Professional Documents
Culture Documents
ated health care utilization in patients with acute pancreatitis (AP). health care utilization on a national level, but also whether ad-
Methods: We identified adult patients with primary discharge diagnosis vances in its management are having an impact on mortality. With
of AP from the National Inpatient Sample database. Patients with chronic increased emphasis on high-value health care, it is important to
pancreatitis and/or pancreatic cancer were excluded. Primary outcomes in- continually gauge current trends in cost of management while en-
cluded age-adjusted incidence of AP and in-hospital mortality based on US suring economic pressure does not result in worse outcomes.
standard population derived from the 2000 census data. Secondary out- We interrogated data from the National (Nationwide) Inpa-
comes were length of stay, inflation-adjusted hospital costs in 2014 US dol- tient Sample (NIS) pertaining to AP from 2001 to 2014 to include
lars, and procedural rates. Subgroup analysis included disease etiologies, rate of primary discharge diagnosis of AP, in-hospital mortality,
age, race, sex, hospital region, hospital size, and institution type. length of stay (LOS), and hospital cost, with subgroup analysis
Results: From 2001 to 2014, the rate of primary discharge diagnosis for within etiology, US geographic region, age, hospital bed size,
AP increased from 65.38 to 81.88 per 100,000 US adults per year. In- and institution type. By analyzing epidemiologic data of AP in
hospital case fatality decreased from 1.68% to 0.69%. Mortality rate is the United States, this study may elucidate and encourage further
higher in patients with AP who are older than 65 years (3.4%). Length of research as to why certain trends exist.
stay decreased, with a median of 3.8 days; cost per hospitalization de-
creased since 2007 from $7602 to $6766 in 2014.
Conclusions: The rate of hospitalization related to AP in the United MATERIALS AND METHODS
States continues to increase. Mortality, length of stay, and cost per hospital-
ization decrease. The increase in volume of hospitalization might contrib- Study Population
ute to an overall increase in health care resource utilization. Hospitalization data were abstracted from the NIS from 2001
Key Words: acute pancreatitis, cost, epidemiology, incidence, mortality to 2014. The NIS is part of a family of databases within the
Healthcare Cost and Utilization Project sponsored by the Agency
(Pancreas 2019;48: 548–554)
for Healthcare Research and Quality. The NIS is the largest pub-
licly available, all-payer inpatient database in the United States
and uses stratified random sampling that, when weighted, estimates
FIGURE 1. National volume and rate of AP per 100,000 US adults (Ptrend < 0.001).
hospital characteristics that included location and teaching status proportions as the design correction. All linear trend analyses
(rural, urban nonteaching, urban teaching), region (Northeast, were conducted using Spearman rank correlation on yearly esti-
Midwest, South, West), and bed size (small, medium, large). We mates (also known as Daniels test for trend). When appropriate,
excluded hospitalizations in which the patient was younger than Wilson confidence intervals (CIs) are provided for categorical
18 years, transferred from another hospital, or died on the day outcomes, whereas CIs for LOS and hospital cost were obtained
of admission; we also excluded hospitalizations with nonpositive from log-linear regression models due to nonnormality and
NIS weights and those missing any patient- or hospital-level charac- heteroscedastic residuals. All analyses accounted for the NIS
teristics. This study was considered exempt research by the Creighton sampling design using updated hospital trend weights. SAS version
University Institutional Review Board (IRBNet ID 1108005-1). 9.4 (SAS Institute, Cary, NC) was used for statistical analysis, and
P < 0.05 indicates statistical significance.
Outcomes
Primary outcomes included year-over-year trends in hospital- RESULTS
izations related to AP incidence per 100,000 US population. Inci- From 2001 to 2014, there was an estimated 2,368,780 hospi-
dence was calculated by dividing the NIS weighted frequency talizations in the United States with a primary discharge diagnosis
estimate by the midyear census estimates for a given year (as pro- of AP (95% CI, 2,298,204–2,439,357). Regarding the etiology of
vided by the US Census Bureau)6 and multiplying that quotient by these AP hospitalizations, 527,653 (22.5%) had an alcohol abuse
100,000. Secondary outcomes included in-hospital mortality, diagnosis, 533,253 (22.7%) had gallstones, 83,836 (3.7%) had
LOS, inflation-adjusted hospital costs in 2014 US dollars,7 and hypertriglyceridemia, and 70,199 (3.0%) had more than one of
procedural rates (see Supplementary Table 1, http://links.lww. these diagnoses; the remaining 1,134,428 (48.3%) hospitaliza-
com/MPA/A714, for procedures and ICD-9-CM codes). For all tions did not have any of the aforementioned diagnoses.
outcomes, we conducted subgroup analyses across AP etiologies
and demographic characteristics. Hospitalizations Related to AP
The national volume of AP-related hospitalizations increased
Statistical Analysis from 138,795 in 2001 to 200,600 in 2014, with overall rate of hos-
Continuous variables are presented as mean ± SE and com- pitalizations due to AP increasing from 65.4 per 100,000 US
pared using linear regression to allow for NIS weighting, whereas adults in 2001 to 81.9 per 100,000 US adults in 2014
categorical variables are presented as frequency and percent and (Ptrend < 0.001; Fig. 1). As shown in Table 1, from 2001 to 2014,
compared using the Rao-Scott χ2 test using the estimated AP with an alcohol abuse diagnosis showed an increase from
2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 Ptrend
Alcohol 15.0 15.4 14.6 15.7 15.3 16.4 16.1 18.4 18.9 21.0 21.2 22.7 22.6 23.4 <0.001
Biliary 15.8 16.4 17.0 17.1 17.1 18.2 17.3 18.0 18.8 20.1 20.3 20.1 19.5 19.3 <0.001
Biliary + ERCP 1.8 1.7 1.6 1.3 1.2 1.1 0.8 0.7 0.6 0.6 0.5 0.5 0.5 0.4 <0.001
Hypertriglyceridemia 2.1 2.4 2.6 2.6 2.9 2.8 3.0 3.6 4.1 4.6 4.8 5.0 4.6 4.9 <0.001
No etiology 33.8 31.2 32.6 35.1 33.6 33.5 31.0 36.2 36.0 38.3 38.3 39.3 38.5 37.6 <0.001
No etiology + ERCP 1.5 1.1 1.0 0.9 0.8 0.7 0.5 0.5 0.4 0.4 0.3 0.4 0.3 0.3 <0.001
Data reported as number of patients per 100,000 US adults.
© 2019 Wolters Kluwer Health, Inc. All rights reserved. www.pancreasjournal.com 549
15.0 to 23.4 per 100,000 US adults, biliary AP increased from and those older than 64 years, which coincided with the higher
15.8 to 19.3 per 100,000 US adults, and AP with hypertriglyc- rates observed in Medicare beneficiaries. Acute pancreatitis with
eridemia increased from 2.1 to 4.9 per 100,000 US adults (all hypertriglyceridemia was more frequent in those aged 35 to 44 years,
Ptrend < 0.001; Table 1). Further, the use of endoscopic retrograde males, Hispanics, the privately insured, and in facilities in the
cholangiopancreatography (ERCP) both in biliary AP and in AP Western United States.
in the absence of the etiologies under study also decreased sub-
stantially (both Ptrend < 0.001). Mortality, LOS, Hospital Cost, and Procedures
As shown in Table 2, AP with an alcohol abuse diagnosis Although the rates of AP-related hospitalizations increased
occurred more frequently in those aged 25 to 54 years, males, from 2001 to 2014, the overall mortality rate in these patients de-
non-Hispanic blacks, and those in which the primary payer was creased from 1.7% to 0.7% (Ptrend < 0.001). As shown in Figure 2,
Medicaid, self-pay, or another form of insurance. Biliary AP similar trends were observed with a concurrent diagnosis of alco-
occurred more frequently in females, those aged 18 to 24 years, hol abuse or gallstones (both Ptrend < 0.001), but no statistically
550 www.pancreasjournal.com © 2019 Wolters Kluwer Health, Inc. All rights reserved.
significant trend was observed in the presence of hypertriglyc- presence of biliary pancreatitis, in those older than 54 years, in
eridemia (Ptrend = 0.503). When considering demographic charac- Hispanics and those of other races, those in the highest income
teristics, mortality rates were higher in whites and Medicare quartile, and in facilities in the Northeast and West (Table 3). An
beneficiaries and continually increase with age with greatest mor- estimated 99,886 hospitalizations included a pancreas-specific
tality in those 65 years or older (Table 2). procedure (95% CI, 96,245–103,528). The procedure rate was
In addition, median LOS decreased from 4.4 days in 2001 to highest in hospitalizations with a concurrent alcohol abuse di-
3.4 days in 2014 (Ptrend < 0.001); similar rates of decrease were agnosis, in those aged 35 to 64 years, males, in those of white
observed within concurrent diagnoses of alcohol abuse, gallstones, or other races, in those living in higher-income zip codes, and
and hypertriglyceridemia (Fig. 3; all Ptrend < 0.001). Median LOS in large or urban teaching facilities (Table 3).
for the entire cohort over the duration of the study was 3.8 days.
Longer lengths of stay were observed in Medicare beneficiaries,
the Northeast region, urban teaching facilities, and facilities with DISCUSSION
large bed sizes (Table 3). This study explores the recent trends of AP in the United
Median hospital cost (in 2014 US dollars) showed a nonlin- States from 2001 to 2014 using a national database of hospital dis-
ear trend as it increased from $6820 in 2001 to $7602 in 2007 charges. Our results show a consistent increase in the rate of hos-
(Ptrend < 0.001 for increase) and then decreased to $6766 in pitalizations related to AP. The general incidence trends observed
2014 (Ptrend < 0.001 for decrease). As shown in Figure 4, a similar in our study are in keeping with other previously reported studies.
increasing followed by decreasing trend was observed within con- Brown et al3 studied NIS trends from 1997 to 2003, which showed
current diagnoses of alcohol abuse and gallstones (Ptrend < 0.001 an increase in frequency of primary discharge diagnosis of AP of
for increase and decrease); the trend for AP with hypertriglyc- 30.2% (from 493 per 100,000 to 642 per 100,000). In another re-
eridemia was mostly flat from 2004 to 2010 (Ptrend < 0.001 for in- view in trends of AP from 1998 to 2006, NIS data from Singla
crease and 0.030 for decrease). In general, costs were higher in the et al8 showed a continued annual increase in incidence of AP, with
© 2019 Wolters Kluwer Health, Inc. All rights reserved. www.pancreasjournal.com 551
TABLE 3. Crude Mortality Rate, LOS, Hospital Cost, and Procedure Rate by Clinical and Demographic Characteristics
incidence rising 29.6% from 1998 to 2006. A study using the Na- Because the NIS database is a stratified random sample, our re-
tional Hospital Discharge Survey, conducted by Fagenholz et al9 from sults are likely indicative of the incidence rate of AP, given that
1988 to 2003, noted the number of cases as rising from 101,000 to a vast majority of patients with AP have to be hospitalized.
201,000 during that study period. A recent study reporting AP data There was an overall trend of decreasing mortality rate from
until 2012 using the NIS without exclusion of chronic pancreatitis 1.7% in 2001 to 0.7% in 2014, and the mortality benefit was seen
or pancreatic malignancy also showed that the rates of AP-related among patients with AP of all etiologies. Overall, biliary pancre-
hospitalization are rising (15% from 2003 to 2012).1 Our study dem- atitis AP was associated with higher mortality compared with
onstrates that the increase in rate of hospitalization due to AP persists. other etiologies. Although the mortality in biliary pancreatitis
552 www.pancreasjournal.com © 2019 Wolters Kluwer Health, Inc. All rights reserved.
did decrease from 2001 to 2007, improvements in subsequent deidentified nature of the data, there was no ability to obtain fur-
years were less notable. The reason for this is unclear. When con- ther clinical information on patients. Instead, this study makes
sidering demographic patterns related to mortality, increasing age use of ICD-9 codes, which are subject to documentation error of
is associated with significant increase in mortality with relative physicians as well as errors and bias by billing personnel.12 Stud-
mortality increasing from 0.1 to 3.4 in the higher age group. It ies determining the accuracy of billing codes for AP are limited,
is very likely that the increase in mortality is due to the presence although a previous abstract in a Veterans Administration hospital
of other comorbid conditions.8,10 Racial comparison indicated setting has indicated a positive predictive value of 62% for ICD-9
that whites were at higher risk of mortality from AP compared coding of inpatient AP as primary diagnosis.13 Another limitation
with blacks or Hispanics. The clinical significance of this is un- of this study is the inability to identify patients and measure trends
clear as a previous study also suggested increased rate of mortality in patients with severe AP, especially because most mortality and
in African Americans with AP.11 Academic medical centers also morbidity in AP occur in severe AP.
appeared to have higher mortality rate, cost, number of proce- Strengths of this study are the sample size and wide geo-
dures, and longer LOS; however, this probably relates to a more graphic distribution of the NIS database. The conclusions gener-
complex patient population being managed in these facilities. ated from the data have the potential to be generalizable as data
In reviewing primary discharge diagnosis of AP by etiology, are derived from across a cross section of population. This con-
the rates of alcoholic, biliary, and hypertriglyceridemia-related trasts with many other epidemiologic studies that include data
AP have all increased. Although AP with hypertriglyceridemia from a small geographic area, such as a single state or multicenter
accounted for only a small portion of all AP discharges, it nearly studies where national generalizability is questionable.
doubled over the 14-year study period. This is probably a reflec- There are several favorable trends with regard to mortality,
tion of the obesity epidemic. Etiology of AP for a sizeable num- LOS, and lower cost per hospitalization observed in patients with
ber of patients was not identified based on information contained AP in the NIS. This is probably a reflection of improvement in the
in this database, probably due to lack of accurate coding. In light management of AP. However, the number of hospitalizations re-
of this, it is difficult to draw accurate conclusions regarding the lated to AP continues to increase that in turn could contribute to
trends of AP etiology in this study. Over the 14-year period, overall increase in health care spending for this condition. This
the total median LOS decreased by 23.3% from 4.4 days in study hopefully will create an impetus for further research that
2001 to 3.4 days in 2014. During this period, the cost per hospital- will help reduce hospital costs and disease burden related to AP.
ization related to AP also reduced to around $6766 in 2014 com-
pared with 2007. However, biliary pancreatitis and AP from REFERENCES
hypertriglyceridemia were associated with higher hospitalization-
1. Peery AF, Crockett SD, Barritt AS, et al. Burden of gastrointestinal, liver,
related cost compared with the rest of the cohort, and it is likely and pancreatic diseases in the United States. Gastroenterology. 2015;149:
due to need for interventions such as cholecystectomy and plasma- 1731–1741.e3.
pheresis in these patients. The driving factor for cost in AP-related
2. Yadav D, Lowenfels AB. Trends in the epidemiology of the first attack of
hospitalizations appears to be LOS as pointed out in a previous
acute pancreatitis: a systematic review. Pancreas. 2006;33:323–330.
study.11 It is plausible that the continued reduction in LOS is con-
tributing to reduction of cost per each hospitalization related to 3. Brown A, Young B, Morton J, et al. Are health related outcomes in acute
AP in recent years. Another significant finding of our study is that pancreatitis improving? an analysis of national trends in the US from 1997
the use of ERCP has decreased consistently over the study period. to 2003. JOP. 2008;9:408–414.
We suspect that this decrease in ERCP usage is related to use of 4. Frey CF, Zhou H, Harvey DJ, et al. The incidence and case-fatality rates of
other noninvasive diagnostic modalities such as magnetic resonance acute biliary, alcoholic, and idiopathic pancreatitis in California,
cholangiopancreatography and endoscopic ultrasound in the man- 1994–2001. Pancreas. 2006;33:336–344.
agement of patients with AP. 5. Banks PA, Bollen TL, Dervenis C, et al. Classification of acute
Limitations of this study are rooted both in its retrospective pancreatitis—2012: revision of the Atlanta Classification and definitions
nature and its dependence on administrative data. Because of the by international consensus. Gut. 2013;62:102–111.
© 2019 Wolters Kluwer Health, Inc. All rights reserved. www.pancreasjournal.com 553
6. Buter A, Imrie C, Carter C, et al. Dynamic nature of early organ 10. Frey C, Zhou H, Harvey D, et al. Co-morbidity is a strong predictor of early
dysfunction determines outcome in acute pancreatitis. Br J Surg. 2002;89: death and multi-organ system failure among patients with acute
298–302. pancreatitis. J Gastrointest Surg. 2007;11:733–742.
7. Johnson CD, Abu-Hilal M. Persistent organ failure during the first week as 11. Fagenholz PJ, Fernández-del Castillo C, Harris NS, et al. Direct medical
a marker of fatal outcome in acute pancreatitis. Gut. 2004;53:1340–1344. costs of acute pancreatitis hospitalizations in the United States. Pancreas.
8. Singla A, Csikesz NG, Simons JP, et al. National hospital volume in acute 2007;35:302–307.
pancreatitis: analysis of the nationwide inpatient sample 1998–2006. 12. O’Malley KJ, Cook KF, Price MD, et al. Measuring diagnoses: ICD code
HPB (Oxford). 2009;11:391–397. accuracy. Health Serv Res. 2005;40:1620–1639.
9. Fagenholz PJ, Castillo CF, Harris NS, et al. Increasing United States 13. Yadav D, Dhir R. How accurate are ICD-9 codes for acute (AP) and chronic
hospital admissions for acute pancreatitis, 1988–2003. Ann Epidemiol. (CP) pancreatitis?—A large VA hospital experience. Pancreas. 2006;33:
2007;17:491–497. 508.abstr.
554 www.pancreasjournal.com © 2019 Wolters Kluwer Health, Inc. All rights reserved.