and addiction treatment allowed charges were identified using the site of service codes for psychiatric and mental health facilities. For inpatient hospital charges, the primary diagnosis codes were used to identify addiction treatment. For outpatient and office claims, Current Procedural Terminology (CPT) codes were used. AH-allowed charges were identified by the primary diagnosis code for the claim and had to be accompanied in the same claim by at least one diagnosis of liver disease. Allowed charges associated with liver transplants were extracted from the AH charges by using the date of the liver transplant and tallying AH charges in the prior year. Allowed charges associated with death were the TAC in the year prior to the year of death. All costs were adjusted to 2013 dollars using the medical portion of the Consumer Price Index. A complete list of codes is included in the appendix.

4

ACCEPTED MANUSCRIPT

Statistical Analysis Demographics. Simple descriptive statistics such as means and proportions were used to describe the patients in the study sample.

SC

RI PT

Costs. Costs during an interval were estimated among those who completed the interval and then adjusted for the percentages of completers, censored subjects, and deaths. Subjects who died during an interval were assumed to have died, on average, at the mid-point of the interval (e.g., at 6 months from the start of an annual interval) and therefore to have incurred 50% of the costs of survivors of that interval. Censored patients (those who exited the interval before its end for reasons other than death) were assumed to accumulate the same average costs as completers, although in another health plan. The average cost during an interval was calculated as the average of the complete costs weighted by the percentage of completers and censored, and of 50% costs weighted by the percentage of deaths.12

M AN U

Liver transplant-related costs were the costs incurred by transplant patients in the 3 quarters prior to the transplant and in the quarter of the transplant. Mortality. Time-to-death was analyzed using Kaplan-Meier survival analysis to compare the mortalities in the Medicare and commercial claims patients; we wanted to assure that these were very similar (see the section on Death Adjustments above. The number of deaths in the commercial claims was adjusted both for the unreported deaths and for administrative censoring of patients who went to another health plan.

TE D

Liver Transplants. Time-to-transplant was analyzed using Kaplan-Meier analysis, while the number of transplants was adjusted for censoring analogously to the adjustment of deaths.

AC C

EP

Hospital Utilization. The clinical outcome of cumulative rehospitalizations and rehospitalization days were analyzed by descriptive statistics. The numbers were adjusted for mortality and censoring analogously to the adjustment of costs, deaths, and transplants. Time-to-firstrehospitalization was analyzed using Kaplan-Meier analysis.

Results

Demographics

In the entire IMS Health Inc. data set, there were 15,496 patients who met criteria for index hospitalization for AH. They had an average age of 54 years and 68% were male. Table 1 shows demographics and major selected comorbidities of these patients. A total of 7637 (49%) completed year 1; 4404 (28%) completed year 2; 2428 (16%) completed year 3; 1117 (7.2%) completed year 4; and 466 (3.0%) completed 5 years of follow-up.

5

ACCEPTED MANUSCRIPT

Table 1: Demographics and Selected Major Comorbidities

RI PT

15,496 54 (11) 68% 56% 21% 21% 10% 9.2% 7.6%

SC

Characteristic Size of Cohort Age (y) (mean, SD) Sex (% Male) Hypertension Chronic Obstructive Pulmonary Disease Diabetes Congestive Heart Failure Cancer Peripheral Vascular Disease

M AN U

Table 2 shows the disposition of patients in this cohort. Of these 15,496 patients, 6517 died in the hospital in the first year (44%). Table 2: Disposition of Patients Patients at Died in Start Hospital 15,496 6,517 7,637 1,885 4,404 460 2,428 46 1,117 28

Year

Remainder 7,637 4,404 2,428 1,117 466

EP

TE D

1 2 3 4 5

Died Elsewhere or Changed Plan 1,342 1,348 1,516 1,265 623

Hospital Utilization

AC C

The average length of the index hospitalization was 9.8 days, and half of the patients stayed longer than 6 days. After the index hospitalization, the adjusted number of hospitalizations over 5 years totaled 39,866, an average of 2.6 per patient. The adjusted number of days in hospital totaled over 257,868 (average: 16.6). Table 3 shows adjusted hospitalizations in the first year were 18,500 with adjusted total hospital days in year 1 of 144,134. There were 257,868 hospital days during the 5 years of follow up time, half of which were in the first year. Hospitalizations decreased over time, as did number of hospital days.

6

ACCEPTED MANUSCRIPT

Table 3: Adjusted Hospitalizations and Hospital Days

M AN U

1 2 3 4 5 TOTAL AVERAGE

Adjusted Hospitalizations 18,500 7,984 5,499 4,154 3,729 39,866 2.6

Adjusted Total Hospital Days 144,134 45,200 29,274 19,293 19,966 257,868 16.6

RI PT

Hospitalizations per 1000 1503 1063 927 760 704 4,957 991

Hospital Days per 1000 11,920 6,002 4,930 3,530 3,770 30,152 6,030

SC

Years after Index Hospitalization

Figure 1 shows time to first re-hospitalization adjusted for deaths (see below) and adjusted for patients who moved to a different health plan. Over half had been re-hospitalized within the first year, and nearly three-fourths by the second year. Average time to first rehospitalization was 7.5 months.

AC C

EP

TE D

Figure 1: Adjusted Time to First Re-hospitalization

Predictions of Deaths in the Commercially Insured Population Since we had to estimate out-of-hospital deaths in the commercial claims, we wanted to test the accuracy of our predictions on the known deaths (i.e., those patients who died in-hospital). In the total AH population, the sensitivity of the prediction equation was 90% (i.e., 90% of the known deaths were successfully predicted). It was 50% specific among those known to be alive

7

ACCEPTED MANUSCRIPT

RI PT

(i.e., 50% of those known to be alive at the chronological end of the data were successfully predicted). Figure 2a shows the actual Medicare survival, the actual commercial claims survival, and the predicted commercial claims survival. There was no evidence of a difference between the predicted commercial survival and the actual Medicare survival (log-rank test p = 0.33). Figure 2b shows the receiver-operator characteristic (ROC) curve; such curves are commonly used to estimate the overall accuracy of a diagnostic or prediction system. The area under the curve, or overall accuracy, was 79%. Figure 2a: AH Patients: Survival in the Figure 2b: ROC Curve for the Mortality Medicare and Adjusted Commercial Claims Prediction

0

.5

1 1.5 Years after Index Hospitalization

source = Adjusted Commercial source = Raw Commercial

M AN U

0.00

Proportion Surviving 0.25 0.50 0.75

SC

1.00

Kaplan-Meier survival estimates

2

source = Medicare

TE D

Mortality

Based on the propensity score we created, the total deaths (known plus predicted) are shown in Table 4 along with the distribution of estimated follow-ups times. We estimate that after 5 years, about 2/3 of the AH population will have died.

EP

Table 4: Adjusted Deaths

AC C

Years FU 0-1 >1 - 2 >2 - 3 >3 - 4 >4 - 5 TOTAL

N at Start of Year 15,496 8,684 6,333 5,534 5,392

Deaths

Percent of Total

6,812 2,351 799 142 187 10,291

44% 15% 5.2% 0.9% 1.2% 66%

8

ACCEPTED MANUSCRIPT

Liver Transplants

1 2 3 4 5 TOTAL

N at Start of Year 15,496 8,684 6,333 5,534 5,392

Probability of Transplant 0.0041 0.0102 0.0154 0.0184 0.0250

Adjusted Transplants 50 77 91 101 132 451

M AN U

Years FU

SC

Table 5: Adjusted Liver Transplants

RI PT

There were 247 observed liver transplants over 5 years. The adjusted number of transplants is shown in Table 5; the adjusted number of transplants is about twice as many as the observed. The adjusted numbers reflect accounting for patients who did not die, but left the health plan, enrolled in other plans and would have been similarly predicted to experience transplant as an outcome as those who stayed in the plan. The cumulative transplant rate was 7.1%. Figure 3 shows the transplant rate in the first year. In the first 6 months after index hospitalization, 20 patients received a liver transplant.

AC C

EP

TE D

Figure 3: Time-to-Liver Transplant

9

ACCEPTED MANUSCRIPT

Costs

RI PT

Total Costs. The total costs include all healthcare costs over time are shown in Table 6. The table shows that the average annual costs decreased over time. Q1 was the most expensive quarter in year 1 (it included the index event), and year 1 was the most expensive of the years, with 37% ($825 million) of the total costs occurring in year 1. Average cost per patient in the first year was $53,269 in the total cohort, with 80% of that incurred in the first quarter ($43,165). Total costs over 5 years were nearly $145,000 per patient. About a third of the costs were accumulated in year 1. (This includes costs after death of $0.) The adjusted total cost over 5 years was over $2.2 billion.

SC

These were the average costs. Many patients incurred costs over $500,000. By the end of the first year, 304 patients had these very high costs. By five years, 588 (3.8% of the cohort) had accumulated these high costs.

FU Time

Adjusted Mean Cost Per Patient $43,165 $53,269 $16,951 $12,106 $10,197 $8,966 $144,654

TE D

EP

Q1 Y1 Y2 Y3 Y4 Y5 TOTAL

N at Start 15,496 12,377 8,684 6,333 5,534 5,392

M AN U

Table 6: Adjusted Total Costs

Adjusted Per Member Per Month Cost $5.76 $1.78 $0.57 $0.40 $0.34 $0.30 $0.97

Adjusted Cost $668,890,209 $825,460,981 $262,669,272 $187,587,593 $158,018,448 $138,929,899 $2,241,556,403

AC C

AH- and Liver-Transplant-Related Costs. Among the patients who received a transplant (including those who received a transplant after 5 years) the mean total costs were about $430,000 in the year leading up to and including the transplant. The liver-transplant-related costs were $279,000. (Transplant costs are a subset of total costs as described in Methods.) Table 7a shows the transplant costs and total costs by length of follow-up among the patients receiving transplants. Patients who were followed for 5 years and had a transplant at any time in that interval incurred total costs over $1 million. This compares to non-transplant patients, who over 5 years averaged under $150,000. The estimated average annual cost after the first year was over $200,000.

10

ACCEPTED MANUSCRIPT

Table 7a: Adjusted Total Costs Over All Follow-up Among Observed Transplant Patients Years of FU

$ 58,353

RI PT

247

Total Allowed $ 340,914 $ 573,792 $ 836,789 $ 700,111 $ 852,171 $ 1,039,536

SC

Transplant-Related $ 211,806 $ 216,870 $ 285,032 $ 228,941 $ 282,487 $ 308,352

$

221,360

M AN U

<1 1 2 3 4 5 Annual Cost after first year of transplant

N 48 62 53 33 20 31

TE D

Table 7b shows the costs associated with AH-related treatment and with liver transplantation in quarter 1 (Q1) and annually thereafter. (The AH-related and liver transplantation-related costs were mutually exclusive subsets of total allowed charges as described in Methods.) The small number of transplant patients explains the low liver transplant means among the entire cohort. The table also shows the percent that these two categories represent of total costs. During Q1, about three-fourths of all costs were AH- or transplant-related. This translates to about twothirds of total costs during the first year. Thereafter, AH- and transplant-related costs hovered around 45% to 65%, but usually close to 50%. Table 7b: Adjusted AH- and Transplant-Related Mean Costs in the Entire Cohort

AC C

Q1 Y1 Y2 Y3 Y4 Y5 TOTAL

AH Treatment $ 31,021 $ 33,881 $ 8,308 $ 6,110 $ 5,169 $ 4,070 $ 88,558

EP

FU Time

Liver Transplant $ 1,636 $ 2,794 $ 1,236 $ 670 $ 527 $ 463 $ 7,325

Percent of Total Costs 75.7% 68.8% 56.3% 56.0% 55.9% 50.6% 66.3%

PMPM Costs $ 4.36 $ 0.98 $ 0.18 $ 0.09 $ 0.07 $ 0.05

Death-Related Costs. Over five years of follow-up, there were 10,291 observed and predicted deaths. The costs surrounding death were similar regardless of how long after the index hospitalization the death occurred: about $50,000 to $60,000. This is 200% to 300% higher than average annual costs of $10,000 to $15,000 for the entire cohort. Table 8 shows the deathrelated costs among patients who died.

11

ACCEPTED MANUSCRIPT

Table 8: Adjusted Death-Related Costs

RI PT

N of Observed and Predicted Deaths

Year of Death

Total Costs in Year Surrounding Death

6812

$ 73,084

2

2351

$ 35,663

3

799

$ 31,743

4

142

$ 92,052

5 Average

187 10,291

SC

1

M AN U

$ 65,881 $ 63,137

Discussion

EP

TE D

Alcohol-related utilization of health care resources, illnesses and mortality are rising. A report in the Proceedings of the National Academy of Sciences found that after decades of improvement, white non-Hispanic men had rising mortality in midlife (ages 45-54 years), with alcohol-related conditions accounting for a significant contribution to this.1 Similarly, another report analyzing data from the Centers for Disease Control and Prevention indicated that mortality rates rose for white women from 1990-2014 in most parts of the U.S. This was predominantly due to drug overdose, heavy drinking, and suicide. Among rural women in their early 50s, this report found a doubling in deaths from cirrhosis since the end of the 20th century.2 Not surprisingly, utilization or health care resources and health care expenditures for alcohol-related diagnoses are increasing.

AC C

AH is an increasingly common, highly morbid and mortal condition. Utilization of healthcare resources due to AH is on the rise, as are estimated costs from the condition.3 Our intention in this burden-of-illness study was to provide estimates for the costs associated with caring with this patient population, who have no effective current treatments. The disease is expensive for payers and carries a high mortality for patients. Current treatments are either not very effective or not widely available, such as liver transplant, which is also extremely expensive. The costs and high mortality reported support the need for development of new treatments. This cohort of 15,496 patients with an index AH event incurred over $2 billion in costs in the 5-year followup period. The first year was the most expensive with a total of $825 million in costs (37%) in that year alone. Most of these first year costs (80%) occurred in the first quarter--the time of the index hospitalization--with an average cost per patient in the first year of $53,269; and $43,165 of that from the first quarter. If we extrapolate the claims counts to an age- and sex-

12

ACCEPTED MANUSCRIPT

RI PT

adjusted overall U.S. population from census data, there would be an estimate of 29,716 people in the U.S. with first hospitalizations for AH annually. The total annual U.S. healthcare costs for this would be $1.5 billion. Our cohort represented only insured patients; therefore our study cannot make conclusions regarding the uninsured population who may require higher health care utilization, incur greater costs and have a worse prognosis

TE D

M AN U

SC

A large study of AH utilizing data from the 2007 Nationwide Healthcare Cost and Utilization Project reported data from over 8 million inpatient admissions, with AH accounting for about 57,000 of those. Mean hospital stay was 6.5 days and death in-hospital occurred in 6.8%. Total charges during hospitalization were over $37,000, for a national total in excess of $2 billion.11 That study included only inpatient costs, while ours looks at outpatient as well. These authors also looked at hospitalizations for AH from 2002 to 2010 as reported in the National Inpatient Sample databases (part of the Healthcare Cost and Utilization Project by the Agency for Healthcare Research and Quality). They reported on 326,403 AH-related hospitalizations (primary and secondary diagnoses) with an average length of stay of 6.1 days and inpatient death rate of 5.8%. The average cost (after adjustment for inflation) increased during that time period by 40.7%: with an average cost related to the primary diagnosis of AH of $46,264 in 2010.10 Our study found an adjusted mean cost in the first year of illness of $53,269, and an average of about 17 days of hospitalization per patient in the first year. The average length of stay for the index hospitalization in our cohort was longer than these reports at 9.8 days, and more than half of the patients stayed longer than 6 days. There were 39,866 hospitalizations, an average of 2.6 per patient. The total number of hospital days was 257,868, with nearly half of these occurring in the first year. Median time to first rehospitalization was about 7.5 months for those who survived (Figure 1 and Table 3).

AC C

EP

AH exacts a high toll from a societal perspective. Patients with the condition tend to be people of middle age in their prime working years (our mean age was 54 years), and the condition negatively affects their productivity in the workforce. This impacts both their families and their employers. It also high cost to insurers due to the high consumption of healthcare resources by these patients. In this study, the average cost of these patients in the year of hospitalization is just under $50,000, including the truncated costs of the estimated 44% who die in the first year. In the years following, costs are in the $10,000 to $15,000 range. Not included in our study were societal costs outside of the healthcare system, e.g. costs to employers of work absenteeism and reduced productivity. Our estimate of a total U.S cost of $1.5 billion per year for this group is, therefore, a conservative estimate of healthcare costs and does not attempt to estimate societal costs. The table below shows how the costs of AH compare with other major conditions. On a perpatient basis, in comparison to diabetes, heart failure, and bipolar disorder, the annual cost of AH is quite expensive. It is in the same ballpark as end-stage renal disease (ESRD), at least in the year of the index hospitalization. On a per-patient basis, it falls between diabetes and ESRD, even with our conservative estimates. By comparison, average health spending in 2015 for

13

ACCEPTED MANUSCRIPT

females and males aged 45-54 years was $4,825 and $5,908, respectively. Spending for females and males aged 55-64 years was $7,850 and $9,489, respectively. 14

Author, Year

Est % U.S. Population with Condition per year

Annual U.S. HealthCare Costs in Billions (Year)

Average Annual per-Patient Costs

Heart Failure

Heidenreich 201315

3% =9.5M*

$21 (2012)

$2200*

Bipolar Disorder

Miller 201416

2.8% = 9.0M

$31 (2009)

$3600*

Diabetes

ADA 201317

7% = 22.3M

$176 (2012)

$7900

AH

This paper

30,000*

$1.5 (2013)

$50,000

$2.3

$29,000

$29.5 (2001)

$75,000*

$5.7 (2015)*

$117,000

SC

M AN U

80,000**

RI PT

Condition

Joyce 200418 0.14 %= 448,000

Pulmonary Arterial Hypertension (PAH)

Sikirica 201419

TE D

End-Stage Renal Disease

0.015% = 49,110*

Of those with PAH:

EP

13% are 65 years or older

AC C

87% are younger than 65

* Incident population over one year. ** Surviving population over 5 years assuming death rates estimated in this study period. Data in this large claim set showed that patients admitted for severe AH have a high mortality, which is consistent with other reports, including the STOPAH trial.7 The majority of the mortality is in the first year but most other literature regarding AH does not report beyond 30or 90-day mortality and there are few that report out to 5 years. Our total population had a mortality rate of 44% in the first year. In the 5-year follow up period, our analysis showed an overall death rate of 66%. Assuming these patients survive their first year, their annual mortality rate averaged 10% over the subsequent 4 years. The data do not allow us to attribute

14

ACCEPTED MANUSCRIPT

RI PT

all of these deaths to AH directly, but clearly there is a very high mortality in individuals who are diagnosed with an index AH event. As a comparison, patients with heart failure—a condition which affects about 3% of the U.S. population and is identified with a well-established high burden on health care—have a 50% mortality in the 3 years after hospitalization.15 It is important for us to note a limitation of this study: in the commercial population, many of the deaths were imputed from our estimate of the propensity score for death, which was 90% sensitive but only 50% specific. Our death rates in the first year are consistent with other reports in AH, however.

EP

TE D

M AN U

SC

There are no treatments that offer significant 90-day mortality benefit in AH. A recent publication reporting 90-day mortality in patients with AH who were treated with extracorporeal liver therapy (“ELAD”) versus controls showed no overall benefit, but subgroup analyses in younger patients without renal failure showed potential benefit. These results are under investigation in a trial examining this group,20 which is potentially promising.21 Whether this therapy will reduce the high mortality from this disease, alter costs, or provide a bridge to transplant remains uncertain. LT for AH remains controversial. There has historically been an almost universal requirement of at least 6 months of sobriety at liver transplant programs, but some LT programs are beginning to perform liver transplantation for AH. The first protocoldriven data on LT transplantation in patients with severe AH were published in 2011.22 Since the a few other single-center reports with small numbers of patients have been published, but LT for severe AH has not yet been widely embraced in the U.S. and there is no standard for accepting these patients for LT. In this group of over 15,000 patients 451 (including imputed) underwent LT over the 5-year follow up; 50 of these were in the first year, and 20 of them in the first 6 months. Therefore, these patients were transplanted prior to achieving the length of sobriety as directed in the historic “six-months rule” and were, therefore, transplanted for AH. The costs associated with LT were between $200,000 and $300,00 per year. Transplant was expensive: Patients who were followed for 5 years and had a transplant at any time in that interval incurred total costs over $1 million in the AH group (Table 7a). This compares to nontransplant patients, who over 5 years averaged under $150,000 (Table 6).

AC C

AH is highly morbid and expensive condition. It is currently without any effective treatment and patients incur multiple hospitalizations and significant morbidity and mortality. It has become apparent from recently published data that costs, morbidity and mortality from alcohol-related diseases are on the rise. Most hospitalizations, costs and mortality occur in the first year. Any contribution to improving the morbidity and mortality for these patients would improve the cost burden as well. The total healthcare burden is lower than most of the comparator conditions only because of the smaller prevalent population, and this, unfortunately, has been rising. Effective treatments for AH are urgently needed.

15

ACCEPTED MANUSCRIPT

References

RI PT

1. Case A, Deaton S. Rising morbidity and mortality in midlife among white non-Hispanic Americans in the 21st century. Proceedings of the National Academy of Sciences of the United States of America. 2015;112(49): 15078-15083. http://www.pnas.org/content/112/49/15078.abstract 2. Achenbach J, Keating D. “A New Divide in American Death.” Washington Post April 10, 2016. Print and on-line: http://www.washingtonpost.com/sf/national/2016/04/10/anew-divide-in-american-death/

M AN U

SC

3. Chen CM, Yoon Y-H, Yi H-y, et al. Alcohol-Related Emergency Department Visits and Hospitalizations and Their Co-occurring Drug-Related, Mental Health, and Injury Conditions in the United States: Findings from the 2006-2010 Nationwide Emergency Department Sample (NEDS) and Nationwide Inpatient Sample (NIS). National Institutes of Health: National Institute on Alcohol Abuse and Alcoholism; September 2013: NIH Publication No. 13-8000. 4. Rehm et al. Global burden of disease and injury and economic cost attributable to alcohol use and alcohol use disorders. Lancet 2009;373(9682):2223-33.

TE D

5. Chen CM, Yi H-y, Falk DE, et al. Alcohol use and alcohol use disorders in the United States: main findings from the 2001-2002 National Epidemiologic Survey on Alcohol Use and Related Conditions (NESARC). National Institutes of Health (U.S.); National Institute on Alcohol Abuse and Alcoholism (U.S.); CSR, Incorporated 2006 Bethesda, MD: National Institutes of Health, National Institute on Alcohol Abuse and Alcoholism; Jan 2006: NIH Publication No. 05-5737. 6. O’Shea RS, Dasarathy S, McCullough AJ. Alcoholic Liver Disease. Practice Guideline Committee of the American Association for the Study of Liver Disease and the Practice Parameters Committee of the American College of Gastroenterology. 2010;5:307-28.

EP

7. Thursz et al. Prednisolone or Pentoxifylline for Alcoholic Hepatitis. NEJM 2015;372: 1619-28.

AC C

8. Moreau R, Jalan R, et al. Acute-on-Chronic Liver Failure Is a Distinct Syndrome That Develops in Patients With Acute Decompensation of Cirrhosis. Gastroenterology 2013;144:1426-1437. 9. Nguyen T, DeShazo J, Thacker L, Puri P, Sanyal A. The worsening profile of alcoholic hepatitis in the United States. Alcohol Clin Exp Res 2016;40(6):1295–1303. 10. Jinjuvadia R, Liangpunsakul S, Translational Research and Evolving Alcoholic Hepatitis Treatment Consortium. Evolving alcoholic hepatitis treatment. Trends in alcoholic hepatitis-related hospitalizations, financial burden and mortality in the United States. J Clin Gastroenterol 2015;49(6):506-11. 11. Liangpunsakul S. Clinical characteristics and mortality of hospitalized alcoholic patients in the United States. J Clin Gastroenterol 2011;45:714-19. 16

ACCEPTED MANUSCRIPT

12. Bang J, Tsiatis AA. Estimating medical costs with censored data. Biometrika 2000; 87(2): 329-343. 13. Medicare Inpatient Hospital Standard Analytical File 2013. Center for Medicare & Medicaid Services (CMS), Baltimore, MD. www.cms.hhs.gov. Accessed June 2017.

RI PT

14. Kaiser Family Foundation Analysis of Medical Expenditure Panel Survey, Agency for Healthcare Research and Quality, U.S. Department of Health and Human Services. www.healthsystmetracker.org. Accessed February 2018.

SC

15. Heidenreich PA, Albert NM, Allen LA, et al. Forecasting the Impact of Heart Failure in the United States: A Policy Statement from the American Heart Association. Circ Heart Fail 2013;6:606-619. 16. Miller S, Dell’Osso B, Ketter TA. The prevalence and burden of bipolar depression. J Affective Disorders 2014;169(S1):S3-S11.

M AN U

17. American Diabetes Association. Economic Costs of Diabetes in the U.S. in 2012. Diabetes Care 2013;36:1033-1046. 18. Joyce AT, Iacoviello JM, Ng S, et al. End-Stage Renal Disease-Associated Managed Care Costs Among Patients With and Without Diabetes. Diabetes Care 2004;27:2829-2835. 19. Sikirica M, Iorga SR, Bancroft T, Potash J. The economic burden of pulmonary arterial hypertension (PAH) in the US of payers and patients. BMC Health Services Research 2014;14:676-86.

TE D

20. Thompson JA, Jones N, Al-Khafaji A, et al. Extracorporeal Cellular Therapy (ELAD) in Severe Alcoholic Hepatitis - A Multinational, Prospective, Controlled, Randomized Trial. Liver Transpl 2017. Epub ahead of print. 21. Morgan, T. ELAD: A Potential Silver Lining? Liver Transpl 2018. Epub ahead of print.

AC C

EP

22. Mathurin P, Moreno C, Samuel D, et al. Early Liver Transplantation for Severe Alcoholic Hepatitis. N Engl J Med 2011;365:1790-1800.

J Thompson, M Martinson and N Martinson have received consulting fees from Vital Therapies, Inc. This paper was written entirely by J Thompson, M Martinson and N Martinson.

17

ACCEPTED MANUSCRIPT

Highlights

•

AC C

EP

TE D

M AN U

SC

•

There is an increasing rate of hospitalizations due to alcohol-related conditions. Medical treatment for AH has limited efficacy and transplantation is the only lifesaving alternative, but it is not widely accepted. The total cost of care of AH outside inpatient care costs are higher on average per year than for heart failure and diabetes patients. Patients with the condition tend to be people of middle age in their prime working years.

RI PT

• •