National Institute on Alcohol Abuse and Alcoholism
Division of Epidemiology and Prevention Research
Alcohol Epidemiologic Data System
SURVEILLANCE REPORT #94
TRENDS IN ALCOHOL‑RELATED MORBIDITY AMONG SHORT‑STAY COMMUNITY HOSPITAL DISCHARGES, UNITED STATES, 1979–2010
Chiung M. Chen, M.A.
Hsiao-ye Yi, Ph.D.
CSR, Incorporated1
Suite 1000
2107 Wilson Boulevard
Arlington, VA 22201
August 2012
U.S. Department of Health and Human Services
Public Health Service
National Institutes of Health
1 CSR, Incorporated, operates the Alcohol
Epidemiologic Data System (AEDS) under Contract No. HHSN267200800023 for the
Division of Epidemiology and Prevention Research, National Institute on Alcohol
Abuse and Alcoholism (NIAAA). Dr. Rosalind A. Breslow serves as NIAAA Project
Officer on the contract.
HIGHLIGHTS
This surveillance report, prepared biennially by the Alcohol Epidemiologic Data System (AEDS), National Institute on Alcohol Abuse and Alcoholism (NIAAA), presents data on alcohol‑related morbidity in the United States from 1979 to 2010. AEDS compiles these statistics on alcohol‑related inpatient stays based on a national sample of hospital discharge episodes from the National Hospital Discharge Survey (NHDS), conducted annually by the National Center for Health Statistics (NCHS). Highlights of general trends and notable findings are listed below. It is important to note that the sample of hospitals was reduced by half beginning in 2008 and the number of diagnosis codes collected by NCHS increased to fifteen in 2010 from seven in the previous years. These changes may artificially affect the trends observed in recent years and potentially increase the numbers and rates of all-listed alcohol-related hospital discharges observed in 2010.
Alcohol-Related Hospital Discharges in 2010
Approximately 390,000 hospital discharge episodes for persons ages 15 and older had a principal (first‑listed) alcohol‑related diagnosis, and approximately 1.9 million discharge episodes had an any (all‑listed) alcohol‑related diagnosis. These figures represent 15.8 principal (first‑listed) and 76.4 any (all‑listed) alcohol‑related discharges per 10,000 population, a decrease in the principal (first‑listed) alcohol‑related discharge rate and an increase in the any (all‑listed) alcohol‑related discharge rate compared with the 2009 rates (17.2 and 68.2, respectively).
Alcoholic psychoses were the largest group (40.6 percent) of principal (first‑listed) diagnoses, followed by cirrhosis of the liver (26.5 percent), alcohol dependence syndrome (24.3 percent), and nondependent abuse of alcohol (8.6 percent).
A substantial difference exists between rates based on principal (first‑listed) and any (all‑listed) diagnoses. Nearly four-fifths (79.3 percent) of alcohol‑related discharge episodes did not have an alcohol‑related principal (first‑listed) diagnosis.
Alcohol‑related diagnoses in decreasing order of severity, as measured by average length of hospital stay, were cirrhosis (5.0 days, with 5.3 days for alcoholic cirrhosis), alcoholic psychoses (4.6 days), alcohol dependence syndrome (4.1 days), and nondependent abuse of alcohol (2.0 days).
General Trends
Hospital discharge rates showed a clear upward trend for both principal (first‑listed) and any (all‑listed) alcoholic psychoses as well as for any (all‑listed) chronic liver disease and cirrhosis and nondependent abuse of alcohol from 1988 to 2010. In contrast, there was a downward trend for principal (first‑listed) alcohol dependence syndrome between 1995 and 2010.
For all alcohol‑related diagnoses, except cirrhosis without mention of alcohol, hospital discharge rates continue to be higher for males than for females. Persons ages 45 to 64 generally have the highest principal (first‑listed) as well as any (all‑listed) alcohol‑related discharge rates, and persons ages 15 to 24 have the lowest alcohol‑related discharge rates.
In terms of percentage shares, alcohol dependence syndrome was the largest group of principal (first‑listed) alcohol‑related diagnoses until 2004. However, its percentage share declined substantially from more than 70 percent in 1979 to 24.3 percent in 2010. The percentage shares of diagnoses for alcoholic psychoses and any cirrhosis increased during the 32 ‑year study period. Alcoholic psychoses outnumbered any cirrhosis since 1994 and has surpassed alcohol dependence syndrome as the largest principal (first‑listed) alcohol‑related diagnoses since 2005.
While there has been relatively little change in the percentage of hospital discharges with principal (first‑listed) mention of an alcohol‑related diagnosis, the proportion of hospital discharges with any (all‑listed) mention of an alcohol‑related diagnosis has increased. The ratio for principal (first‑listed) to any (all‑listed) alcohol‑related discharges declined from 60.9 percent in 1979 to 20.7 percent in 2010.
For alcohol dependence syndrome and cirrhosis, reductions in average length of stay were observed since 1988 but leveled off in the few years before 2010.
INTRODUCTION
This is the eighteenth surveillance report on trends in alcohol‑related morbidity among patients discharged from short‑stay community hospitals in the United States. Prepared by the Alcohol Epidemiologic Data System (AEDS), Division of Epidemiology and Prevention research, National Institute on Alcohol Abuse and Alcoholism (NIAAA), the report updates the trends published in earlier surveillance reports. The findings are intended to be useful to policymakers, health care providers, researchers, and other individuals concerned about the health effects of alcohol abuse.
Data are presented by age and sex, including numbers and population‑based rates for hospital discharges with principal (first‑listed) mention or any (all‑listed) mention of specific diagnoses for chronic diseases resulting from alcohol abuse. Also included are data on the average length of hospital stay for alcohol‑related discharge episodes. Race‑specific data are not reported because a large proportion of discharges do not include race information.
AEDS uses variance estimation procedures recommended by the National Center
for Health Statistics (NCHS) to develop 95‑percent confidence intervals
for each estimate shown in figure 5 through figure 9 based on relative standard
errors (RSEs). The values of all estimates are presented in tables.
DATA
Sources
Hospital discharge data for this report were obtained from the National Hospital Discharge Survey (NHDS), conducted annually by NCHS since 1965. These data are processed by NCHS and made available on public‑use computer files. The report begins with 1979, the year in which the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD‑9‑CM) was implemented. ICD‑9‑CM is a diagnostic coding scheme published by the Commission on Professional and Hospital Activities (1978) and is based on the World Health Organization’s ninth revision of the ICD (1977). Although the Tenth Revision of the ICD (ICD‑10) has been implemented for mortality data beginning with data year 1999, its clinical modification version ICD‑10‑CM for morbidity data will not be implemented until October 1, 2014 (Office of the Secretary, HHS 2012).
The NHDS collects data from a sample of non‑Federal, short‑stay hospitals with six or more beds and an average length of stay fewer than 30 days. The sample is stratified by geographic region and hospital size. The probability of selecting a hospital is directly proportional to its size. Discharge episodes are sampled randomly at each participating hospital; the episode sampling ratio within an individual hospital varies inversely with the probability of hospital selection. NCHS calculates appropriate weights and includes them in the data files to project national estimates from the sample. Descriptions of the NHDS sampling design, data collection procedures, and data collection instruments used during the 1979–1987 period are published elsewhere by NCHS (1977, 2010a, 2010b, 2011, 2012).
In 1988 NCHS implemented a new sample design to (1) provide geographic sampling comparability with other surveys conducted by NCHS, (2) update the sample of hospitals selected for the survey, and (3) maximize the use of data collected through automated systems. This change in the sample design may affect trend data because some differences between NHDS statistics based on the earlier sample (1979–1987) and statistics based on the 1988 sample may be due to sampling variability rather than changes in patterns of hospital utilization (NCHS 2010a, 2010b, 2011, 2012). For example, NCHS compared the old survey results with the new results and produced significantly lower estimates of principal (first‑listed) and any (all‑listed) diagnoses of alcohol dependence in the new series (Haupt and Kozak 1992).
For each hospital discharge episode in the sample, the following items are provided: the patient’s age, sex, race, marital status, and length of stay; the hospital’s size and regional location; and codes for up to seven diagnoses and up to four surgical procedures. In 2010, the number of diagnosis codes collected increased from seven to fifteen and the number of procedure codes collected increased from four to eight (NCHS 2012).
Civilian population data used in calculating hospital discharge rates were estimated by the U.S. Census Bureau and provided by the NCHS.
Limitations
Estimates of alcohol‑related morbidity based on the NHDS sample may underestimate the overall prevalence of such morbidity in the general U.S. population. For example, the NHDS sample does not include Veterans Administration and other Federal hospitals or hospitals where the average length of stay is 30 days or longer. Morbidity among individuals who are not hospitalized (i.e., those who seek outpatient treatment or no treatment) also is not reflected in the NHDS data. Furthermore, the stigma associated with alcohol abuse may lead to some reluctance by health professionals to report an alcohol‑related diagnosis.
To properly interpret the data, the following characteristics of the NHDS should be understood:
The NHDS provides a record for each sampled hospital discharge episode, not for each individual patient; therefore, an unknown portion of discharge episodes may reflect multiple hospital episodes for a single patient in a given year. Because no patient identifiers appear in the NHDS public‑use data files, it is not possible to identify records for different hospital episodes involving the same patients. Consequently, the numbers and rates reported here reflect the incidence of alcohol‑related hospital discharge episodes and not the prevalence of alcohol‑related diagnoses among individual patients.
Because NHDS data are obtained from a sample of hospital discharge episodes, any resulting estimates are subject to sampling error. The reliability of estimates is a function of sample size. NCHS guidelines, based on the NHDS sampling plan, require a minimum of 30 unweighted cases for even marginally acceptable reliability. In this report, data based on fewer than 30 NHDS records are not displayed for table cells. Estimates below 10,000, after weights are applied, may be only marginally reliable.
The NHDS methodology allows for coding up to seven different diagnoses prior to 2010 and fifteen in 2010 for each hospital discharge record. The first of these code positions contains a code for the principal (first‑listed) diagnosis. The remaining six or fourteen positions can be used to code additional diagnoses identified during the patient’s hospital stay. In this report, an any (all‑listed) diagnosis is a diagnosis that appears in any one of the maximum possible code positions for each record; counts of an any (all‑listed) diagnosis are described as counts for any mention of the particular diagnosis. The principal (first‑listed) diagnosis need not be the most serious diagnosis recorded on a discharge record, nor is it necessarily the diagnosis that accounts for the overall length of a patient’s hospital stay.
Numbers and rates based only on principal (first‑listed) diagnoses can be misleading because these data overlook other morbidity that may be diagnosed during the patient’s hospitalization. Therefore, this report presents numbers and rates for principal (first‑listed) as well as any (all‑listed) mentions of alcohol‑related diagnoses. Principal (first‑listed) diagnoses constitute a subset of any (all‑listed) diagnoses. While diagnostic categories based on principal (first‑listed) diagnoses are mutually exclusive, a given discharge may appear in more than one category based on any (all‑listed) diagnoses. However, a hospital discharge with multiple diagnoses in the same category is not counted more than once. For example, one diagnostic category is alcoholic psychoses (ICD‑9‑CM code 291). Under this category are eight subclassifications. A discharge with diagnoses of both alcohol withdrawal delirium (code 291.0) and alcohol withdrawal hallucinosis (code 291.3) would be counted only once under the overall alcoholic psychoses classification even though more than one type of alcoholic psychosis appears on the record.
Since 1985, two data collection procedures have been used in conducting the survey: (1) a manual system of sample selection and data abstraction and (2) an automated method that involves the purchase of data files from abstracting service organizations. An increasing proportion of respondent hospitals have employed the automated method. Prior to 1985, all data were collected manually. In 1987 approximately one‑sixth (17 percent) of sample hospitals used the automated method. But by 2010, 48 percent of these hospitals were using the automated method (NCHS 2012). The variability in data collection procedures may have systematic effects on morbidity trend data.
The change in sample design implemented in 1988 implies a discontinuity in time‑series data. Therefore, the reader must use caution when interpreting trends across the old and the new sample periods.
Because of funding limitations, NHDS reduced the sample size of hospitals by half beginning in 2008. As a result, error estimates for statistics produced by the survey have generally increased, and in some cases, the relative standard errors have doubled.
METHOD
Definitions
A major methodological issue of this report is the specification of the categories of alcohol‑related diagnoses. The level of diagnostic detail defined in the ICD‑9‑CM and available in the NHDS is so great that the most detailed classification of morbidity results in diagnostic categories with very few observations. To minimize the problem of small cell sizes, detailed diagnostic classifications from the NHDS are reported under four major alcohol‑related categories, with three subcategories for chronic liver disease and cirrhosis. These categories (and the associated specific alcohol‑related diagnoses) are listed in the table of definitions below.
Definition of Alcohol‑Related Diagnoses
| Category Used in Report | Classification in ICD‑9‑CM |
| Alcoholic psychoses | 291.0 Alcohol withdrawal delirium |
| Alcohol dependence syndrome | 303.0 Acute alcohol intoxication |
| Nondependent abuse of alcohol | 305.0 Alcohol abuse |
Chronic liver disease and cirrhosis: Alcoholic cirrhosis of the liver |
571.0 Alcoholic fatty liver 571.4 Chronic hepatitis 571.5 Cirrhosis of liver without mention of alcohol |
For chronic liver disease and cirrhosis, the ICD‑9‑CM allows for a distinction between diagnoses with and without mention of alcohol. AEDS has chosen not to emphasize this distinction and has reported all liver cirrhosis in analyses of alcohol‑related morbidity and mortality. This practice was adopted at the recommendation of health professionals and epidemiologists who attended a conference sponsored by AEDS in 1979. In keeping with this practice, this report includes an overall category of chronic liver disease and cirrhosis that does not distinguish between cirrhosis with and without mention of alcohol. For consistency with causes of death reported in other AEDS publications on cirrhosis mortality (Yoon and Yi 2012), this report also includes three subcategories of cirrhosis: (1) alcoholic cirrhosis of the liver, (2) other specified cirrhosis of the liver without mention of alcohol, and (3) unspecified cirrhosis of the liver without mention of alcohol.
This report presents data for the U.S. population ages 15 and older in the following age categories: 15 to 24, 25 to 44, 45 to 64, and 65 and older. Age 15 is below the minimum legal drinking age in all 50 States and the District of Columbia, but survey results show that a large number of adolescents drink alcoholic beverages. For example, data from the NIAAA 2001–2002 National Epidemiologic Survey on Alcohol and Related Conditions indicate that 12.2 percent of current drinkers ages 18 and older in the United States began drinking at age 15 or younger (Alcohol Epidemiologic Data System 2004).
Exclusions
In assessing alcohol‑related diagnoses, discharges of females with deliveries should be excluded because childbirth is not an illness. In a typical year, approximately 12 to 13 percent of all hospital discharges are for delivery. In 2007 this category accounted for 12 percent of all discharges (Hall et al. 2010). Accordingly, to examine the share of all discharges associated with a principal (first‑listed) or an any (all‑listed) alcohol‑related diagnosis in figure 4, the percentages were calculated after excluding from both the numerator and denominator all records coded in the principal (first‑listed) diagnosis as V27, a supplementary ICD‑9‑CM classification for females delivering babies.
Assessment of Statistical Significance
Because data on hospital discharges are based on a sample of all discharges, there is some sampling error in the estimates presented in this report. To assess the statistical significance of apparent differences in the estimates, we have used variance estimation procedures recommended by the NCHS to develop confidence intervals for each estimate.
The confidence intervals presented in figures 5–9 are based on RSEs,
which are standard errors expressed as a percentage of the estimate. NCHS
provides a method for obtaining RSEs in the technical documentation of the
annual National Hospital Discharge Survey summary report for principal
(first‑listed) and any (all‑listed) diagnoses and for length of
stay (NCHS 2010a, 2010b, 2011, 2012). To obtain a 95‑percent confidence
interval, the RSE value was multiplied by the estimate and by 1.96, and the
resulting value was then added to and subtracted from the estimate.
REFERENCES
Alcohol Epidemiologic Data System. Unpublished data from the 2001–2002 National Epidemiologic Survey on Alcohol and Related Conditions, National Institute on Alcohol Abuse and Alcoholism, 2004.
Commission on Professional and Hospital Activities. The International Classification of Diseases, Ninth Revision, Clinical Modification. Ann Arbor, MI, 1978.
Hall, M. J.; DeFrances, C. J.; Williams, S. N.; Golosinskiy, A.; and Schwartzman, A. National Hospital Discharge Survey: 2007 summary. National Health Statistics Reports, No. 29. Hyattsville, MD: National Center for Health Statistics, 2010.
Haupt, B.J., and Kozak, L.J. Estimates From Two Survey Designs: National Hospital Discharge Survey. National Center for Health Statistics. Vital and Health Statistics, Series 13, No. 111. Washington, DC: U.S. Government Printing Office, 1992.
National Center for Health Statistics (W.R. Simmons and G.A. Schnack). Development of the Design of the NCHS Hospital Discharge Survey. Vital and Health Statistics. Series 2, No. 39. U.S. Department of Health, Education and Welfare Publication No. (HRA) 77‑1199. Health Resources Administration. Washington, DC: U.S. Government Printing Office, 1977.
National Center for Health Statistics. 1979–2007 National Hospital Discharge Survey (NHDS) Multi‑Year Public Use Data File Documentation. Hyattsville, MD: NCHS, 2010a.
National Center for Health Statistics. 2008 National Hospital Discharge Survey (NHDS) Public Use Data File Documentation. Hyattsville, MD: NCHS, 2010b.
National Center for Health Statistics. 2009 National Hospital Discharge Survey (NHDS) Public Use Data File Documentation. Hyattsville, MD: NCHS, 2011.
National Center for Health Statistics. 2010 National Hospital Discharge Survey (NHDS) Public Use Data File Documentation. Hyattsville, MD: NCHS, 2012.
Office of the Secretary, HHS. Administrative Simplification: Adoption of a Standard for a Unique Health Plan Identifier; Addition to the National Provider Identifier Requirements; and a Change to the Compliance Date for ICD–10–CM and ICD–10–PCS Medical Data Code Sets. Federal Register 77(74): 22950–23005, 2012.
World Health Organization. Manual of the International Statistical Classification of Diseases, Injuries, and Causes of Death, Ninth Revision. Geneva: World Health Organization, 1977.
Yoon, Y.H., and Yi, H. Surveillance Report #93: Liver Cirrhosis
Mortality in the United States, 1970–2009. Bethesda, MD: National
Institute on Alcohol Abuse and Alcoholism, Division of Epidemiology and
Prevention Research, 2012.
List of Figures
Figure 1. Percent distribution of principal (first‑listed) diagnoses among discharges with principal (first‑listed) mention of an alcohol‑related diagnosis, 2010.
Figure 2. Trends in percent distribution of principal (first‑listed) diagnoses among discharges with principal (first‑listed) mention of an alcohol‑related diagnosis, 1979–2010.
Figure 3. Percent distribution of principal (first‑listed) diagnoses among discharges with any (all‑listed) mention of an alcohol‑related diagnosis, 2010.
Figure 4. Trends in percent of discharges with principal (first‑listed) or any (all‑listed) mention of an alcohol‑related diagnosis among all discharges, 1979–2010.
Figure 5. Rates and 95‑percent confidence intervals for discharges with principal (first‑listed) mention of an alcohol‑related diagnosis, 1979–2010.
Figure 6. Rates and 95‑percent confidence intervals for discharges with any (all‑listed) mention of an alcohol‑related diagnosis, 1979–2010.
Figure 7. Rates and 95‑percent confidence intervals for discharges with principal (first‑listed) mention of specific alcohol‑related diagnoses, 1979–2010.
Figure 8. Rates and 95‑percent confidence intervals for discharges with any (all‑listed) mention of specific alcohol‑related diagnoses, 1979–2010.
Figure 9. Average length of stay and 95‑percent confidence intervals for discharges with principal (first‑listed) mention of specific alcohol‑related diagnoses, 1979–2010.
List of Tables
Table 1. Number and rate of discharges with principal (first‑listed) mention of an alcohol‑related diagnosis for U.S. population ages 15 years and older by sex and age group, 1979–2010.
Table 2. Number and rate of discharges with any (all‑listed) mention of an alcohol‑related diagnosis for U.S. population ages 15 years and older by sex and age group, 1979–2010.
Table 3. Average length of stay (in days) for discharges with principal (first‑listed) mention of an alcohol‑related diagnosis by sex and age group, 1979–2010.
Figure 1. Percent distribution of principal (first‑listed) diagnoses among discharges with principal (first‑listed) mention of an alcohol‑related diagnosis, 2010.
Figure 2. Trends in percent distribution of principal (first‑listed) diagnoses among discharges with principal (first‑listed) mention of an alcohol‑related diagnosis, 1979–2010.
Note: Shaded area represents the period before implementation of a new sample
design that may affect the trend data. Caution should be taken when assessing
differences between the old and new sample design periods.
Data for figure 2 are presented in the following page.
Figure 3. Percent distribution of principal (first‑listed) diagnoses among discharges with any (all‑listed) mention of an alcohol‑related diagnosis, 2010.
Figure 4. Trends in percent of discharges
with principal (first‑listed) or any (all‑listed) mention of an
alcohol‑related diagnosis among all discharges, 1979–2010.
Note: Shaded area represents the period before implementation of a new sample
design that may affect the trend data. Caution should be taken when assessing
differences between the old and new sample design periods.
Data for figure 4 are presented in the following page.
Figure 5. Rates and
95‑percent confidence intervals for discharges with principal
(first‑listed) mention of an alcohol‑related diagnosis,
1979–2010.
[Vertical axes reflect rates per 10,000 population: scale is not uniform for
all graphs]
[Shaded area represents the period before implementation of a new sample design that may affect the trend data. Caution should be taken when assessing differences between the old and new sample design periods.]
Data for figure 5 are presented in Table
1.
Figure 6. Rates and 95‑percent confidence intervals for discharges with any (all‑listed) mention of an alcohol‑related diagnosis, 1979–2010.
[Vertical axes reflect rates per 10,000 population: scale is not uniform for all graphs]
[Shaded area represents the period before implementation of a new sample design that may affect the trend data. Caution should be taken when assessing differences between the old and new sample design periods.]
Data for figure 6 are presented in Table
2.
Figure 7. Rates and 95‑percent confidence intervals for discharges with principal (first‑listed) mention of specific alcohol‑related diagnoses, 1979–2010.
[Vertical axes reflect rates per 10,000 population: scale is not uniform for all graphs]
[Shaded area represents the period before implementation of a new sample design that may affect the trend data. Caution should be taken when assessing differences between the old and new sample design periods.]
Data for figure 7 are presented in Table
1.
Figure 8. Rates and 95‑percent confidence intervals for discharges with any (all‑listed) mention of specific alcohol‑related diagnoses, 1979–2010.
[Vertical axes reflect rates per 10,000 population: scale is not uniform for all graphs]
[Shaded area represents the period before implementation of a new sample design that may affect the trend data. Caution should be taken when assessing differences between the old and new sample design periods.]
Data for figure 8 are presented in Table
2.
Figure 9. Average length of stay and 95‑percent confidence intervals for discharges with principal (first‑listed) mention of specific alcohol‑related diagnoses, 1979–2010.
[Vertical axes reflect average length of stay in days: scale is not uniform for all graphs]
[Shaded area represents the period before implementation of a new sample design that may affect the trend data. Caution should be taken when assessing differences between the old and new sample design periods.]
Data for figure 9 are presented in Table
3.